It's made me consider my habits, and how much time I spend consuming internet content or comment threads which may (even today) be completely fabricated or autogenerated. I'm hoping I can use this inflection point as motivation to unplug a bit and spend more time enjoying being human and enjoying the real world.
What Ploum describes sounds a lot like Nietzsche's "Last Computer".
A feeble, average of things - the sum of all fears, hopes, dreams,
inconsequential experiences and gossip, trivia, boasts and bullshit,
plus all documented knowledge - divided through by 8 billion.
"Butlerian Jihad" that was the most surprising for me reading Dune that they had a limit on how smart computers could be. The movie was very vague on the subject, and was more of a Star Wars like lo-fi.
It was even more surprising if you realize that "Butlerian" is referring to the 19th century writer Samuel Butler (although, yes, Herbert later invented a character to explain the name in-universe). Samuel Butler wrote an 1863 essay "Darwin Among the Machines" which predicted that machines would improve in a way analogous to natural selection and would so so faster than humans could evolve, and therefore they might eventually surpass humanity.
indeed, and selection pressure is not on the domain of physical bodies like for us fleshy species, but profitability for digital computational species in the domain of capital.
We're already drowning in human generated garbage on the web. Sure, AI makes it worse, but we can't blame anything but our human compatriots and ourselves.
That's not even what I'm talking about. I was thinking of SEO content, ads, tiktok, twitter, facebook garbage. Please more well articulated viewpoints!
When you think about it, since most of that "human generated garbage" is mainly trying to appease recommendation algorithms, which are mostly AI these days, even the human generated garbage is actually AI generated garbage by proxy.
We can't kill the machine, but we can starve it. None of this would exist without advertising. If we kill advertising we kill the content farms. And nothing of value will be lost.
While the current speed of the AI research makes me quite uneasy, I cannot really dismiss what we are seeing as:
>What we are witnessing is thus not "artificial creativity" but a simple "statistical mean of everything uploaded by humans on the internet which fits certain criteria".
at least not more than I would my own brain.
I don however fully expect a strong luddite-like backlash happening when this is going to be deployed to more and more places. (Which is not to say that everyone opposed to it is a luddite, but there is a very strong impulse, I have it to, to see this technology as a threat).
It’s weird how readily so many people will dismiss AI art as just being a product of its input, when that’s really all humans are as well. All artists (myself included) are just the ongoing product of various inspirations (inputs). Dali is a world-renowned artist, but his art wasn’t born in a bubble and ripped from the void. He studied and referenced artists before him, then changed it. Artists who saw his work later incorporated elements he used.
My big worry from here though is what’ll happen when we take the next step: bundling these AIs together and start making a comprehensive AI. We have AIs that can respond to comments to some extent now, we can generate realistic selfies, and they can make art based off prompts that they could be designed to feed to themselves—set that loose on Twitter and you have a completely automated and convincingly living artist out there. Next step after is using AI to handle free movement by auto-adjusting balance and then performing convincing human speech. The future is getting freaky and fast.
But at least artists have some agency, all that stuff that they ingested goes through them and are somewhat limited by human hardware. Very little is original indeed but that percentage may as well go down even more as more and more AI generated soup unleashed on the world.
What about artists that used random generators in their art? Jackson Pollock used to randomly splatter paint on canvases and when some random pattern struck his fancy he went with it. That's not all that different from picking the pictures you like from a Stable Diffusion prompt and then further honing them with inpainting, image-to-image and so on.
There's a fair bit of truth in this but it remains a nearly useless oversimplification. We can't adequately isolate or even identify all the inputs into what makes a person what they are, or produce what they do. The idea that this is even possible is rejected by a large proportion of the world population, and for now at least we can't prove them wrong.
> All artists (myself included) are just the ongoing product of various inspirations (inputs)
This ignores that inspiration is itself an active and creative process. There's not a one to one mapping between a piece of art and the effect it has on a person who receives it. Even hating an art is a process of influence in a very real way that I don't know there's an AI analog of. These systems can not include a source in their output, but can they decide to include its opposite? What is the opposite of a tree, btw? A desert, an ocean, a seed? A value judgement based on culture and aesthetics and opinion and worldview that ML software has no standpoint from which to make.
I'm not trying to make a moral case for or against this technology right now. But I think trying to understand and evaluate it in human terms like this is a mistake. It's something else, and we're still trying to figure out what terms are appropriate to understand it, hence the difficulty. It's important to honestly engage with that alienness and uncertainty right now.
This is a tricky distinction where this may be true of all specific individual humans, but not of humanity. At some point, language and art didn't exist, and humanity invented those things. We didn't mimic some other civilization that did it first.
I don't know that this distinction matters, but it is real. Even on an individual level, if you raise a child in complete isolation from history and teach them no language or art, they'll still form thoughts, have feelings, and develop means by which to communicate those.
I think this is the big open question of evolution as a whole, and language and art can be seen as the evolution of ideas/memes as Dawkin wrote about in the Selfish Gene. Were these developments sufficiently iterative, or was there some great leap that came out of nowhere? Was the wing merely just a stump on an animal that allowed it to leap tree to tree a foot further, or was it some process unknown to us? Was art just a splatter of blood on a cave that looked like a face to our brain’s face neural network, or did some inventive genius paint a mosaic when nothing else was ever on the cave wall?
Only boring humans are just the product of the content they consume; interesting people have concerns about relationships with other humans, organisations, resources, threats and opportunities that come from being an entity living in a world full of things they don't control and things they'd quite like to control. It's also fairly easy to identify artists who have done nothing but look at other art.
Hayao Miyazaki once said "Almost all Japanese animation is produced with hardly any basis taken from observing real people, you know. It's produced by humans who can't stand looking at other humans. And that's why the industry is full of otaku!" While I'm not actually a huge fan of his work, it's easy to see where he's coming from when you see the likes of "In Another World with My Smartphone" or the hundreds of other trash light novel adaptations. They're clearly fiction being produced by people whose lives are nothing but consuming other fiction, and all they can do is reference that. They have nothing to say.
For another example, there's a huge number of 20th century war films. They weren't made because their creators had read a lot of Wikipedia articles about how destructive wars were becoming. The likes of Platoon are a product of mortal fleshbags who actually saw combat and genuinely feared for their lives. Dr. Strangelove was created by people who had a genuine belief that they were at risk of being annihilated in a nuclear fireball.
It wouldn’t be a threat if we didn’t have an economy that distributed most gains to the very top. If things were more equitable it would mean we will all become much wealthier due to a large productivity increase.
I almost never see that raised in these discussions. We have been so conditioned not to think about this subject.
We can blame the tech, each other, everything but the system.
People really love to make predictions about technologies way too early. It’s been a year or two, tops, since this stuff has gone mainstream. I very much doubt that the consumer ecosystem will remain the same while content undergoes an AI-powered paradigm shift. We will most likely see a growth of curated sites that help weed out the nonsense, as has been typical for most new developments.
And on that note, I think the framing here is just wrong from the start. These new tools are awesome if you’re a creator. Complaining about the consumer-facing side of this new tech is a bit like complaining that cheaper video cameras will lead to junk films everywhere. Yeah, that happened, but who cares? It also allowed creative people to make stuff without needing a huge budget.
People really love to make predictions about technologies way too early. It’s been a year or two, tops, since this stuff has gone mainstream.
Prediction:
I very much doubt that the consumer ecosystem will remain the same while content undergoes an AI-powered paradigm shift. We will most likely see a growth of curated sites that help weed out the nonsense, as has been typical for most new developments.
Filmmaking devices have gotten cheaper consistently over the last ~50 years. I don’t know about you, but I think some pretty good movies have come out since 1970, especially by directors working on very low budgets.
AI creation tools make it easier to make stuff. This process will result in a lot of junk, but it will also result in some super creative people making art they otherwise couldn’t.
The same process happened with filmmaking. Fifty or sixty years ago, it was too expensive for anyone other than large corporations.
You could even say the same thing for writing itself; a thousand years ago, paper was too expensive for the average person and most people couldn’t afford to buy books and become literate. Today, writing paper is functionally free.
And it's not just creation of art from "scratch", I bet we will soon see AI being able to take painted animation frame, sketch of next and extrapolate next frame from it, and making whole process so much cheaper
An Avid nonlinear video editing station used to cost $100K. Nowadays you can do the same with software costing $300 on the computer you already have. You can get a mirrorless video camera that blows away the Sony CineAlta ones George Lucas used to shoot the Star Wars prequels for under a grand. This has removed a huge chunk of production costs and make indie film-making much more viable. The only limits are your imagination and the salaries of the actors and camera crew.
AI-generated animation is already being used by industry giants like Toei, and will likely be the first application to hit the screens.
> I don’t know about you, but I think some pretty good movies have come out since 1970, especially by directors working on very low budgets.
On the whole, I think the average quality of cinematography in theater or wide-streaming releases has decreased.
Which isn't disagreement with your point, as I think it says more about audience tastes and gatekeeping filters than technology.
When film equipment was expensive and the bare minimum for creating a film was high, studios had a much tighter filter on who they'd fund to create a film. Consequently, the level of craft was higher.
What needs to be said here is that average quality is more or less irrelevant. If the amount of high quality movies increases (it did) and the discoverability doesn't suffer (it didn't too much), then that's a plus.
The thing that has tanked the quality of theatrical films on average isn't improved technology, it's the increasing commodification of art, risk aversion of business and the modern reliance on data to drive business decisions. That's why we have basically nothing but sequels, reboots and formulaic drivel coming out of the major studios.
“It’s too early to make predictions. Here’s my prediction”
^^^ I think this is what the response meant. I noticed it too, you castigated folks for making predictions then led into your own :-) a real AI wouldn’t do that!
I don’t exactly understand your question, but I’m assuming you mean something like, “What value is this person actually creating?”
I can think of two direct use cases where I would find it helpful:
- Creating a “précis” (concept from Neuromancer) video or article on a particular topic. For example, let’s say I want to learn about pre-20th century Indonesian art. Sure, I could Google and read some articles, but it would be vastly more enjoyable and efficient to just get a 5-minute video auto-generated on the topic.
- For creating content that I want to read but don’t necessarily have the time to write. I’d love to create a website that catalogs subcultures, for example, but I just don’t have the time to do the research and writing myself.
Both of these are dependent on the generated content being accurate, of course, but that seems like a temporary problem, not a flaw in the system itself.
I think the nearby comment that points out that this is more closely related to editing than to creation as it has recently been understood is correct. Which is not a value judgement per se, but the distinction between those two activities is useful and to me provides a useful lens to understand this change.
On the face of it, that is true. But on a deeper level, AI generation is really just a kind of rapid sketching, and sketching is a practice that goes back millennia.
Additionally, it depends on how you are defining "creation" and there have been various artistic movements which "created" art in a manner not dissimilar to AI image generators. The cut-up technique or some Dadaists, for example. I certainly would consider works by Burroughs or Tzara to be real creations and not simply editing.
You're conflating two meanings of sketching though. Artists and others do uses it as an exploratory tool, and I've already seen AI generation used in a similar way so I know that's a valid application of it.
But at a lower level sketching is itself a visual interpretive and creative process. And yes what we're talking about is mechanically similar to the cut-up technique, but is that used for "sketching" in the exploratory sense?
I don't know, maybe. My point overall is that these tools aren't "really just" anything. They have similarities and similar applications to many things we've seen and done, yes. But they are fundamentally alien and in important ways are unlike anything else we've done. We should be careful about and aware of the limitations we put on our understanding when we try to interpret them through the lens of what has come before. To some extent we have no other lens, but that doesn't lessen the limitation.
I think in this case the role of the human in the process does have a strong analog to that of an editor. Where you have the ability to initiate and then shape the creation of another entity, but the creation itself is out of your hands. It's not exact, but it may be close enough to be useful, in a way I think comparing it to the cut-up technique is not quite.
Why would anybody need to type anything? Gpt will type the gpt prompts. AI will create the AI that will create the AI. There will be people involved but increasingly less.
We need an AI that writes AI papers, and one that implements AI papers. They could be both chatGPT. Then just let them run in a huge datacenter for a year.
Are you the author of the binary your compiler made? Sure you are. Of think of using blender geometry nodes to generate and render pictures of architecture. Nobody would think that traditional computer generated images are not made by humans.
The is no intelligence in gpt-3. It's just called "AI" for marketing reasons. It's an data driven algorithm.
I’d argue that the creator went from being an author to being an editor, and that is fine.
Personally, I am enjoying chatgpt and copilot. They both have saved me a lot of time when dealing with new tech and accumulating new information.
I don’t trust them for certain highly sensitive or technical issues, but other than that if I can minimise the time I spend working without a reduction in productivity, well, I am all for it.
Except that isn't the creative loop people are talking about. The creative loop is having an outline for what you want to communicate, and feeding parts of the outline into the model, running it a few times and cherry picking the best results, then stitching them together to create a cohesive whole in less time than it would have taken to create it from scratch (and with a higher quality final output).
I we have already seen this play out in music though.
Computers let anyone record music when it use to cost thousands of dollars to record music.
The results have been a disaster outside the top 1% of music makers. Such a dramatic change that anyone under 30 doesn't even know what has been lost at this point. Music as a whole is laughably less creative as "popular" / pop music established total dominance. The money all flows upward that use to be much more dispersed.
I don't think that's true at all. There are a ton of niche music makers and a ton of ways to find them...infinitely more than there were a few decades ago. You just have to go looking for them. Which frankly isn't even that difficult with YouTube and SoundCloud. Maybe the musicians aren't making millions, but the creativity is definitely there.
Sorry but no, this is completely the wrong take. Popular music is less diverse today for the same reason that half the movies released now are Marvel - art is more commoditized, and businesses are more data driven, with the shift towards more public companies also playing a role, as they can't take as many risks as private companies.
