reminds me of a quote in the 1997 movie "Wag the Dog" (context: Stanley Motss, a Hollywood producer played by Dustin Hoffman, helps Conrad Brean, a political spin doctor played by Robert de Niro, to stage a war in order to cover up a presidential sexual-advances-to-underage-girl type scandal two weeks before the election):
Motss: "It's all, you know, thinking ahead. That's what producing is."
Brean: "It's like being a plumber".
Motss: "Yes, like being a plumber: do your job right, nobody should nocice; but when you fuck up, everyone gets full of shit."
The physical labor jobs are only a few years behind due to the ramp-up required in manufacturing for robotics - and price points. But the embodied AIs are coming, in humanoid form, to take all the labor jobs as well.
The Fed is playing a game of chicken with the economy to make it collapse, so we'll have to wait a bit longer because Google laid them off*. Hopefully Tesla's everyday robot attempt will be successful.
Well, power tools will make for even more exciting hallucinations.
Edit: Aren't we discounting the computer power that will be required to have a robot computing and reacting to what will be dozens of inputs all in real time in a mobile bundle?
Power is my biggest concern too. Essentially taking 8B fairly efficient human machines and replicating their power consumption to care for them will be massive.
I _strongly_ doubt that physical labor jobs like plumbing will be replaced within a few years. I spent a year working in welding automation and folks wildly underestimate the difficulty of automating real physical tasks like this. And that's in a weld cell with calibrated physical relationships between every component, laser scanning, stereo cameras, and CAD-modeled parts with reasonable tolerances.
The idea that a robotic solution to "crawling around in a basement and bodging pipes together" is anywhere near happening is frankly laughable. Further, with the current hot AI paradigm being LLMs, it's not even clear how these (very impressive) advancements would get us closer to robotic replacement of the trades.
The physical labour jobs will be solved “soon” after AGI, as robotics is mostly a software/ai/perception problem. Yes, currently robotics hardware kind of sucks, but it sucks because of a lack of investment due to a lack of software. If there were suddenly ~$100B going into robotics like there is going into AI, then many of those hardware issues would be solved very quickly. Especially if you have an AGI hive mind to throw at the problem.
It means the rich no longer needs the poor. If robots also replaces the military it means that the rich no longer needs to fear the poor. With both of those out of the equation things aren't looking great for the poor.
This does not even make the slightest sense, who cares about money when you can subjugate the masses in other ways. It is only a power game at this point.
And if most of the population becomes useless, why would any psychopath keep them around? I’m sure it can’t be that hard to exterminate most of the humanity.
It's not going to happen anytime soon. These language models take several bleeding edge GPU's to run at a reasonable performance. Vision to the likes of how humans perceive detail is another dimension of complexity. There have been groundbreaking advances in machine vision over the past 20 years, yet we still cannot build a robot that folds clothes efficiently [0]. Arguably, this is not only because of vision, but also dexterity, but both are currently of what a human can perform.
But if it happens, likely legislators will regulate AI. Most western governments already provide job procurement programs through being employed by them. Which is a good thing, but HN doesn't like to hear that. The reality is that most people need a job to feel fulfilled and be a healthy member of society. Which could change, but It's not going to happen until this technology falls into the hands of the common people instead of being controlled by large corporations.
One problem is saying anything other than “yep we are all fucked” just comes off as a cope.
I am indeed using ChatGPT for programming using a plugin, and it’s super helpful, but I’m still coding. Still have to know what’s going on. This is a brave new world for sure but my problems can’t currently be solved by firing all my devs and replacing them with AI (yet).
What's a "cope"? It's an intransitive verb, not a noun last I checked.
I don't understand how the universe of all possible ideas other than learned helplessness are somehow rationalizations.
Perhaps there was sanity in thought of diagnosis, but murderous insanity in the prescription of the Ted the Unabomber.
I don't see how the ultra rich, monopolizing technology and the means of production, will ever willingly part with coin except to throw some crumbs to stave off imminent overthrow. If every billionaire is a policy failure, then, in the system of government and commerce controlled as such by billionaires, there can be no lasting or meaningful reform from the ballot or from within.
This means massive, nonviolent revolution to reset the political and economic operating systems is the necessary path forward. It is not the only possible path because people may choose to be meek, apathetic, cowardly, or blame themselves.
It's something we must do if we don't want to careen off a cliff like Will E Coyote before looking down that the Earth's fragile ecosystem is permanently destroyed and the vast majority of humanity except a lucky few are destined to live in favellas.
A cope is an internet comment of coping self-delusion, often written when the author is high on copium. A milder version occurs when the author is under the influence of hopium, but is still generally aware of the desperate reality.
It's zoomer internet slang. From https://en.wiktionary.org/wiki/cope#Noun, "A coping mechanism or self-delusion one clings to in order to endure the hopelessness or despair of existence." Usually seems to be used in a mildly derogatory sense.
I would say no. Broadly speaking, I firmly believe that, to some extent, we're figuring out how to use computers to take the creativity out of what humans like to do outside of work, or rather that it's removing the human experience of creating something from the process outside of simple prompt refinement.
If we get to the point where these models can reliably produces the correct thing on the first (few) tries -- and given the current rate of advancement, I believe that this could be possible -- we've effectively removed the process of creation entirely. Sure you can do things the long way, and I'm sure some people will, but people trend towards instant gratification. For a slightly-flawed analogy, it's like playing a video game with cheats. It becomes boring very quickly because oops! suddenly everything is instant and there's no more reward for getting there.
It starts to feel worse when I see things like brain implants start to be researched as viable consumer products, VR / "metaverse," ... It feels like we as a species are optimising towards a terrible future when you start putting all the bad uses of these technologies together, because legislation won't catch up fast enough. The Matrix was supposed to be a movie, not an instruction manual.
Completely wrong, I enjoy turning wood on a lathe, this has been automated before a million times over, I still think it’s super fun. Logically it’s pretty easy to see that automating something is mostly about convenience and money.
Anyway if it gets boring you can use whatever supernatural, powers your describing to create you something interesting to do.
