AI has been going through hype cycles for about sixty years. I’m an old-ass software developer who’s been through enough hype cycles that I’ve lost my rose-tinted glasses. This will be the continuation of capitalism’s innovations in labor automation that’s been going on for centuries.
Angela Collier: AI does not exist but it will ruin everything anyway
Every time I hear something like this I remember this phrase by Arthur C. Clarke which seems pretty good: “When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong.”
Just because you can see all the ways in which it’s not going to work doesn’t mean that we can’t find a way around it.
Thanks, I’ve never heard this rebuttal before, except in every previous hype cycle.
But a fiction author said it! That means it must be true!
The web3 bubble cratered not ten seconds ago, but here we are again.
But this time my AI ape girlfriend will be able to love me back! (in 10 years after they complete the next firmware update)
And what kind of big discoveries were there on the last hype cycles that make it like what is happening now?
I’d say the Reflexion paper by itself already sets what is happening now apart from any other time, as it shows that inteligence, or something like it, can be achieved with what tech we have, even if it still needs some polishing.
I’m not sure what you’re referring to. Maybe this: “Reflexion: Language Agents with Verbal Reinforcement Learning.” Whatever you meant, I haven’t read it.
Innovations in narrow AI will of course continue, and even accelerate in the short to medium term, given that the Eye of Venture Capital is fixed upon it right now.
But also, there is insane amount of hype, smoke, and mirrors about such innovations right now, and it is coming from the very same capitalist propaganda engine that hyped up blockchain/crypto/NFT/metaverse. It’s very convenient that all those graphics processors that were pumped out for crypto just happen to be transferable to neural networks. What more could the likes of NVIDIA & AMD ask for?
A lot of already insanely rich people in the tech sector stand to collect a whole bunch more by convincing dumb money (the wealthy people & institutions who fund VC firms) to invest in whatever new hype cycle they manufacture. Marc Andreessen of Andreessen Horowitz is not an honest person, he’s a P.T. Barnum for his VC firm. The people who just got kicked off the OpenAI board are of the same unhinged longtermism cult as Sam Bankman-Fried. Sam Altman has a net worth of half a billion; do you take his statements at face value?
If you want some dose of reality, I recommend the Marxist analysis podcast This Machine Kills, which focuses on technology trends. Every other episode is available for free from various podcast streaming outlets, like for instance Podtail.
But also, there is insane amount of hype, smoke, and mirrors about such innovations right now, and it is coming from the very same capitalist propaganda engine that hyped up blockchain/crypto/NFT/metaverse.
I don’t doubt it, I mean, Google themselves put out a presentation about their new AI a few weeks back that was fully faked, but there is still quite a lot of progress in the open source camp with ever better models appearing. I helped a little with one of the open source projects some months back and although that one might not have been the best thing it still had some results that were built upon. Anyway, open source is stil not to the point to threaten OpenAI but there are already many people using it for helping with work, for example.
It’s very convenient that all those graphics processors that were pumped out for crypto just happen to be transferable to neural networks.
Not that transferable from what I’m seeing especially as the best ones are needed for training new models while you don’t need as much to run them locally. What I have seen is more about new chips made for AI/Transformers, and in this case the newest and most advanced ones are being sold to the big tech companies at really high prices. I doubt they would be paying that much if they didn’t think it was needed, although to be fair, as long as it makes a profit it doesn’t need to actually be an advancement or anything good.
Why is this community showing up in my feed? Oh, it’s lemmygrad? Let’s not
over AI here.It’s a marketing buzzword, a meaningless label used to dupe people into supporting tech in hopes that it eventually does all the fantastical things that it is promised. People (like myself) are losing jobs due to this hype, not due to the AI actually being able to do our jobs better. Only because business owners are given empty promises about what it “could” do rather than what it actually “does.”
This comment from UlyssesT sums up my thoughts pretty well (The insults aren’t directed at you at all, it was part of a very long string of comments with a very unpleasant lib, but it’s too funny to not share the whole thing.):
LLMs are not “AI.”
“AI” is a marketing hype label that you have outright swallowed, shat out over a field of credulity, fertilized that field with that corporate hype shit, planted hype seeds, waited for those hype seeds to sprout and blossom, reaped those hype grains, ground those hype grains into hype flour, made a hype cake, baked that hype cake, put candles on it to celebrate the anniversary of ChatGPT’s release, ate the entire cake, licked the tray clean, got an upset stomach from that, then stumbled over, squatted down, and shat it out again to write your most recent reply.
Art AIs are actually competent and shall revolutionize Art-related industries (animation, video games, etc).
Text generators are so far not good enough. But I think it still makes some impacts, mostly for the worse:
- Your customer support will become worse as companies believe text AIs to be substitute to actual humans.
- More AI bots to spam your social networks and emails.
- More people will lose jobs over nothing.
- More shit code written by lazy programmers who don’t bother to double check AI results.
- Another tech bubble.