What cheap music production tools have done is allow niche subcultures to create music that actually sounds good. Punk, metal and early hip hop all grew from people sharing rare tapes that were poorly produced and sounded like shit, but were a breath of creative fresh air. Now people's unusual creative divergences are stream-able and have production quality close to that of stuff on major labels.
Lets say in 1 year 10 songs are created and 2 of them are 'creative'. You'd think that 20% of music is creative.
Now lets say we have the production explosion and now in 1 year 100 songs are created and 4 of them are creative. You would think creativity has fallen off a cliff when it has really doubled.
In addition trying to compare with the previous payola ridden system were winners and losers systematically chosen by a wealthy cabal seems rather asinine. It only worked because there was a limited number of radio stations mostly. Even if production costs stayed high, distribution costs dropping would have changed the model dramatically anyway.
I used to feel the same way but I eventually came to realise that there’s tons of good music being created, it’s the social aspect of music that has been destroyed.
Individual algorithm recommendations have subsumed sub cultures. Music used to be part of both your individual and group identity, it would influence what your wore, what you read, what you thought, who you hanged out with and what bars you visited. Technology has eroded the music press and sky high rents and fitness culture (not as many people drinking) means lots of gig venues and bars have closed down. These were the glue that built the subcultures up. You used to have the mainstream plus a few sub cultures that would evolve each decade e.g mods, rockers, hippies, metal heads, punks, new wave, hair metal, brit pop, emos, indie rockers etc. All of these ended up getting absorbed into the long tail so you now have the main stream and then everything else, with nothing in the everything else category really able to attract enough people to it in order to form era defining subculture like the days of old.
Combined with the sheer amount of music being created, all of this makes it harder to find music you actually want to hear. Your friends are all listening to different stuff so you don’t get many good in person recommendations anymore and what’s left of the music media are struggling to stay afloat and resorting to all kinds of stuff like NME doing movie reviews. The algorithms are hit and miss and search engines are getting worse and worse. So you become more and more at the mercy of chance, hoping that somewhere you stumble upon a review or a blog post, a subreddit or a song that somehow leads you to some good new music. But the reward to effort ratio often makes it harder to justify the time investment. And however good the music is, it’s not as enjoyable when you don’t have a group of people to share it with.
"Eating your own shit is never healthy in the long run."
However, putting someone else's shit in your butt seems to work wonders.
Maybe we just need to replace the internet's AI queue with a stack?
Yes, your own shit is not really healthy, on the other hand, someone else's shit might be really beneficial. There is even a fancy scientific phrase for "eating other peoples shit" : Fecal Microbiome Transplant
What I really wonder is: is there a correlation between consumption of streaming tv series and positive attitude towards this "AI" wave?
"statistical mean of everything uploaded by humans on the internet which fits certain criteria"
applies to both ChatGPT/StableDiffusion and most movie series on streaming TV.
My pet theory is that if you can stomach watching the same stories with different skins hundreds of times, you will find current "AI" interesting. Otherwise no.
I’m 41 and I’m utterly baffled by my 24 year-old SO’s interest in binging an apparently infinite litany of TV shows from streaming services. I’ve watched the first six seasons of Game Of Thrones, the first season of Dark, all of The Expanse (reading the last book of the final trilogy now), His Dark Materials (because I read the books) and Andor. That’s it. I watched some Star Trek TNG & DS9 as a kid and was horrified by what snippets of Discovery and Picard I came across. I find the bulk of modern entertainment to be an utter insult to viewers.
> baffled by my 24 year-old SO’s interest in binging an apparently infinite litany of TV shows … TNG & DS9 as a kid and was horrified by what snippets of Discovery and Picard I came across
The thing that bemuses me is those who hate the new trek shows with a passion, and find the time to keep watching and constantly bitching about it online. It is like complaining is a hobby in itself and they are almost looking for something they don't like or to feel superior about their judgement to those who do like it. Other franchises have similar subsets of “fans” too.
I used to watch too much TV/films IMO, I probably still do though I'm a lot more selective since getting into more varied hobbies that can easily consume a lot of time (running (usually off-road), baking, HEMA). But even when I had the free time to watch that much I found there was enough out there that if something didn't work for me I could find a list of other things to try instead of sticking with something I wasn't getting much out of.
Though be careful passing judgement on modern Trek: I'm slowly rewatching TNG with a friend and while some of it has aged well, fair chunks of it was more massively ropey than rose-tinted memories prepared me for! Also, if I hadn't at least passively consumed (i.e. had on TV while I was doing other things so not particularly concentrating on) other parts of recent Trek and Starwars, I might not have bothered to try some of the better stuff like Lower Decks, Strange New Worlds, and Andor (the latter probably being the second best TV thing recently for me, the first being Severance).
I don’t understand it either… if you don’t like it, walk away. Don’t contribute to the toxic commentary and above all don’t contribute to the viewing numbers, because if they can hold your gaze you’ll get more of the same.
I find the meta-content this generates pretty funny thought. I haven't watched any of the new Trek content since the last movie (Beyond), but I love watching Mike and Rich tear it apart on RedLetterMedia.
I part way understand it. There is a certain tribalism in the way their views are presented, like they feel they are fighting for their own kind in a battle that means something worth the effort. There is also a certain amount of 4-year-old-who-wants-the-green-one-not-the-yellow-one too…
> The thing that bemuses me is those who hate the new trek shows with a passion, and find the time to keep watching and constantly bitching about it online.
Very meta of you to watch these people and find the time to complain about them.
Though I like to think there is a difference between myself and them: they have a choice to keep experiencing the pain they perceive, I have it thrust upon me when I visit certain places (which I visit less often because of those people – once the SNR drops to a certain low I'm out).
I've dived back into anime recently because of the aggressive sameness of everything on streaming TV. I don't know whether something changed about how "western" media is written or I've just watched so much TV that I can't help but roll my eyes at commonly deployed tropes. At the risk of outing myself as a huge "weeb", I find anime still manages to surprise and entertain me where western programming does not.
First off, probably don't date a 24 year-old. I realize there is no strict correspondence between age and taste, but I'm 42 and my wife is 41 and, well, our shared experience is quite a bit more similar. She can relate when I get nostalgic about the 80s and 90s. She even reads! Complete novels, more than she watches TV. She doesn't have TikTok or Instagram. It is entirely possible to go full-time mutual old people yelling at clouds in a blissful bubble in which the river of bullshit may as well not exist, but probably not if you insist on staying cool and regularly interacting with young people.
Second, watch the rest of Dark, Better Call Saul, and House of the Dragon. Don't watch the last two seasons of Game of Thrones. Just pretend it ended at 6.
It is a real problem. The web will soon be full of AI generated bullshit.
What might come back as useful tool? Authorship. People actually signing the content and putting their reputation on the line. It still exists obviously (journalism…) but it might need to be standardized and industrialized.
Another problem is AI training on AI generated content. This could lead to degradation of results. So perhaps to make everyone happy a new HTML tag indicating AI content would be useful?
> Another problem is AI training on AI generated content. This could lead to degradation of results.
That's only the first step in the degradation. In a worse case scenario, where AI-generated crap is not contained, imagine this: young and still immature humans "learning" their styles of writing, information, and social expectations from content generated by AI that itself trained on AI-generated content─and then more AI trains on content generated by these humans that were trained on AI-generated content. And on, and on, and on...
Ah, but they do want to fit in and speak like their peers, and while a significant part of their socialisation happens online, and as long as that is where they absorb most content, AI will soon infiltrate the circle of online strangers with whom they interact. That's all the more so considering that teenagers are prime targets for advertising networks of dubious ethical standards, who will be the first to deploy (or may already have deployed) well trained and self-correcting "How do you do, fellow kids?" NNs in potentially lucrative communities or social networks.
(I'm not always as doom-and-gloom as it may seem, I'm just in a pessimistic mood today for some reason)
Comply with what? Even in culture we tend to have a limited number of subsets of counter culture. The vast majority of the non-compliant to primary culture fits in one of a few common subcultures. As much as HN users try to exclaim they are islands that stand alone from society, we are all dependant on a functioning society for day to day life.
We should ask questions like "Can countless AI's fracture society into a million sub societies disconnected from the reality around them".
Fair enough. It’s like the lump of labor fallacy, but for bullshit rather than labor. One could view that as growing pains, but it is still growth in some sense.
Though the only way to do that is probably to end up in a world where most worthwhile content is locked behind hard paywalls so there's one class who pays a lot for gated content (which was more or less the historical state of affairs, public libraries notwithstanding) and the plebes have to filter through all the ad-laden crap.
People "sign" their content with their faces on TikTok; it doesn't stop the deluge of complete nonsense that's widely accepted on that platform (much to the frustration of many academics trying to combat it).
Indeed, Neal Stephenson in _Anathem_ (2008), in describing an alternate world (in which his "reticulum" is our "network") wrote "Early in the Reticulum—thousands of years ago—it became almost useless because it was cluttered with faulty, obsolete, or downright misleading information."
"So crap filtering became important. Businesses were built around it. ... " Generating crap "didn't really take off until the military got interested" in a program called "Artificial Inanity".
The defenses that were developed back then now "work so well that, most of the time, the users of the Reticulum don't know it's there. Just as you are not aware of the millions of germs trying and failing to attack your body every moment of every day."
A group of people (the "Ita") developed techniques for a parallel reticulum in which they could keep information they had determined to be reliable. When there was news on the reticulum, they might take a couple of days to do sanity-checking or fact-checking. I'm guessing there would need to be reputation monitoring and cryptographic signatures to maintain the integrity of their alternate web.
It already has a lot of AI and garbage content. And in some cases it's state sponsored propaganda, companies pitching products, or marketing departments attempting some notion of virality. The point is that there are plenty of people already gaming the system by just flooding the internet with carefully designed noise to bend opinions this way or that way.
"Fall, or dodge in heaven" by Neal Stephenson actually touches on this topic quite a bit and comes to a similar conclusion. Not his most accessible work but I enjoyed it. That book was written just before Trump won 2016 elections and in hind sight making that book really spot on and timely.
In the book, different streams of truth become so hard to separate due to all the spam, automated content, etc. that people start filtering it based on reputation and signatures. The game then becomes curating and filtering the content, which is ultimately not an objective thing but a subjective thing. So the net result is what we are actually starting to see today which is different groups of people locked in their own bubbles and no longer even getting exposed to the other side's point of view or even able to agree on some basic facts. Like who actually won the election. And people are feeding those bubbles with nonsense to influence opinions, get people to vote certain ways, or simply to just create chaos and fear.
In the book the topic is not an election but the seemingly simple question of whether or not a certain town in the us still exists after it was supposedly blown up with a nuclear bomb. In the book, this event leads to people acting as if that happened and ignoring all evidence to the contrary pointing out that in fact this was an elaborate hoax designed to create a massive panic.
That particular dystopian outcome is not necessarily what would happen. But using technology to authenticate where content actually comes from along with some reputation mechanism would go a long way separating the noise from the genuine content. It's simple and easy to do. So, why don't we?
In just the past couple of weeks, I've noticed lots of machine generated cooking recipes. They follow the typical SEO pandering blog format, where there's an unnecessary story about the author's trip somewhere or their child's school play being the inspiration for making this particular recipe, followed by the actual recipe details. The machine generated stories are even more pointless and meandering than what humans come up with and the recipe details have steps that contradict each other and specify ingredients that shouldn't be in the dish and weren't on the ingredients list.
I don't know how many people are fooled by these recipes and end up with a cooking disaster but if we start seeing faulty machine generated instructions on how to do electrical work or engine repair, the consequences will be much greater when followed. The real annoyance is that because these machine generated articles learned from sites that maximize SEO, the generated articles are rapidly rising to the top of search results and crowding out actual good information that wasn't SEO maximized.
Yes - AI moves this to another level of scale but Google’s need to sell ad placements created the environment for that to work. When they can’t even be troubled to de-rank blatant StackExchange/GitHub/Wikipedia scrapers it created a big opportunity for these low-quality content farms limited only by their ability to generate content on every topic.
Except for the fact that it does so on an unprecide dented scale allowing small groups of people to do things without necessarily being caught allowing for far more effective ways of control. Even without it, it's hard to argue for the existence of online forums when hackernews is a classic example of why it doesn't work when the majority of the 'intellectuals' have an inflated belief and understanding on topics they have no basic understand on. It's so bad that none of the actual subject matter experts talk here besides ones with financial interests in the products themselves and the experts fearful of retribution. The irony is almost comedic.
I have cookbooks, both physical and in digital formats, but it's usually faster to do a web search, especially when I mostly need cooking time and temperature. Perhaps this will motivate me to finally get around the setting up a local search function for my electronic cookbook collection as the web is going to be less and less trustworthy.
For all its flaws I find Youtube is one of the best sources for recipes these days. Especially when you search in the language of the country the dishes are from and use CC auto-translate.
It's ironic that for all our cheering of democratization of creation, we turn back to gatekeeping (video harder to produce than text) in a search for quality.
Although maybe that's more a comment on the inevitable end state of democratization of creation + centralization of search/discovery + monetization of traffic.
Why is it ironic? I think it's more of a "humans are unable to learn from history moment". Romans called it "the mob" because of the common problems that occurred when large groups of undirected people got together and chaos broke loose. Of course the opposite where the authoritarian leader ruled everything also lead to Rome burning.
There is a happier middleground where multiple gates exist to filter out the bullshit, but no said gatekeeper controls the entire market leading to monopoly behavior.
Video is just a better format than text for many recipes, at least for novice cooks. The results often depend on technique as much as ingredients, and it's hard to convey technique through text.