Children who do finger paintings are doing it for fun and for fun only. There are much more experienced artists then them, but it has nothing to with that. It’s to experience life that is what makes it worth living.
I like building a fire in the woods, sitting around it and drinking tea. That’s been done a few times before, it’s still something I hope to do into the future.
> I enjoy turning would on a lathe, this has been automated before a million times over, I still think it’s super fun.
Because it's what you enjoy doing. What if I don't enjoy turning wood on a lathe and just want the end product? I ultimately get someone else to do it for me, because I don't want to, and would rather just have the part immediately. I see people talking lately about the possibilities of generating TV shows or similar. When a never-ending stream of truly novel content is available, will the majority of people really trend towards doing other things themselves instead? Will they even discover those things in the first place? Look at how people talk about their kids and their interactions with YouTube, TikTok, ...
Who cares if people want to plug themselves it to the matrix (endless Netflix) and not turn wood anymore ? I think that’s there decision ?
We’re all having some sort of existential crisis in tech because we are addicted to work, we identify with it. Personally I just look out the window and dream of being outside with no shoes on walking through the forest or along a beautiful non-polluted beach.
My home is here on Earth with other biological creatures and it’s where I’d like to spend my time, it’s all that makes me happy.
Honestly if money wasn’t involved I’d love to just do wood work, build my own house, learn to sail, grow my own food and I’d be happy with that.
It is! But is it really ethical to put the choice there in the first place? Look at how many issues people get from addiction to social media or similar. Scoff at it or not, it can genuinely cause issues for people. Is it really the best idea to push more of that when we already know it's an unhealthy manipulation of a lot of people?
I think you are under estimating the bar for quality content. There is already an endless stream of content mass produced, whether by automated mechanisms or middling creators. You have to truly believe that epic, genre-defining works of art popular in mass media (not a generated static image but an entire movie or TV series) will be created by AI.
Also a lot of art is intertwined with the artist. The musician, actor, and comedian often have a public persona tied in with their creative output. An AI won’t be relatable in the same way.
I can buy mass produced beer at the grocery store. I can also pay 5x the price for artisanal craft beer from my local brewery. I appreciate the latter far more.
> I think you are under estimating the bar for quality content. There is already an endless stream of content mass produced, whether by automated mechanisms or middling creators. You have to truly believe that epic, genre-defining works of art popular in mass media (not a generated static image but an entire movie or TV series) will be created by AI.
My worry is more along the lines of TikTok, YouTube, etc. Look at how scarily good TikTok is at quickly pigeonholing you into your interests. "TikTok but AI-backed and always showing you exactly what you want to see" scares me. I'm not even necessarily worried about 5 or even 10 years down the road from now, but I can see it coming. People already use GPT models for writing papers, books, ... There's Midjourney etc. for image generation from nothing. I'm sure books have already been published this way, and the technology will only get better.
What happens when your TikTok feed is always exactly what you want to see, even though the video itself was just generated a moment before? On its own it may not seem like such a bad thing, but there's also exploration of VR and AR, brain implants, ... Do we really want to live in a world where what we consume is primarily just ML-generated content, tailored specifically to the exact user, to the point where there may not even necessarily be shared experiences per se to bond over? Maybe I'm just fatalistic, but this technology has extremely high potential for abuse, and people seem to just ignore the ethics of it more often than not.
Maybe I’m just unique but I dislike Tik Tok, and the art I consume is primarily movies and non-fiction books.
People have wants and needs. They need food and shelter. They need medicine and healthcare. They want human companionship and community. They want to travel and see and experience new things. They want to get wasted and party the night away. Fall in love and have kids.
Nothing I previously described is going to be radically changed by AI. Entertainment likely will be in fundamental ways, but AI isn’t going to replace flying to Italy and experiencing it for myself.
> Maybe I’m just unique but I dislike Tik Tok, and the art I consume is primarily movies and non-fiction books.
This is the trap I keep seeing people on this site fall into. You don't do it, but that's not the same as nobody does it. What about teenagers currently growing up with it? What about when they have kids who are primarily fed this content? ...
> They want to travel and see and experience new things. [...] AI isn’t going to replace flying to Italy and experiencing it for myself.
GPT and Midjourney and etc. may not do this on their own. But current, active research is ongoing for brain implants, there's all sorts of stuff around VR/AR lately, ... If there's a chip in my brain and a synthetic version being pushed in, what's the difference if it "actually happened," if the experience can be synthetically replicated? I just have to remember that it did.
This may not be possible today, or in 5 years, or even in 10 years. But someone will eventually make it possible if ethics and legislation don't catch up first.
I’m not worried about any of that. If healthcare can be more widespread, effective, and affordable that’s a great thing. And if someone wants to beam stuff into their mind that’s not radically different from watching a movie.
I just don’t think the future is scary just because technology is developing.
> I’m not worried about any of that. If healthcare can be more widespread, effective, and affordable that’s a great thing. And if someone wants to beam stuff into their mind that’s not radically different from watching a movie.
I mean this in the most respectful way possible, this is missing the forest for the trees. On their own none of these things looks bad. Putting them together, given what we already know about human psychology and how much it's been manipulated with technology, it's scary. Look at all the "fake news" commentary around certain American elections. Regardless of your political position, the scale at which that sort of thing was happening should give anyone pause. These tools in their current form already allow for that sort of manipulation on a scale we've never dealt with before.
What happens when these tools are 10x or 100x better? What happens to the generation of kids and teens being raised to think all of this is completely fine, without any of the privacy / ethics / etc. concerns? What happens to their kids? I know this sounds like "but think of the children!!1" but we've already seen what happens to young kids with things like YouTube:
> But child safety advocates and some members of Congress say there’s a problem with the app and corresponding website: an autoplay feature that keeps one video coming after another with no pause or interruption. When one video ends, another video selected by YouTube Kids’ recommendations algorithm automatically plays. Right now, autoplay is on by default, and there is no way to turn off autoplay, so YouTube Kids will continue to feed children algorithmically curated videos that run indefinitely unless someone happens to show up and intervene.
