Comment by stego-tech | Hacker News Reader

Comment by stego-tech | Hacker News Reader

Parent Full thread

The “they’re just jelly that we can do better than they with AI” camp really needs to spend more time hanging around artisans in general, instead of flouncing into comment sections and evangelizing the AI-booster groupthink.

Artists and creators are, broadly, incredibly pissed that their output was used to train these models without compensation or consent by trillion-dollar megacorps and VC-funded startups. That is, and remains, the core grievance. People who already make a pittance by devoting themselves to the creation of art are now forced out of art entirely because programmers just couldn’t be bothered to - GASP - have an original thought and commission someone else to execute it for them.

A distant, but still important, secondary concern is the quality of the slop itself (or lack thereof). Anyone who engages with art sufficiently can see the “seams” in generative content, even in state of the art models: perspectives lack consistency across key frames, anatomy isn’t grounded in reality or bends in ways befitting of a horror movie, geometry and materials that do not “graft” together due to a lack of negative space. These models are garbage because they don’t recognize core artistic concepts, only haphazardly reassemble pieces by prompt.

I challenge the AI crowd to actually go to an art faire, or commission a custom piece of your idea. Have something you had to contribute more than a simple prompt, to. Identify styles you like and artists that work within them. Take the chance to make more human connections and bond over shared creativity.

The artists will thank you, and you’re likely to enjoy the resultant output far more.

At this point, I just assume anyone who advocates for the use of AI is actually just an output from some AI. Given that "human-sounding speech" is the thing that AI is most easily able to produce, and how many different AIs are out there, and how beneficial an army of never-softening commenters can be for any specific pet cause you like, I can't think of why it wouldn't be statistically irresponsible to not assume that.

I've met enough real humans with completely self-important defenses of it that I know that they exist, so I'm willing to at least give them doubt. But the assumption is that they are AI and they need to prove being human. To assume otherwise is unreasonable.

I debated a bit about how to answer this, because I've seen this idea so much after stable diffusion came out. I have a serious answer, and a sarcastic one. I'll go with the serious one. The sarcastic bit was just replacing coders with artists in your text. You can imagine it, I guess :)

Why are "artists" special? Why did you feel the need to type these 4 thoughtful (but overdone imo) paragraphs, defending "artisans"? Why are they special, when compared to coders? Why do the artists get to use ever better tools designed to help them, but when the other side gets the same kinds of tools, it's suddenly faux pas? Is it just "AI hate" or is it something else? Can you at least see the double standards that you apply in your post, as I can see it from outside?

It used to be that games were coded by passionate people. People who knew how to code. They'd painstakingly find ways of making ascii characters do silly things on a screen that wasn't necessarily designed for what they were doing. Later, they started playing with pixels. But they were still coders. So they coded away until the pixels started doing funny things on the screen. You talk about "art"? Hah. THAT was art. The ability and tech knowledge to make those early systems do those things with pixels is something that we just don't see today. And we don't see it, in large, because coders did what coders do and made it simpler for anyone else to do those funny things with pixels on a screen.

At every step of the way coders built software to help other people. They built engines. Then they built harnesses for the designers, animators and so on. Then they built simplified engines. The endless RPG generators, and so on. Then they built "no-code" solutions. Here, friend, you take this piece of code, plug in your art and you have a game! And they were happy to do that, because it was enabling other people to do their thing. With code they wrote. And many of them free of charge!

But now, when suddenly coders have a tool that they can use themselves, to empower them with things that they couldn't previously do, now suddenly there's a problem? Why is one artist's output "art", even if the game code is shit, while the opposite isn't? Why can't a coder enjoy creating a game, with help from tools that do something they simply don't care about? They want to do the logic behind the things moving on the screen, and can't / won't spend time creating the art. Why should they be shunned? Why not enjoy the experience for what it is? Is it just AI hate? If so, perhaps you should disclose it. Dunno, this whole take of yours feels mighty high-horsey for my taste.

> Why are "artists" special?

There is a very entitled class of "artists" out there who are trust fund babies, where mummy and daddy pay for everything, their art career is for show and any money it generates is play money for them. In my personal experience frequently these are the ones that most commonly refer themselves as "artists" and are the loudest shouters on that side of the debate.

Many artists I know that struggled to afford paintbrushes or musical instruments or whatever while developing their craft seem less vocal, they are used to the idea that they are probably unlikely to ever be paid or reconised sufficiently for their work because it was like that for a long time before LLM's came along. Artists in 3D studios for example often work for way less pay than a lot of us would deem acceptable just for the ability to do what they love for a job.

The best artists I know that work professionally (eg. graphic designers etc) are pragmatically using generative AI in productive ways to streamline their workflow.

Not that there's no merit to the "theft" debate at all. I do feel like a lot of people claiming to have had their work "stolen" by big companies probably don't have a portfolio online big enough or noticable enough to have been included in the training data. But for those whose work (eg. prolific and talented indie creators on Sketchfab, Artstation, Soundcloud) they have a point and I feel for them.

I honestly have not idea what you're on about.