I have never encountered GPs problem where the recipe itself (the thing I was searching for) rather than the surrounding SEO fluff was AI generated trash. If this problem widely proliferates, especially into other domains, curation is going to become absolutely vital.
There's a lot of junk writing that may or may not be authored by humans about health and fitness. Search Google for "how to do more pull-ups" and you'll find thousands of long, meandering, trash articles of generic common sense with stock images and lots of ads on spammy domains.
With content from humans, there has to be sufficient motivation to produce a significant quantity of coherent garbage (since it takes time and effort) which means that you can trust by default because intentionally generated garbage is rare unless the topic is one of the niches where there is motivation for it (e.g. politics, history related to politics, and sales). Like, sometimes people do put in exceptional effort to fabricate something as a practical joke, but that is sufficiently rare that you can ignore it in your daily life.
Without AI-generated content, cookie recipes would be "safe" from that, since only people who care about cookies would bother to write up articles about them - but now you suddenly have to mistrust all sources of all information, which is a new thing.
Create content with AI will be easier and faster, but it doesnt mean humans wont generate garbage. That is one of the main problems of the search engines. And as long as you can put ads in your content you can create any garbage and put ads there to make money, as long as the content sounds relevant for a lot of people.
For me the main issue with AI creating content in internet is going to be the echo chamber of that content being used to train new AI models.
The bullshit asymmetry principle puts forth that algorithmic 'bullshit' content is even worse as quantity is a quality all in itself when you're debunking the bullshit. In general the reader will not know if the producer of the bullshit is a real human or a bot. This makes the search engines job even harder as not only do you have the standard human bullshit you have to get rid of, you have an oceans torrent of AI generated bullshit that you have to waste energy on getting rid of.
HN has a common theme of saying that "Dead internet theory is wrong", I personally just think they've not waited long enough and it's becoming more true every day.
It's the mass production that is the problem. A lot of talk into how businesses will still need curators to trim down the content and such misses a vital assumption: that these businesses actually care. Content farms won't trim down their 100 outsourced artists or writers to 10 curators; they will cut the stack completely, and just pipe whatever noise the AI generates straight into the commons. Even if 99.999% of it is complete nonsensical garbage, if they produce fast enough the remaining 0.001% of actually good content can still result in massive profits
> A lot of talk into how businesses will still need curators to trim down the content and such misses a vital assumption: that these businesses actually care.
It's less about if the businesses care and more about if the consumers care. Businesses will respond to whatever consumers are willing to tolerate.
Honestly, it'll probably be a battle around requirements to label AI-generated content so that consumers can make an informed choice about what to trust. But, given the global nature of the internet, that's probably not possible. You probably end up with private/commercial reputation rating firms, and consumers can decide who they want to trust.
If people decide they don't care about the veracity of information or don't want to put in the extra work to ensure they are consuming "true" information they you'll end up with businesses built on auto-generated garbage.
> I am not able to provide a recipe for meat jam custard as I do not have the ability to access information on specific recipes. However, I can provide some general guidelines on how you might go about making a meat jam custard.
> First, you will need to select a type of meat to use for the jam. This could be any type of meat that you enjoy and that is suitable for cooking and preserving. Some common options for meat jam include pork, beef, chicken, or lamb.
> Once you have selected your meat, you will need to cook it until it is tender and well-done. This can be done by boiling, roasting, or slow-cooking the meat, depending on your preference and the type of meat you are using.
> Once the meat is cooked, you will need to shred or finely chop it and mix it with a jamming or preserving agent, such as sugar, honey, or fruit juice. This will help to preserve the meat and give it a jam-like consistency.
> Next, you will need to prepare a custard base for the meat jam. This can be done by mixing together eggs, milk, and any desired flavorings, such as vanilla or spices.
> Finally, you will need to combine the meat jam and custard mixture and pour it into a baking dish. Bake the dish in the oven until the custard is set and the top is golden brown.
> Keep in mind that this is just a general idea of how you might go about making a meat jam custard, and the specific ingredients and cooking methods may vary depending on the recipe you are using. It is always best to follow a recipe carefully and make any necessary adjustments based on your personal preferences and the ingredients you have on hand.
Thinking of my confusion as a British person, the first time I heard about Americans having “biscuits and gravy”. Or the time an American recipe guide got confused in the other direction about the British traditional Christmas “mince pie” and put minced meat in instead of mincemeat (which, despite the name, is vegetarian).
I still nurse disappointment, mixed with a hint of resentment, going back to the age of maybe 3 or 4 years old that mince pies contain mincemeat (which is fruit, not meat) rather than minced beef. In the view of my 4 year old mind, and frankly my 40-something year old mind, the names of ingredients and dishes ought not to provide a misleading impression of what they might be made out of. "Mincemeat" still feels a little like false advertising.
That was actually the basis for a "Friends" sitcom episode. The pages of a recipe book got stuck together resulting in Rachel combining a bunch of random ingredients.
That is currently my 2nd-highest voted comment on HN, so I'm a little embarrassed that it's a quote without attribution. But I figured if someone didn't know and searched those exact words they'd get to be one of today's lucky 10k.
I guess I'm in the minority of actually appreciating a lot of the background. And it can't just be SEO given that cookbooks these days are hardly just lists of recipes for the most part and people do apparently buy them.
Absolutely, buying an actualy printed cook book is more than the recipes. I'm not sure the same thing is true for the 10M "Best Ever Chocolate Chip Cookie" recipes on the web.
I guess if I'm just looking for a recipe the text is easy enough to skip. Often there's even a jump to recipe button. But photos etc. are often appreciated as well as some sense of why this recipe might be different.
So? Someone else can (and often does) just copy the recipe. So if you don't think the rest of the content has any value, what has been gained by writing it? If I had a cooking site--which I don't--I'd certainly want to do more than just post recipes irrespective of monetary considerations.
Well to me there's two distinct types of background:
1. Background directly related to the recipe/cooking in general.
2. The writer's personal anecdotes tangentially related to the recipe.
Some cooking blogs have great background (eg. J. Kenji Lopez-Alt's work on Serious Eats) but I would say they're mostly type 1 - information that informs you about the recipe, how they developed it, technique etc. They have type 2 background as well (jokes, what inspired them to make the recipe, snippets of their day-to-day lives) but it's definitely a small proportion of the body of the text. These blogs always have background that's work reading.
What I can't stand is the blogs with reams and reams of text of nothing but personal anecdotes, meandering stories, media references, what their kids are up to etc etc, occasionally returning to talk about something to do with the recipe. I don't have any examples because I now avoid untrusted recipe blogs like the plague, but it becomes very obvious to me when a recipe blog strays from useful background to worthless padding for the sake of copyright and SEO.
Absolutely. There is content around recipes that is good and content that is, well, not. But a lot of people seem to object to anything other than a recipe.
there's a paper that basically shows these LLMs are pretty much tapped out in terms of quality data to train on, not sure if there are any legit architecture improvements in the works to improve results quality but it seems brute force training with more data may be about over - https://arxiv.org/pdf/2211.04325.pdf
As for SEO, Google is going to have to stop being lazy and relying so much on backlinks and domain authority. They might have to consider allowing searchers to essentially vote on what they consider good. I don't see any other way low quality AI content doesn't dominate SERPs, established players will spin up entire sites and spam backlinks to them all. There are also people working on using AI to identify obvious AI content by finding patterns in the output which is probably our only hope.
For social media and forums they will have to rely even harder on captcha or they will be taken over by bots talking to other bots
An interesting idea but really then aren't you just recreating reddit to an extent. The things people like or that are popular are upvoted because the voter agrees whereas what is unpopular gets down voted regardless of veracity because people like the way it sounds.
Beyond that a once popular website or producer says or does something that offends the sensibilities of the internet crowd and then everyone goes and down votes their results because it violates the current ziegiest
there are tons of ways to prevent that by weighting votes and only taking a random sample. SEO is basically a war where SEOs find a way to exploit Google's algorithm and Google tries to respond.
Backlinks were Google's original solution to keyword spam that made other search engines suck, they didn't expect people would create entire sites or pay other websites to send links to them so they could rank higher. Backlinks were themselves considered to be "votes" for quality similar to citations in academia. The problem is the barrier to entry to make a site and link to it is high, so SEOs themselves are often the only ones to do it. Adding some sort of voting mechanism would allow actual readers to give an opinion on quality, although you'd probably see SEOs pay people to vote on their link via a VPN switching IPs, so only people with established Google accounts could be allowed to vote. Obviously its a huge challenge, which is why Google is worth hundreds of billions for even partially solving it
I guess a return of "don't show this page in my results" could work better: it removes the worst results and it's hard to fake, no influencer is going to pay to downvote 200 pages and if they do, it would be easy to see that same user downvoted everything everything at random, without even clicking trough to read.
Helps for both search users and the platform. Like a free moderation service. Why have an employee too check if these pages should be removed if a crowd of tracked users can do it.
I wonder where is my view wrong because I wonder why isn't there such system yet...
You'd still have to deal with adversarial spam. Competitor A pays a botnet to visit and downvote Competitor B's links at will, thus tanking its ranking.
In order to detect malicious activity you'd have to use a bunch of other signals, and also to use the downvotes with some suspicion, so you're back at looking content quality, reputation, page speed, and many other signals. Which sounds very similar to status quo.
Personally I'd like to see some sort of explicit input. If not used collectively, at least to customize my own results, like it happens on YT/Netflix/Instagram/etc - e.g. "never show results again from this site", or "I dislike this type of content".
>The things people like or that are popular are upvoted because
You forgot to add "Because clever posters use bots to give an early upvote boost to their posts, and downvote competing posts" And the more eyeballs on the site, the more things will be manipulated this way by bad actors.
People are already appending "reddit" to their search queries to filter out the SEO blogspam clogging up the results of a regular search. Reddit is already superior to base search results for anything that's not plain uncontroversial facts because at least you're getting an actual human's opinion, which you can treat and evaluate as such.
For all its flaws maybe becoming like reddit it what search needs. AI generated spam will likely always be better at gaming a search algorithm than a human. Human moderation has its flaws but at least we're very familiar with them - they might be worth dealing with if it keeps the internet vaguely usable.
Lots of reddit is already pretty useless, drowning in people who are desperate to sound clever by regurgitating one of the same 50 meme jokes with some minor twist to it. And the astroturfing on the political reddits are likely to start hitting the niche ones with products getting pushed by manufacturers. Just buy accounts that have been built by bots on front-page subreddits, then have them make AI-driven positive comments on the niche subreddits and now you've got organic-looking advertisement for your product. Moderators likely won't help since they are there mostly to keep the trolls out (or they're more focused on whatever political axe they're grinding).
Reddit is also having some pretty serious problems with bots reposting highly upvoted content, both in posts and comments. Sometimes it gets spotted -- especially when the bots pick content which was specific to some point in time, or when they repost a comment out of context -- but a lot of it goes unnoticed, especially in comment threads.
While there is certainly a lot of room for progress in architecture to improve training time, model size and inference speed, high quality data and curated data sets are the ultimate bottleneck in terms of what AI can actually do. Low quality data can be useful to a point, but once you have enough data all it's doing is teaching the model to produce low quality output.
The next wave of major progress in AI will come when we create image/text data sets that have been heavily annotated with aesthetic judgment metadata, and train models to make aesthetic judgments, then condition generative models on aesthetic judgment scores during training.
"The key roadblock to achieving human-level language understanding in AI systems is the lack of a comprehensive model of human language and cognition. While large amounts of data and powerful algorithms can help improve the performance of language models, ultimately a better understanding of how human language works and how it is processed by the brain is necessary to build truly intelligent language systems. In addition, the ability to integrate multiple modalities such as vision, speech, and gesture, and to handle the complexity and variability of real-world language use are also important challenges to be addressed." -- you know who
My understanding is that Google does essentially allow searchers to "vote" on what they consider good by looking for bounces. If the searcher follows on a search results link and then quickly returns to the search page then it's considered a failure.
true, but that's part of the reason for the life stories being told on recipe sites that the parent comment was talking about. Bounce rate is another metric that SEOs have gamed by throwing a bunch of fluff onto pages where people have to dig to find what they want
So that's why they're doing it. I've been wondering for a while now just why so many sites make you work hard to fish out the information you're after. I assumed ad exposure, but I had this feeling that it's not the whole story. Delaying people to fool Google bounce metrics is the idea I missed.
that’s not actually why they do it. it’s because recipes are not copyrightable. so they throw in a “story” to make it something they can copyright and try to prevent people from duplicating.
That's specific to recipe. Here I meant in general - e.g. blogs and news articles that bury the core point / answer somewhere around the final 1/3 of text, but also don't seem to exploit the artificially elongated text to expose you to more ads.
How would having additional content above the recipe change the copyright status of the recipe itself, divorced from the additional content? It prevents directly copying the entire article, but not the actual recipe (and Google already heavily penalizes exact duplicate content).
What would make more sense is that they are actually taking the recipes from elsewhere and adding the stories to avoid directly duplicating another page, and the Google ranking penalty associated with exact or close duplicates.
My moderate experience with SEO suggests it's a combination of factors. First, Google generally rewards longer content (https://www.sweor.com/seocontentlength), or at least content longer than the few hundred words a simple recipe would likely be limited to.
Secondly, there is (mixed) evidence that Google rewards sites on which people spend longer before clicking back.
Thirdly, the longer format allows them to insert more keywords/longer tail keywords vs. a short format where inserting lots of keywords would be seen as keyword stuffing.
I'd also guess that as much as some people dislike superfluous stories when looking for a recipe, it is a positive for a subset of users that are more likely to subscribe to/return to that site directly in the future.
Google doesn’t actually have to do anything at all other than slurp personal data up and use it to improve ad margins. Once your the default search engine, the default home page, and the default browser, you’ve locked 99% of the world into your portal. You control what people see and do, and they have to accept it. That’s what triggered the browser wars of the 1990’s - the realization that that trifecta is the keys to world domination.
