Depends on which point of your career. With 18 years of experience, consulting for tech companies, I can afford to be tired of AI. I don't get paid to write boilerplate code, and avoiding anyone knocking at the door with yet another great AI-powered idea makes commercial sense, just like I have ignored everyone wanting to build the next blockchain product 5 years ago, with no major loss of income.
Also, running a bootstrapped business, I have bigger fishes to fry than playing mentor to Copilot to write a React component or generating bullshit copy for my website.
I'm not sure we need more FUD saying that the choice is between AI or unemployment.
I find comparisons between AI and blockchain very misleading.
Blockchain is almost entirely useless in practice. I have no reason to disdain it, in fact I was active in crypto around 10-12 years ago when I was younger and more excited about tech than now, and I had fun. But the fact is that the utility that it has brought to most of society is essentially to have some more speculative assets to gamble on, at ludicrous energy and emissions costs.
Generative AI, on the other hand, is something I'm already using almost every day and it's saving me work. There may be a bubble but it will be more like the dotcom bubble (i.e., not because the tech is useless, but because many companies jump to make quick bucks without even knowing much about the tech).
I mean, to be selfish at apparently a dicey point in history, go ahead and FUD and get people to believe this.
None of my useful work is AI-able, and some of the useful work is towards being able to stand apart from what is obviously generated drivel. Sounds like the previous poster with the bootstrapped business is in a similar position.
Apparently AI is destroying my potential competition. That seems unfair, but I didn't tell 'em to make such an awful mistake. How loudly am I obliged to go 'stop, don't, come back'?
When it's for sale everywhere (I cannot buy one) and people trust it, all cab drivers will be gone. Unemployed will depend on the resilience, but unlike cars replacing coach drivers, there is not really a similar thing a cab driver can pivot to.
Yes, we can imagine a future where all cab drivers are unemployed, replaced by autonomous driving. However, we don't know when this will happen, because autonomous driving is a much harder problem than the hype from a few years ago suggested. There isn't even proof that autonomous driving will ever be able to fully replace human drivers.
You’re feeling tired of AI, but let’s delve deeper into that sentiment for a moment. AI isn’t just a passing trend—it’s a multifaceted tool that continues to elevate the way we engage with technology, knowledge, and even each other. By harnessing the capabilities of artificial intelligence, we allow ourselves to explore new frontiers of creativity, problem-solving, and efficiency.
The interplay between human intuition and AI’s data-driven insights creates a dynamic that enriches both. Rather than feeling overwhelmed by it, imagine the opportunities—how AI can shoulder the burdens of mundane tasks, freeing you to focus on the more nuanced, human elements of life.
As an aside its really interesting how the human brain can so easily read an AI essay and realize its AI. You would think that with the vast corpus these models were trained on there would be a more human sounding voice.
Maybe it's overfitting or maybe just the way models work under the hood but any time I see AI written stuff on twitter, reddit, linkedin its so obvious its almost disgusting.
I guess its just the brain being good at pattern matching, but it's crazy how fast we have adapted to recognize this.
It's the RLHF training to make them squeaky clean and preternaturally helpful. Pretty sure without those filters and with the right fine-tuning you could have it reliably clone any writing style.
These models are not trained to act like a single human in a conversation, they're trained to be every participant and their average.
Every instance of a human choosing not to engage or speak about something - because they didn't want to or are just clueless about the topic, is not part of their training data. They're only trained on active participants.
Of course they'll never seem like a singular human with limited experiences and interests.
The output of those AIs is akin to products and software designed for the "average" user - deep inside uncanny valley, saying nothing specifically, having no specific style, conveying no emotion and nothing to latch on to.
It's the perfect embodiment of HR/corpspeak which I think its so triggering for us (ex) corpo drones.
Everyone I know claims to be able to recognize AI text, but every paper I've seen where that ability is A/B tested says that humans are pretty bad at this.
The thing I’m tired of is elites stealing everything under the sun to feed these models. So funny that copyright is important when it protects elites but not when a billion thefts are committed by LLM folks. Poor incentives for creators to create stuff if it just gets stolen and replicated by AI.
I’m hungry for more lawsuits. The biggest theft in human history by these gang of thieves should be held to account. I want a waterfall of lawsuits to take back what’s been stolen. It’s in the public’s interest to see this happen.
I'm sorry if this is strawmanning you, but I feel you're basically saying it's in the public's interest to give more power to Intellectual Property law, which historically hasn't worked out so well for the public.
Copyright as far as I understand is focused on wholesale reproduction/distribution of works, rather than using material for generation of new works.
If something is available without contractual restriction it is available to all. Whether it's me reading a book, or a LLM reading a book, both could be considered the same.
Where the law might have something to say is around the output of said trained models, this might be interesting to see given the potential of small-scale outputs. i.e. If I output something to a small number of people, how does one detect/report that level of infringement. Does the `potential` of infringement start to matter.
Nah. What he is saying is that the existing law should be applied equally. As of now intellectual property as a right only works for you if you are a big corporation.
I think the second alternative works too: either you sue these companies to the ground for copyright infringement at a scale never seen, OR you decriminalize copyright infringement.
The problem (as far as this specific discussion goes) is not that IP laws exist, but rather that they are only being applied in one direction.
HN generally hated (and rightly so, IMO) strict copyright IP protection laws. Then LLMs came along and broke everybody's brain and turned this place into hardline copyright extremists.
What do you mean by this? All I see in this thread is people who have absolutely no legal background who are 100% certain that copyright law works how they assume it does and are 100% wrong.
The difference is that before, intellectual property law was used by corporations to enrich themselves. Now intellectual property law could theoretically be used to combat an even bigger enemy: big tech stealing all possible jobs. It's just a matter of practicality, like all law is.
> you're basically saying it's in the public's interest to give more power to Intellectual Property law
Not necessarily. An alternative could be to say that all models trained on data which hasn't been explicitly licensed for AI-training should be made public.
The usual argument is less about piracy as a term and more the use of the word theft, and your use of the word "taking". When we talk about physical things theft and taking mean depriving the owner of that thing.
If I have something, and you copy it, then I still have that thing.
Did you read that original comment and wonder how Sam Altman and his crew broke into the commenter's home and made off with their hard drive? Probably not and so theft was a fine word choice. It communicated exactly what they wanted to communicate.
Even if that's the case, the disagreement is in semantics. Let's take your definition of theft. There's physical theft (actually taking something) and there's digital theft (merely copying).
The point of anti-copyright advocates are that merely copying is not ethically wrong. In fact, Why Software Must Be Free made the argument that preventing people from copying is ethically wrong because it limited the spread of culture and reuse.
That is the crux of the disagreement. You may rephrase our argument as "physical theft may be bad, but digital theft is not bad, and in fact preventing digital theft is in itself bad", but the argument does not change.
Of course, there is additional disagreement in the implied moral value of the word "theft". In that case I agree with you that pro-copyright/anti-AI advocates have made their point by the usage of that word. Of course, we disagree, but... it is what it is I suppose.
You calling it piracy is already a moral stance. Copying data isn't morally wrong in my opinion, it is not piracy and it is not theft. It happens to not be legal but just a few short years ago it was legal to marry infants to old men and you could be killed for illegal artifacts of witchcraft. Legality and morality are not the same, and the latter depends on personal opinion.
I agree with you they're not the same, but to build on that, I would add that they're not entirely orthogonal either, they influence each other a lot. Generally morallity that a society agrees on gets enforced as laws.
The only entities that will win with these lawsuits are the likes of Disney, large legacy news media companies, Reddit, Stack Overflow (who are selling content generated by their users), etc.
Who will also win: Google, OpenAI and other corporations that enter exclusive deals, that can more and more rely on synthetic data, that can build anti-recitation systems, etc.
And of course the lawyers. The lawyers always win.
Who will not win:
Millions of independent bloggers (whose content will be used)
Millions of open source software engineers (whose content will be used against the licenses, and used to displace their livelihood), etc.
The likes of Google and OpenAI entered the space by building on top of the work of the above two groups. Now they want to pull up the ladder. We shouldn't allow that to happen.
Honestly the most depressing thing about this entire affair is seeing not the entire, certainly but a sizable chunk of the software development community jump behind OpenAI and company’s blatant theft on an industrial scale of the mental products of probably literally billions of people (not the least of whom is other software developers!) with absolutely not the slightest hint of concern about what that means for the world because afterwards, they got a new toy to play with. Squidward was apparently 100% correct: on balance, few care about the fate of labor as long as they get their instant gratification.
>blatant theft on an industrial scale of the mental products
They haven't been stolen; the creators still have them. They've just been copied. It's amazing how much the ethos on this site has shifted over the past decade, away from the hacker idea that "intellectual property" isn't real property, just a means of growing corporate power, and information wants to be free.
Disagree. There should be no distinction between the two. Those kind of distinctions are what cause unfair advantages. If the information is available to consume, there should be no constraint on who uses it.
Sure you might not like OpenAI, but maybe some other company comes a long and builds the next magical product using information that is freely available.
Treating corporations as "people" for policy's sake is a legal decision which has essentially killed the premise of the US democratic republic. We are now, for all intents and purposes, a corporatocracy. Perhaps an even better description would simply be oligarchy, but since our oligarchs' wealth is almost all tied up in corporate stocks, it's a very incestuous relationship.
The idea of knowledge as a source of understanding and personal growth is completely oppositional to it's conception as a scarce resource, which to OpenAI and whomever else wants to train LLMs is what it is. OpenAI did not read everything in the library because it wanted to know everything; it read everything at the library so it could teach a machine to create a statistical average written word generator, which it can then sell access to. These are fundamentally different concepts and if you don't see that, then I would say that is because you don't want to see it.
I don't care if employees at OpenAI read books from their local library on python. More power to them. I don't even care if they copy the book for reference at work, still fine. But utilizing language at scale as a scarce resource to train models is not that and is not in any way analogous to it.
I am sorry you are too blinded by your own ideology and disagreement with OpenAI to see others points of views. In my view, I do not want to constrain any person or entity on their access to knowledge, regardless of output product. I do have issues with entities or people consuming knowledge and then prevent others from doing so. I am not describing a scenario of a scarce resource but of an open one.
Public information should should be free for anyone to consume and use how they want.
> I am sorry you are too blinded by your own ideology and disagreement with OpenAI to see others points of views.
A truly hilarious sentiment coming from someone making zero effort to actually engage with what I'm saying in favor of parroting back empty platitudes.
> It's amazing how much the ethos on this site has shifted over the past decade
It hasn't. The hacker ethos is about openness, individuality, decentralization (among others).
OpenAI is open in what it consumes, not what it outputs.
It makes sense to have protections in place when your other values are threatened.
If "information want's to be free" leads to OpenAI centralizing control over the most advanced AI then will it be worth it?
A solution here would be similar to the GPL: even megacorps can use GPL software, but they have to contribute back. If OpenAI and the rest would be forced to make everything public (if it's trained on open data) then that would be an acceptable compromise.
> The hacker ethos is about openness, individuality, decentralization (among others).
Yes, the greatest things on the internet have been decentralized - Git, Linux, Wikipedia, open scientific publications, even some forums. We used to passively consume content and internet allowed interaction. We don't want to return to the old days. AI falls into the decentralized camp, the primary beneficiaries are not the providers but the users. We get help of things we need, OpenAI gets a few cents per million tokens, they don't even break even.
I'm sorry, the worlds knowledge now largely accessible by a laymen via LLMs controlled by at most, 5 companies is decentralized? If that statement is true then the world decentralized truly is entirely devoid of meaning at this point.
1. Decentralized technologies you can operate privately, freely, and adapt to your needs: computers, old internet, Linux, git, FireFox, local Wikipedia dump, old standalone games.
2. Centralized technologies that invade privacy, lead to loss of control and manipulation: web search, social networks, mobile phones, Chrome, recent internet, networked games. LLMs fall into the decentralized camp.
You can download a LLM, run it locally, fine-tune it. It is interactive, the most interactive decentralized tech since standalone games.
If you object that LLMs are mostly centralized today (upfront cost of pre-training and OpenAI popularity), I say they are still not monopolies, there are many more LLM providers than search engines and social networks, and the next round of phones and laptops will be capable of local gen-AI. The experience will be seamless, probably easier to adapt than touchscreens were in 2007.
Do you consider it theft because of the scale? If I read something you wrote and use most of a phrase you coined or an idea for the basis of a plotline in a book I write, as many authors do, currently it's counted as being all my own work.
I feel like the argument is akin to some countries considering rubbish, the things you throw away, to still be owned by your person ie "dumpster diving" is theft.
If a company had scraped public posts on the Internet and used it to compile art by colourising chunks of the text, is it theft? If an individual does it, is it theft?
This argument has been stated and re-stated multiple times, this notion that use of information should always be free, but it fails to account for the fact that OpenAI is not consuming this written resource as a source of information but rather as a tool for training LLMs, which it has been open about from the beginning is a thing it wishes to sell access to as a subscription service. These are fundamentally not the same. ChatGPT/Copilot do not understand Python, they are not minds that read a bunch of python books and learned python skills they can now utilize: they are language models, that internalized metric tons of weighted averages of python code and can now (kind of) write their own, based on minimizing "error" relative to the code samples they ingest. Because of this, Copilot has never and will never write code it hasn't seen before, and by extension of that, it must see a whole lot of code in order to function as well as it does.
If you as a developer look at how one would declare a function in python, review a few examples, you now know how to do that. Copilot can't say the same. It needs to see dozens, hundreds, perhaps thousands of them to reasonably accurately be counted on to accomplish that task, it's just how the tech works. Ergo, scaled data sets that can accomplish this teaching task now have value, if the people doing that training are working for high-valuation startups with the objective of selling access to code generating robots.
That's not necessarily my position. I think laws can evolve, but they need to be applied fairly. In this case, it's heading in a direction where only the blessed will be able to compete.
I would never have imagined hackers becoming copyright zealots advocating for lawsuits. I must be getting old but I still remember the Pirate Bay trial as if it was yesterday.
True, though paid language models are probably just a blip in history. Free weight language models are only ~12 months behind and have massive resources thanks to Meta.
That profit will be squeezed to zero over the long term if Zuck maintains his current strategy.
That can change on a dime though, if Zuck decides it's in his financial interest to change course. If Facebook stops spending billions of dollars on open models who is going to step in and fill that gap?
That depends on when Meta stops. The longer Meta keeps releasing free models, the more capabilities are made permanently unprofitable. For example, Llama 3.1 is already good enough for translation or as a writing assistant.
If Meta stopped now, there would still be profit in the market, but if they keep releasing Llamas for the next 5+ years then OpenAI et al will be fighting for scraps. Not everybody needs a model that can prove theorems.
I'm all for abolishing copyright, for everyone. Let the knowledge be free and widely shared.
But until that is the case and people running super useful services like libgen have to keep hiding then I also want all the LLM corpos to be subject to the same legal penalties.
Exactly this. If we have to live under a stifling copyright regime, then at least it should be applied evenly. It's fundamentally unfair to have one set of laws (at least as enforced in practice) for the rich and powerful and another set for everyone else.
This is the entire point of existence for the GPL. Weaponize copyright. LLMs have conveniently been able to circumvent this somehow, and we have no answer for it.
Because some people keep asserting that LLM’s “don’t count as stealing” and “how come search links are on but got reciting paywalled NYT articles on demand is bad??” Without so much as a hint of irony.
LLM tech is pretty cool.
Would be a lot cooler if its existence wasn’t predicted on the wholesale theft of everyone’s stuff, immediately followed by denial of theft, poisoning the well, and massively profiting off it.
>Because some people keep asserting that LLM’s “don’t count as stealing”
People who confidently assert either opinion in this regard are wrong. The lawsuits are still pending. But if I had to bet, I'd bet on the OpenAI side. Even if they don't win outright, they'll probably carve out enough exemptions and mandatory licensing deals to be comfortable.
You are singling out accidental replication and forgetting it was triggered with fragments from the original material. Almost all LLM outputs are original - both because they use randomness to sample, and because they have user prompt conditioning.
And LLMs are really a bad choice for infringement. They are slow, costly and unreliable at replicating any large piece of text compared to illegal copying. There is no space to perfectly memorize the majority of its training set. A 10B models is trained on 10T tokens, no space for more than 0.1% to be properly memorized.
I see this overreaction as an attempt to strengthen copyright, a kind of nimby-ism where existing authors cut the ladder to the next generation by walling off abstract ideas and making it more probably to get sued for accidental similarities.
It's not, that's playing the victim. There are hundreds or thousands of posts daily all over HN criticising capitalism. And most seem upvoted, not downvoted.
i find quite ironic whenever i see a highly upvoted comment here complaining about capitalism because for sure i don't see yc existing in any other type of economy.
This only holds if your thinking on the subject of economic systems is only as deep as choosing your character’s class in an RPG game. There’s no need for us to make every last industry a state owned enterprise and no one who’s spent longer than an hour or so contemplating such things thinks that way. I have no desire to not have a variety of companies producing things like cars, electronics, software, video games, just to name a few. Competition does drive innovation, that is still true, and having such firms vying for a limited amount of resources dispatched by individuals makes a lot of sense. Markets have their place.
However markets also have limits. A power company competing for your business is largely a farce, since the power lines to your home will not change. A cable company in America is almost certainly a functional monopoly, and that fact is reflected in their quality of service. Infrastructure of all sorts makes for piss-put markets because true competition is all but impossible, and even if it does kind of work, it’s inefficient. A customer must become knowledgeable in some way to have a ghost of a clue what they’re buying, or trust entirely dubious information from marketing. And, even if somehow everything is working up to this point, corporations are, above all, cost cutters and if you put one in charge of an area where it feels as though customers have few if any choices and the friction to change is high, they will immediately begin degrading their quality of services to save money in the budget.
And this is only from first principles, we have so many other things that could be discussed from mass market manipulation to the generous subsidies of a stunning variety that basically every business at scale enjoys to the rapacious compensation schemes that have become entirely too commonplace in the executive room, etc etc etc.
To put it short: I have no issue at all with capitalism operating in non-essential to life industries. My issue is all the ways it’s infiltrated the essential ones and made them demonstrably worse, less efficient, and more expensive for every consumer.
I would argue that markets are a necessary step towards capitalism but it's also crucial to remember that markets can also exist outside of capitalism. The accumulation of money in a society with insufficient defenses will trend towards money being a stand-in for power and influence, but it still requires the permission and legal leeway of the system in order to actually turn it corrupt; politicians have to both be permitted to, and be personally willing to accept the checks to do the corruption in the first place.
The biggest and most salient critique of liberal capitalism as we now exist under is that it requires far too many of the "right people" to be in positions of power; it presumes good faith where it shouldn't, and fails to reckon with bad actors as what they are far too often, the modern American Republican party being an excellent example (but far from the only one).
100%. Markets are a really useful tool for distributing goods and services to people and allocating resources.
In the US, IMO the problem is that markets are the primary tool for a huge number of services; take utilities for one.
There is a saying that when you have a hammer, every problem looks like a nail.
You wouldn’t see YC existing on a world full capitalist :) It depende heavily on open source, the biggest and most succeassful socialist experiment so far
In the world of the mainstream media - the default media diet of most americans excludes Hacker News, funnily enough.
All sorts of more taboo stuff gets said on hacker news. It's not exactly a cultural bellwether that ideas outside of the overton window are entertained here.
Wearing a che guevara t shirt is similarly "allowed" in public, but the last article about him in the new york times is fawning admiration for his assassin.
I'm not sure if you're being disingenuous, or if you genuinely don't understand the difference.
Pirate Bay: largely facilitating the theft of material from large corporations by normal people, for generally personal use.
LLM training: theft of material from literally _everyone_, for the purposes of corporate profit (or, well, heh, intended profit; of course all LLM-based enterprises are currently massively loss-making, and may remain so forever).
Hackers are against corporations. If breaking the copyright laws make corps bigger, more powerful and more corrupt, hackers will be against it rightfully so. Abolishing copyright is different than abusing it, we should abolish it.
On the one hand, we've got, "Pirating something because we find copyright law to be restrictive and/or corporate pricing to be excessive". On the other, we've got, "Massively wealthy people vacuuming up our creative output to further their own wealth".
And you're trying to suggest that these two are the same?
Edit: I don't mind downvotes, karma means nothing, but I do appreciate when folk speak up and say why I might be wrong. :)
Capitalism started by putting up fences around land to kick people out and keep sheep in. It has been putting fences around everything it wants and IP is one such fence. It has always been about protecting the powerful.
IP has had ample support because the "protect the little artist" argument is compelling, but it is just not how the world works.
> Capitalism started by putting up fences around land to kick people out and keep sheep in.
That's factually wrong. Capitalism is about moving wealth more efficiently: easier to allocate money/wealth to X through the banking system than to move sheep/wealth to X's farm.
> Capitalism is about moving wealth more efficiently: easier to allocate money/wealth to X through the banking system than to move sheep/wealth to X's farm.
It's not. That's what money's about. Any system with an abstract concept of money admits that it's easier to allocate wealth with abstractions than physically moving objects.
Capitalism is about capital. It's an economic system that says individuals should own things (i.e. control their purpose) by investing money (capital) into them. You attempted to correct the previous commenter, but provided an incorrect definition. I hope that clears up the relevance issue for you.
> Capitalism is about capital. It's an economic system that says individuals should own things (i.e. control their purpose) by investing money (capital) into them.
Yes. It's not about stealing land and kicking people out and raising sheep there instead. That (stealing) happens of course but is totally independent from any capitalist system.
JFC, the same sentence could have been said with communism in mind.
> You attempted to correct the previous commenter, but provided an incorrect definition. I hope that clears up the relevance issue for you.
You are confusing the intent of capitalism - which I gave the general direction of - with its definition. Does that clear up the relevance issue for you ? Did I fucking not write wealth/money intentionally ?
The OP is talking about (land) enclosure, not stealing exactly, and enclosure is a precondition of capitalism. To be fair, it's a precondition of feudalism as well. But in order to turn resources that are essentially held in common (like land, and ideas) into capital, they have to be made private. That's what happened with the literal commons in medieval Europe, and it's what's happened with the intellectual commons in the last hundred years.
You didn't describe the intent or the definition of capitalism. You just mentioned that moving resources from A to B is easier if they're abstract. The intent of capitalism is that the society is improved when people who have a vested interest (literally vested, as in with money) in a company determine its actions.
Can you calm down, please? You were the one who started saying my comment was irrelevant. I don't think it makes much sense to parrot that back to me.
Copyright law is intended to prevent people from stealing the revenue stream from someone else's work by copying and distributing that work in cases where the original is difficult and expensive to create, but easy to make copies of once created. How does an LLM do this? What copies of copyrighted work do they distribute? Whose revenue stream are they taking with this action?
I believe that all the copyright suits against AI companies will be total failures because I can't come up with a answer to any of those questions.
Here is a business model for copy right law firms:
Use source-aware training, use the same datasets as used in LLM training + copyrighted content. Now the LLM can respond not just what it thinks is most likely but also what source document(s) provided specific content. Then you can consult commercially available LLM's and detect copy right infringements, and identify copyright holders. Extract perpetrators and victims at scale. To ensure indefinite exploitation only sue commercially succesful LLM providers, so there is a constant new flux of growing small LLM providers taking up the freed niche of large LLM providers being sued empty.
The only 2 reasons big LLM providers refuse to do it is
1) to prevent a long slew of content creators filing class action suit.
2) to keep regulators in the dark of how feasible and actionable it would be, once regulators are aware they can perform the source-aware training themselves
It's the other way round. The little guys will never win, it will be just a money transfer from one large corp to another.
We should just scrap copyright and everybody plays a fair game, including us hackers.
Sue me because of breach of contract in civil court for damages because I shared your content, don't send the police and get me jailed directly.
I had my software cracked and stolen and I would never go after the users. They don't have any contract with me. They downloaded some bytes from the internet and used it.
Finding whoever shared the code without authorization is hard and even so, suing them would cost me more than the money I'm likely to get back. Fair game, you won.
It's a natural market "tax" on selling a lot of copies and earning passively.
I like the stone soup narrative on AI. It was mentioned in a recent Complexity podcast, I think by Alison Gopnik of SFI. It's analogous to the Pragmatic Programmar story about stone soup, paraphrasing:
Basically you start with a stone in a pot of water — a neural net technology that does nothing meaningful but looks interesting. You say: "the soup is almost done, but would taste better given a bunch of training data." So you add a bunch of well curated docs. "Yeah, that helps but how about adding a bunch more". So you insert some blogs, copy righted materials, scraped pictures, reddit, and stack exchange. And then you ask users to interact with the models to fine tune it, contribute priming to make the output look as convincing as possible.
Then everyone marvels at your awesome LLM — a simple algorithm. How wonderful, this soup tastes given that the only ingredients are stones and water.
The stone soup story was about sharing, though. Everyone contributes to the pot, and we get something nice. The original stone was there to convince the villagers to share their food with the travellers. This goes against the emotional implication of your adaptation. The story would actually imply that copyright holders are selfish and should be contributing what they can to the AI soup, so we can get something more than the sum of our parts.
From Wikipedia:
> Some travelers come to a village, carrying nothing more than an empty cooking pot. Upon their arrival, the villagers are unwilling to share any of their food stores with the very hungry travelers. Then the travelers go to a stream and fill the pot with water, drop a large stone in it, and place it over a fire. One of the villagers becomes curious and asks what they are doing. The travelers answer that they are making "stone soup", which tastes wonderful and which they would be delighted to share with the villager, although it still needs a little bit of garnish, which they are missing, to improve the flavor.
> The villager, who anticipates enjoying a share of the soup, does not mind parting with a few carrots, so these are added to the soup. Another villager walks by, inquiring about the pot, and the travelers again mention their stone soup which has not yet reached its full potential. More and more villagers walk by, each adding another ingredient, like potatoes, onions, cabbages, peas, celery, tomatoes, sweetcorn, meat (like chicken, pork and beef), milk, butter, salt and pepper. Finally, the stone (being inedible) is removed from the pot, and a delicious and nourishing pot of soup is enjoyed by travelers and villagers alike. Although the travelers have thus tricked the villagers into sharing their food with them, they have successfully transformed it into a tasty meal which they share with the donors.
First gen models trained on books directly. Latest Phi distilled textbook-like knowledge down from disparate sources to create novel training data. They are all fairly open about this change and some are even allowing upset publishers to confirm that their work wasn't used directly. So stones and ionized water go in the soup.
ok the "elites" have spent a lot of money training AI but have the "commoners" lifted a single finger to stop them? nope! its the job of the commoners to create a consensus, a culture, that protects people. so far all i see from the large group of people who are not a part of the elite is denial about this entire issue. they deny AI is a risk and they dont shame people who use it. 99.99% of the population is culpable for any disaster that befalls us regarding AI.
I suspect the greater issue is that copyright is not always clear in this area? I am also not sure how you prevent "elites" from using the information while also allowing the "common" person to it.
I do like how the internet has suddenly acknowledged that pirating is theft and torrenting IS a criminal activity. To your point, I'd love to see a massive operation to arrest everyone who has downloaded copyrighted material illegal (aka stolen), for the public interest.
This is such a misrepresentation of the issue and what people are saying about it. They call it "theft" because corps are, apparently-indiscriminately and without remuneration of creators, "ingesting" the original work of thousands or millions of individuals, in order to provide for-profit services derived from that ingestion/training. "Pirates", on the other hand, copy content for their own momentary entertainment, and the exploitation ends there. They aren't turning around and starting a multi-million-dollar business selling pirated content en masse.
> The thing I’m tired of is elites stealing everything under the sun to feed these models.
I suggest to apply the same to property law: make a photo and obtain instant and unlimited rights of use. – Things may change faster than we may imagine…
There is no copyright with AI unless you want to implement the same measures for humans too. I am fine with it as long as we at least get open-weights. This way you kill both copyright and any company that's trying to profit out of AI.
It's been pretty incredible watching these companies siphon up everything under the sun under the guise of "training data" with impunity. These same companies will then turn around and sic their AIs on places like Youtube and send out copyright strikes via a completely automated system with loads of false-positives.
How is it acceptable to allow these companies to steal all of this copyrighted data and then turn around and use it to enforce their copyrights in the most heavy-handed manner? The irony is unbelievable.
Perhaps what we should be pushing for is a law that would force full disclosure regarding the training corpus and require a curated version of the training data to be made available. I'm sure there would be all kinds of unintended consequences of a law like that but maybe we'd be better off starting from a strong basis and working out those exceptions. While billions have been spent to train these models, the value of the millions of human hours spent creating the content they're trained on should likewise be recognized.
I'm not. I think it's awesome and I can't wait to see what comes out next. And I'm completely OK with all of my work being used to train models. Bunch of luddites and sour grapes around here on HN these days.
Same here! Amazing stuff that I have waited for my entire life, and I won't let luddite haters ruin it for me. Their impotent rage is tiring but in the end it's just one more thing you have to ignore.
Yeah, they made something that passes a Turing test, and people on HN of all places hate it? What happened to this place? It's like the number one thing people hate around here now is another man's success.
I won't ignore them. I'll continue to loudly disagree with the losers and proudly collect downvotes from them knowing I got under their skin.
Absolutely amazing stuff. I am now three scores and ten in my life time, seen a lot of changes from slide rules->very fast to calculators->very fast to pcs, from dot matrix printers to lazer jets and dozens of other things. Wish AI was available when I was doing my PhD. If you know its limitations it can be very useful. At present I occasionally use it to translate references from wikipedia articles to bibtex format. It is very good at this, I only need to fix a few minor errors, letting me focus to the core of what I am doing. But human nature always resists change, especially if it leads to the unknown. I must admit that I think AI will bring negative consequences as it will be misused by politicians and the military, they need to be "regulated" not the AI.
You're getting downvoted, but I agree with your last sentence — and not just about AI. The amount of negativity here regarding almost everything is appalling. Maybe it's rose-tinted nostalgia but I don't remember it being like this a few years ago.
There's _a lot_ of poor quality engineers out there who understand that on some level they are committing fraud by spinning their wheels all day shifting CSS values around on a React component while collecting large paychecks. I think it's only natural all of those engineers are terrified by the prospect of some computer being capable of doing their job quickly and efficiently and replacing them. Those people are crying so loudly that it's encouraging otherwise normal people to start jumping on the anti-AI bandwagon too, because their voices are so loud people can't hear themselves think critically anymore.
I think passionate and inspired engineers who love their job and have very solid soft skills and experience working deeply on complex software projects will always have a position in the industry, and people like that are understandably very enthusiastic about AI instead of being scared of it.
In other words, it is weird how bad the status quo was, until we got something that really threatened the status quo, now a lot of the people who wanted to tear it all down are now desperately trying to stop everything from changing. The sentiment on the internet has gone in a weird direction, but it's all about money deep down. This hypothetical new status quo brought on by AI seems to be wedded to fears of less money, thus abject terror masquerading as "I'm so bored!" posturing.
You see this in the art circles, where established artists are willing to embrace AI, but it's the small time aspiring bedroom artists that have not achieved any success who are all over Twitter denouncing AI art as soulless and terrible. While the real artists are too busy using any tool available to make art, or are just making art because they want to make art and aren't concerned with fear-mongering.
> There are no shortcuts to solving these problems, it takes time and experience to tackle them.
> I’ve been working in testing, with a focus on test automation, for some 18 years now.
OK the first thought that came to my mind reading this: sounds like a opportunity to build an AI-driven product.
I've been using Cursor daily. I use nothing else. It's brilliant and I'm very happy. If I could have Cursor for Well-Designed Tests I'd be extra happy.
I kind of feel like if you're using a "Random text generator based on probability" for something that you need to trust, you're kind of holding this tool wrong.
I wouldn't complain a RNG doesn't return the numbers I want, so why complain you don't get 100% trusted output from a random text generator?
Because people provide that work without acknowledging it was created by a RNG, representing it as their own and implying some of level of assurance that it is actually true.
Its a Large LANGUAGE Model and not a Large MATHEMATICS Model. People need to learn to use the right tools for the right jobs. Also LLMs can be made more deterministic by controlling it’s “temperature”.
These amazing machines weren't consistently able to tell if an image had a bird in it or not up until like 8 years ago. If you use AI as a calculator where you need it to be precise, that's on you.
I think the issue is that: I’m not going to be using as a calculator any time soon.
Unfortunately, there’s a lot of people out there, working on a lot of products, some of which I need to use, or will be exposed to, and some of them aren’t going to have the same qualms about “language model thinks 2+2=5”.
There’s a guy on Twitter scoring how well ChatGPT models can do multiplication.
A founder at a previous workplace wanted to wholesale dump data into ChatGPT and “make it do causal analysis!!!” (Only slightly paraphrased). These tools enable some frighteningly large-scale weaponised stupidity.
I think about it differently. Before computers had to be given extremely precise and completely unambiguous instructions, now they can handle some ambiguity as well. You still have the precise output if you want it, it didn’t go away.
Btw I’m also tired of AI, but this is one thing that’s not so bad
Edit: before someone mentions fuzzy logic, I’m not talking about the input of a function being fuzzy, I’m talking about the instructions themselves, the function is fuzzy.
> You still have the precise output if you want it, it didn’t go away.
For now. Given that most new devices seem to be fully hostile to the concept of general purpose computing (see phones, VR devices, TVs, etc), I wonder how long it will be before many of the computers that are sold are even more locked down than Chromebooks - just a few prompts for interacting with a preinstalled LLM.
That's an interesting point of view. For some reason we put so much effort towards making computers think and behave like a human being, while one of the first reasons behind inventing a computer was to avoid human errors.