Especially with that second case, a small-enough child won't know that it's bad for them. And as parents resort to these apps more and more, as the tools for generating content at absurd scale get more and more accessible, it starts to not look good. You can argue parental responsibility, but American society -- where a lot of this stuff takes root first -- is built around not making that easy, since so many households have both parents working, sometimes multiple jobs.
You can, sure. But we're in the minority of being users of a fairly niche tech-centric forum. What about when you're a teenager, and all your friends are using <ai-powered app>, so you're left out of <some group experience>? What happens when you're a kid who's raised primarily on this sort of content? This alone could mess up large parts of the relevant generation, but it can get worse past that. A generation raised on this will happily put their kids in front of a screen of it, and the problem compounds. The AI/ML ethics problem isn't just a right-here-right-now problem, but one that'll have effects on all the coming generations.
People vastly overestimate how addictive ML content would be.
First off, what if the content you want to see is simply real stuff that other people are doing? Luscious perfect breasts can likely be generated by ML easily, but a spicy bikini photo of a woman I happen to know personally would easily draw my attention more. Maybe you can fool me by generating such an image, but eventually I’ll only feel any stimulation if I know the image is actually real.
I’ll give you a better example. Consider a movie where a persons head gets blown open, spilling brains all over the place. People might easily watch such a scene and think nothing of it.
But now imagine you watch a snuff film of a person actually getting their brains blown out in real life. For a lot of people this hits differently even though visually it might look the same. Some people specifically seek out these videos because they are real, despite the existence of fake alternatives.
Same goes for text, I don’t give a damn about some GPT generated essay but I will read something someone I know wrote with greater interest. Reading the opinion of an AI is a waste of time.
So as long as people value content made by humans, ML won’t have much to offer besides shallow, empty stimulus. Like watching some random porn and jacking off into oblivion, most people would still rather have actual sex with someone.
> First off, what if the content you want to see is simply real stuff that other people are doing? Luscious perfect breasts can likely be generated by ML easily, but a spicy bikini photo of a woman I happen to know personally would easily draw my attention more. Maybe you can fool me by generating such an image, but eventually I’ll only feel any stimulation if I know the image is actually real.
If we've never met in the real world, how would you ever know "I" wasn't real? ML models will eventually be able to generate effectively-perfect images, video content, voices, etc. People are hooking LLMs up to all sorts of APIs, ML models have demonstrated capability to learn things like gameplay before, etc. We already saw how many people genuinely got attached to Replika and freaked out. What happens when it's effectively impossible to tell it apart from a real human short of meeting it? What happens if we go to the VR-heavy future where it doesn't matter?
> Same goes for text, I don’t give a damn about some GPT generated essay but I will read something someone I know wrote with greater interest.
What if you don't know the difference? If I effectively prompt an LLM with "write X in the style of one of HN user xwdv's friends" and send it to you saying it was written by them, and don't tell you it's from an LLM, then you implicitly trust me that it's real. You don't have to know that it's real, you just have to believe it's real. The same general context applies to snuff films etc.; they're shocking because you believe it's an actual, real human.
> So as long as people value content made by humans, ML won’t have much to offer besides shallow, empty stimulus. Like watching some random porn and jacking off into oblivion, most people would still rather have actual sex with someone.
Real human connection is important, I agree! But we keep building all this technology to minimise direct human-to-human contact and interaction and creativity, and then wonder why everyone's miserable.
There’s tons of models and influencers on instagram with incredible photos and content and yet I find the people I pay attention to the most are just the regular everyday people I know in real life. ML content will just be more noise that I look at for a half second and scroll past.
Without personal connection, an addiction to a social network can’t be reliably sustained.
> Completely wrong, I enjoy turning wood on a lathe, this has been automated before a million times
Ok, but I'm guessing that woodworking is a hobby not your source of income? And also, the lathe itself was an "automation" of earlier methods of carving a furniture leg.
Would you use a slide rule instead of using a calculator? Maybe a few times, but would you do it consistently in a work setting?
I think about it in terms of ceramics. It's nice to go buy a hand-painted cat-themed coffee mug and use it regularly, but also I want the machine-mass-produced stuff to stock my kitchen shelves.
I would say that Ted Kaczynski was rather optimistic about the possibility of stopping technology and the inevitable negative impacts that would have on our lives. I was never that optimistic.
I think this will just encourage the growth of video games for artificial challenges. If AI can program better than me, that might make programming pointless. But if AI can play chess better than me or build better in minecraft, I don’t really care, because I wasn’t playing the game to win, I was playing to play.
Working a long time as a programmer. I have never seen a computer do anything but multiply jobs. This will do the same. IMO the way programming is done will change will need more domain experts and less code monkeys.
I think the implication was meant to be that there will be far more people who are both programmers and domain experts than there currently are programmers?
Domain experts often don't need to do anything "innovative" - they just want to schedule their team's workload or some-such that's been done a million times but the details of their particular situation are important. Most programmers now are doing such things for domain experts as it is. One possibility (I'm not 100% convinced) is that AI would allow the domain expert to do it directly, i.e. become a programmer themselves.
…a “tech lead” for AIs on less important projects that have opened up because the cost of development has been driven down, making bespoke development worthwhile in places where it wasn’t previously.
> Most programmers now are doing such things for domain experts as it is
HN again! ‘Most’ programmers are not doing that; most programmers couldn’t do that if their life depended on it. Or maybe the definition of programmer is different : I am talking about people who have the word programmer, software developer or software engineer in their job title. This, by large, does not mean they can write software, even on a trivial level, compared to what it means here on HN. Most can, are and will be replaced by AI because they simply bad; a handful will remain, which are the good/talented and smart ones. For now at least.
So as can be seen on HN talking about gpt; here many believe we are safe; that’s because the people hanging around here in the bubble are safe for now. But that is because it seems most people here never worked with a 100000 person ‘software development’ outsourcing shop etc; you would definitely say after 5 minutes with ‘senior engineers’ there that they can be replaced by 5 lines of Python, let alone an AI.
I think this applies for as long as humans are actually better than AI. At some point you won’t need a human directing the AI, it can just do any job better.
Do you write code because it's fun? Good, because you'll have a lot more flexibility around that in the future.
Are there parts of coding you don't like? Great, you will probably enjoy what's next.