Art AIs are not competent. They are just a collage generator with a blend tool. They can’t create anything truly new, only an amalgam of pictures they’ve been trained on. They steal art from real artists to do this training. There is no learning process here, the only improvement they can have is through stealing more and more art and getting better at replicating hybrids of various pieces. If they “revolutionise” art related industries, all the art will just be derivative of existing art, there’ll be no soul, no deeper meaning, it will all be empty and shallow.
It’s interesting you mention video games, because I played a game recently that used AI art for character portraits, and they were all lacking any sort of charm, it just vaguely felt like a mugshot image, there was no personality displayed, nothing about their character, just a picture of a person with nothing behind the eyes. The game was a rip-off of another game that literally used characters with circles for heads and dots for eyes and those characters were far more compelling and likeable despite looking technically less proficient. It was an art style with charm, something that art AI can’t actually emulate, it can only try to imitate.
You’re speaking in term of “soul”, I am speaking in term of industry. Industry mass produces. Industry seeks efficiency. Industry seeks profits, not “souls”.
By “Art-related industry”, I am not referring to the few commissions on Pixiv, Twitter, and the likes. These independent artists are not industry.
Ah I see. Yeah, I think we’re in agreement then, it’s much cheaper to have an AI generate a “good enough” bit of media for people to consume than it is to hire artists to actually make something worthwhile. The problem is, those independent artists become industry artists, no one makes furry porn commissions for a living because it’s their lifelong dream. So with fewer artists making art, there’s less art for the AIs to steal from, and they will start to imitate from each other instead of real art. We’ll get the art equivalent of running a sentence through google translate two dozen times and the result will be incomprehensible gibberish. I don’t know if the techbros making this happen actually have a solution for this problem, as it is a societal, not technological one, they can’t just solve it by fine tuning the algorithm. I guess if this AI future comes to pass, except to see every single bit of mass media feeling exactly the same as every other one, just with a vague sci-fi backdrop replaced with a vague fantasy one depending on what trends are projected to be the most profitable.
Now that I think about it, it probably won’t be all that different from how the AAA game industry and Hollywood do things already. Sucks that it means that “furry porn commissions” is the furthest my career can go now though.
There’s nothing that prevents the industry from hiring traditional artists to invent a new art style and feed them to the trainers.
Art AIs can never replace the logic and soul of humans, but they will certainly help copy pasting them easier.
Art AIs don’t make artists obsolete. They will however makes the current method obsolete.
But why not just save a step on that first point and just hire the artists to make the art?
I don’t think this will make the current method “obsolete” so much as it will replace it with a cheaper, less meaningful method, one that can mass produce art the same way industrial machinery can produce chairs instead of master artisans making them by hand, but art isn’t something that benefits from that. Companies can benefit from this in the short term, but I can’t imagine people really loving AI generated content as it would all be empty and derivative, even now people are turning away from a lot of traditional entertainment because it is so formulaic and AI art will just compound this issue, it might be cheaper for companies to make, but that doesn’t mean people will buy it in the same quantities as existing media.
But why not just save a step on that first point and just hire the artists to make the art?
Imagine you are an anime producer, do you want to hire an entire studio of illustrators to draw every single frame to produce some 360 hours of 24fps animated video, or do you want to hire a smaller team of artists, make them draw a few key frames, construct an outline, and use an AI to mass produce 1200 hours of 60fps fluid animation? (Of course, the current tech is not there yet, but it is theoretically possible)
For your second paragraph, I feel like you are defining “art” as “art for the sake of art”. But in reality, arts always have a purpose. If not for entertainment, then for sending a message. Arts for the purpose of profit alone would indeed be hollow and soulless (like NFTs for example). Arts that entertain are less hollow, but their value is fast diminishing (like most mainstream entertainment). The most impactful arts are ones that send messages, they will be studied by academic and scholars for centuries to come for their historical relevancy. Regarding AI generated arts, if it has a purpose, it is art. As you can see, hollow art need not to be AI generated, and AI generated need not be hollow.
I suppose you’re right on that first paragraph there, but again, in this hypothetical future, why would anyone spend the time learning to make art when an AI could produce things far more technically proficient than anything they could while learning? If art is truly a dying field, then I would imagine that artists capable of designing work for the AI to play with would be in short supply.
I’m not at all sure I agree with your final paragraph, “art for art’s sake” is an empty phrase, art is about purpose, meaning, humanity, and AI art cannot have purpose, because AI has no motive to make art, it just makes what it is told to do. Even an artist doing furry porn commissions adds their own unique flair to it, that AI art cannot do. To repeat myself, it can imitate, but it cannot emulate. It cannot understand why something makes art good, and the tech guys making these AI art programs are not artists themselves, so they can make something that looks technically proficient, but has no greater meaning behind it. It is hollow, it will always be hollow, as it has none of the humanity that art contains. At best it can have “meaning” in the same way any big checklist based blockbuster film does, something technically proficient, but not something that sticks with the audience, that allows them to reflect or ponder, just big noise. It’s like pop music, it’s getting increasingly “designed” and “constructed” instead of “created” and the result is something technically proficient, that gets stuck in people’s heads, and is often quite pleasant to listen to, but has no deeper substance, it’s all surface level. People do like it, and plenty of pop stars have huge fan followings who adore their work, even if it isn’t “their” work so much as it was songs designed by committee.