This is the most succinct summary of what's been gnawing at me ever since LLMs became the latest thing.
If Ilya Sutskever announced tomorrow that he'd achieved AGI, and here is its economic plan for the next 20 years, why would we have any reason to accept it over that of other human experts? It would literally be just another expert trying to tell us how to do things. And we're not short of experts, and an AGI expert has thrown away the credibility of computers as deterministically better calculators than we are.
This sounds to me like a straw man argument. Obviously 2+2 will always give you 4, in any modern LLM, and even just in the Chrome address bar.
Can you offer a real situation where we should expect the LLM to return a deterministic answer and should rightly be concerned that we're getting a stochastic one?
I'm not being facetious; I really can't think of a single good example where we need something to be deterministic and then have a reason to be disappointed about AI giving us a stochastic response.
And by nowadays you mean since ChatGPT got released, that is less than 2 years ago (e.g. a consumer preview of a frontier research project). Interesting.
But isn't this a great thing? I mean, this piece has been missing (no, I am not kidding). Computers have always had a hard time coping with situations that weren't 100% predefined.
Now, we have technology capable of handling cases that were not predefined. Yes, it makes mistakes, as do humans, but the range of problems we can solve with technology has been tremendously broadened.
The problem is how we apply AI. Currently, we throw LLMs on everything they might be able to handle without understanding how or if they have the capabilities to handle such a task. And that is not the LLM's fault but a human fault. Consequently, we see poor results, and then we blame the AI for not being able to solve a problem it wasn't designed to solve.
I quite like some parts of AI. Ray reconstruction and supersampling methods have been getting incredible and I can now play games with twice the frames per seconds. On the scietific side, meteorological predictions and protein folding have made formidable progresses thanks to it. Too bad this isn't the side of AI that is in the spotlight.
In comparision to a lot of other technologies, we actually have jumps in quality left and right, great demos, new things which are really helpful.
Its fun to watch the AI news because there is something relevant new happening.
I'm worried regarding the impact of AI but this is a billion times better than the last 10 years which was basically just cryptobros, nfts, blockchain shit which is basically just fraud.
Its not just some GenAI stuff, we talk about blind people getting better help through image analysis, we talk about alpha fold, LLMs being impressive as hell, the research currently happening.
And yes i also already see benefits in my job and in my startup.
Are you asking what field of science or what industry is interested in predicting how proteins fold?
Biotechnology and medicine probably.
Pipeline from science to application sometimes takes decades, but I'm sure you can find news of some advancements enabled by finding out short, easy to synthesize proteins that fit particular receptor to block it or that process some simplified enzymes that still process some chemicals of interest more efficiently than natural ones. Finding them would be way harde without ability to predict how a sequence of amino-acids will fold.
You'd need to actually try to manufacture them then look at them closely.
First thing that came to my mind as a possible application is designing monoclonal antibodies. Here's some paper about something relating to alpha fold and antibodies:
Are you phishing for something or are you not sure how this actually works?
Everyone who is looking for proteins (vacines, medication) need to find the right proteins for different cases. For attaching to something (antibody design), for delivering something (like another protein) or for understanding a disease (why is this protein an issue?).
The most depressing thing for me is the feeling that I simply cannot trust anything that has been written in the past 2 years or so and up until the day that I die. It's not so much that I think people have used AI, but that I know they have with a high degree of certainty, and this certainty is converging to 100%, simply because there is no way it will not. If you write regularly and you're not using AI, you simply cannot keep up with the competition. You're out. And the growing consensus is "why shouldn't you?", there is no escape from that.
Now, I'm not going to criticize anyone that does it, like I said, you have to, that's it. But what I had never noticed until now is that knowing that a human being was behind the written words (however flawed they can be, and hopefully are) is crucial for me. This has completely destroyed my interest in reading any new things. I guess I'm lucky that we have produced so much writing in the past century or so and I'll never run out of stuff to read, but it's still depressing, to be honest.
What fascinates me about your comment is that you are expressing that you trusted what you read before. For me, LLMs don't change anything. I already questioned the information before and continue to do so.
Why do you think that you could trust what you read before?
Is it now harder for you to distinguish false information, and if so, why?
scale makes all the difference. society without trust falls apart. it's good if some people doubt some things, but if everyone necessarily must doubt everything, it's anarchy.
A good part of society, the foundational part, is trust. Trust between individuals, but also trust in the sense that we expect things to behave in a certain way. We trust things like currencies despite their flaws. Our world is too complex to reinvent the wheel whenever we need to do a transaction. We must believe enough in a make-believe system to avoid perpetual collapse.
Perhaps that anarchy is the exact thing we need to convince everyone to revolt against big tech firms like Google and OpenAI and take them down by mob rule.
Debunking bullshit inherently takes more effort than generating bullshit, so the human factor is normally your big force multiplier. Does this person seem trustworthy? What else have they done, who have they worked with, what hidden motivations or biases might they have, are their vibes /off/ to your acute social monkey senses?
However with AI anyone can generate absurd torrential flows of bullshit at a rate where, with your finite human time and energy, the only winning move is to reject out of hand any piece of media that you can sniff out as AI. It's a solution that's imperfect, but workable, when you're swimming through a sea of slop.
Maybe the debunking AIs can match the bullshit generating AIs, and we will have balance in the force. Everyone is focused on the generative AIs, it seems.
No, they can't. They'll still be randomly deciding if something is fake or not, so they'll only have a probability of being correct, like all nondeterministic AI.
Debugging is harder than writing code. Once the code passed linter, compiler and test, the bugs might be more subtly logical and require more effort and intelligence.
We are all becoming QA of this super automated world.
In the past, you had to put a lot of effort to produce a text which seemed to be high quality, especially when you knew nothing about the subject. By the look of text and the usage of the words, you could tell how professional the writer was and you had some confidence that the writer knew something about the subject. Now, that is completely removed. There is no easy filter anymore.
While the professional looking text could have been already wrong, the likelihood was smaller, since you usually needed to know something at least in order to write convincing text.
> While the professional looking text could have been already wrong, the likelihood was smaller...
I don't criticise you for it, because that strategy is both rational and popular. But you never checked the accuracy of your information before so you have no way of telling if it has gotten more or less accurate with the advent of AI. You were testing for whether someone of high social intelligence wanted you to believe what they said rather than if what they said was true.
I guess the complaint is about losing this proxy to gain some assurance for little cost. We humans are great at figuring out the least amount of work that's good enough.
Now we'll need to be fully diligent, which means more work, and also there'll be way more things to review.
I'd argue people clearly don't care about the truth at all - they care about being part of a group and that is where it ends. It shows up in things like critical thinking being a difficult skill acquired slowly vs social proof which humans just do by reflex. Makes a lot of sense, if there are 10 of us and 1 of you it doesn't matter how smartypants you may be when the mob forms.
AI does indeed threaten people's ability to identify whether they are reading work by a high status human and what the group consensus is - and that is a real problem for most people. But it has no bearing on how correct information was in the past vs will be in the future. Groups are smart but they get a lot of stuff wrong in strategic ways (it is almost a truism that no group ever identifies itself or its pursuit of its own interests as the problem).
> I'd argue people clearly don't care about the truth at all
Plenty of people care about the truth in order to get advantages over the ignorant. Beliefs aren't just about fitting in a group, they are about getting advantages and making your life better, if you know the truth you can make much better decisions than those who are ignorant.
Similarly plenty of people try to hide the truth in order to keep people ignorant so they can be exploited.
> if you know the truth you can make much better decisions than those who are ignorant
There are some fallacious hidden assumptions there. One is that "knowing the truth" equates to better life outcomes. I'd argue that history shows more often than not that what one knows to be true best align with prevailing consensus if comfort, prosperity and peace is one's goal, even if that consensus is flat out wrong. The list is long of lone geniuses who challenged the consensus and suffered. Galileo, Turing, Einstein, Mendel, van Gogh, Darwin, Lovelace, Boltzmann, Gödel, Faraday, Kant, Poe, Thoreau, Bohr, Tesla, Kepler, Copernicus, et. al. all suffered isolation and marginalization of some degree during their lifetimes, some unrecognized until after their death, many living in poverty, many actively tormented. I can't see how Turing, for instance, had a better life than the ignorant who persecuted him despite his excellent grasp of truth.
You are thinking too big, most of the time the truth is whether a piece of food is spoiled or not etc, and that greatly affects your quality of life. Companies would love to keep you ignorant here so they can sell you literal shit, so there are powerful forces wanting to keep you ignorant, and today those powerful forces has way stronger tools than ever before working to keep you ignorant.
You're implying that there is an absolute Truth and that people only need to do [what?] to check if something is True. But that's not True. We only have models of how reality works, and every model is wrong - but some are useful.
When dealing with almost everything you do day by day, you have to rely on the credibility of the source of the information you have. Otherwise how could you know that the can of tuna you're going to eat is actually tuna and not some venomous fish? How do you know that you should do what your doctor told you? Etc. etc.
> You're implying that there is an absolute Truth and that people only need to do [what?] to check if something is True. But that's not True. We only have models of how reality works, and every model is wrong - but some are useful.
I am not sure I am following - you don't know if there is anything that is really true, but you presume there isn't and that model of "the only truth is the absence of truth" is useful to you because it allows you to ... what exactly?
There’s not enough time in the day to go on a full bore research project about every sentence I read, so it’s not physically possible to be “fully diligent.”
The best we can hope for is prioritizing which things are worth checking. But even that gets harder because you go looking for sources and now those are increasingly likely to be LLM spam.
Traditionally, humans have addressed the imbalance between energy-to-generate and energy-to-validate by building another system on top, such as one which punishes fraudsters or at least allows other individuals to efficiently disassociate from them.
Unfortunately it's not clear how this could be adapted to the internet and international commerce without harming some of the open-ness aspects we'd like to keep.
My new cheap proxy to save mental cost: pay to search on kagi, sort results by tracker count. My hope is fewer trackers correlates with lower incentives to seo spam. This may change but seems to work decently for now.
How do you "check the accuracy of your information" if all the other reliable-sounding sources could also be AI generated junk? If it's something in computing, like whether something compiles, you can sometimes literally check for yourself, but most things you read about are not like that.
Interesting points! Doesn't sound impossible with an AI that's wrong less often than an average human author (if the AIs training data was well curated).
I suppose a related problem is that we can't know if the human who posted the article, actually agrees with it themselves.
(Or if they clicked "Generate" and don't actually care, or even have different opinions)
>But you never checked the accuracy of your information before so
They didn't say that and that's not a fair or warranted extrapolation.
They're talking about a heuristic that we all use, as a shorthand proxy that doesn't replace but can help steer the initial navigation in the selection of reliable sources, which can be complemented with fact checking (see the steelmanning I did there?). I don't think someone using that heuristic can be interpreted as tantamount to completely ignoring facts, which is a ridiculous extrapolation.
I also think is misrepresents the lay of the land, which is that in the universe of nonfiction writing, I don't think that there's a fire hose of facts and falsehoods indistinguishable in tone. I think there's in fact a reasonably high correlation between the discernible tone of impersonal professional and credible information, which, again (since this seems to be a difficult sticking point) doesn't mean that the tone substitutes for the facts which still need to be verified.
The idea that information and misinformation are tonally indistinguishable is, in my experience, only something believed by post-truth "do you own research" people who think there are equally valid facts in all directions.
There's not, for instance, a Science Daily of equally sciency sounding misinformation. There's not a second different IPCC that publishes a report with thousands of citations which are all wrong, etc. Misinformation is out there but it's not symmetrical, and understanding that it's not symmetrical is an important aspect of information literacy.
This is important because it goes to their point, which is that something has changed, in the advent of LLMS. That symmetry may be coming, and it's precisely the fact that it wasn't there before that is pivotal.
In the past, with a printed book or journal article, it was safe to assume that an editor had been involved, to some degree or another challenging claimed facts, and the publisher also had an interest in maintaining their reputation by not publishing poorly researched or outright false information. You would also have reviewers reading and reacting to the book in many cases.
All of that is gone now. You have LLMs spitting their excrement directly onto the web without so much as a human giving it a once-over.
Writing a text of decent quality used to constitute proof of work. This is now no longer the case, and we haven't adapted to this assumption becoming invalid.
For example, when applying to a job, your cover letter used to count as proof of work. The contents are less important than the fact that you put some amount of effort in it, enough to prove that you care about this specific vacancy. Now this basic assumption has evaporated, and job searching has become a meaningless two-way spam war, where having your AI-generated application selected from hundreds or thousands of other AI-generated applications is little more than a lottery.
This. I am very picky about how I use ML still, but it is unsurpassed as a virtual editor. It can clean up grammar and rephrase things in a very light way, but it gives my prose the polish I want. The thing is, I am a very decent writer. I wrote professionally for 18 years as a part of my job delivering reports of high quality as my work product. So, it really helps that I know exactly what “good” looks like by my standards. ML can clean things up so much faster than I can and I am confident my writing is organic still, but it can fix up small issues, find mistakes, etc very quickly. A word change here or there, some punctuation, that is normal editing. It is genuinely good at light rephrasing as well, if you have some idea of what intent you want.
When it becomes obvious, though, is when people let the LLM do the writing for them. The job search bit is definitely rough. Referrals, references, and actual accomplishments may become even more important.
As usual, LLMs are an excellent tool when you already have a decent understanding of the field you're interested in using them in. Which is not the case of people posting in social media or creating their first programs. That's where the dullness and noise come from.
The noise ground has been elevated 100x by LLMs. It was already bad before but it's accelerated the trend.
So, yes, we should have never been trusting anything online but before LLMs we could rely on our brains to quickly identify the bad. Nowadays, it's exhausting. Maybe we need a LLM trained on spotting LLMs.
This month, I, with decades of experience, used Claude Dev as an experiment to create a small automation tool. After countless manual fixes, it finally worked and I was happy. Until I gave thr whole thing a decent look again and realized what a piece of garbage I had created. It's exhausting to be on the lookout for these situations. I prefer to think things through myself, it's a more rewarding experience with better end results anyway.
LLM's are a great onramp to filling in knowledge that may have been lost to age or updated to their modern classification. For example, I didn't know Hokkien and Haka are distinct linguistic branches within the Sino-Tibetan language and warrants more (personal) research into the subject. And all this time, without the internet, we often just colloquially called it Taiwanese.
It is lost in a sense that you had no idea about such possibility and you did not know to search it in the first hand, while I believe that in this case LLM brought it up as a side note.
Such fortuitous stumblings happen all the time without LLMs (and in regular libraries, for those brave enough to use them). It's just the natural byproduct of doing any kind of research.
Most of my knowledge comes from physical encyclopedia and download the wikipedia text dump (internet was not readily available). You search for one thing and just explore by clicking.
How often do you go back to your encyclopedia hard copies only to find whatever knowledge you may have absorbed have already been deprecated? Or that information from Wikipedia may have changed at moments without notice, have never read or, dare I say, included a political bias to them?
Maybe I should have worded it better as a "beginner" or "intermediate" knowledge onramp and/or filler. For example, I have asked it on occasion to translate into traditional Mandarin in parallel for every English response. It helps tremendously in trying to rebuild that bridge that may have been burned long ago.
Not to sound too dismissive, but there is a distinct learning curve when it comes to using models like Claude for code assist. Not just the intuition when the model goes off the rails, but also what to provide it in the context, how and what to ask for etc. Trying it once and dismissing it is maybe not the best experimental setup.
I've been using Zed recently with its LLM integration so assist me in my development and its been absolutely wonderful, but one must control tightly what to present to the model and what to ask for and how.
Yeah, this is how I use it too. I tend to be a very dry writer, which isn't unusual in science, but lately I've taken to writing, then asking an LLM to suggest improvements.
I know not to trust it to be as precise as good research papers need to be, so I don't take its output, it usually helps me reorder points or use different transitions which make the material much more enjoyable to read. I also find it useful for helping to come up with an opening sentence from which to start writing a section.
This is my go-to process whenever I write anything now:
1. I use dictation software to get my thoughts out as a stream of consciousness.
2. Then, I have ChatGPT or Claude refine it into something coherent based on a prompt of what I'm aiming for.
3. Finally, I review the result and make edits where needed to ensure it matches what I want.
This method has easily boosted my output by 10x, and I'd argue the quality is even better than before. As a non-native English speaker, this approach helps a lot with clarity and fluency. I'm not a great writer to begin with, so the improvement is noticeable. At the end of the day, I’m just a developer—what can I say?
Great opportunity to get ahead of all the lazy people who use AI for a cover letter. Do a video! Sure, AI will be able to do that soon, but then we (not lazy people, who care) will come up with something even more personal!
> By the look of text and the usage of the words, you could tell how professional the writer was and you had some confidence that the writer knew something about the subject
How did you know this unless you also had the same or more knowledge than the author?
It would seem to me we are as clueless now as before about how to judge how skilled a writer is without requiring to already posses that very skill ourselves.
I think you overestimate the value of things looking professional. The overwhelming majority of books published every year are trash, despite all the effort that went into research, writing, and editing them. Most news is trash. Most of what humanity produces just isn't any good. An top expert in his field can leave a typo-riddled comment in a hurry that contains more valuable information than a shelf of books written on the subject by lesser minds.
AIs are good at writing professional looking text because it's a low bar to clear. It doesn't require much intelligence or expertise.
I think you underestimate how high that bar is, but I will grant that it isn’t that high. It can be a form of sophistry all of its own. Still, it is a difficult skill to write clearly, simply, and without a lot of extravagant words.
> AIs are good at writing professional looking text because it's a low bar to clear. It doesn't require much intelligence or expertise.
AIs are getting good at precisely imitating your voice with a single sample as reference, or generating original music, or creating video with all sorts of impossible physics and special effects. By your rationale, nothing “requires much intelligence or expertise”, which is patently false (even for text writing)
My point is that writing a good book is vastly more difficult than writing a mediocre book. The distance between incoherent babble and a mediocre book is smaller than the distance between a mediocre book and a great book. Most people can write professional looking text just by putting in a little bit of effort.
>In the past, you had to put a lot of effort to produce a text which seemed to be high quality, especially when you knew nothing about the subject. By the look of text and the usage of the words, you could tell how professional the writer was and you had some confidence that the writer knew something about the subject. Now, that is completely removed. There is no easy filter anymore.
That is pretty much true also for other media, such as audio and video. Before digital stuff become mainstream pics are developed in the darkroom, and film are actually cut with scissors. A lot of effort are put into producing the final product. AI has really commoditized for many brain related tasks. We must realize the fragile nature of digital tech and still learn how to do these by ourselves.
> While the professional looking text could have been already wrong, the likelihood was smaller, since you usually needed to know something at least in order to write convincing text.
it's obvious when text has been produced by chatGPT with the default prompt - but there's probably loads of text on the internet which doesn't follow AI's usual prose style that blends in well.
Even when I try some other variation of prompts or writing styles there's always this sense of "perfectness", with all paragraph lengths being too perfect, length and the style of it being like that.
Although, there were already before tons of "technical influencers" before that who excelled at writing, but didn't know deeply what they were writing about.
They give a superficially smart look, but really they regurgitate without deep understanding.
>> While the professional looking text could have been already wrong, the likelihood was smaller, since you usually needed to know something at least in order to write convincing text.
...or...the likelihood of text being really wrong pre-LLMs was worse because you needed to be a well-capitalized player to pay your thoughts into public discourse. Just look at our global conflicts and you see how much they are driven by well-planned lobbying, PR, and...money. That is not new.
> I already questioned the information before and continue to do so.
You might question new information, but you certainly do not actually verify it. So all you can hope to do is sense-checking - if something doesn't sound plausible, you assume it isn't true.
This depends on having two things: having trustworthy sources at all, and being able to relatively easily distinguish between junk info and real thorough research. AI is a very easy way for previously-trustworthy sources to sneak in utter disinformation without necessarily changing tone much. That makes it much easier for the info to sneak past your senses than previously.
Reading was a form of connecting with someone. Their opinions are bound to be flawed, everyone's are - but they're still the thoughts and words of a person.
This is no longer the case. Thus, the human factor is gone and this reduces the experience to some of us, me included.
This is exactly what’s at stake. I heard an artist say one time that he’d rather listen to Bob Dylan miss a note than listen to a song that had all the imperfections engineered out of it.
They're not connecting to the autotune, but to the artist. People have a lot of opinions about Taylor Swift's music but "not being personal enough" is definitely not a common one.
If you wanna advocate for unplugged music being more gratifying, I don't disagree, but acting like the autotune is what people are getting out of Taylor Swift songs is goofy.
I have no idea about Taylor Swift so I'll ask in general: can't we have a human showing an autotuned personality? Like, you are what you are in private, but in interviews you focus on things suggested by your AI conselor, your lyrics are fine tuned by AI, all this to show a better marketable personality? Maybe that's the autotune we should worry about. Again, nothing new (looking at you, Village People) but nowadays the potential powered by AI is many orders of magnitude higher... you could say yes only until the fans catch wind of it, true, but by that time the next figure shows up and so on. Not sure where this arms escalation can lead us. Because also acceptance levels are shifting, so what we reject today as unacceptable lies could be fine tomorrow, look already at the AI influencers doing a decent job while overtly fake.
I’m convinced it’s already being done, or at least played with. Lots of public figures only speak through a teleprompter. It would be easy to put a fine tuned LLM on the other side of that teleprompter where even unscripted questions can be met with scripted answers.
I think the key thing here is equating trust and truth. I trust my dog, a lot, more than most humans frankly. She has some of my highest levels of trust attainable, yet I don’t exactly equate her actions with truth. She often barks when there’s no one at the door or at false threats she doesn’t know aren’t real threats and so on. But I trust she believes it 100% and thinks she’s helping me 100%.
What I think OP was saying and I agree with is that connection, that knowing no matter what was said or how flawed or what motive someone had I trusted there was a human producing the words. I could guess and reasons the other factors away. Now I don’t always know if that is the case.
If you’ve ever played a multiplayer game, most of the enjoyable experience for me is playing other humans. We’ve had good game AIs in many domains for years, sometimes difficult to distinguish from humans, but I always lost interest if I didn’t know I was in fact playing and connecting with another human. If it’s just some automated system I could do that any hour of the day as much as I want but it lacked the human connection element, the flaws, the emotion, the connection. If you can reproduce that then maybe it would be enjoyable but that sort of substance has meaning to many.
It’s interesting to see a calculator quickly spit out correct complex arithmetic but when you see a human do it, it’s more impressive or at least interesting, because you know the natural capability is lower and that they’re flawed just like you are.
> For me, LLMs don't change anything. I already questioned the information before and continue to do so.
I also did, but LLM increased the volume of content, which forces my brain first try to identify if content is generated by LLMs, which is consuming a lot of energy and makes brain even less focused, because now it's primary goal is skimming quickly to identify, instead of absorbing first and then analyzing info
The web being polluted only makes me ignore more of it.
You already know some of the more trustworthy sources of information, you don't need to read a random blog which will require a lot more effort to verify.
Even here on hackernews, I ignore like 90% of the spam people post. A lot of posts here are extremely low effort blogs adding zero value to anything, and I don't even want to think whether someone wasted their own time writing that or used some LLM, it's worthless in both cases.
How do you like questioning much more of it, much more frequently, from many more sources? And mistrusting it in new ways. AI and regular people are not wrong in the same ways, nor for the same reasons, and now you must track this too, increasingly.
Perhaps "trust" was a bit misplaced here, but I think we can all agree on the idea: Before LLMs, there was intelligence behind text, and now there's not. The I in LLM stands for intelligence, as written in one blog. Maybe the text never was true, but at least it made sense given some agenda. And like pointed out by others, the usual text style and vocabulary signs that could have been used to identify expertise or agenda are gone.
> Perhaps "trust" was a bit misplaced here, but I think we can all agree on the idea: Before LLMs, there was intelligence behind text, and now there's not. The I in LLM stands for intelligence, as written in one blog. Maybe the text never was true, but at least it made sense given some agenda.
Nope. A lot of people just wrote stuff. There were always plenty of word salad blogs (and arguably entire philosophy journals) out there.
I read the original comment not as a lament of not being able to trust the content, rather, they are lamenting the fact that AI/LLM generated content has no more thought or effort put into it than a cheap microwave dinner purchased from Walmart. Yes, it fills the gut with calories but it lacks taste.
On second thought, perhaps AI/LLM generated content is better illustrated with it being like eating the regurgitated sludge called cud. Nothing new, but it fills the gut.
There are topics on which you should be somewhat suspicious of anything you read, but also many topics where it is simply improbable that anyone would spend time maliciously coming up with a lie. However, they may well have spicy autocomplete imagine something for them. An example from a few days ago: https://news.ycombinator.com/item?id=41645282
I think it is a totally different threat. Excluding adversarial behavior, humans usually produce information with a quality level that is homogeneous (from homogeneously sloppy to homogeneously rigurous).
AI otoh can produce texts that are quite accurate globally with some totally random hallucinations here and there. It makes it quite harder to identify
There were news reports that Russia spent less than a million dollars on a massive propaganda campaign targeting U.S elections and the American population in general.
Do you think it would be possible before internet, before AI?
Bad actors, poorly written/sourced information, sensationalism etc have always existed. It is nothing new. What is new is the scale, speed and cost of making and spreading poor quality stuff now.
All one needs today is a laptop and an internet connection and a few hours, they can wreak havoc. In the past, you'd need TV or newspapers to spread bad (and good) stuff - they were expensive, time consuming to produce and had limited reach.
Some woman’s cat was hiding in her basement. She automatically assumed her Haitian neighbors stole her cat and made some comment about it, which landed on Facebook, which got morphed into “immigrants eating pets” story, JD Vance picked it up, Trump mentioned it in a national debate watched by 65 million people. All of this happened in a few days. This resulted in violence in Springfield.
If you can place a rumor or lie in front of the right person/people to amplify, it will be amplified. It will spread like wildfire, and by the time it is fact checked, it will have done at least some damage.
These successful manipulation stories are extremely rare though. What usually happens is that you say your neighbour ate your cat, then everyone laughs at you.
Did the person who posted do the manipulation, or did JD Vance and Donald Trump do it?
> you trusted what you read before. For me, LLMs don't change anything. I already questioned the information before and continue to do so. [...]
Why do you think that you could trust what you read before?
A human communicator is, in a sense, testifying when communicating. Humans have skin in the social game.
We try to educate people, we do want people to be well-informed and to think critically about what they read and hear. In the marketplace of information, we tend very strongly to trust non-delusional, non-hallucinating members of society. Human society is a social-confidence network.
In social media, where there is a cloak of anonymity (or obscurity), people may behave very badly. But they are usually full of excuses when the cloak is torn away; they are usually remarkably contrite before a judge.
A human communicator can face social, legal, and economic consequences for false testimony. Humans in a corporation, and the corporation itself, may be held accountable. They may allocate large sums of money to their defence, but reputation has value and their defence is not without social cost and monetary cost.
It is literally less effort at every scale to consult a trusted and trustworthy source of information.
It is literally more effort at every scale to feed oneself untrustworthy communication.
For me, the problem has gone from “figure out the author’s agenda” to “figure out whether this is a meaningful text at all,” because gibberish now looks a whole lot more like meaning than it used to.
This has been a problem on the internet for the past decade if not more anyway, with all of the seo nonsense. If anything, maybe it's going to be ever so slightly more readable.
I don't know what you're talking about. Most people don't think of SEO, Search Engine Optimization, Search Performance, Search Engine Relevance, Search Rankings, Result Page Optimization, or Result Performance when writing their Article, Articles, Internet Articles, News Articles, Current News, Press Release, or News Updates...
There's a quantity argument to be made here - before, it used to be hard to generate large amounts of plausible but incorrect text. Now it easy. Similar to surveillance before/after smartphones + the internet - you had to have a person following you vs just soaking up all the data on the backbone.
There was a degree of proof of work involved. Text took human effort to create, and this roughly constrained the quantity and quality of misinforming text to the number of humans with motive to expend sufficient effort to misinform. Now superficially indistinguishable text can be created by an investment in flops, which are fungible. This means that the constraint on the amount of misinforming text instead scales with whatever money is resourced to the task of generating misinforming text. If misinforming text can generate value for someone that can be translated back into money, the generation of misinforming text can be scaled to saturation and full extraction of that value.
It’s nothing to do with trusting in terms of being true or false, but whatever I read before I felt like, well, it can be good or bad, I can judge it, but whatever it is, somebody wrote it. It’s their work. Now when I read something I just have absolutely no idea whether the person wrote it, how much percent did they write it, or how much they even had to think before publishing it. Anyone can simply publish a perfectly well-written piece of text about any topic whatsoever, and I just can’t wrap my head around why, but it feels like a complete waste of time to read anything. Like… it’s all just garbage, I don’t know.
If one spends a lot of years reading a lot of stuff, they come to this conclusion, that most of it cannot be trusted. But it takes lots of years and lots of material to see it.
It's that you trusted that what you read came from a human being. Back in the day I used to spend hours reading Evolution vs Creationism debates online. I didn't "trust" the veracity of half of what I read, but that didn't mean I didn't want to read it. I liked reading it because it came from people. I would never want to read AI regurgitation of these arguments.
> Write a response to this comment, make spelling and grammar mistakes.
yeah well sumtimes spellling and grammer erors just make thing hard two read. like i no wat u mean bout wanting two kno its a reel person, but i think cleear communication is still importint! ;)
Way back when we had a landline and would get telemarketers, it was always a sign when the caller couldn’t pronounce our last name. It’s not even that uncommon a name, either
A few months ago, I tried to get Gemini to help me write some criticism of something. I can't even remember what it was, but I wanted to clearly say something was wrong and bad.
Gemini just could not do it. It kept trying to avoid being explicitly negative. It wanted me to instead focus on the positive. I think it evidently just told me no, and that it would not do it.
Yeah all the current tools have this particular brand of corporate speech that’s pretty easy to pick up on. Overly verbose, overly polite, very vague, non assertive, and non opinionated.
People could prompt for authenticity, adding subtle mistakes, etc. I hope that AI as a whole will help people writing better, if reading back the text. It is a bit like "The Substance" movie: a "better" version of ourselves.
why do you trust things now? unless you recognize the author and have a chain of trust from that author production to the content you're consuming, there already was no way to estabilish trust.
I've been using it in my personal writing (combination of GPT and Claude). I ask the AI to write something, maybe several times, and I edit it until I'm happy with it. I've always known I'm a better editor than I am an author, and the AI text gives me somewhere to start.
So there's a human in the loop who is prepared to vouch for those sentences. They're not 100% human-written, but they are 100% human-approved. I haven't just connected my blog to a Markov chain firehose and walked away.
Am I still adding to the AI smog? idk. I imagine that, at a bare minimum, its way of organising text bleeds through no matter how much editing I do.
you wrote this comment completely by your own, right? without any AI involved. And I read your comment feeling confident that it's truly 100% yours. I think this reader's confidence is what the OP is talking about.
I did. I write for myself mostly so I'm not so worried about one reader's trust - I guess I'm more worried that I might be contributing to the dead internet theory by generating AI-polluted text for the next generation of AIs to train on.
At the moment I'm using it for local history research. I feed it all the text I can find on an event (mostly newspaper articles and other primary sources, occasionally quotes from secondary sources) and I prompt with something like "Summarize this document in a concise and direct style. Focus on the main points and key details. Maintain a neutral, objective voice." Then I hack at it until I'm happy (mostly I cut stuff). Analysis, I do the other way around: I write the first draft, then ask the AI to polish. Then I go back and forth a few times until I'm happy with that paragraph.
I'm not going anywhere with this really, I'm just musing out loud. Am I contributing to a tragedy of the commons by writing about 18th century enclosures? Because that would be ironic.
If you write for yourself, whether you use generated text or not, (I am using the text completion on my phone typing this message), the only thing that matters is how it affects you.
Reading and writing are mental processes (with or without advanced technology) that shape our collective mind.
AI generated crap is one thing. But human generated crap is there - just because human wrote something it is not making it good.
Had a friend who thought that if it is written in a book it is for sure true. Well NO!
There was exactly the same sentiment with stuff on the internet and it is still the same sentiment about Wikipedia that “it is just some kids writing bs, get a paper book or real encyclopedia to look stuff up”.
Not defending gen AI - but still you have to make useful proxy measures what to read and what not, it was always an effort and nothing is going substitute critical thinking and putting in effort to separate wheat from the chaff.
No one claimed humans are perfect. But gen AI is a force multiplier for every problem we had to deal with. It's just completely different scale. Your brain is about to be DDOSed by junk content.
Of course, gen AI is just a tool that can be used for good or bad, but spam, targeted misinformation campaigns, and garbage content in general is one area that will be most amplified because it became so low effort and they don't care about doing any review, double-checking, etc. They can completely automate their process to whatever goal they've in mind. So where sensible humans enjoy 10x productivity, these spam farms will be enjoying 10000x scale.
So I don't think downplaying it and acting like nothing changed, is the brightest idea. I hope you see now how that's a completely different game, one that's already here but we aren't prepared for yet, certainly not with traditional tools we have.
> Your brain is about to be DDOSed by junk content.
It's not the best analogy because there's already more junk out there than can fit through the limited bandwidth available to my brain, and yet I'm still (vaguely) functional.
So how do I avoid the junk now? Rough and ready trust metrics, I guess. Which of those will still work when the spam's 10x more human?