If you answered no to those questions, you are slinging code in the worst way possible, in terms of feelings, per the author's choice of perspectives on coding.
Feelings about coding are by nature--and ought to be--ephemeral, deep, and nuanced. I wouldn't spend too much time asking what the feel of coding will be in the future, in terms of some single feel...unless feeling just really isn't interesting to you.
What I'd ask is:
- Will "what's left" really only resolve to one thing we'll "feel about"? Has that ever happened?
- Do we really think coding is just this "syntax, oft highlighted" experience? To me, it's far more symbolic and psychological than that. If you make yourself a schedule for the day, you are coding. You don't like that? OK, you probably code in some other way--maybe manipulating the boss is more your thing, which is also a form of coding.
- What if we view coders as an interface between humans and code (or even coded results), rather than simple code-typers? As people who have energy to think in a certain way at length, in depth, over time?
To me this is an extremely powerful lens on what will likely happen. Some subjective probability in use here maybe, but maybe that's OK since we've really only discussed subjective feeling thus far...?
I don't think prompt typing will even be the most of it. But I do harbor a stubborn intuition that higher-level logic will be massively involved in what's next. Those places above the current coder's heads where business strategies decide what code is or isn't written will likely need more of the same kind of coder who will have to shift roles a bit, and still sling logic in rigorous ways.
(Oh and speaking of stubborn ideas, I think any boss who feels like it will still figure out ways to let their favorite coder use VB6. It's always important to separate a fear of what may amount to impending nostalgia from the actual question of whether one will still be able to arrange for access to N-years-ago's way of doing things...)
P.S. The physical labor discussion is pretty amusing to me, because here you have people who, by their own indication in career tests, etc. don't even like to use conceptual perspective-taking as a tool on the job, compared with say the sensory experience of physical work.
Anyone who can even string together a couple of LLM-style prompts in a logical manner toward some business-logic goal is going to want to think twice about whether they really understand the ground-floor operations over at the tech worker's hobby-fantasy-phenomenon-zone that is physical labor.
Can you effectively create, and work to, a plan? If so, uh oh--careful what side of yourself you present as career-facing. Planning and conceptualization is not at all a sensory pursuit or any kind of physical labor, and is a common topic for jokes in that world.
Plus, it's not only possible to have a blind spot that conceals from an individual the big-picture, career-scoped value of their own innate talent--it actually happens all the time. A genius of a planner may think, "planning is no big deal, but now...hard labor! Now we are talking." Kaboom, a classic quarterlife or midlife-crisis career issue is formed.
It's one of the biggest traps this career coach has ever heard of, and heard experiences with, again and again...
Tangentially, something that I see come up in all these threads is the notion that knowledge workers unemployed by AI will retrain into some physical labor job. My question is, at that point, who's buying all that labor? Will the workforce just consist of executives and the manual laborers who work for them?
That's what we've always tended to assume would happen, but now it looks like it may not go that way, thus my concern. If knowledge workers could be replaced first, I don't think anyone's going to wait around for robotics to catch up before doing anything.
> AI/Robots should replace physical labor before it replaces knowledge workers.
There's actually a kind of internet famous novel where the exact opposite happens. I don't remember the name of it or the site it's on, but initial premise has AI automating management (starting with pretty menial things like directing workers at fast food restaurant, and working its way up the value chain).
It was pretty interesting, but I only got as far as the protagonist getting interned in a kind of debtors prison once his savings ran out and he couldn't afford to live outside of it (due to all the automation). The prison was theoretically open, but if you wandered too far away, robots would tranq you (after very many "nice" exhortations to return) and you'd wake up back in your bed.
_Some_ knowledge workers may be unemployed. The vast majority? Reduce the overhead on tedium like data entry, writing e-mail, etc. They'll be able to expand what they can do, because AI models like ChatGPT could be trained on company data and used to reduce errors, etc.
If anything, middle managers will be the ones to get squeezed out when they no longer have work to do.
Job interviews: "Programming is about 10% of the job of course, so we are going to ask some situational questions, and after the interview we will talk to your references".
AI Maniacs: "Ah all the jobs are gone, humans are useless!"
On the other hand, I can see in the future we will build complex programs we don't understand via AI as black boxes, and then use tonnes of tests, probably in a similar spirit to aviation industry, to ensure they do the right thing.
> But just stop for a minute and think about what you love about programming.
What I love most is creating stuff. I have little interest in algorithms, o-notation, and other minutiae. I’d love a world where I could just tell the computer what I want to create and it takes care of the minutiae for me. Though, I’m mostly thinking about personal projects. When it comes to writing code for work… well, that is not really in the ven diagram circle of things I love about programming. Maybe I’ll have to do something else for work, though I’m about 10-15 years away from retirement. It will be interesting to see how much things change by then.
> First DALL-E came for the artists, and we laughed, because artists are worthless.
First, this is an annoyingly unnuanced way to start an essay. Second, if at some point AI takes all the pain out of producing stuff (it does all the low level work in all sectors), the only valuable activity that’s left is art.
I think so! I mean, imagine if, when you notice some perf hiccup and you dig and realize you need to store your data in some different way.
Currently, commence a multi-month design cycle, create the migration path, perform the migration in a slow and safe way, soft logging, feature flagging, gradual release, etc., and finally once everything is migrated, remembering to remove the old code. All the meanwhile turning down a bunch of other ideas for improvements you had because you need to get this one thing done. Or worse, interrupting a bunch of half-finished refactorings and leaving them half-finished and having to work around those things until some time that may never come.
Imagine if you could just tell an AI to do this, be reasonably certain it would do the right thing, and it's done and provably correct and manages your deployments and migrations for you. And you could be free to look into the next batch of new things you want to do.
I love coding, but really I love thinking about the ways to do things more than the processes of getting those things done. If there was an AI that could automate that work, that would be incredible.
Kind of a tangential comment, but something I think about semi-regularly is that AI in a way feels like overkill for removing the programming part of software engineering. When you think about it, it's kind of surprising that we still program using text editors and IDEs instead of an interface that's more graphical. For that matter, it's also kind of surprising that GUI nocode builders have yet to really make an impact on the job market. Webflow exists, and has countless predecessors, and yet it and its competitors aren't really an existential threat to web developers.