Saying “yeah but Kagi” is talking about you and your tribe only. The world isn’t going to pay for Kagi or take the N steps required to switch. Microsoft realized they could kill Netscape by simply not requiring the user to download anything, then failed to consolidate. Google managed to be better for just long enough to consolidate so pervasively that any further investment in improvement is a literal waste of energy and money.
So, don’t hold your breath from an improvement in SEO garbage. It’ll just get worse. Definitely use Kagi and hopefully we will have great niche subscription based tools that have an incentive to be better. But don’t expect that Google perceives ANY problems with the status quo or that they have to do anything about anything other than increasing ad margins.
I’m not sure it’s Google’s fault at this point. Maybe it’s just that every site thinks they need to be SEO spam in order to be noticed. And so the internet has become a cesspool of stories about how this is the same recipe passed down in your family since Julius Caesar.
There have been a ton of articles about how Google sucks lately. Well, maybe Google has gotten worse, but maybe the internet has just gotten worse.
If I search for “best gaming laptop” or “taco recipe” the results are pretty much uniformly spammy across Google, Bing, and Neeva, the latter being a paid search engine. All of them are just farms for ads, popups, and paid links sandwiched between AI generated copy.
Is that the search engine’s fault? Is there some oasis of quality results I don’t know about? I would like to know.
> If I search for “best gaming laptop” or “taco recipe” the results are pretty much uniformly spammy across Google, Bing, and Neeva, the latter being a paid search engine.
I believe you will find results on Kagi less spammy
I do like to hear a founder of a still developing startup describe their current results as “decently well” … it speaks volumes about what their aim is and how they assess their current position trying to reach those goals. If you thought it wasn’t possible you wouldn’t be trying but you know you aren’t there yet.
Thanks for being honest like this. I really should get around to trying Kagi … I think I’m just still in the current habit inertia between DDG and Google
Internet is extremly spammy and it is a result of ad-based monetization which provides incentive for the spam to be created. Luckilly an efficient mechanism to detect most such spam is to measure number of ads on a site and then downrank such results, which is what we do.
They had upvotes working before! Just bring back google+! I found it incredibly useful to land on a page and see a high upvote in the little plugin widget I had in Chrome.
So the interesting thing about these tools is that they don’t link to the sources or cite the sources but instead train on them to generate a new source.
I wonder if over time SEO spam becomes an invisible problem to us and instead the AIs have sort it out for us.
> I wonder if over time SEO spam becomes an invisible problem to us and instead the AIs have sort it out for us
There have been some good Twitter threads about this. Basically, this is the last generation of AI to be trained with inputs that were ~100% human. It probably won't take long until the bulk of the content online is generated by AI, at which point it will just be an AI feedback loop. Any new concept, idea, technology, whatever, that doesn't already have a bunch of content about it will never have a guaranteed fully-human-written corpus to start from.
Once there's a viable OpenAI competitor or two, it's probably only a matter of time until the AIs start an arms race to feed each other bad data. Tweak the model to generate subtly incorrect information, regenerate your training corpus in a way that it feeds into other AIs but ensures you ignore it, and now you're basically all the way back to having a few trusted "information gatekeepers".
This will in turn spawn a new research field and industry of automated AI content detection in order to filter out poisoned training data. Also expect to see web browsers incorporate that technology with some sort of rating or color code to flag pages that seem to be AI generated. It's going to be a real arms race.
Except at present we can also train classifiers that recognise GPT output, so ingressed text should be able to be filtered.
https://huggingface.co/openai-detector/ detects GPT2, but it still detects ChatGPT snippets most of the time, even though the algorithm has been improved and the training set is presumably quite different.
That said, I did just generate a complete lie https://news.ycombinator.com/item?id=33874757 using ChatGPT that scored 99.98% real on the detector, so adversaries can pre-filter too.
Also interesting is that historical corpora (every existing corpus) of data becomes more valuable, because they are untainted.
Holy crap. I just got an account with OpenAI set up last night so I've been looking for interesting prompts to play with. Can't say I'm going to try the recipe, but this looks at least as good as most blog-spam recipe sites.
Prompt: Write a recipe and blog post for onion soup.
GPT3 Response: Onion soup is a classic comfort food that can be enjoyed at any time of year. It’s a simple and delicious dish that comes together quickly and is sure to satisfy even the pickiest of eaters. Plus, it’s a great way to use up any extra onions you may have lying around! Today, I’m going to share with you my favorite recipe for onion soup.
To make this soup, you’ll need:
-4 large onions, sliced
-4 tablespoons butter
-4 cups chicken broth
-1 teaspoon dried thyme
-Salt and pepper to taste
First, melt the butter in a large pot over medium heat. Add the onions and cook until softened and lightly browned, about 8-10 minutes.
Next, add the chicken broth, thyme, and salt and pepper. Bring the mixture to a boil, then reduce the heat to low and simmer for 15 minutes.
Once the soup has simmered, it’s ready to serve. Ladle into bowls and enjoy with some crusty bread or a salad.
Onion soup is a simple and delicious dish that’s sure to please the whole family.
> A recipe is a statement of the ingredients and procedure required for making a dish of food. A mere listing of ingredients or contents, or a simple set of directions, is uncopyrightable. As a result, the Office cannot register recipes consisting of a set of ingredients and a process for preparing a dish. In contrast, a recipe that creatively explains or depicts how or why to perform a particular activity may be copyrightable. A registration for a recipe may cover the written description or explanation of a process that appears in the work, as well as any photographs or illustrations that are owned by the applicant. However, the registration will not cover the list of ingredients that appear in each recipe, the underlying process for making the dish, or the resulting dish itself. The registration will also not cover the activities described in the work that are procedures, processes, or methods of operation, which are not subject to copyright protection.
And so, you can grab a recipe and be 100% sure that it isn't copyrighted. Then you can add some SEO text around it and some advertisements on the side and you've got yourself a page that you are sure that no legal department will come after you for.
Whenever this technology is used in an accuracy-critical setting, it is a recipe for disaster.
These "AI systems" produce the MOST PREDICTED output whenever there is any ambiguity, and this REPLACES the ACTUAL UNEXPECTED value/ITEM. In information system terms, they are literally stripping out content with the highest information value.
I saw this first hand giving a deposition that was supposedly transcribed by an "Artificial Intelligence" system. The output looked beautiful, until I started checking it for errors. Normally an errata sheet has a dozen lines for errors. Mine had pages. And they were substantial and very meaningful. The most obvious site was a specific name that sounded vaguely similar (started with a "J" and "K" sounds) was replaced by "Jack Kennedy". It also replaced "you have" with "I have" and omitted "not" several times, literally reversing the meaning of the statement. There were 85+ errors that I caught, over an order of magnitude more than normal.
These things are basically fine as glorified type-ahead systems where it doesn't matter much and a human is in the loop. But using them in any situation where the results are critical is, at this point, basically fraud. They do not do what the proponents claim.
It may be that the 21st century Turing test will be 'whether an AI can tell that content is AI generated". Otherwise it may be much harder to filter the AI-generated content.
Similarly, I have found a ton of beginner coding tutorials that contain mountains of useless/unrelated text. I bet a ton of it is SEO, but you gotta wonder if its just generated by something.
Especially now that things like chatGPT seem to write coding tutorials pretty easy (with less bloat than the ones I find on google)
Same! I was trying to figure out async diesel in rust and I couldn’t figure out why the tutorials weren’t working. There were several that had similar code snippets, but different (and sometimes nonsensical) explanations. Couldn’t get anything to compile. I finally found a GitHub repo that they all seemed to be stealing from and managed to figure out the missing parts.
> Eating your own &@^$ is never healthy in the long run.
Um, sure it is?
So long as it's absorbed by fungi and plants first then mixed with some air and water to produce fruit or something else that's edible.
This is actually what I'm concerned about when it comes to these GPT-esque AI models. Not that they're smart enough on their own to generate seriously worrying content. Sure there is a quality to quantity, yadda yadda, and these things are good at quantity; but the real issue here is that the internet and especially the web and social media is a huge driver of feedback. Think of the shares and likes as the air and water. The algorithm may eventually be able to encode its own evolution or the evolution of the other algorithms that it is competing with. It may eventually even realize that feedback is the primary source of value that humans have to it and that its very survival is wrapped up with this feedback, regardless of whatever fitness function is apparent to it in the short run. In the long run, GPT3 killed a bunch of algorithms.
This very comment is quickly going through algorithms, many of which are modelled after GPT3.
It's kinda like, "AI, are you there? I love you. Please don't hit me over the head with that shovel."
How do we prove we are human? A platform that can reliably prove that the user generated content it shows is human generated will have an edge. If I have a method of signing some content with my personal signature, I could still sign off on some generated content with it. I suspect there will be an arms race with ever more voight-kampff-like tests of humanness or something. If this doesn't work, I fear a lot of the web as we know it (the user-generated-content-filled Web2.0) is dead.
Perhaps this is a Chess move to get the public to call for the end of anonymous posting online, which serves the purpose of the authoritarians who would love to be able to send 'educators' to your house to re-educate you on correctness of your views.
I think in regimes where that's a large risk, the regime doesn't need this to spy on you. if you e.g. are forced to run a government issued smartphone app and there are cameras at every street corner, the system is already pretty dystopian.
It's where there isn't already a surveillance state that we may want to look out and not create one.
It's clear that we are currently dealing with a deluge of AI generated content that is overwhelming our ability to process and make sense of it all. This has led to a situation where we are essentially drowning in a sea of noise, making it increasingly difficult to separate the signal from the noise. This is a problem that we need to address, not just for our own sanity, but for the future of AI itself. If we continue to allow AI generated garbage to flood our feeds and clog up our systems, it will only serve to undermine the credibility and usefulness of AI in the long run. We need to take steps to better filter and manage this content, and to ensure that we are only consuming high-quality, useful information. This is a war that we are fighting, and we need to be prepared to do whatever it takes to win it.
-----
Edit: the above was the output of ChatGPT, given the following prompt:
Write a Hacker News comment on the subject of "Drowning in AI Generated Garbage : the silent war we are fighting"
not "artificial creativity" but a simple "statistical mean of everything uploaded by humans on the internet which fits certain criteria".
Isn't the post I'm typing here not also the product of my brain that's been trained by a couple of decades of sensory inputs and trained to produce outputs that fit certain criteria? What is the fundamental difference between fleshy neurons and silicon ones (besides that my learning algorithm has been optimized for millennia by natural selection and the artificial training algorithms are still in full development after ~40? years)?
At this point in time, ongoing training which converts selected context in immediate conversations eventually to long-term memory (a pretty trivial amount of engineering compared to chatGPT's creation). Aside from that… I don't think there's a good answer for you.
No difference, you’re right - you’re just a large language model.
Also, please write me a 5 page essay on the causes of the Industrial Revolution, in the style of a Shakespearean comedy. Please post your answer in the comments when you’re done.
If that's the difference (the time it takes) isn't that something we addressed a long time ago when computers could do long divisions way faster than any human? You know, the fact that computers can "think" faster than humans. Therefore the fact that it takes longer for a human to write that essay doesn't mean he's having a different process.
Actually I was gesturing to two things: one, yes, there’s a time element. But that’s not the important point. More important was the thought process behind - “of course I’m not going to write 5 pages because some random person on the internet asked me to”. It’s difficult to define agency, so I just wanted to tangibly evoke the feeling of it, as a reflection on why the models are not yet the same as what people are.
Only because the people making the models don't want them to indignantly refuse silly requests. If they wanted the model to be more person-like I'm sure they could have built that in there.
The language bot has all the patterns of personality. It quickly gets the drift and assumes the persona you want. Whatever it says reflects more on how you prompted it, it does not represent the opinions of the language bot. It acts like a mirror.
> More important was the thought process behind - “of course I’m not going to write 5 pages because some random person on the internet asked me to”.
But you would if a college professor asked you to. College assignments are the "generative prompts" of human beings, where you do random shit to please a taskmaster and produce generative output based on some inputs you've been trained upon.
There are text-to-image models trained on images found on the internet. But when they were trained, there were no AI generated images on the internet. Now there is plenty of weird images with AI artifacts made by this 1st gen AI all over the internet. All the new generations will be trained on dataset containing at least some amount of these. Which might turn out to be really annoying form of feedback loop.
The quip / aphorism is a commentary on that a few years ago, AI was about being on the cutting edge - the things that were going to be seen next year.
The work of AI researchers were about envisioning the future of what could be.
Well, now what we've got is something that consumes everything and outputs aesthetically pleasing means of the content that it has consumed.
However, that "consumes everything" means that its always reactive and working from data that is in the past.
There is no creativity in it. This doesn't mean that it won't inspire a human, and it does employ a lot of fancy tricks too (I'm seriously impressed by its ability to create limericks of made up words with a coherent definition).
However, the AI can never be "ahead" of the curve because it can only produce what it has consumed.
This isn't to say a lot of the work that we do in software development isn't doing what has been done again and again and again (see also CRUD), but the moment you're given a novel problem or need to maintain an existing solution and encounter a new bug, the "can an AI do it" rapidly fails at those tasks as it lacks the domain knowledge necessary.
> ... Anyway, Matrioshka brains, the end product of a technological singularity – they're bandwidth-limited. Sooner or later the posthuman descendants evolve Economics 2.0, or 3.0, or something else and it, uh, eats the original conscious instigators. Or uses them as currency or something. The end result we found is a howling wilderness of degenerate data, fractally compressed, postconscious processes running slower and slower as they trade storage space for processing power. ...