But I don’t think everyone is only interested in the shallow surface level versions of media, I do believe people want more. To go back to the furry porn thing again, I have had people genuinely touched by some of the work I’ve created, even in that field, I’ve helped people in a way that AI art cannot do, AI art cannot inspire, it cannot provide meaning, it cannot even be open to interpretation like real art, as the “interpretation” is never “what did the artist mean by this?” but always “what prompts did they use to get that result?” Why should anyone care about an AI art piece someone else “made” when they could just copy paste the prompt and “create” it themselves? Perhaps this could lead to an unexpected result, where if the tech becomes easy enough to use, regular people can create their own animes, entirely without studios, entirely generated by AI, and if that were the case, the studios would struggle to compete, unless they could somehow produce work that is higher quality than what personal use AIs can generate, which would probably loop back around to needing a professional artist’s touch once again.
EDIT: I’m finding this to be a very interesting discussion, I’m not trying to argue with you, I’m just quite enjoying talking about this with you, so I’m hoping I’m not coming across as too abrasive)
From what I’m seeing it seems possible that by 2024-25 there will be quite a lot of unemployment (at least in some countries) due to AIs being able to do more and more jobs or at least massively increasing productivity to the point companies using AI would have a massive advantage, getting rid of the companies still hiring large ammounts of people.
Also, as robots would take longer to appear this unemployment would start with office and WFH workers cascading to less demand for jobs that would support these people and thus to a lot of competition for jobs and ever smaller salaries compounding the situation and leading to massive social changes (with hopefully the people realizing how good an idea taking the means of production is).
Maybe not 2024-2025. Large language models and bots that copy art from the internet are cool and all but they’re not yet capable of matching humans and probably won’t for quite a while.
In some cases they can match and even pass humans, as there have been some research showing it, and even in the cases where it can’t yet do so, having workers using AI could still increase productivity to the point that humans aren’t as necessary for completing the jobs. And if the big corporations don’t start using it startups are likely to arrive to do such jobs, like callcenters which already have AI systems that can do it to some extent and might be fully automated by the time the infrastructure for using AIs in such areas appears.
A calculator can perform mathematical tasks much faster than a human, does this mean a calculator is an AI?
No, but the AIs that already exist that can use calculators and other tools to solve problems are AI. They might not yet be AGI yet but as new systems using LLMs and other techniques develop it shouldn’t take long until they can identify a problem, calculate a solution and act on it by themselves.
But they aren’t identifying problems, they’re algorithms. These aren’t true “AIs” and never will be. They’re just designed to give a response based on input, like a calculator. They might get better at giving that response, but the way the current tech is headed they won’t ever actually create an “AI” just a mechanical Turk designed to look like it passes the Turing test.
Here is a simple example of what AI can already do. It might not be possible to do it all yet in a sinlge system yet, but with current tech as it gets deploy the following should be possible.
*First AI system receives a message from someone that there is a job to be done somewhere. By using it’s “intelect” and a program it sends a robot to solve the problem.
*The robot is an AI system that has AIs that identify the enviroment around the robot, another that works with programs to creat a route, another that actually drives the robot with human like movements on which it was trained.
*The robot finds a path until the problem and then it visually identifies what is what, a 3D model of the enviroment is made and an AI identifies a solution all based on the videos it was trained and the things that all robots have experienced.
*Then the AI that moves the robot interacts with the enviroment in the way that the was deemed to solve the problem and the AIs that were looking at the problem make sure it is fixed.
All these things can already be done by LLMs/AIs to varying degrees and looking at the progress from this year alone it seems likely that all of these things will be doable soon (this decade) and then it’s just a matter of producing more AIs/robots.
But that isn’t AI. That’s my point. Is it automation? Yes. Absolutely. But it isn’t thinking for itself, it is following instructions. It has no ability to deal with problems outside of the areas it is “trained” in. It can’t invent solutions to problems, only follow the instructions for solving a problem it has been given. It doesn’t “think” any more than any other program does.
I’m writing a visual novel game right now. In this game, the characters interact with the player, they respond to questions, have their own preferences and make their own decisions. But no one would think a video game character is actually thinking for themselves. They’re just lines of code made to emulate a conversation with a person.
And you should be very cautious about accepting any claims of “this amazing sci-fi tech will be here within the decade” because the newspapers have been making those claims for as long as we’ve had sci-fi. Someone else in this thread mentioned AI being hyped up for the past 60 years, and that entire time it’s been “just around the corner.” For 60 years. Don’t fall for marketing hype and buzzwords. Are robotics getting more advanced? Yes, they are. Automation is increasing as well. These are not the same thing as actual artificial intelligence though.
They may be able to match and pass humans on specific tasks, but most actual jobs require an element of being able to balance workloads, solve problems and think analytically, humans will still have to play a role for a long time, though we will gradually grow more and more dependent on AI which may(probably will) lead to a few layoffs