His use of the term "shape rotator" kind of helps me articulate my thoughts on the matter. My theory is that attempts to democratize software authorship focus too much on the writing code part, but ultimately do nothing to address the "shape rotator" part, i.e. imagining the problem(s) you're trying to solve as abstract shapes. When I imagine the problem at hand, I construct a kind of mental model of it without really thinking about it all that much.
With that in mind, could our current approach to AI ultimately scale to the point where non-shape-rotators can finally write software as capably, or will it reach a plateau where shape rotators still retain their monopoly on software authorship?
If it's the latter case, then AI might not prove that much of a threat to programming simply because shape rotators seem to really like their text-editors, and really like to program
EDIT: some quick follow-up remarks, but despite feeling that AI art generators are unethical, I also feel at the same time that we're overestimating their potential damage because I think they democratize the wrong thing, which in this case is skill with a paintbrush or similar equivalent, yet they don't really democratize creativity and imagination. I think a lot of AI art is ugly, but I think that's largely due to the fact that the people using AI art generators simply aren't that creative or imaginative. By contrast, people who are creative and imaginative will understand that it's easier to get what they're looking for simply by a more direct, hands-on approach, rather than trying to find "good enough" through a prompt. Sure, people will get good-enough for free in lieu of paying someone more for better, but that's been an issue in the creative industry well before the advent of AI, and often with predictable results. Namely, good-enough doesn't typically create value so much as serve as a placeholder for something deemed essential, i.e. it's good-enough. A restaurant with good-enough branding, or a book with a good-enough book cover, is the same as a restaurant with no-branding or a book with no cover.
When programming is gone, we will be doing other things.
In fact, ChatGPT and other AI developments will probably enable people like me to create things I couldn't before - like video games and software that I have dreamed up but weren't skilled enough to do.
Now apply that to society as a whole, and we'll be able to be more creative with our computers than ever before - leading to a whole slew of new opportunities.
Knowledge workers will need to learn how to play with this new tool and use it to augment their work.
This is not the doomsday you are looking for. hand waves
With your 'video games' example, there will be more video games made. Additionally, this may cause more people going into that field, using GPT as a tool. Thus, wages in that position will come down, as the barrier to entry is lowered, and there will be increased competition.
I think this is going to rhyme with history. TL;DR - It'll be a good thing
When the printing press removed the manual tedium from producing a book, and reduced the cost by orders of magnitude. Literacy went up, and science and the arts flourished. Instead of manually copying texts, authors could write for millions.
When the Thread Cutting Lathe made it possible to make screws accurately and cheaply enough to standardize, the use of them went up, and the industrial revolution happened. Instead of making a screw in a day, a machinist could make thousands.
Once it became possible to store data automatically, the need for file clerks and others manually managing data went down. The management of information by computer made all manner of new business possible. Real time management of inventory and booking of reservations because possible on a vast scale.
GPT and other AI based tools will change the nature of our jobs, but we'll all have new things to do. It'll be less tedious, and a bit more creative, as it always tended to be before.
85 comments
[ 4.3 ms ] story [ 57.9 ms ] threadIdeally, the social status of the "email class" should be lowered as well, and no, I'm not talking about just programmers here.
Learn a Job GPT-4 can't do.
Train for HVAC installation, plumbing, and electrician work at Palo Alto Adult School.
Motss: "It's all, you know, thinking ahead. That's what producing is."
Brean: "It's like being a plumber".
Motss: "Yes, like being a plumber: do your job right, nobody should nocice; but when you fuck up, everyone gets full of shit."
* https://www.wired.com/story/alphabet-layoffs-hit-trash-sorti...
Edit: Aren't we discounting the computer power that will be required to have a robot computing and reacting to what will be dozens of inputs all in real time in a mobile bundle?
The idea that a robotic solution to "crawling around in a basement and bodging pipes together" is anywhere near happening is frankly laughable. Further, with the current hot AI paradigm being LLMs, it's not even clear how these (very impressive) advancements would get us closer to robotic replacement of the trades.
And if most of the population becomes useless, why would any psychopath keep them around? I’m sure it can’t be that hard to exterminate most of the humanity.
But if it happens, likely legislators will regulate AI. Most western governments already provide job procurement programs through being employed by them. Which is a good thing, but HN doesn't like to hear that. The reality is that most people need a job to feel fulfilled and be a healthy member of society. Which could change, but It's not going to happen until this technology falls into the hands of the common people instead of being controlled by large corporations.
[0] https://www.youtube.com/watch?v=2jc6iQZN-uA (Robot is recorded at 4x realtime)
I am indeed using ChatGPT for programming using a plugin, and it’s super helpful, but I’m still coding. Still have to know what’s going on. This is a brave new world for sure but my problems can’t currently be solved by firing all my devs and replacing them with AI (yet).
I don't understand how the universe of all possible ideas other than learned helplessness are somehow rationalizations.
Perhaps there was sanity in thought of diagnosis, but murderous insanity in the prescription of the Ted the Unabomber.
I don't see how the ultra rich, monopolizing technology and the means of production, will ever willingly part with coin except to throw some crumbs to stave off imminent overthrow. If every billionaire is a policy failure, then, in the system of government and commerce controlled as such by billionaires, there can be no lasting or meaningful reform from the ballot or from within.
This means massive, nonviolent revolution to reset the political and economic operating systems is the necessary path forward. It is not the only possible path because people may choose to be meek, apathetic, cowardly, or blame themselves.
It's something we must do if we don't want to careen off a cliff like Will E Coyote before looking down that the Earth's fragile ecosystem is permanently destroyed and the vast majority of humanity except a lucky few are destined to live in favellas.
If we get to the point where these models can reliably produces the correct thing on the first (few) tries -- and given the current rate of advancement, I believe that this could be possible -- we've effectively removed the process of creation entirely. Sure you can do things the long way, and I'm sure some people will, but people trend towards instant gratification. For a slightly-flawed analogy, it's like playing a video game with cheats. It becomes boring very quickly because oops! suddenly everything is instant and there's no more reward for getting there.