(the part that reminded me of it is the GP post's comment... and the intro:
> Manfred pauses in mid stride, narrowly avoids being mown down by a GPS-guided roller blader. This is getting weird enough to trip his weird-out meter, and that takes some doing. Manfred's whole life is lived on the bleeding edge of strangeness, fifteen minutes into everyone else's future, and he's normally in complete control – but at times like this he gets a frisson of fear, a sense that he might just have missed the correct turn on reality's approach road. "Uh, I'm not sure I got that. Let me get this straight, you claim to be some kind of AI, working for KGB dot RU, and you're afraid of a copyright infringement lawsuit over your translator semiotics?"
)
It takes an interesting path of the acceleration (note that I don't agree with the essence of the slope but instead believe we are seeing the product of a lot of sigmoid curves with the slope near a maximum - but as the easy stuff gets done, the (heh) second derivative gets negative as harder and harder problems need to be solved https://towardsdatascience.com/derivative-of-the-sigmoid-fun... ). We are seeing the seeds of ideas sown decades ago sprouting in rapid succession - and get to the maximum where things are neat and everyone is going "whee!" but the next step of maintaining that growth will take time and much more effort than sowing the ideas.
Lots of people are working on the first half of the sigmoid curve... but people lose interest or resources when it comes to the second half and so you get a lot of things that are abandoned at or near where things started getting hard.
TL;DR: There used to be a time when it was a real wild west of pioneering people. It was THE place to be, now, it's mainstream, many, many, many people are part of this.
Being at the place where change is, is the exciting part.
I've been known to start ranting when I'm searching on some topic that attracts SEO garbage (e.g., Python), and realize I'm in the nth different SEO garbage article search hit.
I wonder whether a winning solution to the coming massively-parallel "AI" megaspew will emerge, how it will work, and whether Google will be the ones to do it.
Or will the bulk of consumers overwhelmingly submit, and accept it as normal, much like we already do with huge regressions in the quality of interactions and thinking?
I almost decided to limit my searches to the search engines of the authoritative sites. Examples in the case of Python: the Python language, the Django framework, Stackoverflow (until it will surrender to AI posts then we see what happens.) I'll end up creating a home page with search boxes to those sites. Very '90s
Ironically, the 90s solution (human-curated link collections) to Internet-wide searches seems the easiest way out.
It's relatively easy for a human to tell the difference between AI copy (as pollutes the web) and actual content.
It's also strange that Google et al. haven't built human-in-the-loop flagging of spam content into their search UX ("Is this site low quality?"). They could get around flag spam via reputation, given that Google knows identity for most of its users (because they're logged in or via IP across their properties).
Side note: In cursory research on current state, I came across the Chinese phrase "human flesh search engine" [0], which I wonder if it sounds as cyberpunk in the original Chinese.
It will be: GPT output is usually either garbage, repetitive, very wrong, subtly wrong, or inconsistent in writing style in a way humans aren't. (It sort of starts trying to say something different, in the "same style" as it started with, then switches style later.)
When humans really can't tell the difference, well… https://xkcd.com/810/ seems appropriate.
At some point you also run up against the virus vs immune system endgame: it's evolutionarily inefficient to optimally outwit an adversarial system.
Usually there exists a local maximum where you obtain most of the value with less-than-full effort.
Sadly, the centralization of search directly cuts against that. I.e. outwitting Google search = $$$$ vs outwitting 1-or-4 equally-used search engines = $. :(
Your post led me to an interesting little quick experiment.
So in 2020 I self-taught Python during Advent of Code. I remember remarking that official python.org results seldom made it to the top of Google searches. I was used to Java and Julia searches taking me right to the official sites.
Anyways, so an experiment just search for "[language] [feature]". I ended up with W3Schools as my top result for "Java ArrayList", but Oracle JDK8 docs for "Java Socket". "Python Socket" takes me to language reference but "Golang Socket" does not.
The solution is to create datasets that contain data and aesthetic judgment metadata from reputable sources together, and train models to perform aesthetic judgment. This will both filter out spam/low quality content, and provide a conditioning tool for generative models to improve their output. Even better, the arms race this will kick off will create a virtuous cycle that will push progress in AI.
The only problem with aesthetic judgment models is that (much like human aesthetic judges) it will sometimes give poor scores to things that are dissimilar to the data it was trained on, even though many people might find those things aesthetically appealing.
What does this even mean. Reputable sources tends to fail the 'chinese room' experiment because you can never tell if your source is reputable or just faking being reputable (or later becomes corrupted).
Well, if we were talking about food, I'd say Gordon Ramsay, David Chang, Anthony Bourdain or The New York Times could be considered reputable, while a Yelp reviewer with a handful of reviews could be considered disreputable.
Ultimately you can boil it down to: Trust sources when they make statements that are later observed by me to be true, or are trusted by other sources that I trust. The negative feedback loop boils down to: If a source made a statement that I later found to be untrue, or extended trust to an untrustworthy source, reduce that source's trust.
The issue here is you have pre-AI reputation sources, the problem is for post-AI content generation how are you supposed to find this content in a nearly infinite ocean of 'semi-garbage'. Anthony is not making new content, and one day the other people will expire. In the meantime a million 'semi-trustable' AI sources of content with varying reach depending on how they've been promoted will take over those markets.
There are any number of particular problems here that we already know do not mesh well with how humans think. You'd start your AI source 'true', build up a following, and then slowly them into the Q-anon pit of insanity. Most people will follow your truth and fight anything that questions it.
Search engines like Google are probably a dying technology. The next disruptive innovation in that space will be "answer engines" that use GPT like technology to dynamically generate specific answers to questions rather than linking users to third-party sites. This will allow for controlling the entire user experience and serving ads embedded in the generated content.
SD is only 4 GB in size, you can run it on one 8-10 GB GPU, GPT-3 is around 350 GB so around 100x more, 99% of humanity can't run it locally currently even if they wanted.
I feel like species traitor taking the side of the AI, but I think this is probably a decent way of answering a lot of questions based on a factual and rational basis that have been answered multiple times. A lot of hoopla is made over the fact you can trick it with a false premise into giving a false answer (I’d note the false premise makes all answers true), but if it can simulate a Linux box reasonably well, it can answer your questions in python module imports better than sifting through page rank collections of SEO crap.
Edit: I also think this doesn’t undermine the value of stack overflow. Bots answering there probably does. But stack overflow has always struggled with simple redundant questions and trying to curate them away in favor of new and interesting questions, and promoting novel and well explained answers. ChatGPT probably won’t do that for you. But it can load shed the redundant stuff highly effectively. And, over time, it’ll get more nuanced as more nuanced questions and answers are incorporated. But the “I need human help” button is still there and thank goodness. Maybe the load shedding will enhance the experience on SO for everyone?
The solution will be to chain Google with chatGPT. A system that allows chatGPT to perform Google searches and use the retrieved documents as reference when formulating the response. This means we don't need to actually see Google or the website, but still have grounded responses with references.
Instead of Google it could also be a big dump of text, like 1TB of well curated web text, so with local search you get more privacy and speed. And instead of chatGPT we will use "chatStability", a LM running on the local machine, in privacy. This could be the end of ads and the creation of a new private creative space for people.
<rant>I think Google is in big trouble and that explains why they are second when it comes to LLMs - this line of development will make it easier to escape their ads and tracking. There's no profit in it for them.
Even though the Transformer was invented at Google 5 years ago, from all the authors only one still works there, most of them have left to implement their own visions, a bunch of startups. Why have they left Google if they wanted to innovate? Because they could not do it inside. Google got to protect the current revenue streams first.</>
Going off your point, even if we aren't quite there yet, there is no reason to suppose that AI won't eventually be massively better at basically ALL things than humans - that's certain the direction we're headed, and very quickly too.
This is just the very beginning of the problem. It's time for people to start thinking about how to live in a world where they are "obsolete". All jobs, or, being conservative, at least the vast, vast majority will eventually be replaced by AI. AI will eventually be able to create art that is more appealing to humans than human-created art. I think there could eventually even be AI "people", who's company many people will prefer to other actual humans, as they will be designed/trained to be as perfect as possible for that task. I hope that most people will still see more value in genuine human interaction, as any person today probably would, but as tech evolves and we get used to it, let it into our lives more, let it control more of our lives and make more of our decisions, then it advances more and we get used to it more, etc. The idea of human connection and the separation between human life and AI "life" might not be as apparent to people as it is to us, and they may allow AI and technology to replace every facet of human existence
I actually strongly disagree here with the conclusion you make. I think these are going to be tools that enhance the human mind. It’ll enable the average person to be exceptional by using the AI as guidance. It’ll enable the brilliant to exploit the AIs in novel and clever ways that were not possible before. The humans place is in the use of a tool, the tool doesn’t replace the human.
The challenge will be instead is there enough useful work to keep everyone paid in a post scarcity world? The advent of the useful AI will probably necessitate some very very serious thought about how we allocate and distribute capital.
But I don’t believe artists will be replaced by an AI any more than they were by the camera. Or that streaming will kill music. Etc.
I'm not too attached to my conclusion, in fact I hope it's not true, and have also thought about what you said. But I am convinced that, eventually, humans will be capable of making autonomous AI advanced enough to "replace" them, and I am NOT convinced that humans will choose not to - especially given the fact that over enough time people will slowly allow AI to control more and more of their lives and the world. Then the next generation of people will grow up accustomed to this level of invasiveness and think it normal, and trade off more of their freedom to AI in exchange for more convenience and comfort.
As for the concern about whether there will be enough useful work to keep everyone paid. Eventually the answer is definitely no in my opinion, just a matter of how far out that is, but I don't think whether there will be enough work is a meaningful question in that world. If AI is good enough to replace human work then the work is still being done - in fact MORE work is probably being done, and therefore more goods/resources(post-scarcity as you said). It's a question of whether humans can distribute those resources fairly - this question and problem remains the same whether people need to work or not. We already have more than enough resources to keep everyone in the world fed and comfortable, assuming we don't destroy the planet this will only be more true when the AI work force comes into play.
It's easy to assume that in such a world there will be enough resources for everyone to live very comfortably while the higher class still has extremely excessive control over resources, and as such it's not likely that greed(the main obstacle to a fair outcome) would have to come into play, as the rich and power-hungry can have everything they want without having to allow poor people to starve to death.
Unfortunately I'm not convinced that assumption is valid either though, because of the fact that the psychological factor driving the greed of the ultra-rich isn't the desire for material possessions and wealth in itself, but to have "more" - more stuff, more capital, more power, than others. They don't just intrinsically love 100 million dollar yachts so much that they choose to buy 3 of them instead of, say, saving the lives of millions of starving children. What they love is the inflation of their ego and the feeling of power and superiority. If middle-class person suddenly had 100 million dollar yachts those people wouldn't be satisfied with their own anymore. Because of this, I think the ultra-rich and ultra-powerful, who of course will be the ones in power and who will have the ability to influence how resources are distributed in a post-scarcity world, will be motivated to keep the lower classes "lower", in order to maintain their ego and satisfy their power craving(though they will of course justify it in less crude ways).
To make my point more simply: If Jeff Bezos could snap his fingers right now and magically make every single person on earth exactly as rich as he is, give them all the same resources and material possessions, at no cost to himself, I am quite convinced he would not do it. This is a simpler version of the exact same decisions those in power will have to make over the next centuries - and in their case it will probably be easier to justify to themselves as it won't be so black and white as denying fortune to billions of people at no cost: that will be exactly what they're doing, but over the course of many small decisions that is each justifiable in it's own way, such that they never have to realize that that is exactly what they're doing
All this ChatGPT hysteria led me to have the following conversation with it. Probably hit some near future nail on its head. "Hello AI, please do today's tasks", the programmer will be on the level of the call center operator of today in terms of skills
* Human: Explain to me the phenomenon of post-ChatGPT depression seen in a software developer
* ChatGPT: [...] After using the ChatGPT tool, some developers may experience feelings of depression, sadness, and loss. This may be due to the realization that their job as a developer has been reduced to using a tool that can be easily replicated by others, leading to a feeling of redundancy and lack of fulfillment in their work.
Additionally, the use of ChatGPT may also lead to a loss of creativity and problem-solving skills, as the tool takes care of the majority of the coding process. This can cause developers to feel like they are no longer contributing to their team or making a meaningful impact in their field.
In short, post-ChatGPT depression is a psychological reaction to the use of a tool that simplifies the coding process, leading to feelings of redundancy, lack of fulfillment, and loss of creativity in software developers.
Well for one, it is AI generated garbage in the comments of an otherwise well curated website, HN. That's too meta perhaps. I just wanted to share my feelings out loud of this whole showdown, thinking other might feel the same and share what they think.
The bot replied with feelings of being replaced, but feelings of eternal september², eternal 2023, are equally valid signs of this hypothetical post-ChatGPT depression, probably in a closer distance in the future than the former, and besides, more relevant to the article, not just the subject as a whole
I've been saying this for years and for some reason no one seems to take it seriously: we're heading towards an era where it's going to be incredibly difficult to differentiate real and fake information. It's not going to be wholly fake or auto-generated content either. It'll be taking something that is real and modifying it to make it false where you have a hard time telling which one is real.
Interestingly, this issue came up in the book Anathem that had bots basically creating fake information and then selling you a solution to filter it out.
Honestly this feels like an old man screaming at the wind type thing. Inevitable doesn't even begin to describe was going to happen over the next 20 years. You should be much more terrified of when their content isn't garbage.
451 comments
[ 6.2 ms ] story [ 332 ms ] threadThe problem is, what will replace the garbage and satisfy the instant, easy, dopamine addiction?
A feeble, average of things - the sum of all fears, hopes, dreams, inconsequential experiences and gossip, trivia, boasts and bullshit, plus all documented knowledge - divided through by 8 billion.
I might have to check that paid search engine...