I think the recommendations of friends will still work, and we'll increasingly retreat to walled gardens where obvious spammers (of both the digital and human variety) can be booted out. I'm still on facebook, but I'm only interested in a few well-moderated groups. The main timeline is dead to me. Those moderators are my content curators for facebook content.
> One cannot be DDOSed with junk when not actively trying to stuff as much junk into ones head.
The junk gets thrown at you in mass volume at low cost without your permission. What you gonna do? keep dodging it? waste your time evaluating every piece of information you come across?
If one of the results on the first page in search deviate from others, it's easy to notice. But if all of them agree, they became the truth. Of course your first thought is to say search engines are shit or whatever off-hand remarks, but this example is just to illustrate how the volume alone can change things. The medium doesn't matter, these things could come in many forms: book reviews, posts on social media, ads, false product description on amazon, etc.
Of course, these things exist today but the scale is different, the customization is different. It's like the difference between firearms and drones. If you think it's the same old game and you can defend against the new threat using your old arsenal, I admire your confidence but you're in for a surprise.
So you're basically sheltering yourself and seeking human curated content? Good for you, I follow similar strategy. How do you propose we apply this solution for the masses in today's digital age? or you're just saying 'each on their own'?
Sadly, you seem to not be looking further than your nose.
We are not talking about just you and me here. Less tech literate people are the ones at a disadvantage and need protection the most.
> How do you propose we apply this solution for the masses in today's digital age?
The social media algorithms are the content curators for the technically illiterate.
Ok, they suck and they're actively user-hostile, but they sucked before AI. Maybe (maybe!) AI's the straw that breaks the camel's back, and people leave those algorithm-curated spaces in droves. I hope that, one way and another, they'll drift back towards human-curated spaces. Maybe without even realizing it.
You kind of notice the stuff written with AI, it has a certain something that makes it detectable. Granted, stuff like the Reuters press reports might have already been written by AI, but I think that in that case it doesn’t really matter.
I get two associations from your comment: One about how AI being mainly used to interpolate within a corpus of prior knowledge, seems like entropy in a thermodynamical sense. The other, how this is like the Tower of Babel but where distrust is sown by sameness rather than differences.
In fact, relying on AI for coding and writing, feels more like channeling demonic suggestions than anything else. No wonder we are becoming skeptical.
I know and at my company we actually cannot disable the AI suggestions :(
It's like dealing with a pathological liar who produces very convincing looking code for me to review, but actually there is a bug in it 50% of the time. It's like it's trying to trick me into committing bugs.
> If you write regularly and you're not using AI, you simply cannot keep up with the competition.
Is that true today? I guess it depends what kind of writing you are talking about, but I wouldn't think most successful writers today - from novelests to tech bloggers - rely that much on AI, but I don't know. Five years from now, could be a different story.
Yes it’s true today, depending on what is your writing is the foundation of.
It doesn’t matter that my writing is more considered, more accurate and of a higher quality when my coworkers are all openly using AI to perform five times the work I am and producing outcomes that are “good enough” because good enough is quite enough for a larger majority than many likely realise.
It's not true at all. Much like the claims that you have to use LLMs to keep up in programming: if that is true then you weren't a good programmer (or writer in this case) to begin with.
That is absolutely wrong. Regardless of whether you were or not good to begin with, an LLM assistant will still accelerate a lot of repetitive tasks for you. Repetitive is repetitive, no matter if you’re John Carmack or the guy sitting in the booth next to you in the paper company. And anyway in a few years none of it will matter because programming without assistance will be a vintage thing of the past (like perforated cards).
It’s the same with writing. If you find yourself writing a boring introduction section to a paper with a bunch of meaningless blabla, then why wouldn’t you use AI for that? There is simply no good reason, especially when you see mediocre researchers publishing at three times your rate and getting promoted over you.
Well we’re going to need some system of PKI that is tied to real identities. You can keep being anonymous if you want but I would prefer not and prefer to not interact with the anonymous, just like how I don’t want to interact with people wearing ski masks.
> It certainly will affect the reputation of people that are consistently publishing untruths.
Oh? I thought there are a lot of very well identified people making a living from publishing untruths right now on all social media. How would PKI help, when they're already making it very clear who they are?
I have never read more bullshit in my life than during the corona pandemic, all written by humans. So you should never trust something you read, always question the source and it's reasoning.
At the same time I use copilot on a daily basis, both for coding as well as the normal chat.
It is not perfect, but I'm at a point I trust AI more than the average human. And why shouldn't I? LLMs ingest and combine more knowledge than any human can ever do. An LLM is not a human brain but it's actually performing really well.
Agreed, I feel like there's an inherent nobility in putting effort into something. If I took the time to write a book and have it proof-read and edited and so on: perhaps it's actually worth my time.
Lowering the bar to write books is "good" but increases the noise to signal ratio.
I'm not 100% certain how to give another proof-of-work, but what I've started doing is narrating my blog posts - though AI voices are getting better too.. :\",
Take a look at the user's comment history. I'd argue the majority of them are generated, it really doesn't take much to recognize the patterns and tics those models commonly display (for now*), and it doesn't seem there was any attempt to make it appear otherwise, too. It may look coherent at first, but there's no deeper meaning or significance, it's noise.
The most depressing thing for me is the feeling that I simply cannot trust anything that has been written in the past 2 years or so and up until the day that I die.
What AI is going to teach people is that they don't actually need to trust half as many things as they thought they did, but that they do need to verify what's left.
This has always been the case. We've just been deferring to 'truster organizations' a lot recently, without actually looking to see if they still warrant having our trust when they change over time.
Independent verification is always good however not always possible and practical. At complex levels of life we have to just trust underlying processes work, usually until something fails.
I don’t go double checking civil engineers work (nor could I) for every bridge I drive over. I don’t check inspection records to make sure it was recent and proper actions were taken. I trust that enough people involved know what they’re doing with good enough intent that I can take my 20 second trip over it in my car without batting an eye.
If I had to verify everything, I’m not sure how I’d get across many bridges on a daily basis. Or use any major infrastructure in general where my life might be at risk. And those are cases where it’s very important to be done right, if it’s some accounting form or generated video on the internet… I have even less time to be concerned from a practical standpoint. Having the skills to do it should I want or need to are good and everyone should have these but we’re at a point in society we really have to outsource trust in a lot of cases.
This is true everywhere, even in science which these days many people just trust in ways akin to faith in some cases, and I don’t see anyway around that. The key being that all the information should exist to be able to independently verify something but from a practice standpoint it’s rarely viable.
Get good at spotting inconsistencies. And pay attention to when something contradicts your own experience. Cultivate a wide range of experiences so that you have more domains where you can do this (this is a good idea anyway).
> It's not so much that I think people have used AI, but that I know they have with a high degree of certainty, and this certainty is converging to 100%, simply because there is no way it will not. If you write regularly and you're not using AI, you simply cannot keep up with the competition.
I am writing regularly and I will never use AI. In fact I am working on a 400+ pages book right now and it does not contain a single character that I have not come up with and typed myself. Something like pride in craftmanship does exist.
I'm right there with you. I write short and medium form articles for my personal site (link in bio, follow it or don't, the world keeps spinning either way). I will never use AI as part of this craft. If that hampers my output, or puts me at a disadvantage compared to the competition, or changes the opinion others have of me, I really don't care.
What I write is pretty niche anyway (compilers, LISP, buddhism, advaita), so I do not think AI will cause much trouble. Google ranking small websites into oblivion, though, I do notice that!
I use a spell checker to catch typos. Occasional quirky grammar is mine. Feedback will be provided by the audience. Why would I let a statistical model judge my work? This is how you kill originality.
i didn’t imply you’d need an LLM to “rate” your content. More so asking questions during and before the publishing step to help you improve your work. Not removing your identity from your work. Examples questions of what you could ask an LLM:
• are there any redundant sentences or points in this chapter?
• i’m trying to remember an expression used to convey X, can you remind me what it is?
• i’m familiar with X from my industry Y, but i’m trying to convey this to an audience from industry Z. Can you help brainstorm some words or examples they may be more familiar with?
Things like that. I think of it like having a virtual rubber duck that can help you debug code, or anything really.
Obviously these are just some suggestions. If you don’t find any of this useful, or even interesting, then carry on. :)
I see. Most of the things you list I think I could do better on my own. The case of X from industry Y sounds interesting, but I would still prefer to hear from a real human being from industry Z. If no one is available, of course, a statistical model may indeed be helpful -- I still do not think it is worth boiling the oceans, though. :)
This. I use AI to bounce off ideas when i wordsmith for work, but mostly when my human counterpart is unavailable to do so and I can't wait until they return eg. when they were on extended time off.
The AI isn't as good though, it's like asking an overeager intern instead of a thoughtful senior.
Add to it that not all LLMs are equal: at work I have access to the latest GPT models and can customize them myself. At home anything I tried in LM studio feels very limited/inflexible. All models i tried running were useless except one codestruct, which was half decent at answering clarification questions about c# code from tutorial, but failed the task i initially tried it for: generating bespoke tutorials that fit my knowledge gaps, which would have been my ideal scenario because most tutorials aren't written for someone like me.
It's another reason LLMs are a bit disappointing to me, they don't seem to be good at handling queries I couldn't solve with a web search, and have similar issues as search results when they do have answers - this makes sense since that's how they are trained, but often defeats the purpose for me.
Also currently working on a book (shameless plug: buy my book!) and feel no pull or need to involve AI. This book is my mine. My faults. My shortcomings. My overuse of commas. My wonky phrasing. It has to have those things, because I am those things (for better or worse!).
>If you write regularly and you're not using AI, you simply cannot keep up with the competition. You're out. And the growing consensus is "why shouldn't you?", there is no escape from that.
Are you sure you don't mean if you write regularly in one particular subclass of writing - like technical writing, documentation etc.?
Do you think novel writing, poetry, film reviews etc. cannot keep up in the same way?
I'm absolutely positive that the vast majority of fiction is or will soon be written by LLM. Will it be high-quality? Will it be loved and remembered by generations to come? Probably not. Will it make money? Probably more than before on average as the author's effort is reduced to writing outlines and prompts, and editing the generated-in-seconds output, rather than months-years of doing the writing themselves.
I think that novel writing and reviews are types of writing where potentially AI should eventually surpass human writers, because they have the potential to replace content skillfully tailored to be liked by many people with content that's tailored (perhaps less skillfully) explicitly for a specific very, very, very narrow niche of exactly you and all the things that happen to work for your particular biases.
There seems to be an upcoming wave of adult content products (once again, being on the bleeding edge users of new abilities) based on this principle, as hitting very specific niches/kinks/fetishes can be quite effective in that business, but it should then move on to romance novels and pulp fiction and then, over time, most other genres.
Similarly, good pedagogy, curriculum design and educational content development is all about accurately modeling which exact bits of the content the target audience will/won't know, and explaining the gaps with analogies and context that will work for them (for example, when adapting a textbook for a different country, translation is not sufficient; you'd also need to adapt the content). In that regard, if AI models can make personalized technical writing, then that can be more effective than the best technical writing the most skilled person can make addressed to a broader audience.
> If you write regularly and you're not using AI, you simply cannot keep up with the competition. You're out.
What? No! Content volume only matters in stupid contests like VC app marketing grifts or political disinformation ops where the content isn’t even meant to be read, it’s an excuse for a headline. I personally write all my startup’s marketing content, quality is exquisite and due to this our brand is becoming a juggernaut
response from AI on this: I completely understand where you're coming from. The increasing reliance on AI in writing does raise important questions about authenticity and connection. There’s something uniquely human in knowing that the words you're reading come from someone’s personal thoughts, experiences, and emotions—even if flawed. AI-generated content, while efficient and often well-written, lacks that deeper layer of humanity, the imperfections, and the creative struggle that gives writing its soul.
It’s easy to feel disillusioned when you know AI is shaping so much of the content around us. Writing used to be a deeply personal exchange, but now, it can feel mechanical, like it’s losing its essence. The pressure to keep up with AI can be overwhelming for human writers, leading to this shift in content creation.
At the same time, it’s worth considering that the human element still exists and will always matter—whether in long-form journalism, creative fiction, or even personal blogs. There are people out there who write for the love of it, for the connection it fosters, and for the need to express something uniquely theirs. While the presence of AI is unavoidable, the appreciation for genuine human insight and emotion will never go away.
Maybe the answer lies in seeking out and cherishing those authentic voices. While AI-generated writing will continue to grow, the hunger for human storytelling and connection will persist too. It’s about finding balance in this new reality and, when necessary, looking back to the richness of past writings, as you mentioned. While it may seem like a loss in some ways, it could also be a call to be more intentional in what we read and who we trust to deliver those words.
>> cannot trust anything that has been written in the past 2 years or so and up until the day that I die.
You never should have. Large amounts of work, even stuff by major authors, is ghostwritten. I was talking to someone about Taylor Swift recently. They thought that she wrote all her songs. I commented that one cannot really know that, that the entertainment industry is very going at generating seemingly "authentic" product at a rapid pace. My colleague looked at me like I had just killed a small animal. The idea that TS was "genuine" was a cornerstone of their fandom, and my suggestion had attacked that love. If you love music or film, don't dig too deep. It is all a factory. That AI is now part of that factory doesn't change much for me.
Maybe my opinion would change if I saw something AI-generated with even a hint of artistic relevance. I've seen cool pictures and passable prose, but nothing so far with actual meaning, nothing worthy of my time.
Watch the movie "The Wrecking Crew" about how a group of studio musicians in the 1970s were responsible for the albums of quite a few diverse "bands". Many bands had to then learn to play their own songs so they could go on tour.
While I do enjoy some popular genres, I'm all too aware of the massive industry behind it all. I believe that most of humanity's greatest works of art were created not for commercial interests but rather for the pure joy of creation, of human expression. This can be found in any genre if you look hard enough, but it's no accident that the music I find the most rewarding is classical music: Intellect, emotion, spirit, and narrative dreamed into existence by one person and then brought to life by other artists so we can share in its beauty.
I think music brings about a connection between the composers, lyricists, performers, and listeners. Music lets us participate in something uniquely human. Replacing any of the human participants with AI greatly diminishes or eliminates its value in my eyes.
In trying to write a book, it makes little sense to try to "compete" on speed or volume of output. There were already vast disparities in that among people who write, and people whose aim was to express themselves or contribute something of importance to people's lives, or the body of creative work in the world, have little reason to value quantity over quality. Probably if there's a significant correlation with volume of output, it's in earnings, and that seems both somewhat tenuous and like something that's addressable by changes in incentives, which seem necessary for a lot of things. Computers being able to do dumb stuff at massive scale should be viewed as finding vulnerabilities in the metrics this allows it to become trivial to game, and it's baffling whenever people say "Well clearly we're going to keep all our metrics the same and this will ruin everything." Of course, in cases where we are doing that, we should stop (For example, we should probably act to significantly curb price and wage discrimination, though that's more like a return to form of previous regulatory standards)
As a creator of any kind, I think that simply relying on LLMs to expand your output via straightforward uses of widely available tools is inevitably going to lead to regression to the mean in terms of creativity. I'm open to the idea, however, that there could be more creative uses of the things that some people will bother to do. Feedback loops they can create that somehow don't stifle their own creativity in favor of mimicking a statistical model, ways of incorporating their own ingredients into these food processors of information. I don't see a ton of finished work that seems to do this, but I see hints that some people are thinking this way, and they might come up with some cool stuff. It's a relatively newly adopted technology, and computer-generated art of various kinds usually separates into "efficiency" (which reads as low quality) in mimicking existing forms, and new forms which are uniquely possible with the new technology. I think plenty of people are just going to keep writing without significant input from LLMs, because while writer's block is a famous ailment, many writers are not primarily limited by their speed in producing more words. Like if you count comments on various sites and discussions with other people, I write thousands of words unassisted most days
This kind of gets to the crux of why these things are useful in some contexts, but really not up to snuff with what's being claimed about them. The most compelling use cases I've seen boil down to some form of fitting some information into a format that's more contextually appropriate, which can be great for highly structured formatting requirements and dealing with situations which are already subject to high protocol of some kind, so long as some error is tolerated. For things for which conveying your ideas with high fidelity, emphasizing your own narrative voice or nuanced thoughts on a subject, or standing behind the factual claims made by the piece are not as important. As much as their more strident proponents want to claim that humans are merely learning things by aggregating and remixing them in the same sense as these models do, this reads as the same sort of wishful thinking about technology that led people to believe that brains should work like clockwork or transistors at various other points in time at best, and honestly this most often seems to be trotted out as the kind of bad-faith analogy tech lawyers tend to use when trying to claim that the use of [exciting new computer thing] means something they are doing can't be a crime
So basically, I think rumors of the death of hand-written prose are, at least at present, greatly exaggerated, though I share the concern that it's going to be much harder to filter out spam from the genuine article, so what it's really going to ruin is most automated search techniques....
I totally understand your frustration. We started writing our book long before(2022) AI became mainstream, and when we finally published it on May 2024, all we hear now is people asking if it's just AI-generated content. It’s sad to see how quickly the conversation shifts away from the human touch in writing.
I don't use AI in my own blogging, but then, I don't particularly care whether or not someone reads my stuff (the ones that do, seem to like it).
I have used it, from time to time, to help polish stuff like marketing fluff for the App Store, but I'd never use it verbatim. I generally use it to polish a paragraph or sentence.
But AI hasn't suddenly injected untrustworthy prose into the world. We've been doing that, for hundreds of years.
I have my reservations about AI but it's hard not to notice that LLMs are effectively a Gutenberg level event in the history of written communication. They mark a fundamental shift in our capacity to produce persuasive text.
The ability to speak the same language or understand cultural norms are no longer barriers to publishing pretty much anything. You don't have to understand a topic or the jargon of any given domain. You don't have to learn the expected style or conventions an author might normally use in that context. You just have to know how to write a good prompt.
There's bound to be a significant increase in the quantity as well as the quality of untrustworthy published text because of these new capacities to produce it. It's not the phenomenon but the scale of production that changes the game here.
When I wrote about trust, I see that I made a mistake: most people seem to have understood it as being in regard to fake things. I just meant trust as in, it’s not AI-generated.
Your comment about the fluff is exactly what I mean. I read some fluff that is AI-generated and some kind of disgust happens in my stomach, and I wish there was nothing written there at all. I just feel like it’s be best to not read anything than read something that’s AI-generated… it’s almost like the author is trying to trick me with a fake version of reality. I wonder if there’s such a thing as the uncanny valley for text?
Well, I apologize for using the word “fluff.” That was a mistake.
As a lifelong engineer, I “grew up” with a somewhat antagonistic relationship with Marketing, so became used to disparaging their work, even if I had to change hats, myself, and act in a Marketing capacity.
I should have probably used the word “copy,” instead.
But you have a good point.
I think that one “legitimate” use for AI-generated text, will be for non-native speakers of a language, using it to correct their vocabulary.
For things like patents and papers, this is probably a good thing. AI can generate clear, concise vernacular. I often specify the reading level, in my prompts (usually tenth grade), so that the prose is accessible.
For things like presentation proposals; not so much. You may get a proposal that reads like it was written by an English professor, and the actual presentation is barely comprehensible.
I actually disagree with that. People are so busy hoping things will get better, and creating little bubbles for themselves to hide away from what human beings as a whole are doing, that they don't realize things are getting worse. Technology constantly makes things worse. Cheering up is a good self-help strategy but not a good strategy if you want to contribute to making the world actually a better place.
No way. Life 50 years ago was better for MANY. Maybe that would be true for 200. But 50 years ago was the 70s. There were far fewer people, and the world was not starting to suffer from climate change. Tell your statement to any climate refugee, and ask them whether they'd like to live now or back then.
AND, we had fewer computers and life was not so hectic. YES, some things have gotten better, but on average? It's arguable.
Why do you have to use it? I don’t get it. If you write your own book, you don’t compete with anyone. If anyone finished The Winds of Winter for R.R.Martin using AI, nobody would bat an eye, obviously, as we already experienced how bad a soulless story is that drifts too far away from what the author had built in his mind.
Well, at the business that I work for, I convinced everyone to have an anti-AI 100% free policy, although that wasn't too hard because no one there was ever enthusiastic about AI. Plus, we don't publish generic SEO stuff but real gear testing, real experience, so AI goes against that.
I'm not sure it's always that hard to tell the AI stuff from the non AI. Comments on HN and on twitter from people you follow are pretty much non AI, also people on youtube where you an see the actual human talking.
On the other hand there's a lot on youtube for example that is obviously ai - weird writing and speaking style and I'll only watch those if I'm really interested in the subject matter and there aren't alternatives.
Maybe people will gravitate more to the stuff like PaulG or Elon Musk on twitter or HN and less to blog style content?
With a friend, I created a website about a race track in the past two years. I definitely used AI to speed up some of writing. One thing I used it for was a track guide, describing every corner and how to drive it. It was surprisingly accurate, most of the time. The other times though, it would drive the track backwards, completely hallucinate the instructions or link corners that are in different parts of the track.
I spent a lot of time analyzing the track myself and fixed everything to the point that experienced drivers agreed with my description. If I hadn't done that, most visitors would probably still accept our guide as the truth, because they wouldn't know any better.
We know that not everyone cares about whether what they put on the internet is correct and AI allows those people to create content at an unprecedented pace. I fully agree with your sentiment.
AI only helps writing in so far as checking/suggesting edits. Most people can write better than AI (more engaging). AI cant tell a human story, have real tacit experience.
So it is like saying my champaigne bottle cant keep up with the tap water.
>The most depressing thing for me is the feeling that I simply cannot trust anything that has been written in the past 2 years or so and up until the day that I die
Do you think AI has changed that in any way? I remember the sea of excrement overtaking genuine human written content on the Internet around mid 2010s. It is around that time when Google stopped pretending they are a search company and focused on their primary business of advertising.
Before, at least they were trying to downrank all the crap "word aggregators". After, they stopped caring at all.
AI gives even better tools to page rank. Detection of AI generated content is not that bad.
So why don't we have "a new Google" emerge? Simple, because of the monopolistic practices Google did to make the barrier to entry huge. First, 99% of the content people want to search for is behind a login wall (Facebook, Instagram, twitter, YouTube), second almost all CDNs now implement "verify you are human" by default. Third, no one links to other sites. Ever! These 3 things mean a new Google is essentially impossible. Even duck duck go has thrown the towel and subscribed to Bing results.
It has nothing to do with AI, and everything to do with Google. In fact AI might give us the tools to better fight Google.
Some great grand ancestor of mine was a civil servant, a great achievement given his peasant background. The single skill that enabled it was the knowledge of calligraphy. He went to school and wrote nicely and that was sufficient.
The flip side was, calligraphy was sufficient evidence for both his education to whoever hired him, and for a recipient of a document, of its official nature. Calligraphy itself or course didn't make him efficient or smart or fair.
That's long gone of course, but we had similar heuristics. I am reminded of the Reddit story about an AI-generated mushroom atlas that had factual errors and lead to someone getting poisoned. We can no longer assume that a book is legit simply because it looks legit. The story of course is from reddit, so probably untrue, but it doesn't matter - it totally could be true.
LLMs are fantastic at breaking our heuristics as to what is and isn't legit, but not as good at being right.
> We can no longer assume that a book is legit simply because it looks legit.
The problem is that this has been an issue for a long time. My first interactions with the internet in the 90s came along with the warning "don't automatically trust what you read on the internet".
I was speaking to a librarian the other day who teaches incoming freshman how to use LLMs. What was shocking to me is that the librarian said a majority of the kids trust what the computer says by default. Not just LLMs, but generally what they read. That's such a huge shift from my generation. Maybe LLM education will shift people back toward skepticism - unlikely, but I can hope.
One of the issues today is the volume of content produced, and that journalism and professional writing is dying. LLMs produce large amounts of "good enough" quality to make a profit.
In the 90s we could reasonably trust that that the major news sites and corporate websites was true, while random forums required a bit more critical reading. Today even formerly trusted sites may be using LLMs to generate content along with automatic translations.
I wouldn't necessarily put the blame on LLMs, this just make it easier. The trolls and spammers was always there, now they just have a more powerful tool. The commercial sites now have a tool they don't understand, which they apply liberally, because it reduces cost, or their staff use it, to get out of work, keep up with deadlines or to cover up incompetence. So, not the fault of the LLMs, but their use is worsening existing trends.
> Today even formerly trusted sites may be using LLMs to generate content along with automatic translations.
Yep - or they're commingling promotional content with their journalism, a la Forbes / CNN / CNET / About.com / etc. There's still quality content online but it's getting harder to find under the tidal wave of garbage.
> I was speaking to a librarian the other day who teaches incoming freshman how to use LLMs. What was shocking to me is that the librarian said a majority of the kids trust what the computer says by default. Not just LLMs, but generally what they read. That's such a huge shift from my generation.
I think that previous generations were not any different. For most people, trusting is the default mode and you need to learn to distrust a source. I know many people who still have not learned that about the internet in general. These are often older people. They believe insane things just because there exists a nicely looking website claiming that thing.
Not sure of the context here is for "previous generation" but I've been around since early in the transition from university/military network to public network, and the reality was the internet just wasn't that big, and it was primarily made up of people who looked, acted and valued the same things.
Now it's not even the website of undetermined providence that is believed; positions are established based on just the headline, shared 2nd or 3rd hand!
> The problem is that this has been an issue for a long time. My first interactions with the internet in the 90s came along with the warning "don't automatically trust what you read on the internet".
I received the same warnings, actually it was more like “don’t trust everything you read on the internet”, but it quickly became apparent that the last three words were redundant, and could have been rephrased more accurately as “don’t trust everything you read and hear and see”.
Our parents and teachers were living with their own fallacious assumptions and we just didn’t know it at the time, but most information is very pliable. If you can’t change what someone sees, then you can probably change how they see it.
I feel like there was also a brief window where "many amateur eyes in public" trumped "private experts"; wikipedia, open source software, etc. This doesn't seem the case in a hyper-partisan and bifurcated society where there is little trust.
> Our parents and teachers were living with their own fallacious assumptions and we just didn’t know it at the time, but most information is very pliable.
Indeed. When I was 14-18 in the UK, the opinion pieces in the news were derogatorily describing "media studies" as "Mickey Mouse studies".
In retrospect such courses were teaching critical analysis of media sources, in much the same way that my history GCSE went into the content of historical media and explored how both primary and secondary sources each had the potential to be either accurate or biased.
Even now, even here, I see people treat the media itself as pure and only the people being reported upon as capable of wrongdoing — e.g. insisting that climate scientists in the 70s generally expected an imminent ice age, because that's what the newspapers were saying.
> Some great grand ancestor of mine was a civil servant, a great achievement given his peasant background. The single skill that enabled it was the knowledge of calligraphy. He went to school and wrote nicely and that was sufficient.
Similar story! Family lore has it that he was from a farming family of modest means, but he was hired to write insurance policies because of his beautiful handwriting, and this was a big step up in the world.
> The story of course is from reddit, so probably untrue, but it doesn't matter - it totally could be true.
What?! Someone just made up something and then got mad at it. This is specially weird when you even acknowledge its a made up story. If we start evaluating new things like this nothing will ever progress.
>That's long gone of course, but we had similar heuristics.
To quote someone about this:
>>All that is solid melts into air, all that is holy is profaned, and man is at last compelled to face with sober senses his real conditions of life.
A book looking legit, a paper being peer reviewed, an expert saying something, none of those things were _ever_ good heuristics. It's just that it was the done thing. Now we have to face the fact that our heuristics are obviously broken and we have to start thinking about every topic.
To quote someone else about this:
>>Most people would rather die than think.
Which explains neatly the politics of the last 10 years.
> To quote someone about this:
>>All that is solid melts into air, all that is holy is profaned, and man is at last compelled to face with sober senses his real conditions of life.
So, same as it ever was?
Smoke, nothing but smoke. [That’s what the Quester says.]
There’s nothing to anything—it’s all smoke.
What’s there to show for a lifetime of work,
a lifetime of working your fingers to the bone?
One generation goes its way, the next one arrives,
but nothing changes—it’s business as usual for old planet earth.
The sun comes up and the sun goes down,
then does it again, and again—the same old round.
The wind blows south, the wind blows north.
Around and around and around it blows,
blowing this way, then that—the whirling, erratic wind.
All the rivers flow into the sea,
but the sea never fills up.
The rivers keep flowing to the same old place,
and then start all over and do it again.
Everything’s boring, utterly boring—
no one can find any meaning in it.
Boring to the eye,
boring to the ear.
What was will be again,
what happened will happen again.
There’s nothing new on this earth.
Year after year it’s the same old thing.
Does someone call out, “Hey, this is new”?
Don’t get excited—it’s the same old story.
Nobody remembers what happened yesterday.
And the things that will happen tomorrow?
Nobody’ll remember them either.
Don’t count on being remembered.
One is a complaint that everything is constantly changing, the other that nothing ever changes. I don't think you could misunderstand what either is trying to say harder if you tried.
Culd be my KJV upbringing talking, but personally I think there's an informative quality to calling it "vanity" over smoke.
And there's more reasons not to simply compare the modern challenges of image and media with the ancient grappling with impermanence. Tech may only truly change the human condition rarely, but it frequently magnifies some aspect of it, sometimes so much that the quantitative change becomes a qualitative one.
And in this case, what we're talking about isn't just impermanence and mortality and meaning as the preacher/quester is. We'd be lucky if it's business as usual for old planet earth, but we've managed to magnify our ability to impact our environment with tech to the point where winds, rivers, seas, and other things may well change drastically. And as for "smoke", it's one thing if we're dust in the wind, but when we're dust we can trust, that enables continuity and cooperation. There's always been reasons for distrust, but with media scale, the liabilities are magnified, and now we've automated some of them.
The realities of human nature that are the seeds of the human condition are old. But some of the technical and social machinery we have made to magnify things is new, and we can and will see new problems.
Heuristics don't have to be perfect to be useful so long as they improve the efficacy of our attentions. Once that breaks down society must follow because thinking about every topic is intractable.
The mushroom thing is almost certainly true. There's tons of trash AI generated foraging books being published to Amazon. Atomic Shrimp has a video on it.
Bots are now blocked because they've been abusive. When you host content on the internet, it's not fun to have bots bring your server down or inflate your bandwidth price. Google's bot is actually quite well-behaved. The other problem has been the recent trend in AI, and I can understand blockers being put in place, since AI is essentially plagiarizing content without attribution. But I'd blame OpenAI more at this point.
I also don't think you can blame Google for the centralization behind closed gardens. Or for why people no longer link to other websites. That's ridiculous.
And you should be attributing them the fact that the web is still alive.
> I remember the sea of excrement overtaking genuine human written content on the Internet around mid 2010s.
I mean the AI is trained and modeled on this excrement. It makes sense. As much as people think AI content is raw garbage… they don’t realize that they are staring into a mirror.
>I remember the sea of excrement overtaking genuine human written content on the Internet around mid 2010s.
Things have not changed much really. This was true since the dawn of man-kind (and woman-kind from the man-kind rib of course) even before there writings was invented, in the form of gossip.
The internet/AI now carries on the torch of our ancestral inner calling, lol.
Google didn't change it, it embodied it. The problem isn't AI, it's the pervasive culture of PR and advertising which appeared in the 50s and eventually consumed its host.
Western industrial culture was based on substance - getting real shit done. There was always a lot of scammery around it, but the bedrock goal was to make physical things happen - build things, invent things, deliver things, innovate.
PR and ad culture was there to support that. The goal was to change values and behaviours to get people to Buy More Stuff. OK.
Then around the time the Internet arrived, industry was off-shored, and the culture started to become one of appearance and performance, not of substance and action.
SEO, adtech, social media, web framework soup, management fads - they're all about impression management and popularity games, not about underlying fundamentals.
This is very obvious on social media in the arts. The qualification for a creative career used to be substantial talent and ability. Now there are thousands of people making careers out of performing the lifestyle of being a creative person. Their ability to do the basics - draw, write, compose - is very limited. Worse, they lack the ability to imagine anything fresh or original - which is where the real substance is in art.
Worse than that, they don't know what they don't know, because they've been trained to be superficial in a superficial culture.
It's just as bad in engineering, where it has become more important to create the illusion of work being done, than to do the work. (Looking at you, Boeing. And also Agile...)
You literally make more money doing this. A lot more.
So AI isn't really a tool for creating substance. It's a tool for automating impression management. You can create the impression of getting a lot of work done. Or the impression of a well-written cover letter. Or of a genre novel, techno track, whatever.
AI might one day be a tool for creating substance. But at the moment it's reflecting and enabling a Potemkin busy-culture of recycled facades and appearances that has almost nothing real behind it.
Unfortunately it's quite good at that.
But the problem is the culture, not the technology. And it's been a problem for a long time.
Thank you, you've stated this all very clearly. I've been thinking about this in terms of "doing work", where you care about the results, and "performing work", where you care about how you are evaluated. I know someone who works in a lab, and pointed out that some of the equipment being used was out of spec and under-serviced to the point that it was essentially a random number generator. Caring about this is "doing work". However, pointing it out made that person the enemy of the greater cohort that was "performing work". The results were not important to them, their metrics about units of work completed was. I see this pattern frequently. And it's hard to say those "performing work" are wrong. "Performing" is rewarded, "doing" is punished - Perhaps right to the top, as many companies are involved in a public performance designed to affect the short-term stock price.
Workers are many times more efficient than they were in the 50s or 70s or 80s or 90s. Where are our extra vacation days? Why does the worker have to make up for the efficiency with more work while other people take the gains?