It starts to feel worse when I see things like brain implants start to be researched as viable consumer products, VR / "metaverse," ... It feels like we as a species are optimising towards a terrible future when you start putting all the bad uses of these technologies together, because legislation won't catch up fast enough. The Matrix was supposed to be a movie, not an instruction manual.
Anyway if it gets boring you can use whatever supernatural, powers your describing to create you something interesting to do.
Children who do finger paintings are doing it for fun and for fun only. There are much more experienced artists then them, but it has nothing to with that. It’s to experience life that is what makes it worth living.
I like building a fire in the woods, sitting around it and drinking tea. That’s been done a few times before, it’s still something I hope to do into the future.
I actually think we’ve all gone a little mad ?
Because it's what you enjoy doing. What if I don't enjoy turning wood on a lathe and just want the end product? I ultimately get someone else to do it for me, because I don't want to, and would rather just have the part immediately. I see people talking lately about the possibilities of generating TV shows or similar. When a never-ending stream of truly novel content is available, will the majority of people really trend towards doing other things themselves instead? Will they even discover those things in the first place? Look at how people talk about their kids and their interactions with YouTube, TikTok, ...
We’re all having some sort of existential crisis in tech because we are addicted to work, we identify with it. Personally I just look out the window and dream of being outside with no shoes on walking through the forest or along a beautiful non-polluted beach.
My home is here on Earth with other biological creatures and it’s where I’d like to spend my time, it’s all that makes me happy.
Honestly if money wasn’t involved I’d love to just do wood work, build my own house, learn to sail, grow my own food and I’d be happy with that.
It is! But is it really ethical to put the choice there in the first place? Look at how many issues people get from addiction to social media or similar. Scoff at it or not, it can genuinely cause issues for people. Is it really the best idea to push more of that when we already know it's an unhealthy manipulation of a lot of people?
Also a lot of art is intertwined with the artist. The musician, actor, and comedian often have a public persona tied in with their creative output. An AI won’t be relatable in the same way.
I can buy mass produced beer at the grocery store. I can also pay 5x the price for artisanal craft beer from my local brewery. I appreciate the latter far more.
My worry is more along the lines of TikTok, YouTube, etc. Look at how scarily good TikTok is at quickly pigeonholing you into your interests. "TikTok but AI-backed and always showing you exactly what you want to see" scares me. I'm not even necessarily worried about 5 or even 10 years down the road from now, but I can see it coming. People already use GPT models for writing papers, books, ... There's Midjourney etc. for image generation from nothing. I'm sure books have already been published this way, and the technology will only get better.
What happens when your TikTok feed is always exactly what you want to see, even though the video itself was just generated a moment before? On its own it may not seem like such a bad thing, but there's also exploration of VR and AR, brain implants, ... Do we really want to live in a world where what we consume is primarily just ML-generated content, tailored specifically to the exact user, to the point where there may not even necessarily be shared experiences per se to bond over? Maybe I'm just fatalistic, but this technology has extremely high potential for abuse, and people seem to just ignore the ethics of it more often than not.
People have wants and needs. They need food and shelter. They need medicine and healthcare. They want human companionship and community. They want to travel and see and experience new things. They want to get wasted and party the night away. Fall in love and have kids.
Nothing I previously described is going to be radically changed by AI. Entertainment likely will be in fundamental ways, but AI isn’t going to replace flying to Italy and experiencing it for myself.
This is the trap I keep seeing people on this site fall into. You don't do it, but that's not the same as nobody does it. What about teenagers currently growing up with it? What about when they have kids who are primarily fed this content? ...
> They need medicine and healthcare.
https://news.ycombinator.com/item?id=35319778 "Capabilities of GPT-4 on Medical Challenge Problems"
It will only improve from here.
> They want human companionship and community. [...] Fall in love [...]
https://www.abc.net.au/news/science/2023-03-01/replika-users... "Replika users fell in love with their AI chatbot companions. Then they lost them"
> They want to travel and see and experience new things. [...] AI isn’t going to replace flying to Italy and experiencing it for myself.
GPT and Midjourney and etc. may not do this on their own. But current, active research is ongoing for brain implants, there's all sorts of stuff around VR/AR lately, ... If there's a chip in my brain and a synthetic version being pushed in, what's the difference if it "actually happened," if the experience can be synthetically replicated? I just have to remember that it did.
This may not be possible today, or in 5 years, or even in 10 years. But someone will eventually make it possible if ethics and legislation don't catch up first.
I just don’t think the future is scary just because technology is developing.
I mean this in the most respectful way possible, this is missing the forest for the trees. On their own none of these things looks bad. Putting them together, given what we already know about human psychology and how much it's been manipulated with technology, it's scary. Look at all the "fake news" commentary around certain American elections. Regardless of your political position, the scale at which that sort of thing was happening should give anyone pause. These tools in their current form already allow for that sort of manipulation on a scale we've never dealt with before.
What happens when these tools are 10x or 100x better? What happens to the generation of kids and teens being raised to think all of this is completely fine, without any of the privacy / ethics / etc. concerns? What happens to their kids? I know this sounds like "but think of the children!!1" but we've already seen what happens to young kids with things like YouTube:
https://www.cnbc.com/2018/02/13/youtube-is-causing-stress-an... (2018)
> Mental health experts warn that fear-inducing videos affect brain development in young children.
https://www.vox.com/recode/22412232/youtube-kids-autoplay (2021)
> But child safety advocates and some members of Congress say there’s a problem with the app and corresponding website: an autoplay feature that keeps one video coming after another with no pause or interruption. When one video ends, another video selected by YouTube Kids’ recommendations algorithm automatically plays. Right now, autoplay is on by default, and there is no way to turn off autoplay, so YouTube Kids will continue to feed children algorithmically curated videos that run indefinitely unless someone happens to show up and intervene.
Especially with that second case, a small-enough child won't know that it's bad for them. And as parents resort to these apps more and more, as the tools for generating content at absurd scale get more and more accessible, it starts to not look good. You can argue parental responsibility, but American society -- where a lot of this stuff takes root first -- is built around not making that easy, since so many households have both parents working, sometimes multiple jobs.