First, the "by artists for coders" equivalent always existed! There's tons of free-for-commercial-use art packs and BGM tracks and sound effect packs out there, and more when you add cheaply priced stuff. Will you get hate for using those common assets in a commercial project? Only as much as you'll get for visibly running on RPG Maker!

Which leads into the second - those "no-code" solutions you refer to are a far cry from "just add art". They're really "slightly lower code", relying on heavy scripting to actually shape the faintest approximation of a personal vision out of it. They were never the "by coders for artists" gift you frame them as, any more than Godot or Unity was. They're essentially just a pack of libraries for well-trodden genre boilerplate, used by hobbyist game coders and artists alike.

Artists have always needed to learn to code in order to make their vision for a game into reality. They equally cannot "enjoy creating a game, with help from tools that do something they simply don't care about" unless you want them to - wait for it - AI vibe code the whole thing. Or do you think all the artists nominally against AI art are secretly vibe coding a new wave of games too? Do you even think a vibe-coded game will hew to your expectations for a good game? If not, why?

There's a spectrum of human involvement in producing a thing, and art is possibly the last thing I want to see automate.

In the end, art is about human connection. There's a difference between an print of some generated AI slop found online, a painting made in a Chinese factory for a big store, and the scribble your friend made when they went through depression.

You can make a game with all three process. They are not the same.

While I can see where your argument comes from (because up until about ten years ago, I would’ve written it verbatim myself), I must respectfully disagree with it. Some programmers build tools to help people, but most do not. They build tools for surveillance, engines for advertising, exploit human psychology with patterns and site designs that deliberately hinder users, not help them. Most programmers never contribute to Open Source, but instead go to work for tech conglomerates to make money, because that is what society demands of us and coding - until recently - was a solid path into Middle Class status.

I question the sincerity of this narrative that the AI companies are doing this to “help” us, when their actions say otherwise at every turn. I also question that diffusion models and LLMs “enable” programmers to somehow create things others could do with a pencil, paper, and practice. I question this notion that a human must be able to be entirely self-sufficient through technology rather than cooperative with their fellow man, or that every skill must be commoditized for maximum extraction of wealth instead of respected as expertise within a community. I do not hate AI because to do so would be to hate a hammer, or a screwdriver.

Where the hate in my heart lies are towards those who demand we reduce humanity, its chaos, its ephemeralness, its qualia, to a mathematical model devoid of entropy. I hate that because these people - not the tools themselves - deign themselves superior men who are somehow above or immune to the fundamental force of reality (entropy), devoid of responsibility or accountability for their actions or harms.

A true AI booster should be screaming angry that this compute capacity is being squandered on shitty image generation and chatbots that convince teenagers to commit suicide or psychologists that they’re discovering inter-dimensional communication. These vaunted tools should be used to balance the economy, uplift the populace, hold the powerful to account, mediate disputes, improve outcomes in quality and longevity of life.

They are emphatically not being used in this capacity, and their owners have made it abundantly clear through their repeated actions that said outcomes have never been, and never will be, their intent.

And that is the source of my personal hate.

Hi, I have interacted with a lot of artists, and spent more on commissioning art than is probably financially sensible, as well as playing around with models for writing and image generation and I have a few thoughts on this.

I think, on the whole, the distaste for AI is primarily about a threat to the value of the artist's work. Importantly, I think the idea that this was done by training on their collective work is a bit of a sting on top but it is not the primary reason for the objection. Especially importantly, I think copyright is 100% not a good way to try to mend this issue, because it will primarily enable the parasitic centralization that already plagues the art business, as well as allow for further moat-building by tho ones creating these models (Adobe having already demonstrated this). In my view, a world where the big tech companies have models that only they are legally allowed to train is the worst possible outcome from this tech. I think addressing this either needs to involve some kind of blanket compensation from the big companies (with the important proviso that even their entire valuation spread amongst all the artists in their training set is a relative pittance), or through a general push against AI generation entirely, but from the perspective of the importance of supporting the artists as opposed to leaning on copyright claims which the AI industry can happily navigate if they must.

With regards to quality, Sturgeon's law applies doubly here. The vast majority of AI generated stuff you see will be slop, because it's so easy to make. It is possible to make very good stuff with AI with more effort, but this requires at a minimum some taste and willingness to put thought into what you want to get out of it, and better also some artistic talent. To me the best is when someone engages with it as a tool to achieve a vision as opposed to a perfunctory 'I need to fill some space with something' stock-image type thing (something which humans had already been doing, but were a bit more limited on because of expense and it's hard for someone doing art to not care at least a little bit about making something good even if it's utterly soulless corporate clip-art).

I'll also say it's not universal amongst artists. I know of multiple who are OK with it, and starting to incorporate it into their work. But it's also a somewhat dicey position to take publicly in those circles at the moment, so they're not very visible on the whole. (I suspect this is often dependent on why they got into art: in general the ones who are OK with or actively like AI are the ones who got into art because they wanted to see more art of the kind that they make (insert 'oh boy, two cakes!' meme here). The ones who got into art because they enjoy the process of making the art generally don't like it, though they're not always utterly virulently against it, and the ones who got into art for the status it affords them absolutely hate it. Though of course these are somewhat oversimplified categories)

Reply on news.ycombinator.com