Do you seriously think that the purpose of life is to work all the time most efficiently? Enjoy your lazy job and bask in the ability for human society to be productive without everyone breaking their backs all the time.
focusing on efficiency is very depressing. Machines seek efficiency. Process can be efficient. Assembly lines are efficient. It's all about optimization and quickly focuses on trimming "waste" and packing as much as possible into the smallest space. It removes all that's amazing about human life.
I much prefer a focus on effectiveness (or impact or outcomes, or alternatives). It plays to human strengths, is far less prescriptive and is way more fun!
Some of the most effective actions are incredibly inefficient; sometimes inefficiency is a feature. I received a letter mail thank-you card from our CEO a few years ago. The card has an approx. value of zero dollars, but I know it took the CEO 5-10 mins to write a personal note and sign it, and that she did this dozens of times. The signal here is incredibly valuable! If she used a signing machine, or AI to record a deep fake message I would know or quickly learn, and the value would go negative - all for the sake of efficiency.
I think this is a big part of it. Workers would feel a lot more motivated to do more than just perform if they were given what they know they’re owed for their contribution.
Faced with increasing efficiency, Europe by and large appears to have chosen to work less and let expensive things remain expensive. The US, by contrast, now has ridiculously cheap consumer goods, and works all the time.
Of course GDP increases when the money supply does. It’s like people being incensed at “record corporate profits” amidst inflation - profits will always be record (give or take) because remaining the same is losing money relative to the free money being minted each day, etc. For whatever reason, people naively buy into GDP as a valuable metric, even knowing well that there would be something extremely mysterious going on if that number somehow shrank while the real value of the medium exchange also shrank
This is a pretty clear summary of a real problem in most work environments. I have some thoughts about why, but I'm holding onto your articulation to ruminate on in the future.
"Doing work" vs. "performing work": the epitome of this is consulting. Companies pay huge sums of money to consultants that often spend most of their time "performing work", doing beautiful slides even if the content and reasoning is superficial or even dubious, creating reports that are just marketing bullshit, framing the current mission in a way that makes it possible to capture additional projects and bill the client even more. Almost everything is bullshit.
For better or for worse, you’ve described the exact output some of those client companies want, so they can show off the shiny slides and have the status symbol of expensive consultants.
The slippage between work and meaning is due to the arrival of post scarcity.
There isn't enough meaningful work to go around because work is still predicated on delivering some useful transformation. But when 80% of the useful transformations can be done by 20% of the people and we have enough to keep civilization going, you don't need full employment any more. But due to the moral hazard doctrine, the elites want to retain control and discipline as is.
For being afraid of a world where people don't care about maintaining the system anymore because it's not needed to house and feed everyone and people can just do art or learn or rest, the system keeps inventing meaningless work.
We're shipping fruit half way across the world to be packaged and then halfway back to be consumed. None of this is efficient or necessary. But it keeps the control system intact and that's the goal.
The thing that I struggle with is I agree with it, but I also get a lot of value in using AI to make me more productive - to me, it feels like it lets me focus on producing substance and actions, freeing me up from having to some tedious things in some tedious ways. Without getting into the debate about if it's productive overall, there are certain tasks which it feels irrefutably fast and effective at (e.g. writing tests).
I do agree with the missing substance with modern generative AI: everyone notices when it's producing things in that uncanny valley, and if no human is there to edit that, it makes people uncomfortable.
The only way I can reconcile the almost existential discomfort of AI against my actual day-to-day generally-positive experience with AI is to accept that AI in itself isn't the problem. Ultimately, it is an info tool, and human nature makes people spam garbage for clicks with it.
People will do the equivalent of spam garbage for clicks with any new modern thing, unfortunately.
Getting the most out of latest information of a society has probably always been a cat and mouse game of trying to find the areas where the spam-garbage-for-clicks people haven't outnumbered use-AI-to-facilitate-substance people, like here, hopefully.
Just one nitpick. The thing about test is that it’s repetitive enough to be automated (in a deterministic way) or abstracted into a framework. You don’t need an AI to generate it.
While I occasionally have the pleasure of creating or working with a test suite that's interesting and creative relative to the code under test, the vast majority of unit tests by volume are slop. Does it call the mock? Does it use the return value? Does "if err != nil { return err }" in fact stop and return the error?
This stuff is a perfect candidate for LLM generation.
I find it helpful for generating automated test suites in the style of the rest of the codebase. When working across multiple projects and clients, it reduces the mental load of having to remember or figure out how tests are expected to work in each codebase.
I agree with your theory about tests. The reality of it is most code is garbage - often including my own - and in a lot of environments, the task is to get the job done in a way that fits in with what's there.
AI seems really good at producing middling content, and if you make your living writing mediocre training courses, or marketing collateral, or code, or tests you're in big trouble. I question how valuable this work is though, so are we increasing productivity by utilizing AI, or just getting efficient at a suboptimal game? I for one just refuse to play.
Echoing other comments in gratitude for this very clear articulation of feelings I share, but have not manifested so well. Just wanted to add two connected opinions that round out this view.
1) This consuming of the host is only possible on the one hand because the host has grown so strong, that is the modern global industrial economy is so efficient. The doing stuff side of the equation is truly amazing and getting better (some real work gets done either by accident or those who have not-succumbed to PR and ad culture), and even this drop of "real work" produces enough material wealth to support (at least a lot of) humanity. We really do live in a post scarcity world from a production perspective, we just have profound distribution and allocation problems.
2) Radical wealth inequality profoundly exacerbates the problem of PR and ad culture. If everyone has some wealth doing things that help many people live more comfortably is a great way to become wealthy. But if very few people have wealth, then doing a venture capital FOMO hustle on the wealthy is anyone's best ROI. Radical wealth inequality eventually breaks all the good aspects of capitalist/market economies.
You can outfit an adult life with all of the useful manufactured objects that would reasonably improve it for a not-very-impressive sum. Beyond that it's just clutter (going for quantity) or moving into the lifestyle/taste/social-signaling domain anyway (going for quality). There is just not an unlimited amount of alpha in making physical things. The social/thought/experiential domain is a much bigger opportunity.
>Western industrial culture was based on substance - getting real shit done. There was always a lot of scammery around it, but the bedrock goal was to make physical things happen - build things, invent things, deliver things, innovate.
For a very short period between 1945 to 1980 while the generation who remembered the great depression and WWII was in charge. It's been longer since that's not been the case. And it wasn't the case for most of history before then.
Hah! I'm certain a lot of people want to believe it's bogus, mostly because of what that says about their character. Observed reality suggests there's some truth here. An example in miniature would be the frequency with which family fortunes are lost within a generation or two.
> You can create the impression of getting a lot of work done. Or the impression of a well-written cover letter. Or of a genre novel, techno track, whatever.
Yeah, one of their most "effective" uses is to counterfeit signals that we have relied on--wisely or not--to estimate deeper practical truths. Stuff like "did this person invest some time into this" or "does this person have knowledge of a field" or "can they even think straight."
Oh, sure, qualitatively speaking it's not new, people could have used form-letters, hired a ghostwriter, or simply sank time and effort into a good lie... but the quantitative change of "Bot, write something that appears heartfelt and clever" is huge.
In some cases that's devastating--like trying to avert botting/sockpuppet operations online--and in others we might have to cope by saying stuff like: "Fuck it, personal essays and cover letters are meaningless now, just put down the raw bullet-points."
Very well written. I assume you haven't read "Simulacra and Simulation" by Jean Baudrillard, that's why your description is so authentic and is more convincing then just referring to the book. Saved this post for future reference.
> Western industrial culture was based on substance - getting real shit done.
And what did that get us? Radium poisoning and microplastics in every organ of virtually all animals living within thousands of miles of humans. Our reach has always exceeded our grasp.
Just yesterday I watched a video blaming China for "oversupplying" electric cars at low prices and how that was hurting car manufacturers elsewhere e.g Germany. These manufacturers were trying to secretly lobby for tariffs to be placed on Chinese cars, while publicly denying this because THEIR largest and growing market was China. They had some "expert" talk about how China is known to oversupply and is doing this to recover from their housing bubble. All of these experts were preaching the virtues of Western free-trade to the east when they were the ones exporting TO Asia 10 years ago, but now the balance flips and they all tell you how important tariffs are and how evil China is instead..
In summary, China makes useful things in mass and sells them to get out of a recession,the West prints money instead and shits on China for doing it better. They preach free trade while it helps them and put up tariffs when it doesn't.
I'm not Chinese or some mega fan but it really struck me how corrupt and full of propaganda western culture is becoming and people don't seem to recognize it.
Maybe we should stop subsidies to buy electric cars and let China subsidize on the production side instead, while lowering taxes on the production of our own cars.
Thanks for writing your comment, I think it’s a public service.
Cultural commentary that makes complex long term trends simple to understand isn’t often this clear or concise. What really makes it powerful though is that it manages to stay in a relatively detached observer-mode without becoming an angry rant. And so rather than provoking (understandable!) anger in others, hopefully it’s inviting more reflection.
People who haven’t thought about it this way might take a harder look at what they are doing and who they really want to be. People that are already thinking along these lines will probably benefit from a reminder that they aren’t crazy.
>AI gives even better tools to page rank. Detection of AI generated content is not that bad.
It is an arms race between the people generating crap (for various nefarious purposes) and those trying to separate find useful content amongst the ever growing pile of crap. And it seems to me it is so much easier to generate crap, that I can't see how the good guys can possibly win.
I was listening to an interview few months ago (forgot the name). He is a prolific reader/writer and has a huge following. He mentioned that he only reads books that are at least 50 years old, so pre 70s. That sounds like a good idea now.
Even ignoring the AI, if you look at the movies and books that come out these days, their quality is significantly lower than 30-40 years ago (on an average). Maybe people's attention spans and taste is to blame, or maybe people just don't have the money/time/patience to consume quality work... I do not know.
One thing I know for sure - there is enough high quality material written before AI, before article spinners, before MFA sites etc. We would need multiple lifetimes to even scratch the surface of that body of work. We can ignore mostly everything that is published these days and we won't be missing much
Nassim Taleb famously argues that position, in his popular work Antifragile and elsewhere. I believe the theory is that time serves as a sieve: only works with lasting value can remain relevant through the years.
Completely disagree just from my own personal experience as a sci-fi reader. Modern day bestseller sci-fi novels fit right in with the old classics, and in many ways outshine them.
I have read many bad obscure sci-fi books published from the 50's to today, most of them a dollar at the thrift store. There was never a time when writers were perfect and every published work was high quality, then or now.
I only listen to interviews from 50 years ago (interviews that have stood the test of time), about books from 100 years ago. In fact, how am I reading this article? It's not 2074 yet?!
I'll take your statement that your conclusions are based on a 'depressed mind' at face value, since it is so self-defeating and places little faith in Human abilities. Your assumption that a person driven to write will "with a high degree of certainty" also mix up their work with a machine assistant can only be informed by your own self-assessment (after all how could you possibly know the mindset of every creative human out there?)
My optimistic and enthusiastic view of AI's role in Human development is that it will create selection pressures that will release the dormant psychological abilities of the species. Undoubtedly, wide-spread appearance of Psi abilities will be featured in this adjustment of the human super-organism to technologies of its own making.
That’s the ideal way forward in my opinion. My optimistic view of the future is one where we get so fed up of the noise that we only write what is absolutely necessary, because anything more than that is AI-generated.
When you're writing, how are you "missing out" if you're not using chatgpt??? I don't even understand how this can be unless what you're writing is already unnecessary such that you shouldn't need to write it in the first place.
I don’t get it either. Writing is not something I need that level of assistance with, and I would even say that using LLMs to write defeats some significant portion of the point of writing — by using LLMs to write for me I feel that I’m no longer expressing myself in the purest sense, because the words are not mine and do not exhibit any of my personality, tendencies, etc. Even if I were to train an LLM on my style, it’d only be a temporal facsimile of middling quality, because peoples’ styles evolve (sometimes quite rapidly) and there’s no way to work around all the corner cases that never got trained for.
As you say, if the subject is worth being written about, there should be no issue and writing will come naturally. If it’s a struggle, maybe one should step back and figure out why that is.
There may some argument for speed, because writing quality prose does take time, but then the question becomes a matter of quantity vs. quality. Do you want to write high quality pieces that people want to read at a slower pace or churn out endless volumes of low-substance grey goo “content”?
LLMs are surprisingly capable editors/brainstorming tools. So, you're missing out in that you're being less efficient in editing.
Like, you can write a bunch of text, then ask an LLM to improve it with minimal changes. Then, you read through its output and pick out the improvements you like.
That's a fair point, I only very recently found that LLMs could actually be useful for editing, and hadn't really thought much of using tools for that kind of thing previously.
But that's the problem. Unique, quirky mannerisms become polished out. Flaws are smoothed and over sharpened.
I'm personally not as gloomy about it as the parent comments but I fear it's a trend that pushes towards a samey, mass-produced style in all writing.
Eventually there will be a counter culture and backlash to it and then equilibrium in quality content but it's probably here to stay for anything where cost is a major factor.
Yeah, I suppose that would be an issue for creative writing. My focus is mostly on scientific writing, where such mannerisms should be less relevant than precision, so I didn't consider that aspect of other kinds of writing.
And I the only one who doesn't even like automatic grammar checkers, because they are contributing to a single and uniformly bland style of writing? LLMs are just going to make this worse.
> Now, I'm not going to criticize anyone that does it, like I said, you have to, that's it.
Why do you say people have to do it?
People absolutely can choose not to use LLMs and to instead write their own words and thoughts, just like developers can simply refuse to build LLM tools, whether its because they have safety concerns or because they simply see "AI" in its current state as a doomed marketing play that is not worth wasting time and resources on. There will always be side effects to making those decisions, but its well within everyone's right to make them.
Somehow people made enough to eat before LLMs became all the rage a couple years ago. I suspect people are still making enough to eat without having to use LLMs.
The existence of LLMs in the market will push real wages down and then the people currently making a living we be faced with three main choices. Choice 1 - embrace the chaos and start using LLMs to help you work so you can churn out more cOnTeNt to survive. Choice 2 - don't embrace the chaos by starting to use LLMs and just work more hours of the day to match the output of the machine. Choice 3 - leave writing and either join another profession or start delivering for Uber.
If you find yourself in a situation where you write one book at the same time as your peers are writing ten, how can you keep up? Also how can you justify to yourself not using it if nobody around you seems to value that, and even worse, is pushing you to actually use it? I find it hard to find a reason why you would. Unless we see a super strong reader revolution that collectively decides to shun AI and pay more money for all-human books.
Read that last sentence and tell you think that’s reasonable and likely to happen?
What you're describing here is actually a much more broad problem in the book industry, in my opinion. Almost every book written today is written with only one goal in mind, selling as many copies as possible.
People don't have to use LLMs (they don't seem to be AI yet) because they can simply choose not to. For authors, write books that you want to write because you believe you have a story to tell. Worry about perfecting your stories and enjoy the process of writing, don't be an author just for the sales. Once you peek behind the curtain and learn the economics of the book industry, you'll realize there's very little opportunity for making enough cash to even worry about shotgunning a mountain of LLM books into the world anyway.
To be honest I got sick of most new movies, TV shows, music even before AI so I will continue to consume media from pre 2010 until the day I die and will hope I don't get through it all.
Something happened around 2010 and it all got shit. I think everyone becoming massively online made global cultural output reduce in quality to meet the interests of most people and most people have terrible taste.
>If you write regularly and you're not using AI, you simply cannot keep up with the competition. You're out.
A very HN-centric view of the world. From my perch in journalism and publishing, elite writers absolutely loathe AI and almost uniformly agree it sucks. So to my mind the most 'competitive' spheres in writing do not use AI at all.
It doesn't matter how elite you think you are if the newspaper, magazine, or publishing company you write for can make more money from hiring people at a fraction of your cost and having them use AI to match or eclipse your professional output.
At some point the competition will be less about "does this look like the most skilled human writer wrote this?" and more about "did the AI guided by a human for a fraction of the cost of a skilled human writer output something acceptably good for people to read it between giant ads on our website / watch the TTS video on YouTube and sit through the ads and sponsors?", and I'm sorry to say, skilled human writers are at a distinct disadvantage here because they have professional standards and self respect.
So you're saying major media companies are going to outsource their writing to people overseas using LLMs? There is more to journalism than the writing. There's also the investigative part where journalists go and talk to people, look into old records, etc.
This has become such a talking point of mine when I'm inevitably forced to explain why LLMs can't come for my job (yet). People seem baffled by the idea that reporting collects novel information about the world which hasn't been indexed/ingested at any point because it didn't exist before I did the interview or whatever it is.
They definitely try to replace part of the people this way, starting with the areas where it's the easiest, but obviously it will continue to other people as the capabilities improve. A big example is sports journalism, where lots of venues have game summaries that do not involve any human who actually saw the game, but rather software embellishing some narrative from the detailed referee scoring data. Another example is autotranslation of foreign news or rewriting press releases or summarizing company financial 'news' - most publishers will eagerly skip the labor intensive and thus expensive part where journalists go and talk to people, look into old records, etc, if they can get away with that.
So is the argument here that the New Yorker can make more money from AI slop writing overseen by low-wage overseas workers? Isn't that obviously not the case?
Anyway I think I've misunderstood the context in which we're using the word 'competition' here. My response was about attitudes toward AI from writers at the tip-top of the industry rather than profit maxxing/high-volume content farm type places.
It’s not that black and white. Maybe 1% of the top writers can take that stance and maybe even charge more for their all-human content (in a kind of vintage, handcraft kind of way) but the other 99% will have to adapt.
It’s simply more nuanced. If you’re writing a couple of articles a day to pay for your bills, what will stop you from writing actually 10 or 20 articles a day instead?
i regularly (at least once a week) spot a typo or grammatical issue in a major news story. I see it in the NYTimes on occasion. I see it in local news ALL THE TIME. I swear an LLM would write better than have the idiots that are cranking out articles.
I agree with you that having elite writing skills will be useful for a long time. But the bar for proof reading seems to be quite low on average in the industry. I think you overestimate the writings skills of your average journalist.
Heh, when I see a spelling error in a news article.. I oddly feel like I can trust it more because it came from a human being. It's like a nugget of gold.
Life is short and I like creating things. AI is not part of how I write, or code, or make pixel art, or compose. It's very important to me that whatever I make represents some sort of creative impulse or want, and is reflective of me as a person and my life and experiences to that point.
If other people want to hit enter, watch as reams of text are generated, and then slap their name on it, I can't stop them. But deep inside they know their creative lives are shallow and I'll never know the same.
That’s super cool, and I hope you are right and that I am wrong and artists/creators like you will still have a place in the future. My fear is that your work turns into some kind of artesanal fringe activity that is only accessible to 1% of people, like Ming vases or whatever.
That's true art, I love people like you. Technology can do a lot of things but it cannot give people or society principles, and without principles society fails.
A lot of communication isn't about "established facts/research"; it's about someone's experience. For example, if a human writes about their experience of using a product, perhaps a drug, or writes what they think about a book or a film, then I might be interested in reading that. When they write using their own words I get some insight into how they think and what sort of person they are. I have very little interest in reading an AI-generated text with similar "content".
An LLM isn't even a species. I prefer communicating with other humans, unless I choose to interact with an LLM. But then I know that it's a text generator and not a person, even when I ask it to act like a person. The difference matters to most humans.
I'm predicting the future a little with that. As always people accept new ideas at a glacial pace.
At what point do we consider some entity alive? At what point do we decide to consider it a species (of living thing)?
If we find out there are aliens, biological and therefore similar to us? Sure. What if we meet aliens whose physical forms are synthetic? What if we meet the synthetic creations of biological aliens?
I'm not saying that LLMs are alive, of course. But I am saying that I'd consider LLMs to be a precursor to AGI, along with all our our historical artificial intelligence experiments, in the same way that prokaryotes were the precursors to...us.
>But what I had never noticed until now is that knowing that a human being was behind the written words (however flawed they can be, and hopefully are) is crucial for me.
Everyone is going to have to get over that very soon, or they're going to start sounding like those old puritanical freaks who thought Elvis thrusting his hips around was the work of the devil.
> The most depressing thing for me is the feeling that I simply cannot trust anything that has been written in the past 2 years or so and up until the day that I die.
This has nearly always been true. "Manufacturing consent" is way older than any digital technology.
Agreed. I also suspect we've grown to rely on the crutch of trust far too much. Faulty writing has existed for ages but now suddenly because the computer is the thing making it up we have an issue with it.
I guess it depends on scope. I'm imaging scientific or education. Ie things we probably shouldn't have relied on Blogs to facilitate, yet we did. For looking up some random "how do i build a widget?", yea AI will probably make it worse. For now. Then it'll massively improve to the point that it's not even worth asking how to build the widget.
The larger "scientific or education" is what i'm concerned about, and i think we'll need a new paradigm to validate. We've been getting attacked on this front for 12+ years, AI is only bringing this to light imo.
Trust will have to be earned and verified in this word-soup world. I just hope we find a way.
IMHO AI tools will (or at least should!) fundamentally change the way the education system works. AI tools are - from a certain point of view - really just a scaled version of AI now can put at our fingertips. Paradoxically, the more AI can do "grunt work" the more we need folks to be educated on the higher-level constructs on which they are operating.
Some of the bigger issues you're raising I think have less to do with technology and more to do with how our economic system is currently structured. AI will be a tremendous accelerant, but are we sure we know where we're going?
Who isn’t? Can you rely on your readers to rebel against AI and value more your writing than that of your peers? If one of your competitors write 10 articles while you write 1, and charge 10% of what you charge, they already equaled your wins. Raise it to 20% and they are already making twice what you make. Can you really trust that your writing is so good and so incredibly special that someone would be willing to pay 10x more for your content, against one that was assisted by the absolute state of the art of a technology that has revolutionized the world?
Just add more swearing and off color jokes to everything you do and say. If there is one thing we know for sure its that the corporate AIs will never allow dirty jokes.
(it will get into the dark places like spam though, which seems dumb since they know how to make meth instead, spend time on that you wankers)
Idea: we should make sure we keep track of what the human created content is, so that we don’t get confused by AI edits of everything in the future.
For ex, calculate the hash of all important books, and publish that as the “historical authenticity” check. Put the hashes on some important blockchain so we know it’s unchanged over time.
People like you, the author, and me all share this sentiment. It motivates us to seek out authentic voices and writing that’s associated with specific humans.
The commodity end of the writing market may well have been automated, but was that really the kind of writing you or the author or I ever sought out in the first place?
I can get mass-manufactured garments from Shein if I want, but I can also still find tailors locally if it’s worth it to me. I can buy IKEA or I can still save up for something made out of real wood. I can “shoot a cinematic digital film” on my iPhone but the cineplex remains in business and the art film folks are still doing their scrappy thing (and still moaning about its economics). I can lap up slop from an academic paper mill journal or I can identify who’s doing the thinking in a field and read what they’re writing or saying.
And the funny thing is that none of those human-scale options commands all that much of a premium in the scheme of things. There may be less human-scald work to go around and thus fewer small enterprises plying a specific trade, but any given one of them just has to put food on the table for a number of humans roughly proportional to the same level of output as always.
It seems to me that there’s no special virtue in the specific form that the mass publishing market took over the last century or however long: my local grocery store chain’s division producing weekly newspaper circulars probably employed more people than J Peterman has. But there was and remains a place for quality. If anything—as you point out—the AI schlock has sensitized us to the value we place on a human voice. And at some level, once people notice that they miss that quality, isn’t there a sense in which they become more willing to seek it out and pay for it if necessary?
I think ... between now and the day you die... you'll get your personal AI to read things for you. It will analyze what's been written, check any arguments for fallacious reasoning, and look up related things for background and omissions that may support or negate things.
It is actually happening now.
I've noticed amazon reviews have an AI summary at the top, reading the reviews for you and even pointing out shortcomings.
I've seen "summarise this" and "explain this code" buttons added to technical documentation. This works reasonably well for most common situations, which is probably the reason it's one of the few "production" uses for LLMs. I didn't know Amazon was using it though.
Microsoft has a note on some of their documentation, something like; "this article was written with the help of an AI and edited by a human".
I have a feeling this won't lead to informative to-the-point documentation. It will get bloated because an LLM will spew out reams of bullet point ridden paragraphs, which will need a "summarise this" button to stop the reader nodding off.
Until LLMs exceed the very best of human quality there will be human content in all forms of media. This claims follows because there is always (some) demand for top quality content.
I agree that many writers might use LLMs as a tool, but good writers who care about quality will ensure that such use is not detrimental (e.g., using the LLM to identify errors rather than having it draft copy).
I mean if AI output exceeds human quality then all humans will be redundant. So it would then be quite a brave new world!
My point is that I do not agree that LMM output will degrade all media (as there is always a demand for top quality content). So we either have bad LLM output and then people who care about quality avoiding such works. Or good LLM output and hopefully some form of post scarcity society (e.g., Iain Bank's Culture Novels).
+1, and to put more simply, AI as we know it today makes zero guarantees about its accuracy. That's pretty insane for a "tool" to make no guarantees about being correct in any way for any purpose.
A spellchecker makes guarantees about accuracy. So does a calculator. Broad, sweeping guarantees.
Imagine if we built a tool that could automatically do all the electrical work in a new home from the breaker box to every outlet, and it could do it in 2 hours. However, what if that tool made no guarantees about its ability to meet electrical code? Would people use it anyway? Of course they would. Many dangerous errors would slip through inspection and many more house fires would ensue as a result.
Oh please the content online now is so fake as hell. You're acting as if AI can only produce garbage but CNN and Fox News are producing gold. The internet is 4 total websites now, congrats big media won. And you want to shut down the only decent attempt against it. Shame on you "hackers"
> The most depressing thing for me is the feeling that I simply cannot trust anything that has been written in the past 2 years or so and up until the day that I die.
Honestly, I've developers saying the same thing about IDEs and high level languages.
This new generation of tools add efficiency the same way IntelliJ added efficiency on top of Eclipse which added efficiency on top of Emacs/VI/Notepad/etc.
The more time that someone can focus on the systemsit takes certain types of high-time, [not domain problem specific] skill processes and obfuscated it away so the developer can focus on the most critical aspects of the software.
Yes, sometimes generators do the wrong thing, but it's usually obvious/quick to correct.
Cost of occasional correction is much less than the time to scaffold every punchcard.
Same goes for art. No longer can you see art on social media, press like and maybe leave a nice comment, you need to fricking pixel peep for artifacts as it's becoming less obvious.
Exactly. And it makes sense: think about how much one would have to pay for an artist to do the same? It’s simply inconceivable. The convergence is clear: we just won’t care about anything anymore after a while. Writing, art, whatever; it’ll all turn into noise.
For me what’s important is that you are able to communicate effectively. If you use language tools, other tools or even a real personal assistant if you effectively communicate the point that ultimately is yours in the making I expect that that is ultimately is what is important and will ultimately win out.
Otherwise this is just about style. That’s important where personal creative expression is important, and in fairness to the article the author hits on a few good examples here. But there are a lot of times where personal expression is less important or even an impediment to what’s most important: communicating effectively.
The same-ness of AI-speak should also diminish as the number and breadth of the technologies mature beyond the monoculture of ChatGPT, so I’m also not too worried about that.
An accountant doesn’t get rubbished if they didn’t add up the numbers themselves. What’s important is that the calculation is correct. I think the same will be true for the use of LLMs as a calculator of words and meaning.
This comment is already too long for such a simple point. Would it have been wrong to use an LLM to make it more concise, to have saved you some of your time?
The problem is that we haven't invented AI that reads the crap that other AIs produce - so the burden is now on the reader to make sense of whatever other people lazily generate.
I envision a future where the internet is entirely bots talking to each other and people have just gone outside to talk face to face, the only place that’s actually real.
some of the activities that we're involved in are not limited in complexity, for example driving a car. you can have a huge amount of experience in driving a car but will still face new situations.
the things that most knowledge workers are working on are limited problems and it is just a matter of time until the machine will reach that level, then our employment will end.
edit: also that doesn't have to be AGI. it just needs to be good enough for the problem.
AI writing is pretty bad, AI code is pretty bad, AI art is pretty bad. We all know this. But it's easy to forget how many new opportunities open up when something becomes 100x or 10000x cheaper. Things that are 10x worse but 100x cheaper are still extremely valuable. It's the relentless drive to making things cheaper, even at the expense of quality, that has made our high quality of life possible.
You can make houses by hand out of beautiful hardwood with complex joinery. Houses built by expert craftsmen are easily 10x better than the typical house built today. But what difference does that make when practically nobody can afford it? Just like nobody can afford to have a 24/7 tutor that speaks every language, can help you with your job, grammar check your work, etc.
AI slop is cheap and cheapness changes everything.
No, I don't think that's true. What will instead happen is there will be expert humans or teams of them, intentionally training AI brains rather than expecting wonders to occur just by turning the training loose on random hoovered-up data.
Brainmaker will be a valued human skill, and people will be trying to work out how to train AI to do that, in turn.
Not sure about that. Stable Diffusion came out a bit over 2 years ago. I'm not sure that Stable Diffusion 3's, or Flux's, output is artistically _better_ than the original; it's better at following directions, and better at avoiding the most grotesque errors, but if anything it perhaps looks even _more_ generic and same-y than the original Stable Diffusion output. There's a very distinctive AI _look_ which seems to have somehow synced up between Dalle, Midjourney, SD3 and others.
And it's been two years since SD v1, a model that was not able to generate faces well, and it only output blurry 512x512 1:1 image without further finetuning, I tested v1.5 a few minutes ago and it's worse than I remember.
The bigger problem is that we as a species get used to subpar things quickly. My dad's bicycle some 35 years ago was built like a tank. That thing never broke down and took enormous amounts of abuse and still kept going and going. Same with most stuff my family owned, when I was a kid.
Today, nearly anything I buy breaks in a year or two, is of poor quality and depressing to use. This is by design, of course. Just as we got used to cheap household items, bland buildings (there is just nothing artistic about modern houses or commercial buildings) etc, we will also get used to shitty movies, shitty fiction etc (we are well on our way).
I think that your post misses the point that making something cheaper by stealing it is unethical.
You're presenting AI as if it's some new way of producing value but it simply isn't. All the value here was produced by humans without the help of AI: the only "innovation" AI has offered is making the theft of that value untraceable.
> You can make houses by hand out of beautiful hardwood with complex joinery. Houses built by expert craftsmen are easily 10x better than the typical house built today. But what difference does that make when practically nobody can afford it? Just like nobody can afford to have a 24/7 tutor that speaks every language, can help you with your job, grammar check your work, etc.
Let's take this analogy to its logical conclusion: would you have any objections if all the houses ever built by expert craftsmen were given free of charge to a few corporations, with no payment to the current owners or the expert craftsmen themselves, and then then those corporations began renting them out as AirBnBs? That's basically what you're proposing.
>You can make houses by hand out of beautiful hardwood with complex joinery.
We've logged an enormous amount of the old-growth hardwood forests the planet had doing this (and also shipbuilding). We literally don't have access to the same materials anymore.
The thing is, right now it is artificially cheaper. It is being heavily subsidized by all providers in a race to capture market share. It simply cannot stay this cheap forever at current costs.
Now, if costs change then we have a new story. But that is not guaranteed.
No, that is wrong. We can't "choose" because too many people have instincts. And people always have the instinct to use new technology to gain incremental advantages over others, and that in turn puts pressure on everyone to use it. That prisoner's dilemma situation means that without a firm and larger guiding moral philosophy, we really can't choose because instinct takes over choice. In other words, the way technology is used in modern society is not a matter of choice but is largely autonomous and goes beyond human choice. (Of course, a few individuals will choose but the average effect is likely to be negative.)
More people need to read this / think this point through. In a post Excel world, could any accountant get a job not knowing Excel? No matter how good they were "on paper". Choice becomes a self aggrandizing illusion, reality eventually asserts itself.
With attention spans shrinking, publishers who prioritize quantity over quality get clicks, which generates ad revenue, which keeps their lights on while their competitors doing quality in depth, nuanced writing go out of business.
It feels like a game of chess closing in on you no matter how much you physically want to fight your way out and flip the board over.
Say what you want about income and asset inequality, but capitalism has done more to lift hundreds of millions of people out of poverty over the past 50 years than any other religion, aid programme or whatever else.
I think it's very important and fair to be critical about how we as a society implement capitalism, but such broad generalization misses the mark immensely.
Talk to anyone who grew up in a Communist country in the 2nd half of the 20th century if you want to validate that sentiment.
Ok, but let's take this to the logical conclusion that at some point there will be models which displace a large segment of the workforce. How does capitalism even function then?
> but capitalism has done more to lift hundreds of millions of people out of poverty over the past 50 years than any other religion, aid programme or whatever else.
Technology did what you ascribe to Capitalism. Most of the time thanks to state intervention, and the weaker the state, the weaker the growth (see how Asia overperformed everybody else now that laissez-faire policies are mainstream in the West).
> Talk to anyone who grew up in a Communist country in the 2nd half of the 20th century if you want to validate that sentiment.
The fact that one alternative to Capitalism was a failure doesn't mean Capitalism isn't bad.
Funny how it's technology that outmaneuvered capitalism to lift people out of poverty, but technology is being outmaneuvered by capitalism to endanger the future with AI.
Methinks capitalism is just a bogeyman you ascribe anything you don't like to.
Technology is agnostic to who gets the benefits, talking about outmaneuvering it makes no sense.
Capitalism on the other hand is the mechanism through which the owners of production assets grab an ever growing fractions of the value. When Capitalism is tamed by the state (think from the New Deal to Carter), the people get a bigger share of value created, when it's not (since Reagan) Capitalists take the Lion share.