First off, what if the content you want to see is simply real stuff that other people are doing? Luscious perfect breasts can likely be generated by ML easily, but a spicy bikini photo of a woman I happen to know personally would easily draw my attention more. Maybe you can fool me by generating such an image, but eventually I’ll only feel any stimulation if I know the image is actually real.
I’ll give you a better example. Consider a movie where a persons head gets blown open, spilling brains all over the place. People might easily watch such a scene and think nothing of it.
But now imagine you watch a snuff film of a person actually getting their brains blown out in real life. For a lot of people this hits differently even though visually it might look the same. Some people specifically seek out these videos because they are real, despite the existence of fake alternatives.
Same goes for text, I don’t give a damn about some GPT generated essay but I will read something someone I know wrote with greater interest. Reading the opinion of an AI is a waste of time.
So as long as people value content made by humans, ML won’t have much to offer besides shallow, empty stimulus. Like watching some random porn and jacking off into oblivion, most people would still rather have actual sex with someone.
If we've never met in the real world, how would you ever know "I" wasn't real? ML models will eventually be able to generate effectively-perfect images, video content, voices, etc. People are hooking LLMs up to all sorts of APIs, ML models have demonstrated capability to learn things like gameplay before, etc. We already saw how many people genuinely got attached to Replika and freaked out. What happens when it's effectively impossible to tell it apart from a real human short of meeting it? What happens if we go to the VR-heavy future where it doesn't matter?
> Same goes for text, I don’t give a damn about some GPT generated essay but I will read something someone I know wrote with greater interest.
What if you don't know the difference? If I effectively prompt an LLM with "write X in the style of one of HN user xwdv's friends" and send it to you saying it was written by them, and don't tell you it's from an LLM, then you implicitly trust me that it's real. You don't have to know that it's real, you just have to believe it's real. The same general context applies to snuff films etc.; they're shocking because you believe it's an actual, real human.
> So as long as people value content made by humans, ML won’t have much to offer besides shallow, empty stimulus. Like watching some random porn and jacking off into oblivion, most people would still rather have actual sex with someone.
Real human connection is important, I agree! But we keep building all this technology to minimise direct human-to-human contact and interaction and creativity, and then wonder why everyone's miserable.
Without personal connection, an addiction to a social network can’t be reliably sustained.
Ok, but I'm guessing that woodworking is a hobby not your source of income? And also, the lathe itself was an "automation" of earlier methods of carving a furniture leg.
Would you use a slide rule instead of using a calculator? Maybe a few times, but would you do it consistently in a work setting?
The train departed with Tiktok on this one. I'd say we have nothing to lose with the AI.
Domain experts often don't need to do anything "innovative" - they just want to schedule their team's workload or some-such that's been done a million times but the details of their particular situation are important. Most programmers now are doing such things for domain experts as it is. One possibility (I'm not 100% convinced) is that AI would allow the domain expert to do it directly, i.e. become a programmer themselves.
…a “tech lead” for AIs on less important projects that have opened up because the cost of development has been driven down, making bespoke development worthwhile in places where it wasn’t previously.
HN again! ‘Most’ programmers are not doing that; most programmers couldn’t do that if their life depended on it. Or maybe the definition of programmer is different : I am talking about people who have the word programmer, software developer or software engineer in their job title. This, by large, does not mean they can write software, even on a trivial level, compared to what it means here on HN. Most can, are and will be replaced by AI because they simply bad; a handful will remain, which are the good/talented and smart ones. For now at least.
So as can be seen on HN talking about gpt; here many believe we are safe; that’s because the people hanging around here in the bubble are safe for now. But that is because it seems most people here never worked with a 100000 person ‘software development’ outsourcing shop etc; you would definitely say after 5 minutes with ‘senior engineers’ there that they can be replaced by 5 lines of Python, let alone an AI.
This really depends on how you look at code.
Do you write code because it's fun? Good, because you'll have a lot more flexibility around that in the future.
Are there parts of coding you don't like? Great, you will probably enjoy what's next.
If you answered no to those questions, you are slinging code in the worst way possible, in terms of feelings, per the author's choice of perspectives on coding.
Feelings about coding are by nature--and ought to be--ephemeral, deep, and nuanced. I wouldn't spend too much time asking what the feel of coding will be in the future, in terms of some single feel...unless feeling just really isn't interesting to you.
What I'd ask is:
- Will "what's left" really only resolve to one thing we'll "feel about"? Has that ever happened?
- Do we really think coding is just this "syntax, oft highlighted" experience? To me, it's far more symbolic and psychological than that. If you make yourself a schedule for the day, you are coding. You don't like that? OK, you probably code in some other way--maybe manipulating the boss is more your thing, which is also a form of coding.
- What if we view coders as an interface between humans and code (or even coded results), rather than simple code-typers? As people who have energy to think in a certain way at length, in depth, over time?
To me this is an extremely powerful lens on what will likely happen. Some subjective probability in use here maybe, but maybe that's OK since we've really only discussed subjective feeling thus far...?
I don't think prompt typing will even be the most of it. But I do harbor a stubborn intuition that higher-level logic will be massively involved in what's next. Those places above the current coder's heads where business strategies decide what code is or isn't written will likely need more of the same kind of coder who will have to shift roles a bit, and still sling logic in rigorous ways.
(Oh and speaking of stubborn ideas, I think any boss who feels like it will still figure out ways to let their favorite coder use VB6. It's always important to separate a fear of what may amount to impending nostalgia from the actual question of whether one will still be able to arrange for access to N-years-ago's way of doing things...)
P.S. The physical labor discussion is pretty amusing to me, because here you have people who, by their own indication in career tests, etc. don't even like to use conceptual perspective-taking as a tool on the job, compared with say the sensory experience of physical work.
Anyone who can even string together a couple of LLM-style prompts in a logical manner toward some business-logic goal is going to want to think twice about whether they really understand the ground-floor operations over at the tech worker's hobby-fantasy-phenomenon-zone that is physical labor.