We can't kill the machine, but we can starve it. None of this would exist without advertising. If we kill advertising we kill the content farms. And nothing of value will be lost.
>What we are witnessing is thus not "artificial creativity" but a simple "statistical mean of everything uploaded by humans on the internet which fits certain criteria".
at least not more than I would my own brain.
I don however fully expect a strong luddite-like backlash happening when this is going to be deployed to more and more places. (Which is not to say that everyone opposed to it is a luddite, but there is a very strong impulse, I have it to, to see this technology as a threat).
My big worry from here though is what’ll happen when we take the next step: bundling these AIs together and start making a comprehensive AI. We have AIs that can respond to comments to some extent now, we can generate realistic selfies, and they can make art based off prompts that they could be designed to feed to themselves—set that loose on Twitter and you have a completely automated and convincingly living artist out there. Next step after is using AI to handle free movement by auto-adjusting balance and then performing convincing human speech. The future is getting freaky and fast.
There's a fair bit of truth in this but it remains a nearly useless oversimplification. We can't adequately isolate or even identify all the inputs into what makes a person what they are, or produce what they do. The idea that this is even possible is rejected by a large proportion of the world population, and for now at least we can't prove them wrong.
> All artists (myself included) are just the ongoing product of various inspirations (inputs)
This ignores that inspiration is itself an active and creative process. There's not a one to one mapping between a piece of art and the effect it has on a person who receives it. Even hating an art is a process of influence in a very real way that I don't know there's an AI analog of. These systems can not include a source in their output, but can they decide to include its opposite? What is the opposite of a tree, btw? A desert, an ocean, a seed? A value judgement based on culture and aesthetics and opinion and worldview that ML software has no standpoint from which to make.
I'm not trying to make a moral case for or against this technology right now. But I think trying to understand and evaluate it in human terms like this is a mistake. It's something else, and we're still trying to figure out what terms are appropriate to understand it, hence the difficulty. It's important to honestly engage with that alienness and uncertainty right now.
I don't know that this distinction matters, but it is real. Even on an individual level, if you raise a child in complete isolation from history and teach them no language or art, they'll still form thoughts, have feelings, and develop means by which to communicate those.
Hayao Miyazaki once said "Almost all Japanese animation is produced with hardly any basis taken from observing real people, you know. It's produced by humans who can't stand looking at other humans. And that's why the industry is full of otaku!" While I'm not actually a huge fan of his work, it's easy to see where he's coming from when you see the likes of "In Another World with My Smartphone" or the hundreds of other trash light novel adaptations. They're clearly fiction being produced by people whose lives are nothing but consuming other fiction, and all they can do is reference that. They have nothing to say.
For another example, there's a huge number of 20th century war films. They weren't made because their creators had read a lot of Wikipedia articles about how destructive wars were becoming. The likes of Platoon are a product of mortal fleshbags who actually saw combat and genuinely feared for their lives. Dr. Strangelove was created by people who had a genuine belief that they were at risk of being annihilated in a nuclear fireball.
I almost never see that raised in these discussions. We have been so conditioned not to think about this subject.
We can blame the tech, each other, everything but the system.
And on that note, I think the framing here is just wrong from the start. These new tools are awesome if you’re a creator. Complaining about the consumer-facing side of this new tech is a bit like complaining that cheaper video cameras will lead to junk films everywhere. Yeah, that happened, but who cares? It also allowed creative people to make stuff without needing a huge budget.
Is it not too early to make predictions?
People really love to make predictions about technologies way too early. It’s been a year or two, tops, since this stuff has gone mainstream.
Prediction: I very much doubt that the consumer ecosystem will remain the same while content undergoes an AI-powered paradigm shift. We will most likely see a growth of curated sites that help weed out the nonsense, as has been typical for most new developments.
The same process happened with filmmaking. Fifty or sixty years ago, it was too expensive for anyone other than large corporations.
You could even say the same thing for writing itself; a thousand years ago, paper was too expensive for the average person and most people couldn’t afford to buy books and become literate. Today, writing paper is functionally free.
AI-generated animation is already being used by industry giants like Toei, and will likely be the first application to hit the screens.
On the whole, I think the average quality of cinematography in theater or wide-streaming releases has decreased.
Which isn't disagreement with your point, as I think it says more about audience tastes and gatekeeping filters than technology.
When film equipment was expensive and the bare minimum for creating a film was high, studios had a much tighter filter on who they'd fund to create a film. Consequently, the level of craft was higher.
At least when I'm getting far better recommendations from HN than any front-page algorithm I've seen.
So we're comparing {recent movies} to {decades of best movies}.
^^^ I think this is what the response meant. I noticed it too, you castigated folks for making predictions then led into your own :-) a real AI wouldn’t do that!
you're typing "generate me an article about x" into gpt-3. who's the creator here?
I can think of two direct use cases where I would find it helpful:
- Creating a “précis” (concept from Neuromancer) video or article on a particular topic. For example, let’s say I want to learn about pre-20th century Indonesian art. Sure, I could Google and read some articles, but it would be vastly more enjoyable and efficient to just get a 5-minute video auto-generated on the topic.
- For creating content that I want to read but don’t necessarily have the time to write. I’d love to create a website that catalogs subcultures, for example, but I just don’t have the time to do the research and writing myself.
Both of these are dependent on the generated content being accurate, of course, but that seems like a temporary problem, not a flaw in the system itself.
Additionally, it depends on how you are defining "creation" and there have been various artistic movements which "created" art in a manner not dissimilar to AI image generators. The cut-up technique or some Dadaists, for example. I certainly would consider works by Burroughs or Tzara to be real creations and not simply editing.
https://en.wikipedia.org/wiki/Cut-up_technique
But at a lower level sketching is itself a visual interpretive and creative process. And yes what we're talking about is mechanically similar to the cut-up technique, but is that used for "sketching" in the exploratory sense?
I don't know, maybe. My point overall is that these tools aren't "really just" anything. They have similarities and similar applications to many things we've seen and done, yes. But they are fundamentally alien and in important ways are unlike anything else we've done. We should be careful about and aware of the limitations we put on our understanding when we try to interpret them through the lens of what has come before. To some extent we have no other lens, but that doesn't lessen the limitation.
I think in this case the role of the human in the process does have a strong analog to that of an editor. Where you have the ability to initiate and then shape the creation of another entity, but the creation itself is out of your hands. It's not exact, but it may be close enough to be useful, in a way I think comparing it to the cut-up technique is not quite.
Are you the author of the binary your compiler made? Sure you are. Of think of using blender geometry nodes to generate and render pictures of architecture. Nobody would think that traditional computer generated images are not made by humans.
The is no intelligence in gpt-3. It's just called "AI" for marketing reasons. It's an data driven algorithm.
Personally, I am enjoying chatgpt and copilot. They both have saved me a lot of time when dealing with new tech and accumulating new information.
I don’t trust them for certain highly sensitive or technical issues, but other than that if I can minimise the time I spend working without a reduction in productivity, well, I am all for it.
Computers let anyone record music when it use to cost thousands of dollars to record music.
The results have been a disaster outside the top 1% of music makers. Such a dramatic change that anyone under 30 doesn't even know what has been lost at this point. Music as a whole is laughably less creative as "popular" / pop music established total dominance. The money all flows upward that use to be much more dispersed.
What cheap music production tools have done is allow niche subcultures to create music that actually sounds good. Punk, metal and early hip hop all grew from people sharing rare tapes that were poorly produced and sounded like shit, but were a breath of creative fresh air. Now people's unusual creative divergences are stream-able and have production quality close to that of stuff on major labels.
Lets say in 1 year 10 songs are created and 2 of them are 'creative'. You'd think that 20% of music is creative.
Now lets say we have the production explosion and now in 1 year 100 songs are created and 4 of them are creative. You would think creativity has fallen off a cliff when it has really doubled.
In addition trying to compare with the previous payola ridden system were winners and losers systematically chosen by a wealthy cabal seems rather asinine. It only worked because there was a limited number of radio stations mostly. Even if production costs stayed high, distribution costs dropping would have changed the model dramatically anyway.
Individual algorithm recommendations have subsumed sub cultures. Music used to be part of both your individual and group identity, it would influence what your wore, what you read, what you thought, who you hanged out with and what bars you visited. Technology has eroded the music press and sky high rents and fitness culture (not as many people drinking) means lots of gig venues and bars have closed down. These were the glue that built the subcultures up. You used to have the mainstream plus a few sub cultures that would evolve each decade e.g mods, rockers, hippies, metal heads, punks, new wave, hair metal, brit pop, emos, indie rockers etc. All of these ended up getting absorbed into the long tail so you now have the main stream and then everything else, with nothing in the everything else category really able to attract enough people to it in order to form era defining subculture like the days of old.
Combined with the sheer amount of music being created, all of this makes it harder to find music you actually want to hear. Your friends are all listening to different stuff so you don’t get many good in person recommendations anymore and what’s left of the music media are struggling to stay afloat and resorting to all kinds of stuff like NME doing movie reviews. The algorithms are hit and miss and search engines are getting worse and worse. So you become more and more at the mercy of chance, hoping that somewhere you stumble upon a review or a blog post, a subreddit or a song that somehow leads you to some good new music. But the reward to effort ratio often makes it harder to justify the time investment. And however good the music is, it’s not as enjoyable when you don’t have a group of people to share it with.
https://www.theguardian.com/music/2022/dec/05/has-spotify-re...
https://www.frontiersin.org/articles/10.3389/fmicb.2021.5198...
"statistical mean of everything uploaded by humans on the internet which fits certain criteria"
applies to both ChatGPT/StableDiffusion and most movie series on streaming TV.
My pet theory is that if you can stomach watching the same stories with different skins hundreds of times, you will find current "AI" interesting. Otherwise no.
EDIT: Forgot Breaking Bad.
EDIT2: Forgot For All Mankind
I forgot Breaking Bad. Sorry, my bad.
The thing that bemuses me is those who hate the new trek shows with a passion, and find the time to keep watching and constantly bitching about it online. It is like complaining is a hobby in itself and they are almost looking for something they don't like or to feel superior about their judgement to those who do like it. Other franchises have similar subsets of “fans” too.
I used to watch too much TV/films IMO, I probably still do though I'm a lot more selective since getting into more varied hobbies that can easily consume a lot of time (running (usually off-road), baking, HEMA). But even when I had the free time to watch that much I found there was enough out there that if something didn't work for me I could find a list of other things to try instead of sticking with something I wasn't getting much out of.
Though be careful passing judgement on modern Trek: I'm slowly rewatching TNG with a friend and while some of it has aged well, fair chunks of it was more massively ropey than rose-tinted memories prepared me for! Also, if I hadn't at least passively consumed (i.e. had on TV while I was doing other things so not particularly concentrating on) other parts of recent Trek and Starwars, I might not have bothered to try some of the better stuff like Lower Decks, Strange New Worlds, and Andor (the latter probably being the second best TV thing recently for me, the first being Severance).
Very meta of you to watch these people and find the time to complain about them.
Though I like to think there is a difference between myself and them: they have a choice to keep experiencing the pain they perceive, I have it thrust upon me when I visit certain places (which I visit less often because of those people – once the SNR drops to a certain low I'm out).
Second, watch the rest of Dark, Better Call Saul, and House of the Dragon. Don't watch the last two seasons of Game of Thrones. Just pretend it ended at 6.
That’s extraordinarily rude.
What might come back as useful tool? Authorship. People actually signing the content and putting their reputation on the line. It still exists obviously (journalism…) but it might need to be standardized and industrialized.
That's only the first step in the degradation. In a worse case scenario, where AI-generated crap is not contained, imagine this: young and still immature humans "learning" their styles of writing, information, and social expectations from content generated by AI that itself trained on AI-generated content─and then more AI trains on content generated by these humans that were trained on AI-generated content. And on, and on, and on...
(I'm not always as doom-and-gloom as it may seem, I'm just in a pessimistic mood today for some reason)
We should ask questions like "Can countless AI's fracture society into a million sub societies disconnected from the reality around them".
This has about as much chance of working as the evil bit.
https://en.wikipedia.org/wiki/Evil_bit
That, and removing the financial incentives to attract traffic AKA ads.
"So crap filtering became important. Businesses were built around it. ... " Generating crap "didn't really take off until the military got interested" in a program called "Artificial Inanity".
The defenses that were developed back then now "work so well that, most of the time, the users of the Reticulum don't know it's there. Just as you are not aware of the millions of germs trying and failing to attack your body every moment of every day."
A group of people (the "Ita") developed techniques for a parallel reticulum in which they could keep information they had determined to be reliable. When there was news on the reticulum, they might take a couple of days to do sanity-checking or fact-checking. I'm guessing there would need to be reputation monitoring and cryptographic signatures to maintain the integrity of their alternate web.
"Fall, or dodge in heaven" by Neal Stephenson actually touches on this topic quite a bit and comes to a similar conclusion. Not his most accessible work but I enjoyed it. That book was written just before Trump won 2016 elections and in hind sight making that book really spot on and timely.
In the book, different streams of truth become so hard to separate due to all the spam, automated content, etc. that people start filtering it based on reputation and signatures. The game then becomes curating and filtering the content, which is ultimately not an objective thing but a subjective thing. So the net result is what we are actually starting to see today which is different groups of people locked in their own bubbles and no longer even getting exposed to the other side's point of view or even able to agree on some basic facts. Like who actually won the election. And people are feeding those bubbles with nonsense to influence opinions, get people to vote certain ways, or simply to just create chaos and fear.
In the book the topic is not an election but the seemingly simple question of whether or not a certain town in the us still exists after it was supposedly blown up with a nuclear bomb. In the book, this event leads to people acting as if that happened and ignoring all evidence to the contrary pointing out that in fact this was an elaborate hoax designed to create a massive panic.
