The AI revolution is rotten to the core [video] | Hacker News Reader

The AI revolution is rotten to the core [video]

youtube.com

29 pointsdownboots26 comments

I'm not a native English speaker and youtube was very bad at transcribing videos, now thanks to YouTube speech-to-text model and translation model powered by neural networks the subtitles are actually very good and people can enjoy the content on how "The AI revolution is rotten to the core" from around the world, thanks AI!

My whole job is generative AI research at this point, so I am obviously biased. And yes there were quite a number of technical points that fall flat for me, like the stuff about hidden layers, the comparison with photography (you can color grade AI art in Photoshop too), the usual "no understanding because it just outputs the next word" argument, the time wasted by focusing on short-term technical limitations, etc. But despite all that I was very compelled by the discussion of how AI tech is primarily being used to further squeeze everything enjoyable out human life and maximize corporate profits. It helps put words to my growing dissatisfaction with how my field is being monetized. And it isn't enough to say that people will keep doing enjoyable stuff as a hobby. I made a career out of doing creative programming work that I love, and will really struggle to sleep at night if I help to take that opportunity away from someone else.

I mean, sure, but that's not to be blamed on ML-anything. It's explicitly the human problem, because against all odds, it turns out that we have been the paperclip optimizers all along.

Yeah, it can go awry, but so far I don't see the problem. And while I appreciate that certain people will actually suffer for the fact that they get rationalized away, I'd rather superfluous workers get to live on a hypothetical monthly allowance (huh, wonder if we will get THAT right) and do whatever than not allow, say, quadriplegics and the other many disabled people around us to finally tap into their creativity and churn out art, speech, even experience a motor-sensory immediacy they have long ago lost any appreciation for.

The benefits so massively, so drastically outweigh the negatives, starting with more fulfilling artistic creation (at any granularity you want, people pretend like it's only text-to-picture or nothing at all, which is such a silly misunderstanding of what we will enable ourselves to do), round-the-clock therapy, tutoring, mindfulness instruction and help... you name it. Some foul fields and workplaces (yeah, it's not the AI revolution, it's all of humanity and their history with being pricks towards humanity itself that is "rotten to the core") will have to be demolished and cauterized before it can get better, but it's going to still be an overwhelming net positive. Of course, that is if it/we manage to actually amputate the gangrenous limbs of capitalism and bad incentives, bad workplace culture, atrocious treatment of fellow humans - just name it, chances are we have been doing it so, so very wrong.

I reserve skepticism, it's a huge gamble regardless. But just in case people are wondering, we're already beyond the point where this is just merely handy, at this stage, a Hail Mary is pretty much our only option... and there won't be another one like this here "rotten" ML-revolution.

More hand-wringing over technological progress, as usual.

Are artists going to stop producing art, or will they use AI as a tool to produce entirely new styles and forms of art?

I see it as similar to the hand made vs factory made debate. There’s room for both, and the vast majority are happy with the factory product.

Generative AI will make it easier to get new ideas to the masses affordably. That could yield some really interesting results. There will still be a demand for art like Miyazaki animations or similarly painstakingly human generated works of art, and I expect it’ll live on as well.

Thanks to machines, I can afford to own some crazy technology relatively affordably. It’s okay that it’s not made by a human, because it allows this human to engage life in more ways. I can build things with wood that would be virtually impossible for me 200 years ago, but I can do it in my spare time over weekends. I can own an aquarium with plants and animals from South America, which is an immensely gratifying hobby for me, providing endless appreciation for life and nature when I’m away from it.

I think generative AI will eventually accomplish something similar for people. They will be abstracted from the experience of life in some sense, but ultimately may allow them to experience it more in other ways. It could awaken in many people the idea that they can imagine and bring things to life. Letting their ideas out and into the light could be a massively liberating experience! How many people would probably love drawing if only they could realize that they truly do have ideas and visions worth bringing to life? That it really can bring them joy?

There’s plenty of room for doom and terror, but that’s too obvious. Other forms of automation in the past were no different. What I find more interesting is how to find the good in it, because like it or not, it’s here.

Yeah, my take as well. I think "Her" portrayed it nicely; you can still get your human-written poem or whatever nonsense you want to be made by hand, doesn't mean everyone has to dispatch atrociously meaningless or repetitive tasks. What, do people think VFX grunts love rotoscoping mundane subjects all day long? People hate boilerplate work, and getting rid of it will be a huge boost for enjoyment, too.