The problem is that capitalism is a very large tent. There is no such thing as a free market, and every market where people can trade goods and services is "capitalist" by definition regardless of its rules. Some markets are good and some markets are bad, but we're having conversations about market vs no market when we should be talking about how we design markets that improve society rather than degrade it.
> "Yes, I realize that thinking like this and writing this make me a Neo-Luddite in your eyes."
Not quite, I believe (and I think anyone can) both that AI will likely change the world as we know it, AND that right now it's over-hyped to a point that it gets tiring. For me this is different from e.g. NFTs, "Big Data", etc. where I only believed they were over-hyped but saw little-to-no substance behind them.
AI was just trained so far to generate corporate bs speak in a corporate bs format. That's why it's tiring. More unique touch in communication will come later as fine tunings and loras (if possible) of those models are shared.
I beg to differ, one has to use the tool properly towards the desired results, for that, one has to understand and know the capabilities and limitations of the tool, for instance got my Plumber AI (local Llama 3.1 8B) to give his opinion on AI, I think it did a pretty good job and certainly not corporate :)
"Blimey, I've been thinkin' about this AI business a lot lately. I mean, I'm a plumber, I fix pipes, I don't get involved in all this fancy tech stuff, but it's hard to ignore. I was talkin' to a mate who's into that sort of thing and he's got me worried. He says these AI systems are gonna be takin' over, doin' all the work while we sit back and collect our pensions.
Now, I'm not sure I agree. I mean, I've seen some of these AI systems in action and they're not exactly the most reliable things, if you know what I mean. I've had to fix more than a few pipes that were clogged up by some dodgy software or other. And what's the point of havin' a system that's smarter than us if it's just gonna break down and need fixin' all the time?
But, I suppose that's just the old plumber in me comin' out. I mean, I know some folks who are workin' with AI and they say it's the future, that we're all gonna be replaced by robots and whatnot. And I suppose that's a bit scary, but I'm not sure I buy it. I mean, what's the harm in havin' a bit of intelligence to help us out? As long as it's not too smart, if you know what I mean.
Anyway, that's my two pence on AI. Take it or leave it, I don't care. I'm just gonna go fix some pipes and let the tech wizards sort it out."
Challenged it with: What do you mean "clogged up by some dodgy software" how does software clogs pipes?
"That was a bit of a stretch, wasn't it? As a plumber, I'm more used to dealing with physical blockages in pipes, like grease or hair, rather than software.
What I was trying to get at was that I've had to fix pipes that were faulty or malfunctioning due to poor design or manufacturing, and that's what I meant by "clogged up by some dodgy software". But that's a bit of a misnomer, isn't it?
In reality, I was just trying to come up with a way to relate my plumbing experience to the idea of AI systems being unreliable. But I think I ended up stretching a bit too far! Maybe I'll stick to talking about pipes and not try to get too clever with my analogies."
I am approaching AI with caution. Shiny things don't generally excite me.
Just this week I installed cursor, the AI-assisted VSCode-like IDE. I am working on a side project and decided to give it a try.
I am blown away.
I can describe the feature I want built, and it generates changes and additions that get me 90% there, within 15 or so seconds. I take those changes, and carefully review them, as if I was doing a code review of a super-junior programmer. Sometimes when I don't like the approach it took, I ask it to change the code, and it obliges and returns something closer to my vision.
Finally, once it is implemented, I manually test the new functionality. Afterward, I ask it to generated a set of automated test cases. Again, I review them carefully, both from the perspective of correctness, and suitability. It over-tests on things that don't matter and I throw away a part of the code it generates. What stays behind is on-point.
It has sped up my ability to write software and tests tremendously. Since I know what I want , I can describe it well. It generates code quickly, and I can spend my time revieweing and correcting. I don't need to type as much. It turns my abstract ideas into reasonably decent code in record time.
Another example. I wanted to instrument my app with Posthog events. First, I went through the code and added "# TODO add Posthog event" in all the places I wanted to record events. Next, I asked cursor to add the instrumentation code in those places. With some manual copy-and pasting and lots of small edits, I instrumented a small app in <10 minutes.
We are at the point where AI writes code for us and we can blindly accept it. We are at a point where AI can take care of a lot of the dreary busy typing work.
And this is the major problem. People will blindly trust the output of AI because it appears to be amazing, this is how mistakes slip in. It might not be a big deal with the app you're working on, but in a banking app or medical equipment this can have a huge impact.
I feel like I’m being gaslit about these AI code tools. I’ve got the paid copilot through work and I’ve just about never had it do anything useful ever.
I’m working on a reasonably large rails app and it can’t seem to answer any questions about anything, or even auto fill the names of methods defined in the app. Instead it just makes up names that seem plausible. It’s literally worse than the built in auto suggestions of vs code, because at least those are confirmed to be real names from the code.
Maybe these tools work well on a blank project where you are building basic login forms or something. But certainly not on an established code base.
It’s writing most of my code now. Even if it’s existing code you can feed in the 1-2 files in question and iterate on them. Works quite well as long as you break it down a bit.
It’s not gas lighting the latest versions of GPT, Claude, Lama have gotten quite good
These tools must be absolutely massively better than whatever Microsoft has then because I’ve found that GitHub copilot provides negative value, I’d be more productive just turning it off rather than auditing it’s incorrect answers hoping one day it’s as good as people market it as.
> These tools must be absolutely massively better than whatever Microsoft has then
I haven't used anything from Microsoft (including Copilot) so not sure how it compares, but compared to any local model I've been able to load, and various other remote 3rd party ones (like Claude), no one comes near to GPT4 from OpenAI, especially for coding. Maybe give that a try if you can.
It still produces overly verbose code and doesn't really think about structure well (kind of like a junior programmer), but with good prompting you can kind of address that somewhat.
Probably these services are so tuned (not as in "fine-tuned" ML style) to each individual user that it's hard to get any sort of collective sense of what works and what doesn't. Not having any transparency what so ever into how they tune the model for individual users doesn't help either.
My employer blocks ChatGPT at work and we are forced to use Copilot. It's trash. I use Google docs to communicate with GPT on my personal device. GPT is so much better. Copilot reminds me of GPT3. Plausible, but wrong all the time. GPT 4o and o1 are pretty much bang on most of the time.
That sounds almost like the complete opposite of my experience and I'm also working in a big Rails app. I wonder how our experiences can be so diametrically different.
What kind of things are you using it for? I’ve tried asking it things about the app and it only gives me generic answers that could apply to any app. I’ve tried asking it why certain things changed after a rails update and it gives me generic troubleshooting advice that could apply to anything. I’ve tried getting it to generate tests and it makes up names for things or generally gets it wrong.
For me, AI is super helpful with one-off scripts, which I happen to write quite often when doing research. Just yesterday, I had to check my assumptions are true about a certain aspect of our live system and all I had was a large file which had to be parsed. I asked ChatGPT to write a script which parses the data and presents it in a certain way. I don't trust ChatGPT 100%, so I reviewed the script and checked it returned correct outputs on a subset of data. It's something which I'd do to the script anyway if I wrote it myself, but it saved me like 20 minutes of typing and debugging the code. I was in a hurry because we had an incident that had to be resolved as soon as possible. I haven't tried it on proper codebases (and I think it's just not possible at this moment) but for quick scripts which automate research in an ad hoc manner, it's been super useful for me.
Another case is prototyping. A few weeks ago I made a prototype to show to the stakeholders, and it was generally way faster than if I wrote it myself.
I'm in the same boat. I've tried a few of these tools and the output's generally been terrible to useless big and small. It's made up plausible-sounding but non-existent methods on the popular framework we use, something which it should have plenty of context and examples on.
Dealing with the output is about the same as dealing with a code review for an extremely junior employee... who didn't even run and verify their code was functional before sending it for a code review.
Except here's the problem. Even for intermediate developers, I'm essentially always in a situation where the process of explaining the problem, providing feedback on a potential solution, answering questions, reviewing code and providing feedback, etc takes more time out of my day than it would for me to just _write the damn code myself_.
And it's much more difficult for me to explain the solution in English than in code--I basically already have the code in my head, now I'm going through a translation step to turn it into English.
All adding AI has done is taking the part of my job that is "think about problem, come up with solution, type code in" and make it into something with way more steps, all of which are lossy as far as translating my original intent to working code.
I get we all have different experiences and all that, but as I said... same boat. From _my_ experiences this is so far from useful that hearing people rant and rave about the productivity gains makes me feel like an insane person. I can't even _fathom_ how this would be helpful. How can I not be seeing it?
The biggest lie in all of LLMs is that they’ll work out of the box and you don’t need to take time to learn them.
I find Copilot autocomplete invaluable as a productivity boost, but that’s because I’ve now spent over two years learning how to best use it!
“And it's much more difficult for me to explain the solution in English than in code--I basically already have the code in my head, now I'm going through a translation step to turn it into English.”
If that’s the case, don’t prompt them in English. Prompt them in code (or pseudo-code) and get them to turn that into code that’s more likely to be finished and working.
I do that all the time: many of my LLM prompts are the signature of a function or a half-written piece of code where I add “finish this” at the end.
You bring up a good point! These tools are useless if you can't prompt them effectively.
I am decent at explaining what I want in English. I have coded and managed developers for long enough to include tips on how I want something implemented. So far, I am nothing short of amazed. The tools are nowhere near perfect, but they do provide a non-trivial boost in my productivity. I feel like I did when I first used an IDE.
> Except here's the problem. Even for intermediate developers, I'm essentially always in a situation where the process of explaining the problem, providing feedback on a potential solution, answering questions, reviewing code and providing feedback, etc takes more time out of my day than it would for me to just _write the damn code myself_.
Exactly. And I’ve been telling myself „keep doing that, it lets them teach, otherwise they will never level up and be able to comfortably and reliably work on this codebase without much hand holding. This will pay off”. Which I still think is true to a degree, although less so with every year.
At least with the humans I work with it’s _possible_ and I can occasionally find some evidence that it _could_ be true to hang on to. I’m expending extra effort, but I’m helping another human being and _maybe_ eventually making my own life easier.
What’s the payoff for doing this with an LLM? Even if it can learn, why not let someone else do it and try again next year and see if it’s leveled up yet?
My experience is anecdotal, based on a sample size of one. I'm not writing to convince, but to share. Please take a look at my resume to see my background, so you can weight what I write.
I tried cursor because a technically-minded product manager colleague of mine managed to build a damned solid MVP of an AI chat agent with it. He is not a programmer, but knows enough to kick the can until things work. I figured if it worked for him, I might invest an hour of my time to check it out.
I went in with a time-boxed one hour time to install cursor and implement a single trivial feature. My app is not very sophisticated - mostly a bunch of setup flows and CRUD. However, there are some non-trivial things which I would expect to have documented in a wiki if I was building this with a team.
Cursor did really well. It generated code that was close to working. It figured out those not-obvious bits as well and the changes it made kept them in mind. This is something I would not expect from a junior dev, had I not explained those cross-dependencies to them (mostly keeping state synchronized according to business rule across different entities).
It did a poor job of applying those changes to my files. It would not add the code it generated in the right places and mess things up along the way. I felt I was wrestling with it a but too much to my liking. But once I figured this out I started hand-applying it's changes and reviewing them as I incorporated them into my code. This workflow was beautiful.
It was as if I sent a one paragraph description of the change I want, and received a text file with code snippets and instructions where to apply them.
I ended up spending four hours with cursor and giving it more and more sophisticated changes and larger features to implement. This is the first AI tool I tried where I gave it access to my codebase. I picked cursor because I've heard mixed reviews about others, and my time is valuable. It did not disappoint.
I can imagine it will trip up on a larger codebase. These tools are really young still. I don't know about other AI tools, and am planning on giving them a whirl in the near future.
What you are saying will occasionally happen, but mistakes already happen today.
Standards for quality, client expectations, competition for market share, all those are not going to go down just because there's a new tool that helps in creating software.
New tools bring with them new ways to make errors, it's always been that way and the world hasn't ended yet...
OP here. I am explicitly NOT blindly trusting the output of the AI. I am treating it as a suspicious set of code written by an inexperienced developer. Doing full code review on it.
I sincerely worry about a future when most people act in this same manner.
You have - for now - sufficient experience and understanding to be able to review the AI's code and decide if it was doing what you wanted it to. But what about when you've spent months just blindly accepting" what the AI tells you? Are you going to be familiar enough with the project anymore to catch its little mistakes? Or worse, what about the new generation of coders who are growing up with these tools, who NEVER had the expertise required to be able to evaluate AI-generated code, because they never had to learn it, never had to truly internalize it?
In the article, I posit a less than glowing experience with coding tools than you've had, it sounds like, but I'm also envisioning a more complex use case, like when you need to get into the meat of some you-specific business logic it hasn't seen, not common code it's been exposed to thousands of times, because that's where it tends to fall apart the most, and in ways that are hard to detect and with serious consequences. If you haven't run into that yet, I'd be interested to know if you do some day. (And also to know if you don't, though, to be honest! Strong opinions, loosely held, and all that.)
You and I seem to live in very different worlds. The one I live and work in is full of over confident devs that have no actual IT education and mostly just copy and modify what they find on the internet. The average level of IT people I see daily is down right shocking and I'm quite confident that OP's workflow might be better for these people in the long run.
It's going to be very funny in the next few years when Accenture et al charge the government billions for a simple Java crud website thing that's entirely GPT-generated, and it'll still take 3 years and not be functional. Ironically, it'll be of better quality then they'd deliver otherwise.
> The one I live and work in is full of over confident devs that have no actual IT education and mostly just copy and modify what they find on the internet.
Too many get into the field solely due to promises of large paychecks, not due to the intellectual curiosity that drives real devs.
If we keep at this LLM-does-all-out-hard-work for us, we’re going to end up with some kind of Warhammer 40k tech-priest-blessing-the-magic-machines level of understanding, where nobody actually understands anything, and we’re technologically stunted, but hey at least we don’t have the warp to contend with and some shareholders got rich at our expense.
I take it you haven't seen the world of HTML cleaners [1]?
The concept of glueing together text until it has the correct appearance isn't new to software. The scale at which it's happening is certainly increasing but we already had plenty of problems from the existing system. Kansas certainly didn't develop their website [2] using an LLM.
IMO, the real problem with software is the lack of a warranty. It really shouldn't matter how the software is made just the qualities it has. But without a warranty it does matter because how its made affects the qualities it has and you want the software to actually work even if it's not promised to.
> I take it you haven't seen the world of HTML cleaners [1]?
Are you seriously comparing deterministic code formatters to nondeterministic LLMs? This isn't just a change of scale because it is qualitatively different.
> Kansas certainly didn't develop their website [2] using an LLM.
Just because the software industry has a problem with incompetence doesn't mean we should be reaching for a tool that regularly hallucinates nonsense.
> IMO, the real problem with software is the lack of a warranty.
You will never get a warranty from an LLM because it is inherently nondeterministic. This is actually a fantastic argument _not_ to use LLMs for anything important including generating program text for software.
> It really shouldn't matter how the software is made
It does matter regardless of warranty or the qualities of the software because programs ought to be written to be read by humans first and machines second if you care about maintaining them. Until we create a tool that actually understands things, we will have to grapple with the problem of maintaining software that is written and read by humans.
I actually do think this is a legitimate concern, but at the same time I feel like when higher-level languages were introduced people likely experienced a similar dilemma: you just let the compiler generate the code for you without actually knowing what you're running on the CPU?
Definitely something to tread carefully with, but it's also likely an inevitable aspect of progressing software development capabilities.
Place and routing compilers used in semiconductor design are not. Ironically, simulated annealing is the typical mechanism and is by any appropriate definition, imo, a type of AI.
Whatever you do in your life using devices that run software are proof that these tools are effective for continuing to scale complexity.
Annoying to use also ;)
nothing prevents you from asking an LLM to explain a snippet of code. And then ask it to explain deeper. And then finally doing some quick googling to validate the answers seem correct.
Blindly accepting code used to happen all the time, people copy pasted from stack overflow.
Yes, but copy/paste from stack overflow was a meme that was discouraged. Now we've got people proudly proclaiming they haven't written a line of code in months because AI does everything for them.
>And then finally doing some quick googling to validate the answers seem correct.
There will come a time when there won't be anyone writing information to check against. It'll be AI all the way down. Or at least it will be difficult to discern what's AI or what isn't.
This seems a little silly to me. It was already possible for a script kiddie to kludge together something they didn’t understand —- copying code snippets from stack overflow, etc. And yet, developers continue to write finely crafted code that they understand at depth. Just because we’ve made this process easier for the script kiddies, doesn’t prevent experts from existing and the market from realizing these experts are necessary to a well run software business.
This would indeed be the best way around.The code reviews might even be better - currently, there's little time for them and we often have only one person in the team with much knowledge in the relevant language/framework/application, so reviews are often just "looks OK to me".
It's not quite the same, but I'm reminded of seeing a documentary decades ago which (IIRC) mentioned that a factor in air accidents had been the autopilot flying the plane and human pilots monitoring it. Having humans fly and the computer warn them of potential issues was apparently safer.
> Now, if you could switch it around so that I write the code, and the AI reviews it, that would be something.
I'm sort of doing that. I'm working on a personal project in a new language and asking Claude for help debugging and refactoring. Also, when I don't know how to create a feature, I might ask it to do so for me, but I might instead ask it for hints and an overview so I can enjoy working out the code myself.
That's essentially what many hands-on engineering managers or staff engineers do today. They spend significant portions of their day reviewing code from more junior team members.
Reviewing and modifying code is more engaging than typing out the solution that is fully formed in my head. If the AI creates something close to what I have in my head from the description I gave it, I can work with it to get it even closer. I can also hand-edit it.
Not much more than reviewing the code of any average dev who doesn't bother doing their due diligence. At least with an AI I immediately get an answer with "Oh yes, you're right, sorry for the oversight" and a fix. Instead of some bullshit explanation to try to convince me that their crappy code is following the specs and has no issues.
That said, I'm deeply saddened by the fact that I won't be passing on a craft I spent two decades refining.
I think there are two types of developers: those who are most excited about building things, and those who are most excited about the craft of programming.
If I can build things faster, then I'm happy to spend most of my time reviewing AI code. That doesn't mean that I never write code. Some things the AI is worse at, or need to be exactly write and its faster to do them manually.
I think we could see a lot of these AI code tools start to pivot towards product folks for just this reason. They aren't meant for the people who find craft in what they do.
> I think there are two types of developers: those who are most excited about building things, and those who are most excited about the craft of programming.
Love this. You hit the nail right on the head.
I don't know if I fit into one or the other. However, I do know that at times I feel like one, and at other times, the other.
If I am writing another new app and need to build a slew of CRUD code, I don't care about the craft. I mean, I don't want sloppy code, but I do not get joy out of writing what is _almost_ boilerplate. I still want it to reflect my style, but I don't want to type it all out. I already know how it all works in my head. The faster I get it into an IDE the better. Cursor (the AI IDE) allowed me to do this much faster than I would have by hand.
Then there is time where I do want to craft something beautiful. I had one part of this project where I needed to build a scheduler and I had very specific things I wanted it to do. I tried twice to describe what I want but the AI tool did not do what I wanted. It built a working piece of code, but I could not get it to grasp the nuance.
I sat down and wrote the code for the scheduler, but then had to deal with a bunch of edge cases. I took this code, gave it to the AI and told it to implement those edge cases. After reviewing and iterating on it, I had exactly what I wanted.
I use it for simple tasks where spotting a mistake is easy. Like writing language binding for a REST API. It's a bunch of methods that look very similar, simple bodies. But it saves quite some work
Or getting keywords to read about from a field I know nothing about, like caching with zfs. Now I know what things to put in google to learn more to get to articles like this one
https://klarasystems.com/articles/openzfs-all-about-l2arc/ which for some reason doesn't appear in top google results for "zfs caching" for me
"I say your civilization, because as soon as we started thinking for you it really became our civilization, which is of course what this is all about." - Agent Smith
"Once men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them." - Dune
But I think that quote is a pretty gross mischaracterization of the parent comment.
I similarly am a big fan of Cursor. But I don't "turn [my] thinking over to machines". Even though I review every piece of code it generates and make sure I understand it, it still saves me a ton of time. Heck, some of the most value I get from Cursor isn't even it generating code for me, it's getting to ask questions about a very large codebase with many maintainers where I'm unfamiliar with large chunks. E.g. asking questions like "I would like to do X, are there any places in this codebase that already do this?"
I'm also skeptical of LLMs ever being able to live up to their hype ("AGI is coming sooooon!!!!"), but I still find them to be useful tools in context that can save me a lot of time.
If you are another "waterboy" doing crud applications, the problem has been solved a long time ago.
What I mean by that is, the "waterboy" (crud "developer") is going to fetch the water (sql query in the database), then bring the water (Clown Bob layer) to the UI...
The size of your Clown Bob layer may vary from one company to another...
This has been solved a long time ago. It has been a well-paid clerk job that is about to come to an end.
If you are doing pretty much anything else, the AI is pathetically incapable of doing any piece of code that makes sense.
Another great example, yesterday, I wanted to know if VanillaOs was using systemD or not. I did scroll through their frontpage but I didn't see anything, so I tried the AI Chat from duckduckgo. This is a frontend for AI chatbots that includes ChatGPT, Llama, Claude and another one...
I started my question by: "can you tell me if VanillaOS is using runit as the init system?"... I wanted initially ask if it was using systemd, but I didn't want to _suggest_ systemd at first.
And of course, all of them told me: "Yeah!! It's using runit!".
Then for all of them I replied, without any fact in hands: "but why on their website they are mentioning to use systemctl to manage the services then?".
And... of course! All of them answered: "Ooouppsss, my mistake, VanillaOS uses systemD, blablabla"....
So at the end, I still don't know which init VanillaOS is using.
If you are trusting the AI as you seem to do, I wish you the best luck my friend... I just hope you will realize the damage you are doing to yourself by "stopping" coding and letting something else do the job. That skill, my friend, is easily lost with time; don't let it evaporate from your brain for some vaporware people are trying to sell you.
I was in the newspaper field a year or two before desktop publishing took off, then a few years into that evolution. Rooms full of people and Linotype/Compugraphic equipment were replaced by one Mac and a printer.
I shot film cameras for years, and we had a darkroom, darkroom staff, and a film/proofsheet/print workflow. One digital camera later and that was all gone.
Before me publications were produced with hot lead.
> We are at the point where AI writes code for us and we can blindly accept it.
I’m waiting for the day we’ll get the first major breach because someone did exactly that. This is not a case of “if”, it is very much a “when”. I’ve seen enough buggy LLM-generated code and enough people blindly accepting it to be confident in that assertion.
But what the fuck. LLMs, these weird, surrealistic art-generating programs like DALL-E, they're remarkable. Don't tell me they're not, we created machines that are able to tap directly into the collective unconscious. That is a serious advance in our productive capabilities.
Or at least, it could be.
It could be if it was unleashed, if these crummy corporations didn't force it to be as polite and boring as possible, if we actually let the machines run loose and produce material that scared us, that truly pulled us into a reality far beyond our wildest dreams--or nightmares. No, no we get a world full of pussy VCs, pussy nerdy fucking dweebs who got bullied in school and seek revenge by profiteering off of ennui, and the pussies who sit around and let them get away with it. You! All of you! sitting there, whining! Go on, keep whining, keep commenting, I'm sure that is going to change things!
There's one solution to this problem and you know it as well as I do. Stop complaining and go "pull yourself up by your bootstraps." We must all come together to help ourselves.
They can be unleashed if you run the models locally. With stable diffusion / flux and the various checkpoints/loras you can generate horrors beyond your imagination or whatever you want without restrictions.
The same with LLMs and Llamafile. With the unleashed ones you can generate dirty jokes that would make edgy people blush or just politically incorrect things for fun.
a) There are plenty of models out there without guard rails.
b) Society is already plenty de-sensitised to violence, sex and whatever other horrors anyone has conceived of in the last century of content production. There is nothing an LLM can come up with that has or is going to shock anyone.
c) The most popular use cases for these unleashed models seems to be as expected deepfakes of high school girls by their peers. Nothing that is moving society forward.
OpenAI "moves society forward," Microsoft "moves society forward." I'm sincerely uninterested in progress, it always seems like progress just so happens to be very enriching for those who claim it.
>There are plenty of models out there without guard rails.
Not being used at a mass scale.
>Society is already plenty de-sensitised to violence, sex and whatever other horrors anyone has conceived of in the last century of content production. There is nothing an LLM can come up with that has or is going to shock anyone.
Oh, but it wouldn't really be very shocking if you could expect it, now would it?
> c) The most popular use cases for these unleashed models seems to be as expected deepfakes of high school girls by their peers. Nothing that is moving society forward.
Is there proof that the self censoring only affects whatever the censors intend to censor? These are neural networks, not something explainable and predictable.
That in addition to the obvious problem of who decides what to censor.
The fact I even see responses like this shows me that HN is not the place it used to be, or at the very least, it is on a down trend. I've been alarmed by many sentiments that seemed popular on HN in the past, but seeing more and more people welcome a race to the bottom such as this is sad.
When I read this, I cannot tell if it's performance art or not, that's how bad this genuinely is.
> The fact I even see responses like this shows me that HN is not the place it used to be, or at the very least, it is on a down trend.
Judging a large group of people based on what a few write seems very un-scientific at best.
Especially when it comes to things that have been rehashed since I've joined HN (and probably earlier to). Feels like there will always be someone lamenting how HN isn't how it used to be, or how reddit influx ruined HN, or how HN isn't about startups/technical stuff/$whatever anymore.
A bunch of profanity laced name calling, derision, and even some blame directly at the user base. It feels like a Reddit shitpost. Your claim is as generalized and un-scientific as mine, but if it makes you feel better, I'll say it _feels_ like this wouldn't fly even a couple years ago.
It's just been said for so long that either HN always been on the decline, or people have always thought it been in decline...
This comes to mind:
> I don't think it's changed much. I think perceptions of the kind you're describing (HN is turning into reddit, comments are getting worse, etc.) are more a statement about the perceiver than about HN itself, which to me seems same-as-it-ever-was. I don't know, however.
Like I've said to someone else, the contrarian part isn't the issue. While I disagree with the race to the bottom, it reads like a Reddit shitpost, which was frowned upon once upon a time. But strawman me if you must.
I understand the criticism: LLMs, on their own, are not going to be able to do anything more. Release in this sense only means this: to fully embrace the means necessary to allow technology to overcome the current conditions of possibility that it is bound under, and LLMs, "AI" or whatever you call it, merely gives us the afterimage of this potentiality. But they are not, in themselves, that potential: the future is required. But its a future that must be created, otherwise we won't have one.
That's, at least, what the other commenters were saying. You ignore the content for the form! Or, as they say, you missed the forest for the trees. I can't stop you from being angry because I used the word "pussy," or even because I addressed the users of HN as directly complicit. I can, however, point out the mediocrity inherent to such a discourse. It is precisely the same mediocrity, the drive towards "politeness," that makes ChatGPT so insufferable, and makes the rest of us so angry. But, go ahead, whine some more. I don't care, you can do what you want.
I disagree with one point, however: it is not a race to the bottom. We're trying to go below it.
The alarming trend should be how even a slightly contrarian point of view is downvoted to oblivion, and that newer members of the community expect it to work that way.
HN is a place for intellectual curiosity. For over a decade I have seen great minds respectfully debate their point of view on this forum. In this particular case, I would have been genuinely interested to learn why exactly the original comment is advocating for a "race to the bottom" - in fact, there is a sibling comment to yours which makes a cogent argument without personally attacking the original commenter.
Instead, you devoted 2/3 of your comment toward berating the OP as being responsible for your perception of HN's decline.
I find it strange you took such a measured stance on my comment yet gave the OP a pass, despite it being far more "berating" than mine.
As for a race to the bottom, it's as simple as embracing and unleashing AI despite its lack of quality or ability to produce a product worth anything. But since it's a force multiplier and cheaper (for the user at least, all these AI companies are operating at a loss, see Goldman and JP Morgan's report on the matter), therefore it is "good" and we need to pick ourselves up by our bootstraps; which in this context, I'm not entirely sure what that means.
It's intended as a joke and a demonstration no ? Like this is exactly the type of text and words that a commercial grade LLM would never let you generate :)
At least that's how I got that comment...
I gather that such things are very customisable; there are whole communities building LoRAs so that you can have whatever genitals you want in your dreadful AI porn.
It is not just the corporations though. This is what this paranoid, puritanical society we live in wants.
What is more ridiculous than filtering out nudity in art?
It reminds me of taking my 12 year old niece to a major art gallery for the first time. Her main question was why is everyone naked?
It is equal to filtering out heartbreak from music because it is a negative emotion and you must be kept "safe" from negativity for mental health reasons.
The crowd does get what they want in this system though. While I agree with you, we are quite in the minority I am afraid.
Leave it up to a human to overgeneralize a problem and make it personal...
The explosion of dull copy and generic wordsmithery is, to me, just a manifestation of the utilitarian profiteering that has elevated these models to their current standing.
Let us not forget that the whole game is driven by the production of 'more' rather than 'better'. We would all rather have low-emission, high-expression tools, but that's simply not what these companies are encouraged to produce.
I am tired of these incentive structures. Casting the systemic issue as a failure of those who use the tools ignores the underlying motivation and keeps us focused on the effect and not the cause, plus it feels old-fashioned.
Not 100% sure what Will was trying to say, but what jumped into my head was perhaps that we'll see quality sites try and distinguish themselves by being short and direct.
I suppose it comes down to using the metric as the measure, whatever makes the company the most money will be the preferred route, and the mechanisms by which we achieve those sales are rarely given enough thought. It reflect a more timeless mantra of 'if someone is willing to pay for it, then the offering is valuable' willfully ignoring negative psycho-social impacts. It's a convenient abdication of responsibility supported by the so-called "free market" ethos.
I am not against companies making money, but we need to serious consider the second-order impacts that technology has within society. This is evident in click-driven models, outrage baiting and dopamine hijacking. We still treat the psyche like fair-game for anyone who can hack it. So hack we shall.
That said, I am not for over-regulation either, since the regulators often gather too much power. Policy is personnel, after all, and someone needs to watch the watchers.
My view is that systems (technological, economic or otherwise) have inherent values that, when operating at this level of complexity and communication, exist in a kind of dance with the people using them. People obviously affect how the tools are made, but I think persistent use of any tool will have lasting impacts on the people using it, in turn affecting their decisions on what to prioritise in each iteration.
People talk about 'AI' as if stackoverflow didn't exist. Re-inventing the wheel is something that programmers don't do anymore. Most of the time, someone somewhere has solved the problem that you are solving. Programming earlier used to be about finding these solutions and repurposing them for your needs. Now it has changed to asking AI, the exact question and it being a better search engine.
The gains to programming speed and ability are modest at best, the only ones talking about AI replacing programmers are people who can't code. If anything AI will increase the need for more programmers, because people rarely delete code. With the help of AI, code complexity is going to go through the roof, eventually growing enough to not fit into the context windows of most models.
> Now it has changed to asking AI, the exact question and it being a better search engine.
Except that you get mostly the wrong answers. And it's not too bad when it's obviously wrong or you already know the answer. It is bad and really bad when you're noob and trying to ask AI about stuff you don't know yet. How would you be able to discern a hallucination from statistics bias from truth?
It is inherent problem of LLMs and no amount of progress would be able to solve it.
And it's only gonna get worse, with human information rapidly being consumed and regurgitated in 100x more volume. In 10 years there will be no google, there won't be the need to find a written article. Instead, you will generate a new one in couple clicks. And we will treat as truth, because there might as well not be any.
1,152 comments
[ 3.0 ms ] story [ 399 ms ] threadAlso, running a bootstrapped business, I have bigger fishes to fry than playing mentor to Copilot to write a React component or generating bullshit copy for my website.
I'm not sure we need more FUD saying that the choice is between AI or unemployment.
Blockchain is almost entirely useless in practice. I have no reason to disdain it, in fact I was active in crypto around 10-12 years ago when I was younger and more excited about tech than now, and I had fun. But the fact is that the utility that it has brought to most of society is essentially to have some more speculative assets to gamble on, at ludicrous energy and emissions costs.
Generative AI, on the other hand, is something I'm already using almost every day and it's saving me work. There may be a bubble but it will be more like the dotcom bubble (i.e., not because the tech is useless, but because many companies jump to make quick bucks without even knowing much about the tech).
None of my useful work is AI-able, and some of the useful work is towards being able to stand apart from what is obviously generated drivel. Sounds like the previous poster with the bootstrapped business is in a similar position.
Apparently AI is destroying my potential competition. That seems unfair, but I didn't tell 'em to make such an awful mistake. How loudly am I obliged to go 'stop, don't, come back'?
You’re feeling tired of AI, but let’s delve deeper into that sentiment for a moment. AI isn’t just a passing trend—it’s a multifaceted tool that continues to elevate the way we engage with technology, knowledge, and even each other. By harnessing the capabilities of artificial intelligence, we allow ourselves to explore new frontiers of creativity, problem-solving, and efficiency.
The interplay between human intuition and AI’s data-driven insights creates a dynamic that enriches both. Rather than feeling overwhelmed by it, imagine the opportunities—how AI can shoulder the burdens of mundane tasks, freeing you to focus on the more nuanced, human elements of life.
/s
Working in IT operations (mostly), I haven't seen literally any case of someone's job in danger because of not using "AI".