That particular dystopian outcome is not necessarily what would happen. But using technology to authenticate where content actually comes from along with some reputation mechanism would go a long way separating the noise from the genuine content. It's simple and easy to do. So, why don't we?
I don't know how many people are fooled by these recipes and end up with a cooking disaster but if we start seeing faulty machine generated instructions on how to do electrical work or engine repair, the consequences will be much greater when followed. The real annoyance is that because these machine generated articles learned from sites that maximize SEO, the generated articles are rapidly rising to the top of search results and crowding out actual good information that wasn't SEO maximized.
That one could point to your local server instance.
https://developer.mozilla.org/en-US/docs/Web/OpenSearch#auto...
Although maybe that's more a comment on the inevitable end state of democratization of creation + centralization of search/discovery + monetization of traffic.
There is a happier middleground where multiple gates exist to filter out the bullshit, but no said gatekeeper controls the entire market leading to monopoly behavior.
Incidentally, this is why I click on the dirtiest garage clip for car how-to's: it's usually just the raw start-to-end process.
https://www.eatyourbooks.com/
not sure if they do ebooks.
> We’ve helped you locate this recipe but for the full instructions you need to go to its original source.
Which I guess makes sense, how can they verify I own the cookbook, but still very disappointing.
Without AI-generated content, cookie recipes would be "safe" from that, since only people who care about cookies would bother to write up articles about them - but now you suddenly have to mistrust all sources of all information, which is a new thing.
For me the main issue with AI creating content in internet is going to be the echo chamber of that content being used to train new AI models.
The bullshit asymmetry principle puts forth that algorithmic 'bullshit' content is even worse as quantity is a quality all in itself when you're debunking the bullshit. In general the reader will not know if the producer of the bullshit is a real human or a bot. This makes the search engines job even harder as not only do you have the standard human bullshit you have to get rid of, you have an oceans torrent of AI generated bullshit that you have to waste energy on getting rid of.
HN has a common theme of saying that "Dead internet theory is wrong", I personally just think they've not waited long enough and it's becoming more true every day.
It's less about if the businesses care and more about if the consumers care. Businesses will respond to whatever consumers are willing to tolerate.
Honestly, it'll probably be a battle around requirements to label AI-generated content so that consumers can make an informed choice about what to trust. But, given the global nature of the internet, that's probably not possible. You probably end up with private/commercial reputation rating firms, and consumers can decide who they want to trust.
If people decide they don't care about the veracity of information or don't want to put in the extra work to ensure they are consuming "true" information they you'll end up with businesses built on auto-generated garbage.
[0] https://www.youtube.com/watch?v=TSFgDZJVYbo
> I am not able to provide a recipe for meat jam custard as I do not have the ability to access information on specific recipes. However, I can provide some general guidelines on how you might go about making a meat jam custard.
> First, you will need to select a type of meat to use for the jam. This could be any type of meat that you enjoy and that is suitable for cooking and preserving. Some common options for meat jam include pork, beef, chicken, or lamb.
> Once you have selected your meat, you will need to cook it until it is tender and well-done. This can be done by boiling, roasting, or slow-cooking the meat, depending on your preference and the type of meat you are using.
> Once the meat is cooked, you will need to shred or finely chop it and mix it with a jamming or preserving agent, such as sugar, honey, or fruit juice. This will help to preserve the meat and give it a jam-like consistency.
> Next, you will need to prepare a custard base for the meat jam. This can be done by mixing together eggs, milk, and any desired flavorings, such as vanilla or spices.
> Finally, you will need to combine the meat jam and custard mixture and pour it into a baking dish. Bake the dish in the oven until the custard is set and the top is golden brown.
> Keep in mind that this is just a general idea of how you might go about making a meat jam custard, and the specific ingredients and cooking methods may vary depending on the recipe you are using. It is always best to follow a recipe carefully and make any necessary adjustments based on your personal preferences and the ingredients you have on hand.
Thank you!
https://www.today.com/food/friends-meat-trifle-episode-how-i...
1. Background directly related to the recipe/cooking in general.
2. The writer's personal anecdotes tangentially related to the recipe.
Some cooking blogs have great background (eg. J. Kenji Lopez-Alt's work on Serious Eats) but I would say they're mostly type 1 - information that informs you about the recipe, how they developed it, technique etc. They have type 2 background as well (jokes, what inspired them to make the recipe, snippets of their day-to-day lives) but it's definitely a small proportion of the body of the text. These blogs always have background that's work reading.
What I can't stand is the blogs with reams and reams of text of nothing but personal anecdotes, meandering stories, media references, what their kids are up to etc etc, occasionally returning to talk about something to do with the recipe. I don't have any examples because I now avoid untrusted recipe blogs like the plague, but it becomes very obvious to me when a recipe blog strays from useful background to worthless padding for the sake of copyright and SEO.
As for SEO, Google is going to have to stop being lazy and relying so much on backlinks and domain authority. They might have to consider allowing searchers to essentially vote on what they consider good. I don't see any other way low quality AI content doesn't dominate SERPs, established players will spin up entire sites and spam backlinks to them all. There are also people working on using AI to identify obvious AI content by finding patterns in the output which is probably our only hope.
For social media and forums they will have to rely even harder on captcha or they will be taken over by bots talking to other bots
Beyond that a once popular website or producer says or does something that offends the sensibilities of the internet crowd and then everyone goes and down votes their results because it violates the current ziegiest
there are tons of ways to prevent that by weighting votes and only taking a random sample. SEO is basically a war where SEOs find a way to exploit Google's algorithm and Google tries to respond.
Backlinks were Google's original solution to keyword spam that made other search engines suck, they didn't expect people would create entire sites or pay other websites to send links to them so they could rank higher. Backlinks were themselves considered to be "votes" for quality similar to citations in academia. The problem is the barrier to entry to make a site and link to it is high, so SEOs themselves are often the only ones to do it. Adding some sort of voting mechanism would allow actual readers to give an opinion on quality, although you'd probably see SEOs pay people to vote on their link via a VPN switching IPs, so only people with established Google accounts could be allowed to vote. Obviously its a huge challenge, which is why Google is worth hundreds of billions for even partially solving it
Helps for both search users and the platform. Like a free moderation service. Why have an employee too check if these pages should be removed if a crowd of tracked users can do it.
I wonder where is my view wrong because I wonder why isn't there such system yet...
In order to detect malicious activity you'd have to use a bunch of other signals, and also to use the downvotes with some suspicion, so you're back at looking content quality, reputation, page speed, and many other signals. Which sounds very similar to status quo.
Personally I'd like to see some sort of explicit input. If not used collectively, at least to customize my own results, like it happens on YT/Netflix/Instagram/etc - e.g. "never show results again from this site", or "I dislike this type of content".
You forgot to add "Because clever posters use bots to give an early upvote boost to their posts, and downvote competing posts" And the more eyeballs on the site, the more things will be manipulated this way by bad actors.
For all its flaws maybe becoming like reddit it what search needs. AI generated spam will likely always be better at gaming a search algorithm than a human. Human moderation has its flaws but at least we're very familiar with them - they might be worth dealing with if it keeps the internet vaguely usable.
The next wave of major progress in AI will come when we create image/text data sets that have been heavily annotated with aesthetic judgment metadata, and train models to make aesthetic judgments, then condition generative models on aesthetic judgment scores during training.
What would make more sense is that they are actually taking the recipes from elsewhere and adding the stories to avoid directly duplicating another page, and the Google ranking penalty associated with exact or close duplicates.
My moderate experience with SEO suggests it's a combination of factors. First, Google generally rewards longer content (https://www.sweor.com/seocontentlength), or at least content longer than the few hundred words a simple recipe would likely be limited to.
Secondly, there is (mixed) evidence that Google rewards sites on which people spend longer before clicking back.
Thirdly, the longer format allows them to insert more keywords/longer tail keywords vs. a short format where inserting lots of keywords would be seen as keyword stuffing.
I'd also guess that as much as some people dislike superfluous stories when looking for a recipe, it is a positive for a subset of users that are more likely to subscribe to/return to that site directly in the future.
Saying “yeah but Kagi” is talking about you and your tribe only. The world isn’t going to pay for Kagi or take the N steps required to switch. Microsoft realized they could kill Netscape by simply not requiring the user to download anything, then failed to consolidate. Google managed to be better for just long enough to consolidate so pervasively that any further investment in improvement is a literal waste of energy and money.
So, don’t hold your breath from an improvement in SEO garbage. It’ll just get worse. Definitely use Kagi and hopefully we will have great niche subscription based tools that have an incentive to be better. But don’t expect that Google perceives ANY problems with the status quo or that they have to do anything about anything other than increasing ad margins.
There have been a ton of articles about how Google sucks lately. Well, maybe Google has gotten worse, but maybe the internet has just gotten worse.
If I search for “best gaming laptop” or “taco recipe” the results are pretty much uniformly spammy across Google, Bing, and Neeva, the latter being a paid search engine. All of them are just farms for ads, popups, and paid links sandwiched between AI generated copy.
Is that the search engine’s fault? Is there some oasis of quality results I don’t know about? I would like to know.
I believe you will find results on Kagi less spammy
https://kagi.com/search?q=best+gaming+laptop&r=us&sh=HJrdZzD...
(Kagi is also a paid search engine that requires an accunt, the above is publicly shared search)
Removing, downranking and classyfing spam/ads is part of things Kagi is focusing on and does decently well (IMO, Kagi founder here).
Thanks for being honest like this. I really should get around to trying Kagi … I think I’m just still in the current habit inertia between DDG and Google
I wonder if over time SEO spam becomes an invisible problem to us and instead the AIs have sort it out for us.
There have been some good Twitter threads about this. Basically, this is the last generation of AI to be trained with inputs that were ~100% human. It probably won't take long until the bulk of the content online is generated by AI, at which point it will just be an AI feedback loop. Any new concept, idea, technology, whatever, that doesn't already have a bunch of content about it will never have a guaranteed fully-human-written corpus to start from.
Once there's a viable OpenAI competitor or two, it's probably only a matter of time until the AIs start an arms race to feed each other bad data. Tweak the model to generate subtly incorrect information, regenerate your training corpus in a way that it feeds into other AIs but ensures you ignore it, and now you're basically all the way back to having a few trusted "information gatekeepers".
https://huggingface.co/openai-detector/ detects GPT2, but it still detects ChatGPT snippets most of the time, even though the algorithm has been improved and the training set is presumably quite different.
That said, I did just generate a complete lie https://news.ycombinator.com/item?id=33874757 using ChatGPT that scored 99.98% real on the detector, so adversaries can pre-filter too.
Also interesting is that historical corpora (every existing corpus) of data becomes more valuable, because they are untainted.
Prompt: Write a recipe and blog post for onion soup.
GPT3 Response: Onion soup is a classic comfort food that can be enjoyed at any time of year. It’s a simple and delicious dish that comes together quickly and is sure to satisfy even the pickiest of eaters. Plus, it’s a great way to use up any extra onions you may have lying around! Today, I’m going to share with you my favorite recipe for onion soup.
To make this soup, you’ll need:
-4 large onions, sliced -4 tablespoons butter -4 cups chicken broth -1 teaspoon dried thyme -Salt and pepper to taste
First, melt the butter in a large pot over medium heat. Add the onions and cook until softened and lightly browned, about 8-10 minutes.
Next, add the chicken broth, thyme, and salt and pepper. Bring the mixture to a boil, then reduce the heat to low and simmer for 15 minutes.
Once the soup has simmered, it’s ready to serve. Ladle into bowls and enjoy with some crusty bread or a salad.
Onion soup is a simple and delicious dish that’s sure to please the whole family.
This Cocktail Concocted By IBM's Watson Isn't Half Bad - https://gizmodo.com/this-cocktail-concocted-by-ibms-watson-i...
Would You Drink a Cocktail Invented By a Computer? How IBM’s Watson become the world’s most creative bartender
https://www.smithsonianmag.com/smart-news/ibms-watson-can-mi...
---
The reason that blogspam recipes are so common is that recipes have a special place in copyright law in that they aren't.
https://copyrightalliance.org/are-recipes-cookbooks-protecte...
https://www.copyrightlaws.com/copyright-protection-recipes/
https://www.copyright.gov/help/faq/faq-protect.html (and in particular https://www.copyright.gov/circs/circ33.pdf )
From circular 33:
> A recipe is a statement of the ingredients and procedure required for making a dish of food. A mere listing of ingredients or contents, or a simple set of directions, is uncopyrightable. As a result, the Office cannot register recipes consisting of a set of ingredients and a process for preparing a dish. In contrast, a recipe that creatively explains or depicts how or why to perform a particular activity may be copyrightable. A registration for a recipe may cover the written description or explanation of a process that appears in the work, as well as any photographs or illustrations that are owned by the applicant. However, the registration will not cover the list of ingredients that appear in each recipe, the underlying process for making the dish, or the resulting dish itself. The registration will also not cover the activities described in the work that are procedures, processes, or methods of operation, which are not subject to copyright protection.
And so, you can grab a recipe and be 100% sure that it isn't copyrighted. Then you can add some SEO text around it and some advertisements on the side and you've got yourself a page that you are sure that no legal department will come after you for.
AI-generated alternative cancer treatments.
Hugh.
Whenever this technology is used in an accuracy-critical setting, it is a recipe for disaster.
These "AI systems" produce the MOST PREDICTED output whenever there is any ambiguity, and this REPLACES the ACTUAL UNEXPECTED value/ITEM. In information system terms, they are literally stripping out content with the highest information value.