There truly is room for doom and terror, but none of those wannabe essayists even remotely scratch the reality of it all; they are almost exclusively trying to dunk on tech that will help disabled, impaired, destitute folks to even remotely catch up to the rest of those who are living a relatively cushy life. Just look at therapy and what the sheer possibilities are, it's so incredibly callous in my mind to pretend like it's all going to be so bad when we have tangible proof that this is going to be positively changing billions of lives in the shortest amount of time imaginable.

There is so, so much good here, and I'm getting kind of tired of the narrow-minded naysayers. I mean, they're kind of few, all things considered, and they'll instantly vanish the moment we get really competent agents doing anything... but it's still exhausting being aware of all this all the time. I guess AI/ML can improve my own personal life by just helping me sidestep all the bickering and all these awful non-arguments I keep stunlocking myself on, which would be pretty funny to me.

You don't need AI to make art. You need AI to take some low-fi tasks off your plate so you can make more art.

You don't need AI to make art like you don't need a camera to make art. But it's likely you can use new tools to make entirely new types of art.

It's also likely new tools will make certain types of art a commodity. People used to need to paint portraits, now they can just take a photo. There will still be some demand for the hand crafted art that AI can produce but it will be diminished.

I expected this to be a criticism of OpenAI and the whole generative AI ilk, and while that's somewhat related, this video mostly focuses on the shortcomings of AI generated worlds, art, content and dialogue in the popular Minecraft game. The speaker does delve a bit into the shortcomings of generative and other AI as applied to the real world as well later though.

Skip ahead to 3:39, the Minecraft part ends there. Everything up to that point was generated by ChatGPT (as explained later on). Don't worry, the rest of the video is about AI

Sort of about AI.

It’s mostly about law, copywrite, bias, and becsuse these big companies don’t know how to use it yet that it probably sucks.

I disagree on some things and really don’t care on others. We’re due for a change from historical copywrite concepts, the old models are starting to develop holes.

Honestly if using AI to summarize this hour+ video means being able to efficiently grok the content without suffering through dozens of mindnumbing YT ads, then it's not really that rotten.

Agreed, like any technology, AI is going to enable both good and bad solutions. I certainly feel that generative AI has made life easier in some ways for me, summarizing long text is a super useful one.

I do agree with the author that generative AI in gaming and other content lacks the polish that a human creator brings to the table, but there are certainly some good use cases for AI.

It’s really simple. If you want to sell your art, AI can produce it for you. If you like making real art AI can recommend which watercolors, brushes and clay to buy. Art as a product will be the domain of AI. The fun and awesomeness of standing in your cluttered atelier and creating something with your hands is for humans, enjoy.

Not if AI steals your job, then your backup job, then everyone's plan C has 1000x more applicants than openings, then rent doubles again, you can't afford art supplies, but you can watch YouTube videos of other people doing art (with ads) and ChatGPT can tell anything you want about art supplies you'll never be able to afford again.

sure, but at this point, people are going to just freaking demolish server farms by hand

I'm not sure how we'll handle the income problem, but once the shoe drops, and drop it will, very soon, I have a feeling that core existential needs will "magically" align with policymaking, such that people will at least be happier than they are today. I mean, this is the big one, the biggest unknown in all this. Not some future skynet nonsense, us humans always trying to keep others down because of our fragile, hilarious little egos. If we can overcome ourselves and make sure that we all get to eat and enjoy proper healthcare (uh, only way is up with that one anyway), people will change their minds right quick.

High school project?

Nice video. Seems to be getting the HN commentariat predictably flustered...

Did you post this as a joke?

Too long; didn't watch.

Didn’t miss much. It’s very long and mostly complains about copywrite and applying traditional ethics.

Traditional ethics? Is there some modern non-traditional ethics these days as well?

Read it together, traditional copywrite ethics.

But what's the traditional part? Training models with other people's work without credit or pay is ok by some new form of ethics now? It's almost like you want to say it's unethical to do so, but want to side step the issue altogether at the same time.

The way we have traditionally thought about copywrite has not necessarily taken into account the immediacy and effortlessness of derivatives.

It’s just different now, and we don’t really understand what this is going to mean going forward.

The irony being that it would have massively benefitted from a huge range of ML tools, be it the crusty audio or the atrocious pacing. But I mean... same deal as every day, pop-luddites squeezing their audiences for subscribers (another bout of irony because they accuse "AI" to do the same, somehow). It's still fascinating and good to see that we get some lengthy essays about this, I can't wait for us to discuss how well it's aged in one or two years or so.

Reply on news.ycombinator.com