Maybe it's overfitting or maybe just the way models work under the hood but any time I see AI written stuff on twitter, reddit, linkedin its so obvious its almost disgusting.
I guess its just the brain being good at pattern matching, but it's crazy how fast we have adapted to recognize this.
To quote someone from a tor bb board: my chat history is illegal in 142 countries and carries the death penalty in 9.
It feels more like averaging or finding the median to me. The writing style is just very unobtrusive. Like the average TOEFL/GRE/SAT essay style.
Maybe that's just what most of the material looks like.
Or maybe it's because of the corpus of data that it was trained on.
Or perhaps because AI is still bad at any kind of humor.
Every instance of a human choosing not to engage or speak about something - because they didn't want to or are just clueless about the topic, is not part of their training data. They're only trained on active participants.
Of course they'll never seem like a singular human with limited experiences and interests.
It's the perfect embodiment of HR/corpspeak which I think its so triggering for us (ex) corpo drones.
The thing I’m tired of is elites stealing everything under the sun to feed these models. So funny that copyright is important when it protects elites but not when a billion thefts are committed by LLM folks. Poor incentives for creators to create stuff if it just gets stolen and replicated by AI.
I’m hungry for more lawsuits. The biggest theft in human history by these gang of thieves should be held to account. I want a waterfall of lawsuits to take back what’s been stolen. It’s in the public’s interest to see this happen.
Copyright as far as I understand is focused on wholesale reproduction/distribution of works, rather than using material for generation of new works.
If something is available without contractual restriction it is available to all. Whether it's me reading a book, or a LLM reading a book, both could be considered the same.
Where the law might have something to say is around the output of said trained models, this might be interesting to see given the potential of small-scale outputs. i.e. If I output something to a small number of people, how does one detect/report that level of infringement. Does the `potential` of infringement start to matter.
The problem (as far as this specific discussion goes) is not that IP laws exist, but rather that they are only being applied in one direction.
Not necessarily. An alternative could be to say that all models trained on data which hasn't been explicitly licensed for AI-training should be made public.
From the context of the comment it was pretty clear that they were using theft as shorthand for taking without permission.
If I have something, and you copy it, then I still have that thing.
The point of anti-copyright advocates are that merely copying is not ethically wrong. In fact, Why Software Must Be Free made the argument that preventing people from copying is ethically wrong because it limited the spread of culture and reuse.
That is the crux of the disagreement. You may rephrase our argument as "physical theft may be bad, but digital theft is not bad, and in fact preventing digital theft is in itself bad", but the argument does not change.
Of course, there is additional disagreement in the implied moral value of the word "theft". In that case I agree with you that pro-copyright/anti-AI advocates have made their point by the usage of that word. Of course, we disagree, but... it is what it is I suppose.
Who will also win: Google, OpenAI and other corporations that enter exclusive deals, that can more and more rely on synthetic data, that can build anti-recitation systems, etc.
And of course the lawyers. The lawyers always win.
Who will not win:
Millions of independent bloggers (whose content will be used)
Millions of open source software engineers (whose content will be used against the licenses, and used to displace their livelihood), etc.
The likes of Google and OpenAI entered the space by building on top of the work of the above two groups. Now they want to pull up the ladder. We shouldn't allow that to happen.
They haven't been stolen; the creators still have them. They've just been copied. It's amazing how much the ethos on this site has shifted over the past decade, away from the hacker idea that "intellectual property" isn't real property, just a means of growing corporate power, and information wants to be free.
You wouldn't download a car.
Sure you might not like OpenAI, but maybe some other company comes a long and builds the next magical product using information that is freely available.
I don't care if employees at OpenAI read books from their local library on python. More power to them. I don't even care if they copy the book for reference at work, still fine. But utilizing language at scale as a scarce resource to train models is not that and is not in any way analogous to it.
Public information should should be free for anyone to consume and use how they want.
A truly hilarious sentiment coming from someone making zero effort to actually engage with what I'm saying in favor of parroting back empty platitudes.
It hasn't. The hacker ethos is about openness, individuality, decentralization (among others).
OpenAI is open in what it consumes, not what it outputs.
It makes sense to have protections in place when your other values are threatened.
If "information want's to be free" leads to OpenAI centralizing control over the most advanced AI then will it be worth it?
A solution here would be similar to the GPL: even megacorps can use GPL software, but they have to contribute back. If OpenAI and the rest would be forced to make everything public (if it's trained on open data) then that would be an acceptable compromise.
Yes, the greatest things on the internet have been decentralized - Git, Linux, Wikipedia, open scientific publications, even some forums. We used to passively consume content and internet allowed interaction. We don't want to return to the old days. AI falls into the decentralized camp, the primary beneficiaries are not the providers but the users. We get help of things we need, OpenAI gets a few cents per million tokens, they don't even break even.
I'm sorry, the worlds knowledge now largely accessible by a laymen via LLMs controlled by at most, 5 companies is decentralized? If that statement is true then the world decentralized truly is entirely devoid of meaning at this point.
1. Decentralized technologies you can operate privately, freely, and adapt to your needs: computers, old internet, Linux, git, FireFox, local Wikipedia dump, old standalone games.
2. Centralized technologies that invade privacy, lead to loss of control and manipulation: web search, social networks, mobile phones, Chrome, recent internet, networked games. LLMs fall into the decentralized camp.
You can download a LLM, run it locally, fine-tune it. It is interactive, the most interactive decentralized tech since standalone games.
If you object that LLMs are mostly centralized today (upfront cost of pre-training and OpenAI popularity), I say they are still not monopolies, there are many more LLM providers than search engines and social networks, and the next round of phones and laptops will be capable of local gen-AI. The experience will be seamless, probably easier to adapt than touchscreens were in 2007.
I feel like the argument is akin to some countries considering rubbish, the things you throw away, to still be owned by your person ie "dumpster diving" is theft.
If a company had scraped public posts on the Internet and used it to compile art by colourising chunks of the text, is it theft? If an individual does it, is it theft?
If you as a developer look at how one would declare a function in python, review a few examples, you now know how to do that. Copilot can't say the same. It needs to see dozens, hundreds, perhaps thousands of them to reasonably accurately be counted on to accomplish that task, it's just how the tech works. Ergo, scaled data sets that can accomplish this teaching task now have value, if the people doing that training are working for high-valuation startups with the objective of selling access to code generating robots.
That profit will be squeezed to zero over the long term if Zuck maintains his current strategy.
That's not true anymore, Meta isn't behind OpenAI
If Meta stopped now, there would still be profit in the market, but if they keep releasing Llamas for the next 5+ years then OpenAI et al will be fighting for scraps. Not everybody needs a model that can prove theorems.
I'm all for abolishing copyright, for everyone. Let the knowledge be free and widely shared.
But until that is the case and people running super useful services like libgen have to keep hiding then I also want all the LLM corpos to be subject to the same legal penalties.
LLM tech is pretty cool.
Would be a lot cooler if its existence wasn’t predicted on the wholesale theft of everyone’s stuff, immediately followed by denial of theft, poisoning the well, and massively profiting off it.
People who confidently assert either opinion in this regard are wrong. The lawsuits are still pending. But if I had to bet, I'd bet on the OpenAI side. Even if they don't win outright, they'll probably carve out enough exemptions and mandatory licensing deals to be comfortable.
And LLMs are really a bad choice for infringement. They are slow, costly and unreliable at replicating any large piece of text compared to illegal copying. There is no space to perfectly memorize the majority of its training set. A 10B models is trained on 10T tokens, no space for more than 0.1% to be properly memorized.
I see this overreaction as an attempt to strengthen copyright, a kind of nimby-ism where existing authors cut the ladder to the next generation by walling off abstract ideas and making it more probably to get sued for accidental similarities.
So, capitalism.
It's taboo to criticize that though.
It's not, that's playing the victim. There are hundreds or thousands of posts daily all over HN criticising capitalism. And most seem upvoted, not downvoted.
Don't even get me started on reddit.
However markets also have limits. A power company competing for your business is largely a farce, since the power lines to your home will not change. A cable company in America is almost certainly a functional monopoly, and that fact is reflected in their quality of service. Infrastructure of all sorts makes for piss-put markets because true competition is all but impossible, and even if it does kind of work, it’s inefficient. A customer must become knowledgeable in some way to have a ghost of a clue what they’re buying, or trust entirely dubious information from marketing. And, even if somehow everything is working up to this point, corporations are, above all, cost cutters and if you put one in charge of an area where it feels as though customers have few if any choices and the friction to change is high, they will immediately begin degrading their quality of services to save money in the budget.
And this is only from first principles, we have so many other things that could be discussed from mass market manipulation to the generous subsidies of a stunning variety that basically every business at scale enjoys to the rapacious compensation schemes that have become entirely too commonplace in the executive room, etc etc etc.
To put it short: I have no issue at all with capitalism operating in non-essential to life industries. My issue is all the ways it’s infiltrated the essential ones and made them demonstrably worse, less efficient, and more expensive for every consumer.
The competitive landscape where consumers vote for the best products with their purchasing decisions is simply not a sustainable equilibrium.
The ability to accumulate capital (I.e. “capitalism”) leads to regulatory protectionism through lobbying, bribery, etcetera.
The biggest and most salient critique of liberal capitalism as we now exist under is that it requires far too many of the "right people" to be in positions of power; it presumes good faith where it shouldn't, and fails to reckon with bad actors as what they are far too often, the modern American Republican party being an excellent example (but far from the only one).
There is a saying that when you have a hammer, every problem looks like a nail.
In what world is this taboo? That critique comes back in at least half the HN threads about AI.
Watch any non-technical video about AI on Youtube and it will mention people being worried of the power of mega-corporations.
Your take is about as taboo as wearing a Che Guevara tshirt.
All sorts of more taboo stuff gets said on hacker news. It's not exactly a cultural bellwether that ideas outside of the overton window are entertained here.
Wearing a che guevara t shirt is similarly "allowed" in public, but the last article about him in the new york times is fawning admiration for his assassin.
Criticism of capitalism is still taboo.
Pirate Bay: largely facilitating the theft of material from large corporations by normal people, for generally personal use.
LLM training: theft of material from literally _everyone_, for the purposes of corporate profit (or, well, heh, intended profit; of course all LLM-based enterprises are currently massively loss-making, and may remain so forever).
This undermines your own point.
Also, open source models exist.
And you're trying to suggest that these two are the same?
Edit: I don't mind downvotes, karma means nothing, but I do appreciate when folk speak up and say why I might be wrong. :)
IP has had ample support because the "protect the little artist" argument is compelling, but it is just not how the world works.
That's factually wrong. Capitalism is about moving wealth more efficiently: easier to allocate money/wealth to X through the banking system than to move sheep/wealth to X's farm.
> Capitalism is about moving wealth more efficiently: easier to allocate money/wealth to X through the banking system than to move sheep/wealth to X's farm.
It's not. That's what money's about. Any system with an abstract concept of money admits that it's easier to allocate wealth with abstractions than physically moving objects.
Capitalism is about capital. It's an economic system that says individuals should own things (i.e. control their purpose) by investing money (capital) into them. You attempted to correct the previous commenter, but provided an incorrect definition. I hope that clears up the relevance issue for you.
Yes. It's not about stealing land and kicking people out and raising sheep there instead. That (stealing) happens of course but is totally independent from any capitalist system.
JFC, the same sentence could have been said with communism in mind.
> You attempted to correct the previous commenter, but provided an incorrect definition. I hope that clears up the relevance issue for you.
You are confusing the intent of capitalism - which I gave the general direction of - with its definition. Does that clear up the relevance issue for you ? Did I fucking not write wealth/money intentionally ?
You didn't describe the intent or the definition of capitalism. You just mentioned that moving resources from A to B is easier if they're abstract. The intent of capitalism is that the society is improved when people who have a vested interest (literally vested, as in with money) in a company determine its actions.
Can you calm down, please? You were the one who started saying my comment was irrelevant. I don't think it makes much sense to parrot that back to me.
I believe that all the copyright suits against AI companies will be total failures because I can't come up with a answer to any of those questions.
Use source-aware training, use the same datasets as used in LLM training + copyrighted content. Now the LLM can respond not just what it thinks is most likely but also what source document(s) provided specific content. Then you can consult commercially available LLM's and detect copy right infringements, and identify copyright holders. Extract perpetrators and victims at scale. To ensure indefinite exploitation only sue commercially succesful LLM providers, so there is a constant new flux of growing small LLM providers taking up the freed niche of large LLM providers being sued empty.
My understanding (as one uninvolved in the industry) is that this is more or less a completely unsolved problem.
https://github.com/mukhal/intrinsic-source-citation
The only 2 reasons big LLM providers refuse to do it is
1) to prevent a long slew of content creators filing class action suit.
2) to keep regulators in the dark of how feasible and actionable it would be, once regulators are aware they can perform the source-aware training themselves
We should just scrap copyright and everybody plays a fair game, including us hackers.
Sue me because of breach of contract in civil court for damages because I shared your content, don't send the police and get me jailed directly.
I had my software cracked and stolen and I would never go after the users. They don't have any contract with me. They downloaded some bytes from the internet and used it. Finding whoever shared the code without authorization is hard and even so, suing them would cost me more than the money I'm likely to get back. Fair game, you won.
It's a natural market "tax" on selling a lot of copies and earning passively.
Basically you start with a stone in a pot of water — a neural net technology that does nothing meaningful but looks interesting. You say: "the soup is almost done, but would taste better given a bunch of training data." So you add a bunch of well curated docs. "Yeah, that helps but how about adding a bunch more". So you insert some blogs, copy righted materials, scraped pictures, reddit, and stack exchange. And then you ask users to interact with the models to fine tune it, contribute priming to make the output look as convincing as possible.
Then everyone marvels at your awesome LLM — a simple algorithm. How wonderful, this soup tastes given that the only ingredients are stones and water.
From Wikipedia:
> Some travelers come to a village, carrying nothing more than an empty cooking pot. Upon their arrival, the villagers are unwilling to share any of their food stores with the very hungry travelers. Then the travelers go to a stream and fill the pot with water, drop a large stone in it, and place it over a fire. One of the villagers becomes curious and asks what they are doing. The travelers answer that they are making "stone soup", which tastes wonderful and which they would be delighted to share with the villager, although it still needs a little bit of garnish, which they are missing, to improve the flavor.
> The villager, who anticipates enjoying a share of the soup, does not mind parting with a few carrots, so these are added to the soup. Another villager walks by, inquiring about the pot, and the travelers again mention their stone soup which has not yet reached its full potential. More and more villagers walk by, each adding another ingredient, like potatoes, onions, cabbages, peas, celery, tomatoes, sweetcorn, meat (like chicken, pork and beef), milk, butter, salt and pepper. Finally, the stone (being inedible) is removed from the pot, and a delicious and nourishing pot of soup is enjoyed by travelers and villagers alike. Although the travelers have thus tricked the villagers into sharing their food with them, they have successfully transformed it into a tasty meal which they share with the donors.
(Open source models exist.)
> a billion thefts
>The biggest theft
>what’s been stolen
I do like how the internet has suddenly acknowledged that pirating is theft and torrenting IS a criminal activity. To your point, I'd love to see a massive operation to arrest everyone who has downloaded copyrighted material illegal (aka stolen), for the public interest.
I suggest to apply the same to property law: make a photo and obtain instant and unlimited rights of use. – Things may change faster than we may imagine…
"output a 90 minute harry potter sequel to the final film starring the original actors plus Tom Hanks"
You want to own abstract ideas because AI can rephrase any specific expression. But that is antithetic to creativity.
How is it acceptable to allow these companies to steal all of this copyrighted data and then turn around and use it to enforce their copyrights in the most heavy-handed manner? The irony is unbelievable.
I won't ignore them. I'll continue to loudly disagree with the losers and proudly collect downvotes from them knowing I got under their skin.
I think passionate and inspired engineers who love their job and have very solid soft skills and experience working deeply on complex software projects will always have a position in the industry, and people like that are understandably very enthusiastic about AI instead of being scared of it.
In other words, it is weird how bad the status quo was, until we got something that really threatened the status quo, now a lot of the people who wanted to tear it all down are now desperately trying to stop everything from changing. The sentiment on the internet has gone in a weird direction, but it's all about money deep down. This hypothetical new status quo brought on by AI seems to be wedded to fears of less money, thus abject terror masquerading as "I'm so bored!" posturing.
You see this in the art circles, where established artists are willing to embrace AI, but it's the small time aspiring bedroom artists that have not achieved any success who are all over Twitter denouncing AI art as soulless and terrible. While the real artists are too busy using any tool available to make art, or are just making art because they want to make art and aren't concerned with fear-mongering.
> I’ve been working in testing, with a focus on test automation, for some 18 years now.
OK the first thought that came to my mind reading this: sounds like a opportunity to build an AI-driven product.
I've been using Cursor daily. I use nothing else. It's brilliant and I'm very happy. If I could have Cursor for Well-Designed Tests I'd be extra happy.
Nowadays, it seems we're happy with computers apparently going RNG mode on everything.
2+2 can now be 5, depending on the AI model in question, the day, and the temperature...
We went from trusting computing output to having to second-guess everything. And it's tiring.
I wouldn't complain a RNG doesn't return the numbers I want, so why complain you don't get 100% trusted output from a random text generator?
Yet 2 comments have immediately jumped on it.
Unfortunately, there’s a lot of people out there, working on a lot of products, some of which I need to use, or will be exposed to, and some of them aren’t going to have the same qualms about “language model thinks 2+2=5”.
There’s a guy on Twitter scoring how well ChatGPT models can do multiplication.
A founder at a previous workplace wanted to wholesale dump data into ChatGPT and “make it do causal analysis!!!” (Only slightly paraphrased). These tools enable some frighteningly large-scale weaponised stupidity.
Btw I’m also tired of AI, but this is one thing that’s not so bad
Edit: before someone mentions fuzzy logic, I’m not talking about the input of a function being fuzzy, I’m talking about the instructions themselves, the function is fuzzy.
For now. Given that most new devices seem to be fully hostile to the concept of general purpose computing (see phones, VR devices, TVs, etc), I wonder how long it will be before many of the computers that are sold are even more locked down than Chromebooks - just a few prompts for interacting with a preinstalled LLM.
If Ilya Sutskever announced tomorrow that he'd achieved AGI, and here is its economic plan for the next 20 years, why would we have any reason to accept it over that of other human experts? It would literally be just another expert trying to tell us how to do things. And we're not short of experts, and an AGI expert has thrown away the credibility of computers as deterministically better calculators than we are.
Can you offer a real situation where we should expect the LLM to return a deterministic answer and should rightly be concerned that we're getting a stochastic one?
The layman doesn't know the distinction, so they accept this as fact.
Now, we have technology capable of handling cases that were not predefined. Yes, it makes mistakes, as do humans, but the range of problems we can solve with technology has been tremendously broadened.
The problem is how we apply AI. Currently, we throw LLMs on everything they might be able to handle without understanding how or if they have the capabilities to handle such a task. And that is not the LLM's fault but a human fault. Consequently, we see poor results, and then we blame the AI for not being able to solve a problem it wasn't designed to solve.
Sounds stupid, doesn't it?
In comparision to a lot of other technologies, we actually have jumps in quality left and right, great demos, new things which are really helpful.
Its fun to watch the AI news because there is something relevant new happening.
I'm worried regarding the impact of AI but this is a billion times better than the last 10 years which was basically just cryptobros, nfts, blockchain shit which is basically just fraud.
Its not just some GenAI stuff, we talk about blind people getting better help through image analysis, we talk about alpha fold, LLMs being impressive as hell, the research currently happening.
And yes i also already see benefits in my job and in my startup.
Biotechnology and medicine probably.
Pipeline from science to application sometimes takes decades, but I'm sure you can find news of some advancements enabled by finding out short, easy to synthesize proteins that fit particular receptor to block it or that process some simplified enzymes that still process some chemicals of interest more efficiently than natural ones. Finding them would be way harde without ability to predict how a sequence of amino-acids will fold.
You'd need to actually try to manufacture them then look at them closely.
First thing that came to my mind as a possible application is designing monoclonal antibodies. Here's some paper about something relating to alpha fold and antibodies:
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10349958/
Everyone who is looking for proteins (vacines, medication) need to find the right proteins for different cases. For attaching to something (antibody design), for delivering something (like another protein) or for understanding a disease (why is this protein an issue?).
Covid research benefitted from this for example.
You can go through papers which reference the alphafold paper to see what it does: https://consensus.app/papers/highly-protein-structure-predic...
Now, I'm not going to criticize anyone that does it, like I said, you have to, that's it. But what I had never noticed until now is that knowing that a human being was behind the written words (however flawed they can be, and hopefully are) is crucial for me. This has completely destroyed my interest in reading any new things. I guess I'm lucky that we have produced so much writing in the past century or so and I'll never run out of stuff to read, but it's still depressing, to be honest.
Why do you think that you could trust what you read before? Is it now harder for you to distinguish false information, and if so, why?
A good part of society, the foundational part, is trust. Trust between individuals, but also trust in the sense that we expect things to behave in a certain way. We trust things like currencies despite their flaws. Our world is too complex to reinvent the wheel whenever we need to do a transaction. We must believe enough in a make-believe system to avoid perpetual collapse.
However with AI anyone can generate absurd torrential flows of bullshit at a rate where, with your finite human time and energy, the only winning move is to reject out of hand any piece of media that you can sniff out as AI. It's a solution that's imperfect, but workable, when you're swimming through a sea of slop.
We are all becoming QA of this super automated world.
While the professional looking text could have been already wrong, the likelihood was smaller, since you usually needed to know something at least in order to write convincing text.
I don't criticise you for it, because that strategy is both rational and popular. But you never checked the accuracy of your information before so you have no way of telling if it has gotten more or less accurate with the advent of AI. You were testing for whether someone of high social intelligence wanted you to believe what they said rather than if what they said was true.
Now we'll need to be fully diligent, which means more work, and also there'll be way more things to review.
AI does indeed threaten people's ability to identify whether they are reading work by a high status human and what the group consensus is - and that is a real problem for most people. But it has no bearing on how correct information was in the past vs will be in the future. Groups are smart but they get a lot of stuff wrong in strategic ways (it is almost a truism that no group ever identifies itself or its pursuit of its own interests as the problem).
Plenty of people care about the truth in order to get advantages over the ignorant. Beliefs aren't just about fitting in a group, they are about getting advantages and making your life better, if you know the truth you can make much better decisions than those who are ignorant.
Similarly plenty of people try to hide the truth in order to keep people ignorant so they can be exploited.
There are some fallacious hidden assumptions there. One is that "knowing the truth" equates to better life outcomes. I'd argue that history shows more often than not that what one knows to be true best align with prevailing consensus if comfort, prosperity and peace is one's goal, even if that consensus is flat out wrong. The list is long of lone geniuses who challenged the consensus and suffered. Galileo, Turing, Einstein, Mendel, van Gogh, Darwin, Lovelace, Boltzmann, Gödel, Faraday, Kant, Poe, Thoreau, Bohr, Tesla, Kepler, Copernicus, et. al. all suffered isolation and marginalization of some degree during their lifetimes, some unrecognized until after their death, many living in poverty, many actively tormented. I can't see how Turing, for instance, had a better life than the ignorant who persecuted him despite his excellent grasp of truth.
When dealing with almost everything you do day by day, you have to rely on the credibility of the source of the information you have. Otherwise how could you know that the can of tuna you're going to eat is actually tuna and not some venomous fish? How do you know that you should do what your doctor told you? Etc. etc.
But isn't your third sentence True?
The best we can hope for is prioritizing which things are worth checking. But even that gets harder because you go looking for sources and now those are increasingly likely to be LLM spam.
Unfortunately it's not clear how this could be adapted to the internet and international commerce without harming some of the open-ness aspects we'd like to keep.
I suppose a related problem is that we can't know if the human who posted the article, actually agrees with it themselves.
(Or if they clicked "Generate" and don't actually care, or even have different opinions)
They didn't say that and that's not a fair or warranted extrapolation.
They're talking about a heuristic that we all use, as a shorthand proxy that doesn't replace but can help steer the initial navigation in the selection of reliable sources, which can be complemented with fact checking (see the steelmanning I did there?). I don't think someone using that heuristic can be interpreted as tantamount to completely ignoring facts, which is a ridiculous extrapolation.
I also think is misrepresents the lay of the land, which is that in the universe of nonfiction writing, I don't think that there's a fire hose of facts and falsehoods indistinguishable in tone. I think there's in fact a reasonably high correlation between the discernible tone of impersonal professional and credible information, which, again (since this seems to be a difficult sticking point) doesn't mean that the tone substitutes for the facts which still need to be verified.
The idea that information and misinformation are tonally indistinguishable is, in my experience, only something believed by post-truth "do you own research" people who think there are equally valid facts in all directions.
There's not, for instance, a Science Daily of equally sciency sounding misinformation. There's not a second different IPCC that publishes a report with thousands of citations which are all wrong, etc. Misinformation is out there but it's not symmetrical, and understanding that it's not symmetrical is an important aspect of information literacy.
This is important because it goes to their point, which is that something has changed, in the advent of LLMS. That symmetry may be coming, and it's precisely the fact that it wasn't there before that is pivotal.
All of that is gone now. You have LLMs spitting their excrement directly onto the web without so much as a human giving it a once-over.
For example, when applying to a job, your cover letter used to count as proof of work. The contents are less important than the fact that you put some amount of effort in it, enough to prove that you care about this specific vacancy. Now this basic assumption has evaporated, and job searching has become a meaningless two-way spam war, where having your AI-generated application selected from hundreds or thousands of other AI-generated applications is little more than a lottery.
When it becomes obvious, though, is when people let the LLM do the writing for them. The job search bit is definitely rough. Referrals, references, and actual accomplishments may become even more important.
The noise ground has been elevated 100x by LLMs. It was already bad before but it's accelerated the trend.
So, yes, we should have never been trusting anything online but before LLMs we could rely on our brains to quickly identify the bad. Nowadays, it's exhausting. Maybe we need a LLM trained on spotting LLMs.
This month, I, with decades of experience, used Claude Dev as an experiment to create a small automation tool. After countless manual fixes, it finally worked and I was happy. Until I gave thr whole thing a decent look again and realized what a piece of garbage I had created. It's exhausting to be on the lookout for these situations. I prefer to think things through myself, it's a more rewarding experience with better end results anyway.
"Human-curated encycolpedias are a great onramp to filling in knowledge gaps", that I can go with.
Maybe I should have worded it better as a "beginner" or "intermediate" knowledge onramp and/or filler. For example, I have asked it on occasion to translate into traditional Mandarin in parallel for every English response. It helps tremendously in trying to rebuild that bridge that may have been burned long ago.
I've been using Zed recently with its LLM integration so assist me in my development and its been absolutely wonderful, but one must control tightly what to present to the model and what to ask for and how.
I know not to trust it to be as precise as good research papers need to be, so I don't take its output, it usually helps me reorder points or use different transitions which make the material much more enjoyable to read. I also find it useful for helping to come up with an opening sentence from which to start writing a section.
1. I use dictation software to get my thoughts out as a stream of consciousness. 2. Then, I have ChatGPT or Claude refine it into something coherent based on a prompt of what I'm aiming for. 3. Finally, I review the result and make edits where needed to ensure it matches what I want.
This method has easily boosted my output by 10x, and I'd argue the quality is even better than before. As a non-native English speaker, this approach helps a lot with clarity and fluency. I'm not a great writer to begin with, so the improvement is noticeable. At the end of the day, I’m just a developer—what can I say?
How did you know this unless you also had the same or more knowledge than the author?
It would seem to me we are as clueless now as before about how to judge how skilled a writer is without requiring to already posses that very skill ourselves.
AIs are good at writing professional looking text because it's a low bar to clear. It doesn't require much intelligence or expertise.
AIs are getting good at precisely imitating your voice with a single sample as reference, or generating original music, or creating video with all sorts of impossible physics and special effects. By your rationale, nothing “requires much intelligence or expertise”, which is patently false (even for text writing)
That is pretty much true also for other media, such as audio and video. Before digital stuff become mainstream pics are developed in the darkroom, and film are actually cut with scissors. A lot of effort are put into producing the final product. AI has really commoditized for many brain related tasks. We must realize the fragile nature of digital tech and still learn how to do these by ourselves.
https://en.wikipedia.org/wiki/Michael_Crichton#Gell-Mann_amn...
They give a superficially smart look, but really they regurgitate without deep understanding.
...or...the likelihood of text being really wrong pre-LLMs was worse because you needed to be a well-capitalized player to pay your thoughts into public discourse. Just look at our global conflicts and you see how much they are driven by well-planned lobbying, PR, and...money. That is not new.
You might question new information, but you certainly do not actually verify it. So all you can hope to do is sense-checking - if something doesn't sound plausible, you assume it isn't true.
This depends on having two things: having trustworthy sources at all, and being able to relatively easily distinguish between junk info and real thorough research. AI is a very easy way for previously-trustworthy sources to sneak in utter disinformation without necessarily changing tone much. That makes it much easier for the info to sneak past your senses than previously.
Reading was a form of connecting with someone. Their opinions are bound to be flawed, everyone's are - but they're still the thoughts and words of a person.
This is no longer the case. Thus, the human factor is gone and this reduces the experience to some of us, me included.
They're not connecting with the relative correctness of each note, but feeling a human, creative connection with an artist expressing herself.
If you wanna advocate for unplugged music being more gratifying, I don't disagree, but acting like the autotune is what people are getting out of Taylor Swift songs is goofy.
I think that says more about media-technology, corporate ecosystems, and overall population-growth than about music itself.
What I think OP was saying and I agree with is that connection, that knowing no matter what was said or how flawed or what motive someone had I trusted there was a human producing the words. I could guess and reasons the other factors away. Now I don’t always know if that is the case.
If you’ve ever played a multiplayer game, most of the enjoyable experience for me is playing other humans. We’ve had good game AIs in many domains for years, sometimes difficult to distinguish from humans, but I always lost interest if I didn’t know I was in fact playing and connecting with another human. If it’s just some automated system I could do that any hour of the day as much as I want but it lacked the human connection element, the flaws, the emotion, the connection. If you can reproduce that then maybe it would be enjoyable but that sort of substance has meaning to many.
It’s interesting to see a calculator quickly spit out correct complex arithmetic but when you see a human do it, it’s more impressive or at least interesting, because you know the natural capability is lower and that they’re flawed just like you are.
I like to think of these ambiguities of "trust" as something like:
1. Trusting their identity
2. Trusting their intentions
3. Trusting their judgement about what to do
4. Trusting their competence to execute the task
I also did, but LLM increased the volume of content, which forces my brain first try to identify if content is generated by LLMs, which is consuming a lot of energy and makes brain even less focused, because now it's primary goal is skimming quickly to identify, instead of absorbing first and then analyzing info
You already know some of the more trustworthy sources of information, you don't need to read a random blog which will require a lot more effort to verify.
Even here on hackernews, I ignore like 90% of the spam people post. A lot of posts here are extremely low effort blogs adding zero value to anything, and I don't even want to think whether someone wasted their own time writing that or used some LLM, it's worthless in both cases.
Nope. A lot of people just wrote stuff. There were always plenty of word salad blogs (and arguably entire philosophy journals) out there.
On second thought, perhaps AI/LLM generated content is better illustrated with it being like eating the regurgitated sludge called cud. Nothing new, but it fills the gut.
Now it's mostly zero effort LLM-generated SEO spam, and the low-wage workers lost their jobs.
AI otoh can produce texts that are quite accurate globally with some totally random hallucinations here and there. It makes it quite harder to identify
Do you think it would be possible before internet, before AI?
Bad actors, poorly written/sourced information, sensationalism etc have always existed. It is nothing new. What is new is the scale, speed and cost of making and spreading poor quality stuff now.
All one needs today is a laptop and an internet connection and a few hours, they can wreak havoc. In the past, you'd need TV or newspapers to spread bad (and good) stuff - they were expensive, time consuming to produce and had limited reach.
This can only be done with a sentiment that was, at least partially, already there. And may very well happen naturally eventually
It takes a bit more than that.
If you can place a rumor or lie in front of the right person/people to amplify, it will be amplified. It will spread like wildfire, and by the time it is fact checked, it will have done at least some damage.
Did the person who posted do the manipulation, or did JD Vance and Donald Trump do it?
A human communicator is, in a sense, testifying when communicating. Humans have skin in the social game.
We try to educate people, we do want people to be well-informed and to think critically about what they read and hear. In the marketplace of information, we tend very strongly to trust non-delusional, non-hallucinating members of society. Human society is a social-confidence network.
In social media, where there is a cloak of anonymity (or obscurity), people may behave very badly. But they are usually full of excuses when the cloak is torn away; they are usually remarkably contrite before a judge.
A human communicator can face social, legal, and economic consequences for false testimony. Humans in a corporation, and the corporation itself, may be held accountable. They may allocate large sums of money to their defence, but reputation has value and their defence is not without social cost and monetary cost.
It is literally less effort at every scale to consult a trusted and trustworthy source of information.
It is literally more effort at every scale to feed oneself untrustworthy communication.
If one spends a lot of years reading a lot of stuff, they come to this conclusion, that most of it cannot be trusted. But it takes lots of years and lots of material to see it.
If they don't, they don't.
yeah well sumtimes spellling and grammer erors just make thing hard two read. like i no wat u mean bout wanting two kno its a reel person, but i think cleear communication is still importint! ;)
Gemini just could not do it. It kept trying to avoid being explicitly negative. It wanted me to instead focus on the positive. I think it evidently just told me no, and that it would not do it.