Can you effectively create, and work to, a plan? If so, uh oh--careful what side of yourself you present as career-facing. Planning and conceptualization is not at all a sensory pursuit or any kind of physical labor, and is a common topic for jokes in that world.
Plus, it's not only possible to have a blind spot that conceals from an individual the big-picture, career-scoped value of their own innate talent--it actually happens all the time. A genius of a planner may think, "planning is no big deal, but now...hard labor! Now we are talking." Kaboom, a classic quarterlife or midlife-crisis career issue is formed.
It's one of the biggest traps this career coach has ever heard of, and heard experiences with, again and again...
There's actually a kind of internet famous novel where the exact opposite happens. I don't remember the name of it or the site it's on, but initial premise has AI automating management (starting with pretty menial things like directing workers at fast food restaurant, and working its way up the value chain).
It was pretty interesting, but I only got as far as the protagonist getting interned in a kind of debtors prison once his savings ran out and he couldn't afford to live outside of it (due to all the automation). The prison was theoretically open, but if you wandered too far away, robots would tranq you (after very many "nice" exhortations to return) and you'd wake up back in your bed.
it actually ends on a positive note, it's trying to compare two possible ways society could respond to the automation revolution.
Yes, that's it!
> it actually ends on a positive note, it's trying to compare two possible ways society could respond to the automation revolution.
Yeah, but I stopped reading at that point. Partially because I had something else to do, and partially because it seemed like a utopian fantasy.
_Some_ knowledge workers may be unemployed. The vast majority? Reduce the overhead on tedium like data entry, writing e-mail, etc. They'll be able to expand what they can do, because AI models like ChatGPT could be trained on company data and used to reduce errors, etc.
If anything, middle managers will be the ones to get squeezed out when they no longer have work to do.
No code solutions have existed for years (e.g. Excel) and yet data scientists are still getting paid in the six figures.
AI Maniacs: "Ah all the jobs are gone, humans are useless!"
On the other hand, I can see in the future we will build complex programs we don't understand via AI as black boxes, and then use tonnes of tests, probably in a similar spirit to aviation industry, to ensure they do the right thing.
What I love most is creating stuff. I have little interest in algorithms, o-notation, and other minutiae. I’d love a world where I could just tell the computer what I want to create and it takes care of the minutiae for me. Though, I’m mostly thinking about personal projects. When it comes to writing code for work… well, that is not really in the ven diagram circle of things I love about programming. Maybe I’ll have to do something else for work, though I’m about 10-15 years away from retirement. It will be interesting to see how much things change by then.
First, this is an annoyingly unnuanced way to start an essay. Second, if at some point AI takes all the pain out of producing stuff (it does all the low level work in all sectors), the only valuable activity that’s left is art.
Currently, commence a multi-month design cycle, create the migration path, perform the migration in a slow and safe way, soft logging, feature flagging, gradual release, etc., and finally once everything is migrated, remembering to remove the old code. All the meanwhile turning down a bunch of other ideas for improvements you had because you need to get this one thing done. Or worse, interrupting a bunch of half-finished refactorings and leaving them half-finished and having to work around those things until some time that may never come.
Imagine if you could just tell an AI to do this, be reasonably certain it would do the right thing, and it's done and provably correct and manages your deployments and migrations for you. And you could be free to look into the next batch of new things you want to do.
I love coding, but really I love thinking about the ways to do things more than the processes of getting those things done. If there was an AI that could automate that work, that would be incredible.
His use of the term "shape rotator" kind of helps me articulate my thoughts on the matter. My theory is that attempts to democratize software authorship focus too much on the writing code part, but ultimately do nothing to address the "shape rotator" part, i.e. imagining the problem(s) you're trying to solve as abstract shapes. When I imagine the problem at hand, I construct a kind of mental model of it without really thinking about it all that much.
With that in mind, could our current approach to AI ultimately scale to the point where non-shape-rotators can finally write software as capably, or will it reach a plateau where shape rotators still retain their monopoly on software authorship?
If it's the latter case, then AI might not prove that much of a threat to programming simply because shape rotators seem to really like their text-editors, and really like to program
EDIT: some quick follow-up remarks, but despite feeling that AI art generators are unethical, I also feel at the same time that we're overestimating their potential damage because I think they democratize the wrong thing, which in this case is skill with a paintbrush or similar equivalent, yet they don't really democratize creativity and imagination. I think a lot of AI art is ugly, but I think that's largely due to the fact that the people using AI art generators simply aren't that creative or imaginative. By contrast, people who are creative and imaginative will understand that it's easier to get what they're looking for simply by a more direct, hands-on approach, rather than trying to find "good enough" through a prompt. Sure, people will get good-enough for free in lieu of paying someone more for better, but that's been an issue in the creative industry well before the advent of AI, and often with predictable results. Namely, good-enough doesn't typically create value so much as serve as a placeholder for something deemed essential, i.e. it's good-enough. A restaurant with good-enough branding, or a book with a good-enough book cover, is the same as a restaurant with no-branding or a book with no cover.
In fact, ChatGPT and other AI developments will probably enable people like me to create things I couldn't before - like video games and software that I have dreamed up but weren't skilled enough to do.
Now apply that to society as a whole, and we'll be able to be more creative with our computers than ever before - leading to a whole slew of new opportunities.
Knowledge workers will need to learn how to play with this new tool and use it to augment their work.
This is not the doomsday you are looking for. hand waves
When the printing press removed the manual tedium from producing a book, and reduced the cost by orders of magnitude. Literacy went up, and science and the arts flourished. Instead of manually copying texts, authors could write for millions.
When the Thread Cutting Lathe made it possible to make screws accurately and cheaply enough to standardize, the use of them went up, and the industrial revolution happened. Instead of making a screw in a day, a machinist could make thousands.
Once it became possible to store data automatically, the need for file clerks and others manually managing data went down. The management of information by computer made all manner of new business possible. Real time management of inventory and booking of reservations because possible on a vast scale.
GPT and other AI based tools will change the nature of our jobs, but we'll all have new things to do. It'll be less tedious, and a bit more creative, as it always tended to be before.
Project management, architects, business Analysts, on-premises devops/infrastructure.