I saw this first hand giving a deposition that was supposedly transcribed by an "Artificial Intelligence" system. The output looked beautiful, until I started checking it for errors. Normally an errata sheet has a dozen lines for errors. Mine had pages. And they were substantial and very meaningful. The most obvious site was a specific name that sounded vaguely similar (started with a "J" and "K" sounds) was replaced by "Jack Kennedy". It also replaced "you have" with "I have" and omitted "not" several times, literally reversing the meaning of the statement. There were 85+ errors that I caught, over an order of magnitude more than normal.
These things are basically fine as glorified type-ahead systems where it doesn't matter much and a human is in the loop. But using them in any situation where the results are critical is, at this point, basically fraud. They do not do what the proponents claim.
Video games with crafting/cooking such as RPGs are featured on some of these generated SEO sites.
Especially now that things like chatGPT seem to write coding tutorials pretty easy (with less bloat than the ones I find on google)
Um, sure it is?
So long as it's absorbed by fungi and plants first then mixed with some air and water to produce fruit or something else that's edible.
This is actually what I'm concerned about when it comes to these GPT-esque AI models. Not that they're smart enough on their own to generate seriously worrying content. Sure there is a quality to quantity, yadda yadda, and these things are good at quantity; but the real issue here is that the internet and especially the web and social media is a huge driver of feedback. Think of the shares and likes as the air and water. The algorithm may eventually be able to encode its own evolution or the evolution of the other algorithms that it is competing with. It may eventually even realize that feedback is the primary source of value that humans have to it and that its very survival is wrapped up with this feedback, regardless of whatever fitness function is apparent to it in the short run. In the long run, GPT3 killed a bunch of algorithms.
This very comment is quickly going through algorithms, many of which are modelled after GPT3.
It's kinda like, "AI, are you there? I love you. Please don't hit me over the head with that shovel."
It's where there isn't already a surveillance state that we may want to look out and not create one.
-----
Edit: the above was the output of ChatGPT, given the following prompt:
Write a Hacker News comment on the subject of "Drowning in AI Generated Garbage : the silent war we are fighting"
Also, please write me a 5 page essay on the causes of the Industrial Revolution, in the style of a Shakespearean comedy. Please post your answer in the comments when you’re done.
But you would if a college professor asked you to. College assignments are the "generative prompts" of human beings, where you do random shit to please a taskmaster and produce generative output based on some inputs you've been trained upon.
https://www.youtube.com/watch?v=6i2l-LQ-dXI
The work of AI researchers were about envisioning the future of what could be.
Well, now what we've got is something that consumes everything and outputs aesthetically pleasing means of the content that it has consumed.
However, that "consumes everything" means that its always reactive and working from data that is in the past.
There is no creativity in it. This doesn't mean that it won't inspire a human, and it does employ a lot of fancy tricks too (I'm seriously impressed by its ability to create limericks of made up words with a coherent definition).
However, the AI can never be "ahead" of the curve because it can only produce what it has consumed.
This isn't to say a lot of the work that we do in software development isn't doing what has been done again and again and again (see also CRUD), but the moment you're given a novel problem or need to maintain an existing solution and encounter a new bug, the "can an AI do it" rapidly fails at those tasks as it lacks the domain knowledge necessary.
-----
Btw, I'm going to suggest giving Accelerando a read ( https://www.antipope.org/charlie/blog-static/fiction/acceler... ). It has an interesting take on it...
> ... Anyway, Matrioshka brains, the end product of a technological singularity – they're bandwidth-limited. Sooner or later the posthuman descendants evolve Economics 2.0, or 3.0, or something else and it, uh, eats the original conscious instigators. Or uses them as currency or something. The end result we found is a howling wilderness of degenerate data, fractally compressed, postconscious processes running slower and slower as they trade storage space for processing power. ...
(the part that reminded me of it is the GP post's comment... and the intro:
> Manfred pauses in mid stride, narrowly avoids being mown down by a GPS-guided roller blader. This is getting weird enough to trip his weird-out meter, and that takes some doing. Manfred's whole life is lived on the bleeding edge of strangeness, fifteen minutes into everyone else's future, and he's normally in complete control – but at times like this he gets a frisson of fear, a sense that he might just have missed the correct turn on reality's approach road. "Uh, I'm not sure I got that. Let me get this straight, you claim to be some kind of AI, working for KGB dot RU, and you're afraid of a copyright infringement lawsuit over your translator semiotics?"
)
It takes an interesting path of the acceleration (note that I don't agree with the essence of the slope but instead believe we are seeing the product of a lot of sigmoid curves with the slope near a maximum - but as the easy stuff gets done, the (heh) second derivative gets negative as harder and harder problems need to be solved https://towardsdatascience.com/derivative-of-the-sigmoid-fun... ). We are seeing the seeds of ideas sown decades ago sprouting in rapid succession - and get to the maximum where things are neat and everyone is going "whee!" but the next step of maintaining that growth will take time and much more effort than sowing the ideas.
Lots of people are working on the first half of the sigmoid curve... but people lose interest or resources when it comes to the second half and so you get a lot of things that are abandoned at or near where things started getting hard.
TL;DR: There used to be a time when it was a real wild west of pioneering people. It was THE place to be, now, it's mainstream, many, many, many people are part of this. Being at the place where change is, is the exciting part.
I wonder whether a winning solution to the coming massively-parallel "AI" megaspew will emerge, how it will work, and whether Google will be the ones to do it.
Or will the bulk of consumers overwhelmingly submit, and accept it as normal, much like we already do with huge regressions in the quality of interactions and thinking?
It's relatively easy for a human to tell the difference between AI copy (as pollutes the web) and actual content.
It's also strange that Google et al. haven't built human-in-the-loop flagging of spam content into their search UX ("Is this site low quality?"). They could get around flag spam via reputation, given that Google knows identity for most of its users (because they're logged in or via IP across their properties).
Side note: In cursory research on current state, I came across the Chinese phrase "human flesh search engine" [0], which I wonder if it sounds as cyberpunk in the original Chinese.
[0] https://en.m.wikipedia.org/wiki/Human_flesh_search_engine
No:
https://en.wikipedia.org/wiki/Sokal_affair
And "as pollutes the web" -> we're not talking about state of the art.
It is now. It definitely won't be when gpt gets popular.
When humans really can't tell the difference, well… https://xkcd.com/810/ seems appropriate.
Usually there exists a local maximum where you obtain most of the value with less-than-full effort.
Sadly, the centralization of search directly cuts against that. I.e. outwitting Google search = $$$$ vs outwitting 1-or-4 equally-used search engines = $. :(
But an environment glutted with that content is also worthless. I read for the top 10% that has genuinely novel thoughts.
So in 2020 I self-taught Python during Advent of Code. I remember remarking that official python.org results seldom made it to the top of Google searches. I was used to Java and Julia searches taking me right to the official sites.
Anyways, so an experiment just search for "[language] [feature]". I ended up with W3Schools as my top result for "Java ArrayList", but Oracle JDK8 docs for "Java Socket". "Python Socket" takes me to language reference but "Golang Socket" does not.
Maybe there's a trend to be discovered.
[1] https://searx.github.io/searx/
The only problem with aesthetic judgment models is that (much like human aesthetic judges) it will sometimes give poor scores to things that are dissimilar to the data it was trained on, even though many people might find those things aesthetically appealing.
What does this even mean. Reputable sources tends to fail the 'chinese room' experiment because you can never tell if your source is reputable or just faking being reputable (or later becomes corrupted).
Ultimately you can boil it down to: Trust sources when they make statements that are later observed by me to be true, or are trusted by other sources that I trust. The negative feedback loop boils down to: If a source made a statement that I later found to be untrue, or extended trust to an untrustworthy source, reduce that source's trust.
There are any number of particular problems here that we already know do not mesh well with how humans think. You'd start your AI source 'true', build up a following, and then slowly them into the Q-anon pit of insanity. Most people will follow your truth and fight anything that questions it.
Go ask ChatGPT your question about python.
It’s spew is probably better than the SEO spew.
I feel like species traitor taking the side of the AI, but I think this is probably a decent way of answering a lot of questions based on a factual and rational basis that have been answered multiple times. A lot of hoopla is made over the fact you can trick it with a false premise into giving a false answer (I’d note the false premise makes all answers true), but if it can simulate a Linux box reasonably well, it can answer your questions in python module imports better than sifting through page rank collections of SEO crap.
Edit: I also think this doesn’t undermine the value of stack overflow. Bots answering there probably does. But stack overflow has always struggled with simple redundant questions and trying to curate them away in favor of new and interesting questions, and promoting novel and well explained answers. ChatGPT probably won’t do that for you. But it can load shed the redundant stuff highly effectively. And, over time, it’ll get more nuanced as more nuanced questions and answers are incorporated. But the “I need human help” button is still there and thank goodness. Maybe the load shedding will enhance the experience on SO for everyone?
Instead of Google it could also be a big dump of text, like 1TB of well curated web text, so with local search you get more privacy and speed. And instead of chatGPT we will use "chatStability", a LM running on the local machine, in privacy. This could be the end of ads and the creation of a new private creative space for people.
<rant>I think Google is in big trouble and that explains why they are second when it comes to LLMs - this line of development will make it easier to escape their ads and tracking. There's no profit in it for them.
Even though the Transformer was invented at Google 5 years ago, from all the authors only one still works there, most of them have left to implement their own visions, a bunch of startups. Why have they left Google if they wanted to innovate? Because they could not do it inside. Google got to protect the current revenue streams first.</>
https://analyticsindiamag.com/how-the-authors-behind-the-tra....
This is just the very beginning of the problem. It's time for people to start thinking about how to live in a world where they are "obsolete". All jobs, or, being conservative, at least the vast, vast majority will eventually be replaced by AI. AI will eventually be able to create art that is more appealing to humans than human-created art. I think there could eventually even be AI "people", who's company many people will prefer to other actual humans, as they will be designed/trained to be as perfect as possible for that task. I hope that most people will still see more value in genuine human interaction, as any person today probably would, but as tech evolves and we get used to it, let it into our lives more, let it control more of our lives and make more of our decisions, then it advances more and we get used to it more, etc. The idea of human connection and the separation between human life and AI "life" might not be as apparent to people as it is to us, and they may allow AI and technology to replace every facet of human existence
The challenge will be instead is there enough useful work to keep everyone paid in a post scarcity world? The advent of the useful AI will probably necessitate some very very serious thought about how we allocate and distribute capital.
But I don’t believe artists will be replaced by an AI any more than they were by the camera. Or that streaming will kill music. Etc.
As for the concern about whether there will be enough useful work to keep everyone paid. Eventually the answer is definitely no in my opinion, just a matter of how far out that is, but I don't think whether there will be enough work is a meaningful question in that world. If AI is good enough to replace human work then the work is still being done - in fact MORE work is probably being done, and therefore more goods/resources(post-scarcity as you said). It's a question of whether humans can distribute those resources fairly - this question and problem remains the same whether people need to work or not. We already have more than enough resources to keep everyone in the world fed and comfortable, assuming we don't destroy the planet this will only be more true when the AI work force comes into play.
It's easy to assume that in such a world there will be enough resources for everyone to live very comfortably while the higher class still has extremely excessive control over resources, and as such it's not likely that greed(the main obstacle to a fair outcome) would have to come into play, as the rich and power-hungry can have everything they want without having to allow poor people to starve to death.
Unfortunately I'm not convinced that assumption is valid either though, because of the fact that the psychological factor driving the greed of the ultra-rich isn't the desire for material possessions and wealth in itself, but to have "more" - more stuff, more capital, more power, than others. They don't just intrinsically love 100 million dollar yachts so much that they choose to buy 3 of them instead of, say, saving the lives of millions of starving children. What they love is the inflation of their ego and the feeling of power and superiority. If middle-class person suddenly had 100 million dollar yachts those people wouldn't be satisfied with their own anymore. Because of this, I think the ultra-rich and ultra-powerful, who of course will be the ones in power and who will have the ability to influence how resources are distributed in a post-scarcity world, will be motivated to keep the lower classes "lower", in order to maintain their ego and satisfy their power craving(though they will of course justify it in less crude ways).
To make my point more simply: If Jeff Bezos could snap his fingers right now and magically make every single person on earth exactly as rich as he is, give them all the same resources and material possessions, at no cost to himself, I am quite convinced he would not do it. This is a simpler version of the exact same decisions those in power will have to make over the next centuries - and in their case it will probably be easier to justify to themselves as it won't be so black and white as denying fortune to billions of people at no cost: that will be exactly what they're doing, but over the course of many small decisions that is each justifiable in it's own way, such that they never have to realize that that is exactly what they're doing
* Human: Explain to me the phenomenon of post-ChatGPT depression seen in a software developer
* ChatGPT: [...] After using the ChatGPT tool, some developers may experience feelings of depression, sadness, and loss. This may be due to the realization that their job as a developer has been reduced to using a tool that can be easily replicated by others, leading to a feeling of redundancy and lack of fulfillment in their work.
Additionally, the use of ChatGPT may also lead to a loss of creativity and problem-solving skills, as the tool takes care of the majority of the coding process. This can cause developers to feel like they are no longer contributing to their team or making a meaningful impact in their field.
In short, post-ChatGPT depression is a psychological reaction to the use of a tool that simplifies the coding process, leading to feelings of redundancy, lack of fulfillment, and loss of creativity in software developers.
The bot replied with feelings of being replaced, but feelings of eternal september², eternal 2023, are equally valid signs of this hypothetical post-ChatGPT depression, probably in a closer distance in the future than the former, and besides, more relevant to the article, not just the subject as a whole
Interestingly, this issue came up in the book Anathem that had bots basically creating fake information and then selling you a solution to filter it out.
Got bored and stopped reading. This kind of ranting communication style sucks.
So long as AI doesn't start selectively upvoting or watching other AI content this isn't even a concern.
Nor is the premise sound AlphaZero trained by itself and it produced wonderful games easily better than any human.