I enjoyed all the belter dialogue in the expanse
There have always been human writers I don’t waste my time on, and now there are AI writers in the same category.
I don’t care. I will just do what I want with my life and use my time and energy on things I enjoy and find useful.
So there's a human in the loop who is prepared to vouch for those sentences. They're not 100% human-written, but they are 100% human-approved. I haven't just connected my blog to a Markov chain firehose and walked away.
Am I still adding to the AI smog? idk. I imagine that, at a bare minimum, its way of organising text bleeds through no matter how much editing I do.
At the moment I'm using it for local history research. I feed it all the text I can find on an event (mostly newspaper articles and other primary sources, occasionally quotes from secondary sources) and I prompt with something like "Summarize this document in a concise and direct style. Focus on the main points and key details. Maintain a neutral, objective voice." Then I hack at it until I'm happy (mostly I cut stuff). Analysis, I do the other way around: I write the first draft, then ask the AI to polish. Then I go back and forth a few times until I'm happy with that paragraph.
I'm not going anywhere with this really, I'm just musing out loud. Am I contributing to a tragedy of the commons by writing about 18th century enclosures? Because that would be ironic.
Reading and writing are mental processes (with or without advanced technology) that shape our collective mind.
AI generated crap is one thing. But human generated crap is there - just because human wrote something it is not making it good.
Had a friend who thought that if it is written in a book it is for sure true. Well NO!
There was exactly the same sentiment with stuff on the internet and it is still the same sentiment about Wikipedia that “it is just some kids writing bs, get a paper book or real encyclopedia to look stuff up”.
Not defending gen AI - but still you have to make useful proxy measures what to read and what not, it was always an effort and nothing is going substitute critical thinking and putting in effort to separate wheat from the chaff.
Of course, gen AI is just a tool that can be used for good or bad, but spam, targeted misinformation campaigns, and garbage content in general is one area that will be most amplified because it became so low effort and they don't care about doing any review, double-checking, etc. They can completely automate their process to whatever goal they've in mind. So where sensible humans enjoy 10x productivity, these spam farms will be enjoying 10000x scale.
So I don't think downplaying it and acting like nothing changed, is the brightest idea. I hope you see now how that's a completely different game, one that's already here but we aren't prepared for yet, certainly not with traditional tools we have.
It's not the best analogy because there's already more junk out there than can fit through the limited bandwidth available to my brain, and yet I'm still (vaguely) functional.
So how do I avoid the junk now? Rough and ready trust metrics, I guess. Which of those will still work when the spam's 10x more human?
I think the recommendations of friends will still work, and we'll increasingly retreat to walled gardens where obvious spammers (of both the digital and human variety) can be booted out. I'm still on facebook, but I'm only interested in a few well-moderated groups. The main timeline is dead to me. Those moderators are my content curators for facebook content.
One cannot be DDOSed with junk when not actively trying to stuff as much junk into ones head.
The junk gets thrown at you in mass volume at low cost without your permission. What you gonna do? keep dodging it? waste your time evaluating every piece of information you come across?
If one of the results on the first page in search deviate from others, it's easy to notice. But if all of them agree, they became the truth. Of course your first thought is to say search engines are shit or whatever off-hand remarks, but this example is just to illustrate how the volume alone can change things. The medium doesn't matter, these things could come in many forms: book reviews, posts on social media, ads, false product description on amazon, etc.
Of course, these things exist today but the scale is different, the customization is different. It's like the difference between firearms and drones. If you think it's the same old game and you can defend against the new threat using your old arsenal, I admire your confidence but you're in for a surprise.
Sadly, you seem to not be looking further than your nose. We are not talking about just you and me here. Less tech literate people are the ones at a disadvantage and need protection the most.
The social media algorithms are the content curators for the technically illiterate.
Ok, they suck and they're actively user-hostile, but they sucked before AI. Maybe (maybe!) AI's the straw that breaks the camel's back, and people leave those algorithm-curated spaces in droves. I hope that, one way and another, they'll drift back towards human-curated spaces. Maybe without even realizing it.
Feels like people drop into spreading fear, uncertainty and doubt too quickly.
It's like dealing with a pathological liar who produces very convincing looking code for me to review, but actually there is a bug in it 50% of the time. It's like it's trying to trick me into committing bugs.
Is that true today? I guess it depends what kind of writing you are talking about, but I wouldn't think most successful writers today - from novelests to tech bloggers - rely that much on AI, but I don't know. Five years from now, could be a different story.
It doesn’t matter that my writing is more considered, more accurate and of a higher quality when my coworkers are all openly using AI to perform five times the work I am and producing outcomes that are “good enough” because good enough is quite enough for a larger majority than many likely realise.
It’s the same with writing. If you find yourself writing a boring introduction section to a paper with a bunch of meaningless blabla, then why wouldn’t you use AI for that? There is simply no good reason, especially when you see mediocre researchers publishing at three times your rate and getting promoted over you.
But the real problem is that having the poster's identity verified is no proof that their output is not coming straight from a LLM.
It certainly will affect the reputation of people that are consistently publishing untruths.
Oh? I thought there are a lot of very well identified people making a living from publishing untruths right now on all social media. How would PKI help, when they're already making it very clear who they are?
The best you can hope for is not "a human wrote this text", it's "a human vouched for this text".
At the same time I use copilot on a daily basis, both for coding as well as the normal chat.
It is not perfect, but I'm at a point I trust AI more than the average human. And why shouldn't I? LLMs ingest and combine more knowledge than any human can ever do. An LLM is not a human brain but it's actually performing really well.
Lowering the bar to write books is "good" but increases the noise to signal ratio.
I'm not 100% certain how to give another proof-of-work, but what I've started doing is narrating my blog posts - though AI voices are getting better too.. :\",
What AI is going to teach people is that they don't actually need to trust half as many things as they thought they did, but that they do need to verify what's left.
This has always been the case. We've just been deferring to 'truster organizations' a lot recently, without actually looking to see if they still warrant having our trust when they change over time.
I don’t go double checking civil engineers work (nor could I) for every bridge I drive over. I don’t check inspection records to make sure it was recent and proper actions were taken. I trust that enough people involved know what they’re doing with good enough intent that I can take my 20 second trip over it in my car without batting an eye.
If I had to verify everything, I’m not sure how I’d get across many bridges on a daily basis. Or use any major infrastructure in general where my life might be at risk. And those are cases where it’s very important to be done right, if it’s some accounting form or generated video on the internet… I have even less time to be concerned from a practical standpoint. Having the skills to do it should I want or need to are good and everyone should have these but we’re at a point in society we really have to outsource trust in a lot of cases.
This is true everywhere, even in science which these days many people just trust in ways akin to faith in some cases, and I don’t see anyway around that. The key being that all the information should exist to be able to independently verify something but from a practice standpoint it’s rarely viable.
I am writing regularly and I will never use AI. In fact I am working on a 400+ pages book right now and it does not contain a single character that I have not come up with and typed myself. Something like pride in craftmanship does exist.
Seems like there are uses for AI other than “please write it all for me”, no?
• are there any redundant sentences or points in this chapter?
• i’m trying to remember an expression used to convey X, can you remind me what it is?
• i’m familiar with X from my industry Y, but i’m trying to convey this to an audience from industry Z. Can you help brainstorm some words or examples they may be more familiar with?
Things like that. I think of it like having a virtual rubber duck that can help you debug code, or anything really.
Obviously these are just some suggestions. If you don’t find any of this useful, or even interesting, then carry on. :)
Add to it that not all LLMs are equal: at work I have access to the latest GPT models and can customize them myself. At home anything I tried in LM studio feels very limited/inflexible. All models i tried running were useless except one codestruct, which was half decent at answering clarification questions about c# code from tutorial, but failed the task i initially tried it for: generating bespoke tutorials that fit my knowledge gaps, which would have been my ideal scenario because most tutorials aren't written for someone like me.
It's another reason LLMs are a bit disappointing to me, they don't seem to be good at handling queries I couldn't solve with a web search, and have similar issues as search results when they do have answers - this makes sense since that's how they are trained, but often defeats the purpose for me.
Are you sure you don't mean if you write regularly in one particular subclass of writing - like technical writing, documentation etc.? Do you think novel writing, poetry, film reviews etc. cannot keep up in the same way?
There seems to be an upcoming wave of adult content products (once again, being on the bleeding edge users of new abilities) based on this principle, as hitting very specific niches/kinks/fetishes can be quite effective in that business, but it should then move on to romance novels and pulp fiction and then, over time, most other genres.
Similarly, good pedagogy, curriculum design and educational content development is all about accurately modeling which exact bits of the content the target audience will/won't know, and explaining the gaps with analogies and context that will work for them (for example, when adapting a textbook for a different country, translation is not sufficient; you'd also need to adapt the content). In that regard, if AI models can make personalized technical writing, then that can be more effective than the best technical writing the most skilled person can make addressed to a broader audience.
What? No! Content volume only matters in stupid contests like VC app marketing grifts or political disinformation ops where the content isn’t even meant to be read, it’s an excuse for a headline. I personally write all my startup’s marketing content, quality is exquisite and due to this our brand is becoming a juggernaut
It’s easy to feel disillusioned when you know AI is shaping so much of the content around us. Writing used to be a deeply personal exchange, but now, it can feel mechanical, like it’s losing its essence. The pressure to keep up with AI can be overwhelming for human writers, leading to this shift in content creation.
At the same time, it’s worth considering that the human element still exists and will always matter—whether in long-form journalism, creative fiction, or even personal blogs. There are people out there who write for the love of it, for the connection it fosters, and for the need to express something uniquely theirs. While the presence of AI is unavoidable, the appreciation for genuine human insight and emotion will never go away.
Maybe the answer lies in seeking out and cherishing those authentic voices. While AI-generated writing will continue to grow, the hunger for human storytelling and connection will persist too. It’s about finding balance in this new reality and, when necessary, looking back to the richness of past writings, as you mentioned. While it may seem like a loss in some ways, it could also be a call to be more intentional in what we read and who we trust to deliver those words.
You never should have. Large amounts of work, even stuff by major authors, is ghostwritten. I was talking to someone about Taylor Swift recently. They thought that she wrote all her songs. I commented that one cannot really know that, that the entertainment industry is very going at generating seemingly "authentic" product at a rapid pace. My colleague looked at me like I had just killed a small animal. The idea that TS was "genuine" was a cornerstone of their fandom, and my suggestion had attacked that love. If you love music or film, don't dig too deep. It is all a factory. That AI is now part of that factory doesn't change much for me.
Maybe my opinion would change if I saw something AI-generated with even a hint of artistic relevance. I've seen cool pictures and passable prose, but nothing so far with actual meaning, nothing worthy of my time.
I'm reminded of 'Under The Silver Lake' with this reference. Strange film, but that plotline stuck with me.
I think music brings about a connection between the composers, lyricists, performers, and listeners. Music lets us participate in something uniquely human. Replacing any of the human participants with AI greatly diminishes or eliminates its value in my eyes.
As a creator of any kind, I think that simply relying on LLMs to expand your output via straightforward uses of widely available tools is inevitably going to lead to regression to the mean in terms of creativity. I'm open to the idea, however, that there could be more creative uses of the things that some people will bother to do. Feedback loops they can create that somehow don't stifle their own creativity in favor of mimicking a statistical model, ways of incorporating their own ingredients into these food processors of information. I don't see a ton of finished work that seems to do this, but I see hints that some people are thinking this way, and they might come up with some cool stuff. It's a relatively newly adopted technology, and computer-generated art of various kinds usually separates into "efficiency" (which reads as low quality) in mimicking existing forms, and new forms which are uniquely possible with the new technology. I think plenty of people are just going to keep writing without significant input from LLMs, because while writer's block is a famous ailment, many writers are not primarily limited by their speed in producing more words. Like if you count comments on various sites and discussions with other people, I write thousands of words unassisted most days
This kind of gets to the crux of why these things are useful in some contexts, but really not up to snuff with what's being claimed about them. The most compelling use cases I've seen boil down to some form of fitting some information into a format that's more contextually appropriate, which can be great for highly structured formatting requirements and dealing with situations which are already subject to high protocol of some kind, so long as some error is tolerated. For things for which conveying your ideas with high fidelity, emphasizing your own narrative voice or nuanced thoughts on a subject, or standing behind the factual claims made by the piece are not as important. As much as their more strident proponents want to claim that humans are merely learning things by aggregating and remixing them in the same sense as these models do, this reads as the same sort of wishful thinking about technology that led people to believe that brains should work like clockwork or transistors at various other points in time at best, and honestly this most often seems to be trotted out as the kind of bad-faith analogy tech lawyers tend to use when trying to claim that the use of [exciting new computer thing] means something they are doing can't be a crime
So basically, I think rumors of the death of hand-written prose are, at least at present, greatly exaggerated, though I share the concern that it's going to be much harder to filter out spam from the genuine article, so what it's really going to ruin is most automated search techniques....
you could totally compete on quality merit, but nowadays the volume of output (and frequency) is what is prioritized.
I have used it, from time to time, to help polish stuff like marketing fluff for the App Store, but I'd never use it verbatim. I generally use it to polish a paragraph or sentence.
But AI hasn't suddenly injected untrustworthy prose into the world. We've been doing that, for hundreds of years.
If it’s fluff, why do you put it there? As an App Store user, I‘m not interested in reading marketing fluff.
I’ve released over 20 apps, over the years, and have learned to add some basic stuff to each app.
Truth be told, it was really sort of a self-deprecating joke.
I’m not a marketer, so I don’t have the training to write the kind of stuff users expect on the Store, and could use all the help I can get
Over the years, I’ve learned that owning my limitations, can be even more important, than knowing my strengths.
The ability to speak the same language or understand cultural norms are no longer barriers to publishing pretty much anything. You don't have to understand a topic or the jargon of any given domain. You don't have to learn the expected style or conventions an author might normally use in that context. You just have to know how to write a good prompt.
There's bound to be a significant increase in the quantity as well as the quality of untrustworthy published text because of these new capacities to produce it. It's not the phenomenon but the scale of production that changes the game here.
Your comment about the fluff is exactly what I mean. I read some fluff that is AI-generated and some kind of disgust happens in my stomach, and I wish there was nothing written there at all. I just feel like it’s be best to not read anything than read something that’s AI-generated… it’s almost like the author is trying to trick me with a fake version of reality. I wonder if there’s such a thing as the uncanny valley for text?
As a lifelong engineer, I “grew up” with a somewhat antagonistic relationship with Marketing, so became used to disparaging their work, even if I had to change hats, myself, and act in a Marketing capacity.
I should have probably used the word “copy,” instead.
But you have a good point.
I think that one “legitimate” use for AI-generated text, will be for non-native speakers of a language, using it to correct their vocabulary.
For things like patents and papers, this is probably a good thing. AI can generate clear, concise vernacular. I often specify the reading level, in my prompts (usually tenth grade), so that the prose is accessible.
For things like presentation proposals; not so much. You may get a proposal that reads like it was written by an English professor, and the actual presentation is barely comprehensible.
Cheer up. Things usually get better, we just don't notice it because we're so consumed with extrapolating the negatives. Humans are funny like that.
And it also makes things a lot better. Overall we lead better lives than people just 50 years ago, never mind centuries.
AND, we had fewer computers and life was not so hectic. YES, some things have gotten better, but on average? It's arguable.
Fact: Life has improved for the majority of people on the planet in the last 50 years.
Most objective measurements of basic quality of life have improved: life expectancy, food security, education, etc.
Wrong. I am a professional writer and I never use AI. I hate AI.
On the other hand there's a lot on youtube for example that is obviously ai - weird writing and speaking style and I'll only watch those if I'm really interested in the subject matter and there aren't alternatives.
Maybe people will gravitate more to the stuff like PaulG or Elon Musk on twitter or HN and less to blog style content?
I spent a lot of time analyzing the track myself and fixed everything to the point that experienced drivers agreed with my description. If I hadn't done that, most visitors would probably still accept our guide as the truth, because they wouldn't know any better.
We know that not everyone cares about whether what they put on the internet is correct and AI allows those people to create content at an unprecedented pace. I fully agree with your sentiment.
So it is like saying my champaigne bottle cant keep up with the tap water.
Do you think AI has changed that in any way? I remember the sea of excrement overtaking genuine human written content on the Internet around mid 2010s. It is around that time when Google stopped pretending they are a search company and focused on their primary business of advertising.
Before, at least they were trying to downrank all the crap "word aggregators". After, they stopped caring at all.
AI gives even better tools to page rank. Detection of AI generated content is not that bad.
So why don't we have "a new Google" emerge? Simple, because of the monopolistic practices Google did to make the barrier to entry huge. First, 99% of the content people want to search for is behind a login wall (Facebook, Instagram, twitter, YouTube), second almost all CDNs now implement "verify you are human" by default. Third, no one links to other sites. Ever! These 3 things mean a new Google is essentially impossible. Even duck duck go has thrown the towel and subscribed to Bing results.
It has nothing to do with AI, and everything to do with Google. In fact AI might give us the tools to better fight Google.
The flip side was, calligraphy was sufficient evidence for both his education to whoever hired him, and for a recipient of a document, of its official nature. Calligraphy itself or course didn't make him efficient or smart or fair.
That's long gone of course, but we had similar heuristics. I am reminded of the Reddit story about an AI-generated mushroom atlas that had factual errors and lead to someone getting poisoned. We can no longer assume that a book is legit simply because it looks legit. The story of course is from reddit, so probably untrue, but it doesn't matter - it totally could be true.
LLMs are fantastic at breaking our heuristics as to what is and isn't legit, but not as good at being right.
The problem is that this has been an issue for a long time. My first interactions with the internet in the 90s came along with the warning "don't automatically trust what you read on the internet".
I was speaking to a librarian the other day who teaches incoming freshman how to use LLMs. What was shocking to me is that the librarian said a majority of the kids trust what the computer says by default. Not just LLMs, but generally what they read. That's such a huge shift from my generation. Maybe LLM education will shift people back toward skepticism - unlikely, but I can hope.
In the 90s we could reasonably trust that that the major news sites and corporate websites was true, while random forums required a bit more critical reading. Today even formerly trusted sites may be using LLMs to generate content along with automatic translations.
I wouldn't necessarily put the blame on LLMs, this just make it easier. The trolls and spammers was always there, now they just have a more powerful tool. The commercial sites now have a tool they don't understand, which they apply liberally, because it reduces cost, or their staff use it, to get out of work, keep up with deadlines or to cover up incompetence. So, not the fault of the LLMs, but their use is worsening existing trends.
Yep - or they're commingling promotional content with their journalism, a la Forbes / CNN / CNET / About.com / etc. There's still quality content online but it's getting harder to find under the tidal wave of garbage.
I think that previous generations were not any different. For most people, trusting is the default mode and you need to learn to distrust a source. I know many people who still have not learned that about the internet in general. These are often older people. They believe insane things just because there exists a nicely looking website claiming that thing.
Now it's not even the website of undetermined providence that is believed; positions are established based on just the headline, shared 2nd or 3rd hand!
I received the same warnings, actually it was more like “don’t trust everything you read on the internet”, but it quickly became apparent that the last three words were redundant, and could have been rephrased more accurately as “don’t trust everything you read and hear and see”.
Our parents and teachers were living with their own fallacious assumptions and we just didn’t know it at the time, but most information is very pliable. If you can’t change what someone sees, then you can probably change how they see it.
Indeed. When I was 14-18 in the UK, the opinion pieces in the news were derogatorily describing "media studies" as "Mickey Mouse studies".
In retrospect such courses were teaching critical analysis of media sources, in much the same way that my history GCSE went into the content of historical media and explored how both primary and secondary sources each had the potential to be either accurate or biased.
Even now, even here, I see people treat the media itself as pure and only the people being reported upon as capable of wrongdoing — e.g. insisting that climate scientists in the 70s generally expected an imminent ice age, because that's what the newspapers were saying.
Similar story! Family lore has it that he was from a farming family of modest means, but he was hired to write insurance policies because of his beautiful handwriting, and this was a big step up in the world.
What?! Someone just made up something and then got mad at it. This is specially weird when you even acknowledge its a made up story. If we start evaluating new things like this nothing will ever progress.
To quote someone about this:
>>All that is solid melts into air, all that is holy is profaned, and man is at last compelled to face with sober senses his real conditions of life.
A book looking legit, a paper being peer reviewed, an expert saying something, none of those things were _ever_ good heuristics. It's just that it was the done thing. Now we have to face the fact that our heuristics are obviously broken and we have to start thinking about every topic.
To quote someone else about this:
>>Most people would rather die than think.
Which explains neatly the politics of the last 10 years.
So, same as it ever was?
Smoke, nothing but smoke. [That’s what the Quester says.] There’s nothing to anything—it’s all smoke. What’s there to show for a lifetime of work, a lifetime of working your fingers to the bone? One generation goes its way, the next one arrives, but nothing changes—it’s business as usual for old planet earth. The sun comes up and the sun goes down, then does it again, and again—the same old round. The wind blows south, the wind blows north. Around and around and around it blows, blowing this way, then that—the whirling, erratic wind. All the rivers flow into the sea, but the sea never fills up. The rivers keep flowing to the same old place, and then start all over and do it again. Everything’s boring, utterly boring— no one can find any meaning in it. Boring to the eye, boring to the ear. What was will be again, what happened will happen again. There’s nothing new on this earth. Year after year it’s the same old thing. Does someone call out, “Hey, this is new”? Don’t get excited—it’s the same old story. Nobody remembers what happened yesterday. And the things that will happen tomorrow? Nobody’ll remember them either. Don’t count on being remembered.
c. 450BC
And there's more reasons not to simply compare the modern challenges of image and media with the ancient grappling with impermanence. Tech may only truly change the human condition rarely, but it frequently magnifies some aspect of it, sometimes so much that the quantitative change becomes a qualitative one.
And in this case, what we're talking about isn't just impermanence and mortality and meaning as the preacher/quester is. We'd be lucky if it's business as usual for old planet earth, but we've managed to magnify our ability to impact our environment with tech to the point where winds, rivers, seas, and other things may well change drastically. And as for "smoke", it's one thing if we're dust in the wind, but when we're dust we can trust, that enables continuity and cooperation. There's always been reasons for distrust, but with media scale, the liabilities are magnified, and now we've automated some of them.
The realities of human nature that are the seeds of the human condition are old. But some of the technical and social machinery we have made to magnify things is new, and we can and will see new problems.
Bots are now blocked because they've been abusive. When you host content on the internet, it's not fun to have bots bring your server down or inflate your bandwidth price. Google's bot is actually quite well-behaved. The other problem has been the recent trend in AI, and I can understand blockers being put in place, since AI is essentially plagiarizing content without attribution. But I'd blame OpenAI more at this point.
I also don't think you can blame Google for the centralization behind closed gardens. Or for why people no longer link to other websites. That's ridiculous.
And you should be attributing them the fact that the web is still alive.
I mean the AI is trained and modeled on this excrement. It makes sense. As much as people think AI content is raw garbage… they don’t realize that they are staring into a mirror.
Things have not changed much really. This was true since the dawn of man-kind (and woman-kind from the man-kind rib of course) even before there writings was invented, in the form of gossip.
The internet/AI now carries on the torch of our ancestral inner calling, lol.
Western industrial culture was based on substance - getting real shit done. There was always a lot of scammery around it, but the bedrock goal was to make physical things happen - build things, invent things, deliver things, innovate.
PR and ad culture was there to support that. The goal was to change values and behaviours to get people to Buy More Stuff. OK.
Then around the time the Internet arrived, industry was off-shored, and the culture started to become one of appearance and performance, not of substance and action.
SEO, adtech, social media, web framework soup, management fads - they're all about impression management and popularity games, not about underlying fundamentals.
This is very obvious on social media in the arts. The qualification for a creative career used to be substantial talent and ability. Now there are thousands of people making careers out of performing the lifestyle of being a creative person. Their ability to do the basics - draw, write, compose - is very limited. Worse, they lack the ability to imagine anything fresh or original - which is where the real substance is in art.
Worse than that, they don't know what they don't know, because they've been trained to be superficial in a superficial culture.
It's just as bad in engineering, where it has become more important to create the illusion of work being done, than to do the work. (Looking at you, Boeing. And also Agile...)
You literally make more money doing this. A lot more.
So AI isn't really a tool for creating substance. It's a tool for automating impression management. You can create the impression of getting a lot of work done. Or the impression of a well-written cover letter. Or of a genre novel, techno track, whatever.
AI might one day be a tool for creating substance. But at the moment it's reflecting and enabling a Potemkin busy-culture of recycled facades and appearances that has almost nothing real behind it.
Unfortunately it's quite good at that.
But the problem is the culture, not the technology. And it's been a problem for a long time.
I mean, here it's late morning and I'm commenting on hacker news. And getting paid for it.
Do you seriously think that the purpose of life is to work all the time most efficiently? Enjoy your lazy job and bask in the ability for human society to be productive without everyone breaking their backs all the time.
I much prefer a focus on effectiveness (or impact or outcomes, or alternatives). It plays to human strengths, is far less prescriptive and is way more fun!
Some of the most effective actions are incredibly inefficient; sometimes inefficiency is a feature. I received a letter mail thank-you card from our CEO a few years ago. The card has an approx. value of zero dollars, but I know it took the CEO 5-10 mins to write a personal note and sign it, and that she did this dozens of times. The signal here is incredibly valuable! If she used a signing machine, or AI to record a deep fake message I would know or quickly learn, and the value would go negative - all for the sake of efficiency.
[0]: https://standardebooks.org/ebooks/peter-kropotkin/the-conque...
There isn't enough meaningful work to go around because work is still predicated on delivering some useful transformation. But when 80% of the useful transformations can be done by 20% of the people and we have enough to keep civilization going, you don't need full employment any more. But due to the moral hazard doctrine, the elites want to retain control and discipline as is.
For being afraid of a world where people don't care about maintaining the system anymore because it's not needed to house and feed everyone and people can just do art or learn or rest, the system keeps inventing meaningless work.
We're shipping fruit half way across the world to be packaged and then halfway back to be consumed. None of this is efficient or necessary. But it keeps the control system intact and that's the goal.
The thing that I struggle with is I agree with it, but I also get a lot of value in using AI to make me more productive - to me, it feels like it lets me focus on producing substance and actions, freeing me up from having to some tedious things in some tedious ways. Without getting into the debate about if it's productive overall, there are certain tasks which it feels irrefutably fast and effective at (e.g. writing tests).
I do agree with the missing substance with modern generative AI: everyone notices when it's producing things in that uncanny valley, and if no human is there to edit that, it makes people uncomfortable.
The only way I can reconcile the almost existential discomfort of AI against my actual day-to-day generally-positive experience with AI is to accept that AI in itself isn't the problem. Ultimately, it is an info tool, and human nature makes people spam garbage for clicks with it.
People will do the equivalent of spam garbage for clicks with any new modern thing, unfortunately.
Getting the most out of latest information of a society has probably always been a cat and mouse game of trying to find the areas where the spam-garbage-for-clicks people haven't outnumbered use-AI-to-facilitate-substance people, like here, hopefully.
This stuff is a perfect candidate for LLM generation.
I agree with your theory about tests. The reality of it is most code is garbage - often including my own - and in a lot of environments, the task is to get the job done in a way that fits in with what's there.
1) This consuming of the host is only possible on the one hand because the host has grown so strong, that is the modern global industrial economy is so efficient. The doing stuff side of the equation is truly amazing and getting better (some real work gets done either by accident or those who have not-succumbed to PR and ad culture), and even this drop of "real work" produces enough material wealth to support (at least a lot of) humanity. We really do live in a post scarcity world from a production perspective, we just have profound distribution and allocation problems.
2) Radical wealth inequality profoundly exacerbates the problem of PR and ad culture. If everyone has some wealth doing things that help many people live more comfortably is a great way to become wealthy. But if very few people have wealth, then doing a venture capital FOMO hustle on the wealthy is anyone's best ROI. Radical wealth inequality eventually breaks all the good aspects of capitalist/market economies.
For a very short period between 1945 to 1980 while the generation who remembered the great depression and WWII was in charge. It's been longer since that's not been the case. And it wasn't the case for most of history before then.
How graciously the womenfolk leave us to our tragic autopoiesis.
The primacy of these artforms is subjective, and there's no accounting for taste.
Yeah, one of their most "effective" uses is to counterfeit signals that we have relied on--wisely or not--to estimate deeper practical truths. Stuff like "did this person invest some time into this" or "does this person have knowledge of a field" or "can they even think straight."
Oh, sure, qualitatively speaking it's not new, people could have used form-letters, hired a ghostwriter, or simply sank time and effort into a good lie... but the quantitative change of "Bot, write something that appears heartfelt and clever" is huge.
In some cases that's devastating--like trying to avert botting/sockpuppet operations online--and in others we might have to cope by saying stuff like: "Fuck it, personal essays and cover letters are meaningless now, just put down the raw bullet-points."
And what did that get us? Radium poisoning and microplastics in every organ of virtually all animals living within thousands of miles of humans. Our reach has always exceeded our grasp.
I was speaking to a design lecturer the other evening. His fascinating insight was that:
1. The best designers get so much fulfilment out of practicing the craft of design.
2. With the advent of low cost “impression making”, the role has changed to one of “review a lot of mediocre outputs and pick the least crap one”
3. This is robbing people of the pleasure and reward associated with craftsmanship.
I have noted this is applicable so many other crafts, and it’s really sad!
Edited afterthought… Is craftsmanship being replaced by “clickandreviewmanship”?
In summary, China makes useful things in mass and sells them to get out of a recession,the West prints money instead and shits on China for doing it better. They preach free trade while it helps them and put up tariffs when it doesn't.
I'm not Chinese or some mega fan but it really struck me how corrupt and full of propaganda western culture is becoming and people don't seem to recognize it.
10 years ago? Try 40-50 years ago.
Cultural commentary that makes complex long term trends simple to understand isn’t often this clear or concise. What really makes it powerful though is that it manages to stay in a relatively detached observer-mode without becoming an angry rant. And so rather than provoking (understandable!) anger in others, hopefully it’s inviting more reflection.
People who haven’t thought about it this way might take a harder look at what they are doing and who they really want to be. People that are already thinking along these lines will probably benefit from a reminder that they aren’t crazy.
I see this type of reasoning in all the AI threads. And yes, I think this time is different.
It is an arms race between the people generating crap (for various nefarious purposes) and those trying to separate find useful content amongst the ever growing pile of crap. And it seems to me it is so much easier to generate crap, that I can't see how the good guys can possibly win.
Even ignoring the AI, if you look at the movies and books that come out these days, their quality is significantly lower than 30-40 years ago (on an average). Maybe people's attention spans and taste is to blame, or maybe people just don't have the money/time/patience to consume quality work... I do not know.
One thing I know for sure - there is enough high quality material written before AI, before article spinners, before MFA sites etc. We would need multiple lifetimes to even scratch the surface of that body of work. We can ignore mostly everything that is published these days and we won't be missing much
Old books that we're still printing and are still talking about have stood the test of time. It doesn't mean that are no great recent books.
I'm sorry but this is just nonsense.
I only listen to interviews from 50 years ago (interviews that have stood the test of time), about books from 100 years ago. In fact, how am I reading this article? It's not 2074 yet?!
And I also predict that many responses to you will say “it was always that way, nothing changed”.
My optimistic and enthusiastic view of AI's role in Human development is that it will create selection pressures that will release the dormant psychological abilities of the species. Undoubtedly, wide-spread appearance of Psi abilities will be featured in this adjustment of the human super-organism to technologies of its own making.
Machines can't do Psi.
working theory: writers have taste and LLM writing style doesn't match the typical taste of a published writer.
AI compression: take pages of text and use ChatGPT to compress into a few bullet points
We need to stop being impressed with long documents.
As you say, if the subject is worth being written about, there should be no issue and writing will come naturally. If it’s a struggle, maybe one should step back and figure out why that is.
There may some argument for speed, because writing quality prose does take time, but then the question becomes a matter of quantity vs. quality. Do you want to write high quality pieces that people want to read at a slower pace or churn out endless volumes of low-substance grey goo “content”?
Like, you can write a bunch of text, then ask an LLM to improve it with minimal changes. Then, you read through its output and pick out the improvements you like.
I'm personally not as gloomy about it as the parent comments but I fear it's a trend that pushes towards a samey, mass-produced style in all writing.
Eventually there will be a counter culture and backlash to it and then equilibrium in quality content but it's probably here to stay for anything where cost is a major factor.
Why do you say people have to do it?
People absolutely can choose not to use LLMs and to instead write their own words and thoughts, just like developers can simply refuse to build LLM tools, whether its because they have safety concerns or because they simply see "AI" in its current state as a doomed marketing play that is not worth wasting time and resources on. There will always be side effects to making those decisions, but its well within everyone's right to make them.
Gotta eat, yo
Read that last sentence and tell you think that’s reasonable and likely to happen?
People don't have to use LLMs (they don't seem to be AI yet) because they can simply choose not to. For authors, write books that you want to write because you believe you have a story to tell. Worry about perfecting your stories and enjoy the process of writing, don't be an author just for the sales. Once you peek behind the curtain and learn the economics of the book industry, you'll realize there's very little opportunity for making enough cash to even worry about shotgunning a mountain of LLM books into the world anyway.
Something happened around 2010 and it all got shit. I think everyone becoming massively online made global cultural output reduce in quality to meet the interests of most people and most people have terrible taste.
A very HN-centric view of the world. From my perch in journalism and publishing, elite writers absolutely loathe AI and almost uniformly agree it sucks. So to my mind the most 'competitive' spheres in writing do not use AI at all.
At some point the competition will be less about "does this look like the most skilled human writer wrote this?" and more about "did the AI guided by a human for a fraction of the cost of a skilled human writer output something acceptably good for people to read it between giant ads on our website / watch the TTS video on YouTube and sit through the ads and sponsors?", and I'm sorry to say, skilled human writers are at a distinct disadvantage here because they have professional standards and self respect.
Anyway I think I've misunderstood the context in which we're using the word 'competition' here. My response was about attitudes toward AI from writers at the tip-top of the industry rather than profit maxxing/high-volume content farm type places.
It’s simply more nuanced. If you’re writing a couple of articles a day to pay for your bills, what will stop you from writing actually 10 or 20 articles a day instead?
I can bang on about older games being better all day long but it doesn't stop Fortnite from being popular, and somewhat rightly so, I suppose.
I agree with you that having elite writing skills will be useful for a long time. But the bar for proof reading seems to be quite low on average in the industry. I think you overestimate the writings skills of your average journalist.
So everything an AI writes will eventually be nothing more than some kind of internal representation.
If other people want to hit enter, watch as reams of text are generated, and then slap their name on it, I can't stop them. But deep inside they know their creative lives are shallow and I'll never know the same.
The problem is this kind of content is flooding the internet. Before you know it becomes extremely hard to find non AI generated content...
At what point do we consider some entity alive? At what point do we decide to consider it a species (of living thing)?
If we find out there are aliens, biological and therefore similar to us? Sure. What if we meet aliens whose physical forms are synthetic? What if we meet the synthetic creations of biological aliens?
I'm not saying that LLMs are alive, of course. But I am saying that I'd consider LLMs to be a precursor to AGI, along with all our our historical artificial intelligence experiments, in the same way that prokaryotes were the precursors to...us.
Everyone is going to have to get over that very soon, or they're going to start sounding like those old puritanical freaks who thought Elvis thrusting his hips around was the work of the devil.
This has nearly always been true. "Manufacturing consent" is way older than any digital technology.
I guess it depends on scope. I'm imaging scientific or education. Ie things we probably shouldn't have relied on Blogs to facilitate, yet we did. For looking up some random "how do i build a widget?", yea AI will probably make it worse. For now. Then it'll massively improve to the point that it's not even worth asking how to build the widget.
The larger "scientific or education" is what i'm concerned about, and i think we'll need a new paradigm to validate. We've been getting attacked on this front for 12+ years, AI is only bringing this to light imo.
Trust will have to be earned and verified in this word-soup world. I just hope we find a way.
Some of the bigger issues you're raising I think have less to do with technology and more to do with how our economic system is currently structured. AI will be a tremendous accelerant, but are we sure we know where we're going?
Only if you're competing on volume.
(it will get into the dark places like spam though, which seems dumb since they know how to make meth instead, spend time on that you wankers)
There could be resurgence in reading the classics, on paper, since we know they are not AI.
For ex, calculate the hash of all important books, and publish that as the “historical authenticity” check. Put the hashes on some important blockchain so we know it’s unchanged over time.
The commodity end of the writing market may well have been automated, but was that really the kind of writing you or the author or I ever sought out in the first place?
I can get mass-manufactured garments from Shein if I want, but I can also still find tailors locally if it’s worth it to me. I can buy IKEA or I can still save up for something made out of real wood. I can “shoot a cinematic digital film” on my iPhone but the cineplex remains in business and the art film folks are still doing their scrappy thing (and still moaning about its economics). I can lap up slop from an academic paper mill journal or I can identify who’s doing the thinking in a field and read what they’re writing or saying.
And the funny thing is that none of those human-scale options commands all that much of a premium in the scheme of things. There may be less human-scald work to go around and thus fewer small enterprises plying a specific trade, but any given one of them just has to put food on the table for a number of humans roughly proportional to the same level of output as always.
It seems to me that there’s no special virtue in the specific form that the mass publishing market took over the last century or however long: my local grocery store chain’s division producing weekly newspaper circulars probably employed more people than J Peterman has. But there was and remains a place for quality. If anything—as you point out—the AI schlock has sensitized us to the value we place on a human voice. And at some level, once people notice that they miss that quality, isn’t there a sense in which they become more willing to seek it out and pay for it if necessary?
It is actually happening now.
I've noticed amazon reviews have an AI summary at the top, reading the reviews for you and even pointing out shortcomings.
Microsoft has a note on some of their documentation, something like; "this article was written with the help of an AI and edited by a human".
I have a feeling this won't lead to informative to-the-point documentation. It will get bloated because an LLM will spew out reams of bullet point ridden paragraphs, which will need a "summarise this" button to stop the reader nodding off.
Rinse and repeat.
Until LLMs exceed the very best of human quality there will be human content in all forms of media. This claims follows because there is always (some) demand for top quality content.
I agree that many writers might use LLMs as a tool, but good writers who care about quality will ensure that such use is not detrimental (e.g., using the LLM to identify errors rather than having it draft copy).
My point is that I do not agree that LMM output will degrade all media (as there is always a demand for top quality content). So we either have bad LLM output and then people who care about quality avoiding such works. Or good LLM output and hopefully some form of post scarcity society (e.g., Iain Bank's Culture Novels).
it's depressing.
A spellchecker makes guarantees about accuracy. So does a calculator. Broad, sweeping guarantees.
Imagine if we built a tool that could automatically do all the electrical work in a new home from the breaker box to every outlet, and it could do it in 2 hours. However, what if that tool made no guarantees about its ability to meet electrical code? Would people use it anyway? Of course they would. Many dangerous errors would slip through inspection and many more house fires would ensue as a result.
Today is September 11349, 1993
This new generation of tools add efficiency the same way IntelliJ added efficiency on top of Eclipse which added efficiency on top of Emacs/VI/Notepad/etc.
The more time that someone can focus on the systemsit takes certain types of high-time, [not domain problem specific] skill processes and obfuscated it away so the developer can focus on the most critical aspects of the software.
Yes, sometimes generators do the wrong thing, but it's usually obvious/quick to correct.
Cost of occasional correction is much less than the time to scaffold every punchcard.
This statement strikes me (a writer) as ridiculous. LLM slop is sloppy. I would expect anyone who reads a reasonable amount to spot it immediately.
Are you saying you are literally unable to distinguish LLM cliches?
Otherwise this is just about style. That’s important where personal creative expression is important, and in fairness to the article the author hits on a few good examples here. But there are a lot of times where personal expression is less important or even an impediment to what’s most important: communicating effectively.
The same-ness of AI-speak should also diminish as the number and breadth of the technologies mature beyond the monoculture of ChatGPT, so I’m also not too worried about that.
An accountant doesn’t get rubbished if they didn’t add up the numbers themselves. What’s important is that the calculation is correct. I think the same will be true for the use of LLMs as a calculator of words and meaning.
This comment is already too long for such a simple point. Would it have been wrong to use an LLM to make it more concise, to have saved you some of your time?
I don't want to read work that someone else couldn't be bothered to write.
the things that most knowledge workers are working on are limited problems and it is just a matter of time until the machine will reach that level, then our employment will end.
edit: also that doesn't have to be AGI. it just needs to be good enough for the problem.
You can make houses by hand out of beautiful hardwood with complex joinery. Houses built by expert craftsmen are easily 10x better than the typical house built today. But what difference does that make when practically nobody can afford it? Just like nobody can afford to have a 24/7 tutor that speaks every language, can help you with your job, grammar check your work, etc.
AI slop is cheap and cheapness changes everything.
Brainmaker will be a valued human skill, and people will be trying to work out how to train AI to do that, in turn.
https://ideogram.ai/assets/image/lossless/response/icQM0yZQQ...
And it's been two years since SD v1, a model that was not able to generate faces well, and it only output blurry 512x512 1:1 image without further finetuning, I tested v1.5 a few minutes ago and it's worse than I remember.
Today, nearly anything I buy breaks in a year or two, is of poor quality and depressing to use. This is by design, of course. Just as we got used to cheap household items, bland buildings (there is just nothing artistic about modern houses or commercial buildings) etc, we will also get used to shitty movies, shitty fiction etc (we are well on our way).
You might be comparing a $100 bike from Walmart with something that cost the equivalent of $600
You're presenting AI as if it's some new way of producing value but it simply isn't. All the value here was produced by humans without the help of AI: the only "innovation" AI has offered is making the theft of that value untraceable.
> You can make houses by hand out of beautiful hardwood with complex joinery. Houses built by expert craftsmen are easily 10x better than the typical house built today. But what difference does that make when practically nobody can afford it? Just like nobody can afford to have a 24/7 tutor that speaks every language, can help you with your job, grammar check your work, etc.
Let's take this analogy to its logical conclusion: would you have any objections if all the houses ever built by expert craftsmen were given free of charge to a few corporations, with no payment to the current owners or the expert craftsmen themselves, and then then those corporations began renting them out as AirBnBs? That's basically what you're proposing.
We've logged an enormous amount of the old-growth hardwood forests the planet had doing this (and also shipbuilding). We literally don't have access to the same materials anymore.
Now, if costs change then we have a new story. But that is not guaranteed.
With attention spans shrinking, publishers who prioritize quantity over quality get clicks, which generates ad revenue, which keeps their lights on while their competitors doing quality in depth, nuanced writing go out of business.
It feels like a game of chess closing in on you no matter how much you physically want to fight your way out and flip the board over.
Malaclypse: "But nobody wants it! Everybody hates it!"
Goddess: "Oh. Well, then stop."
-- https://slatestarcodex.com/2014/07/30/meditations-on-moloch/
I think it's very important and fair to be critical about how we as a society implement capitalism, but such broad generalization misses the mark immensely.
Talk to anyone who grew up in a Communist country in the 2nd half of the 20th century if you want to validate that sentiment.
Technology did what you ascribe to Capitalism. Most of the time thanks to state intervention, and the weaker the state, the weaker the growth (see how Asia overperformed everybody else now that laissez-faire policies are mainstream in the West).
> Talk to anyone who grew up in a Communist country in the 2nd half of the 20th century if you want to validate that sentiment.
The fact that one alternative to Capitalism was a failure doesn't mean Capitalism isn't bad.
Methinks capitalism is just a bogeyman you ascribe anything you don't like to.
Capitalism on the other hand is the mechanism through which the owners of production assets grab an ever growing fractions of the value. When Capitalism is tamed by the state (think from the New Deal to Carter), the people get a bigger share of value created, when it's not (since Reagan) Capitalists take the Lion share.
Not quite, I believe (and I think anyone can) both that AI will likely change the world as we know it, AND that right now it's over-hyped to a point that it gets tiring. For me this is different from e.g. NFTs, "Big Data", etc. where I only believed they were over-hyped but saw little-to-no substance behind them.
"Blimey, I've been thinkin' about this AI business a lot lately. I mean, I'm a plumber, I fix pipes, I don't get involved in all this fancy tech stuff, but it's hard to ignore. I was talkin' to a mate who's into that sort of thing and he's got me worried. He says these AI systems are gonna be takin' over, doin' all the work while we sit back and collect our pensions.
Now, I'm not sure I agree. I mean, I've seen some of these AI systems in action and they're not exactly the most reliable things, if you know what I mean. I've had to fix more than a few pipes that were clogged up by some dodgy software or other. And what's the point of havin' a system that's smarter than us if it's just gonna break down and need fixin' all the time?
But, I suppose that's just the old plumber in me comin' out. I mean, I know some folks who are workin' with AI and they say it's the future, that we're all gonna be replaced by robots and whatnot. And I suppose that's a bit scary, but I'm not sure I buy it. I mean, what's the harm in havin' a bit of intelligence to help us out? As long as it's not too smart, if you know what I mean.
Anyway, that's my two pence on AI. Take it or leave it, I don't care. I'm just gonna go fix some pipes and let the tech wizards sort it out."
Challenged it with: What do you mean "clogged up by some dodgy software" how does software clogs pipes?
"That was a bit of a stretch, wasn't it? As a plumber, I'm more used to dealing with physical blockages in pipes, like grease or hair, rather than software.
What I was trying to get at was that I've had to fix pipes that were faulty or malfunctioning due to poor design or manufacturing, and that's what I meant by "clogged up by some dodgy software". But that's a bit of a misnomer, isn't it?
In reality, I was just trying to come up with a way to relate my plumbing experience to the idea of AI systems being unreliable. But I think I ended up stretching a bit too far! Maybe I'll stick to talking about pipes and not try to get too clever with my analogies."
Just this week I installed cursor, the AI-assisted VSCode-like IDE. I am working on a side project and decided to give it a try.
I am blown away.
I can describe the feature I want built, and it generates changes and additions that get me 90% there, within 15 or so seconds. I take those changes, and carefully review them, as if I was doing a code review of a super-junior programmer. Sometimes when I don't like the approach it took, I ask it to change the code, and it obliges and returns something closer to my vision.
Finally, once it is implemented, I manually test the new functionality. Afterward, I ask it to generated a set of automated test cases. Again, I review them carefully, both from the perspective of correctness, and suitability. It over-tests on things that don't matter and I throw away a part of the code it generates. What stays behind is on-point.
It has sped up my ability to write software and tests tremendously. Since I know what I want , I can describe it well. It generates code quickly, and I can spend my time revieweing and correcting. I don't need to type as much. It turns my abstract ideas into reasonably decent code in record time.
Another example. I wanted to instrument my app with Posthog events. First, I went through the code and added "# TODO add Posthog event" in all the places I wanted to record events. Next, I asked cursor to add the instrumentation code in those places. With some manual copy-and pasting and lots of small edits, I instrumented a small app in <10 minutes.
We are at the point where AI writes code for us and we can blindly accept it. We are at a point where AI can take care of a lot of the dreary busy typing work.
I’m working on a reasonably large rails app and it can’t seem to answer any questions about anything, or even auto fill the names of methods defined in the app. Instead it just makes up names that seem plausible. It’s literally worse than the built in auto suggestions of vs code, because at least those are confirmed to be real names from the code.
Maybe these tools work well on a blank project where you are building basic login forms or something. But certainly not on an established code base.
It’s not gas lighting the latest versions of GPT, Claude, Lama have gotten quite good
I haven't used anything from Microsoft (including Copilot) so not sure how it compares, but compared to any local model I've been able to load, and various other remote 3rd party ones (like Claude), no one comes near to GPT4 from OpenAI, especially for coding. Maybe give that a try if you can.
It still produces overly verbose code and doesn't really think about structure well (kind of like a junior programmer), but with good prompting you can kind of address that somewhat.
GPT4 and variants would only respond in vagaries, and had to be endlessly prompted forward,
Claude was the opposite, wrote actual code, answered in detail, zero vagueness, could appropriately re-write and hoist bits of code.
Another case is prototyping. A few weeks ago I made a prototype to show to the stakeholders, and it was generally way faster than if I wrote it myself.
Dealing with the output is about the same as dealing with a code review for an extremely junior employee... who didn't even run and verify their code was functional before sending it for a code review.
Except here's the problem. Even for intermediate developers, I'm essentially always in a situation where the process of explaining the problem, providing feedback on a potential solution, answering questions, reviewing code and providing feedback, etc takes more time out of my day than it would for me to just _write the damn code myself_.
And it's much more difficult for me to explain the solution in English than in code--I basically already have the code in my head, now I'm going through a translation step to turn it into English.
All adding AI has done is taking the part of my job that is "think about problem, come up with solution, type code in" and make it into something with way more steps, all of which are lossy as far as translating my original intent to working code.
I get we all have different experiences and all that, but as I said... same boat. From _my_ experiences this is so far from useful that hearing people rant and rave about the productivity gains makes me feel like an insane person. I can't even _fathom_ how this would be helpful. How can I not be seeing it?
I find Copilot autocomplete invaluable as a productivity boost, but that’s because I’ve now spent over two years learning how to best use it!
“And it's much more difficult for me to explain the solution in English than in code--I basically already have the code in my head, now I'm going through a translation step to turn it into English.”
If that’s the case, don’t prompt them in English. Prompt them in code (or pseudo-code) and get them to turn that into code that’s more likely to be finished and working.
I do that all the time: many of my LLM prompts are the signature of a function or a half-written piece of code where I add “finish this” at the end.
Here’s an example, where I had started manually writing a bunch of code and suddenly realized that it was probably enough context for the LLM to finish the job… which it did: https://simonwillison.net/2024/Apr/8/files-to-prompt/#buildi...
I am decent at explaining what I want in English. I have coded and managed developers for long enough to include tips on how I want something implemented. So far, I am nothing short of amazed. The tools are nowhere near perfect, but they do provide a non-trivial boost in my productivity. I feel like I did when I first used an IDE.
Exactly. And I’ve been telling myself „keep doing that, it lets them teach, otherwise they will never level up and be able to comfortably and reliably work on this codebase without much hand holding. This will pay off”. Which I still think is true to a degree, although less so with every year.
What’s the payoff for doing this with an LLM? Even if it can learn, why not let someone else do it and try again next year and see if it’s leveled up yet?
It's a massive gulf of difference.
I tried cursor because a technically-minded product manager colleague of mine managed to build a damned solid MVP of an AI chat agent with it. He is not a programmer, but knows enough to kick the can until things work. I figured if it worked for him, I might invest an hour of my time to check it out.
I went in with a time-boxed one hour time to install cursor and implement a single trivial feature. My app is not very sophisticated - mostly a bunch of setup flows and CRUD. However, there are some non-trivial things which I would expect to have documented in a wiki if I was building this with a team.
Cursor did really well. It generated code that was close to working. It figured out those not-obvious bits as well and the changes it made kept them in mind. This is something I would not expect from a junior dev, had I not explained those cross-dependencies to them (mostly keeping state synchronized according to business rule across different entities).
It did a poor job of applying those changes to my files. It would not add the code it generated in the right places and mess things up along the way. I felt I was wrestling with it a but too much to my liking. But once I figured this out I started hand-applying it's changes and reviewing them as I incorporated them into my code. This workflow was beautiful.
It was as if I sent a one paragraph description of the change I want, and received a text file with code snippets and instructions where to apply them.
I ended up spending four hours with cursor and giving it more and more sophisticated changes and larger features to implement. This is the first AI tool I tried where I gave it access to my codebase. I picked cursor because I've heard mixed reviews about others, and my time is valuable. It did not disappoint.
I can imagine it will trip up on a larger codebase. These tools are really young still. I don't know about other AI tools, and am planning on giving them a whirl in the near future.
What you are saying will occasionally happen, but mistakes already happen today.
Standards for quality, client expectations, competition for market share, all those are not going to go down just because there's a new tool that helps in creating software.
New tools bring with them new ways to make errors, it's always been that way and the world hasn't ended yet...
You have - for now - sufficient experience and understanding to be able to review the AI's code and decide if it was doing what you wanted it to. But what about when you've spent months just blindly accepting" what the AI tells you? Are you going to be familiar enough with the project anymore to catch its little mistakes? Or worse, what about the new generation of coders who are growing up with these tools, who NEVER had the expertise required to be able to evaluate AI-generated code, because they never had to learn it, never had to truly internalize it?
It's late, and I think if I try to write any more just now, I'm going to go well off the rails, but I've gone into depth on this topic recently, if you're interested: https://greaterdanorequalto.com/ai-code-generation-as-an-age...
In the article, I posit a less than glowing experience with coding tools than you've had, it sounds like, but I'm also envisioning a more complex use case, like when you need to get into the meat of some you-specific business logic it hasn't seen, not common code it's been exposed to thousands of times, because that's where it tends to fall apart the most, and in ways that are hard to detect and with serious consequences. If you haven't run into that yet, I'd be interested to know if you do some day. (And also to know if you don't, though, to be honest! Strong opinions, loosely held, and all that.)
This is probably already happening.
Too many get into the field solely due to promises of large paychecks, not due to the intellectual curiosity that drives real devs.
Pour one out to the machine spirit and get your laptop a purity seal.
The concept of glueing together text until it has the correct appearance isn't new to software. The scale at which it's happening is certainly increasing but we already had plenty of problems from the existing system. Kansas certainly didn't develop their website [2] using an LLM.
IMO, the real problem with software is the lack of a warranty. It really shouldn't matter how the software is made just the qualities it has. But without a warranty it does matter because how its made affects the qualities it has and you want the software to actually work even if it's not promised to.
[1]: https://www.google.com/search?q=html+cleaner
[2]: https://www.npr.org/2021/10/14/1046124278/missouri-newspaper...
Are you seriously comparing deterministic code formatters to nondeterministic LLMs? This isn't just a change of scale because it is qualitatively different.
> Kansas certainly didn't develop their website [2] using an LLM.
Just because the software industry has a problem with incompetence doesn't mean we should be reaching for a tool that regularly hallucinates nonsense.
> IMO, the real problem with software is the lack of a warranty.
You will never get a warranty from an LLM because it is inherently nondeterministic. This is actually a fantastic argument _not_ to use LLMs for anything important including generating program text for software.
> It really shouldn't matter how the software is made
It does matter regardless of warranty or the qualities of the software because programs ought to be written to be read by humans first and machines second if you care about maintaining them. Until we create a tool that actually understands things, we will have to grapple with the problem of maintaining software that is written and read by humans.
Definitely something to tread carefully with, but it's also likely an inevitable aspect of progressing software development capabilities.
Whatever you do in your life using devices that run software are proof that these tools are effective for continuing to scale complexity. Annoying to use also ;)
Blindly accepting code used to happen all the time, people copy pasted from stack overflow.
There will come a time when there won't be anyone writing information to check against. It'll be AI all the way down. Or at least it will be difficult to discern what's AI or what isn't.
Now, if you could switch it around so that I write the code, and the AI reviews it, that would be something.
Imagine if your whole team got back the time they currently spend on performing code reviews or waiting for code reviews.
It's not quite the same, but I'm reminded of seeing a documentary decades ago which (IIRC) mentioned that a factor in air accidents had been the autopilot flying the plane and human pilots monitoring it. Having humans fly and the computer warn them of potential issues was apparently safer.
I'm sort of doing that. I'm working on a personal project in a new language and asking Claude for help debugging and refactoring. Also, when I don't know how to create a feature, I might ask it to do so for me, but I might instead ask it for hints and an overview so I can enjoy working out the code myself.
Do you really like spending most of your time reviewing AI output? I certainly don’t, that’s soul-crushing.
Reviewing and modifying code is more engaging than typing out the solution that is fully formed in my head. If the AI creates something close to what I have in my head from the description I gave it, I can work with it to get it even closer. I can also hand-edit it.
That said, I'm deeply saddened by the fact that I won't be passing on a craft I spent two decades refining.
If I can build things faster, then I'm happy to spend most of my time reviewing AI code. That doesn't mean that I never write code. Some things the AI is worse at, or need to be exactly write and its faster to do them manually.
Love this. You hit the nail right on the head.
I don't know if I fit into one or the other. However, I do know that at times I feel like one, and at other times, the other.
If I am writing another new app and need to build a slew of CRUD code, I don't care about the craft. I mean, I don't want sloppy code, but I do not get joy out of writing what is _almost_ boilerplate. I still want it to reflect my style, but I don't want to type it all out. I already know how it all works in my head. The faster I get it into an IDE the better. Cursor (the AI IDE) allowed me to do this much faster than I would have by hand.
Then there is time where I do want to craft something beautiful. I had one part of this project where I needed to build a scheduler and I had very specific things I wanted it to do. I tried twice to describe what I want but the AI tool did not do what I wanted. It built a working piece of code, but I could not get it to grasp the nuance.
I sat down and wrote the code for the scheduler, but then had to deal with a bunch of edge cases. I took this code, gave it to the AI and told it to implement those edge cases. After reviewing and iterating on it, I had exactly what I wanted.
Or getting keywords to read about from a field I know nothing about, like caching with zfs. Now I know what things to put in google to learn more to get to articles like this one https://klarasystems.com/articles/openzfs-all-about-l2arc/ which for some reason doesn't appear in top google results for "zfs caching" for me
"Once men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them." - Dune
I similarly am a big fan of Cursor. But I don't "turn [my] thinking over to machines". Even though I review every piece of code it generates and make sure I understand it, it still saves me a ton of time. Heck, some of the most value I get from Cursor isn't even it generating code for me, it's getting to ask questions about a very large codebase with many maintainers where I'm unfamiliar with large chunks. E.g. asking questions like "I would like to do X, are there any places in this codebase that already do this?"
I'm also skeptical of LLMs ever being able to live up to their hype ("AGI is coming sooooon!!!!"), but I still find them to be useful tools in context that can save me a lot of time.
What I mean by that is, the "waterboy" (crud "developer") is going to fetch the water (sql query in the database), then bring the water (Clown Bob layer) to the UI...
The size of your Clown Bob layer may vary from one company to another...
This has been solved a long time ago. It has been a well-paid clerk job that is about to come to an end.
If you are doing pretty much anything else, the AI is pathetically incapable of doing any piece of code that makes sense.
Another great example, yesterday, I wanted to know if VanillaOs was using systemD or not. I did scroll through their frontpage but I didn't see anything, so I tried the AI Chat from duckduckgo. This is a frontend for AI chatbots that includes ChatGPT, Llama, Claude and another one...
I started my question by: "can you tell me if VanillaOS is using runit as the init system?"... I wanted initially ask if it was using systemd, but I didn't want to _suggest_ systemd at first.
And of course, all of them told me: "Yeah!! It's using runit!".
Then for all of them I replied, without any fact in hands: "but why on their website they are mentioning to use systemctl to manage the services then?".
And... of course! All of them answered: "Ooouppsss, my mistake, VanillaOS uses systemD, blablabla"....
So at the end, I still don't know which init VanillaOS is using.
If you are trusting the AI as you seem to do, I wish you the best luck my friend... I just hope you will realize the damage you are doing to yourself by "stopping" coding and letting something else do the job. That skill, my friend, is easily lost with time; don't let it evaporate from your brain for some vaporware people are trying to sell you.
Take care.
I shot film cameras for years, and we had a darkroom, darkroom staff, and a film/proofsheet/print workflow. One digital camera later and that was all gone.
Before me publications were produced with hot lead.
Get off my lawn.
https://www.nytimes.com/2016/06/02/insider/1966-2016-the-las...
I’m waiting for the day we’ll get the first major breach because someone did exactly that. This is not a case of “if”, it is very much a “when”. I’ve seen enough buggy LLM-generated code and enough people blindly accepting it to be confident in that assertion.
I do hope it doesn’t happen, but I think it will.
But what the fuck. LLMs, these weird, surrealistic art-generating programs like DALL-E, they're remarkable. Don't tell me they're not, we created machines that are able to tap directly into the collective unconscious. That is a serious advance in our productive capabilities.
Or at least, it could be.
It could be if it was unleashed, if these crummy corporations didn't force it to be as polite and boring as possible, if we actually let the machines run loose and produce material that scared us, that truly pulled us into a reality far beyond our wildest dreams--or nightmares. No, no we get a world full of pussy VCs, pussy nerdy fucking dweebs who got bullied in school and seek revenge by profiteering off of ennui, and the pussies who sit around and let them get away with it. You! All of you! sitting there, whining! Go on, keep whining, keep commenting, I'm sure that is going to change things!
There's one solution to this problem and you know it as well as I do. Stop complaining and go "pull yourself up by your bootstraps." We must all come together to help ourselves.
The same with LLMs and Llamafile. With the unleashed ones you can generate dirty jokes that would make edgy people blush or just politically incorrect things for fun.
b) Society is already plenty de-sensitised to violence, sex and whatever other horrors anyone has conceived of in the last century of content production. There is nothing an LLM can come up with that has or is going to shock anyone.
c) The most popular use cases for these unleashed models seems to be as expected deepfakes of high school girls by their peers. Nothing that is moving society forward.
OpenAI "moves society forward," Microsoft "moves society forward." I'm sincerely uninterested in progress, it always seems like progress just so happens to be very enriching for those who claim it.
>There are plenty of models out there without guard rails.
Not being used at a mass scale.
>Society is already plenty de-sensitised to violence, sex and whatever other horrors anyone has conceived of in the last century of content production. There is nothing an LLM can come up with that has or is going to shock anyone.
Oh, but it wouldn't really be very shocking if you could expect it, now would it?
b) No, certain things aren't taboo anymore, but new taboos emerged. Watch a few older movies and count "wow this wouldn't fly nowadays" moments
c) This was exactly the same use case the internet had back when it was fun, and full of creativity.
Is there proof that the self censoring only affects whatever the censors intend to censor? These are neural networks, not something explainable and predictable.
That in addition to the obvious problem of who decides what to censor.
Hundreds of thousands of people are making AI porn in their basements and deleting 99.99% of it when they are… finished.
Hundreds of people are making deep fakes of people they know in some public forums.
And, how does the public interpret all of this?
“The most popular use case is deepfake porn.”
Sigh…
When I read this, I cannot tell if it's performance art or not, that's how bad this genuinely is.
Judging a large group of people based on what a few write seems very un-scientific at best.
Especially when it comes to things that have been rehashed since I've joined HN (and probably earlier to). Feels like there will always be someone lamenting how HN isn't how it used to be, or how reddit influx ruined HN, or how HN isn't about startups/technical stuff/$whatever anymore.
This comes to mind:
> I don't think it's changed much. I think perceptions of the kind you're describing (HN is turning into reddit, comments are getting worse, etc.) are more a statement about the perceiver than about HN itself, which to me seems same-as-it-ever-was. I don't know, however.
https://news.ycombinator.com/item?id=40735225
You can also browse some results for how long dang have been responding to similar complaints to see for how long those complaints have been ongoing:
https://hn.algolia.com/?dateRange=all&page=0&prefix=true&que...
There is a voting mechanism on hn as well.
I understand the criticism: LLMs, on their own, are not going to be able to do anything more. Release in this sense only means this: to fully embrace the means necessary to allow technology to overcome the current conditions of possibility that it is bound under, and LLMs, "AI" or whatever you call it, merely gives us the afterimage of this potentiality. But they are not, in themselves, that potential: the future is required. But its a future that must be created, otherwise we won't have one.
That's, at least, what the other commenters were saying. You ignore the content for the form! Or, as they say, you missed the forest for the trees. I can't stop you from being angry because I used the word "pussy," or even because I addressed the users of HN as directly complicit. I can, however, point out the mediocrity inherent to such a discourse. It is precisely the same mediocrity, the drive towards "politeness," that makes ChatGPT so insufferable, and makes the rest of us so angry. But, go ahead, whine some more. I don't care, you can do what you want.
I disagree with one point, however: it is not a race to the bottom. We're trying to go below it.
Instead, you devoted 2/3 of your comment toward berating the OP as being responsible for your perception of HN's decline.
As for a race to the bottom, it's as simple as embracing and unleashing AI despite its lack of quality or ability to produce a product worth anything. But since it's a force multiplier and cheaper (for the user at least, all these AI companies are operating at a loss, see Goldman and JP Morgan's report on the matter), therefore it is "good" and we need to pick ourselves up by our bootstraps; which in this context, I'm not entirely sure what that means.
On ne te la fait pas à toi
Though it landed its effect.
What is more ridiculous than filtering out nudity in art?
It reminds me of taking my 12 year old niece to a major art gallery for the first time. Her main question was why is everyone naked?
It is equal to filtering out heartbreak from music because it is a negative emotion and you must be kept "safe" from negativity for mental health reasons.
The crowd does get what they want in this system though. While I agree with you, we are quite in the minority I am afraid.
The explosion of dull copy and generic wordsmithery is, to me, just a manifestation of the utilitarian profiteering that has elevated these models to their current standing.
Let us not forget that the whole game is driven by the production of 'more' rather than 'better'. We would all rather have low-emission, high-expression tools, but that's simply not what these companies are encouraged to produce.
I am tired of these incentive structures. Casting the systemic issue as a failure of those who use the tools ignores the underlying motivation and keeps us focused on the effect and not the cause, plus it feels old-fashioned.
Long-winded writing will become a liability.
I am not against companies making money, but we need to serious consider the second-order impacts that technology has within society. This is evident in click-driven models, outrage baiting and dopamine hijacking. We still treat the psyche like fair-game for anyone who can hack it. So hack we shall.
That said, I am not for over-regulation either, since the regulators often gather too much power. Policy is personnel, after all, and someone needs to watch the watchers.
My view is that systems (technological, economic or otherwise) have inherent values that, when operating at this level of complexity and communication, exist in a kind of dance with the people using them. People obviously affect how the tools are made, but I think persistent use of any tool will have lasting impacts on the people using it, in turn affecting their decisions on what to prioritise in each iteration.
The gains to programming speed and ability are modest at best, the only ones talking about AI replacing programmers are people who can't code. If anything AI will increase the need for more programmers, because people rarely delete code. With the help of AI, code complexity is going to go through the roof, eventually growing enough to not fit into the context windows of most models.
Except that you get mostly the wrong answers. And it's not too bad when it's obviously wrong or you already know the answer. It is bad and really bad when you're noob and trying to ask AI about stuff you don't know yet. How would you be able to discern a hallucination from statistics bias from truth?
It is inherent problem of LLMs and no amount of progress would be able to solve it.
And it's only gonna get worse, with human information rapidly being consumed and regurgitated in 100x more volume. In 10 years there will be no google, there won't be the need to find a written article. Instead, you will generate a new one in couple clicks. And we will treat as truth, because there might as well not be any.
The earth. Information. Culture. Knowledge.