AI Art feels kinda shitty.....

Everything else
51

This is so fucking not accurate!!! You have no fucking idea how artists learn!

I’m so god damned sick of some not artist saying “this is how artists learn” when it just fucking isn’t.

QT3, Reddit, Gamefound the whole damned internet is suddenly full of people who are suddenly “experts” on how artists learn even though they have gotten it all wrong.

I’ve tried to patiently explain why it’s not a 1 to 1 comparison, but the pro AI crowd just doesn’t want to bother listening and learning.

I am fed the fuck up at this point.

52

My guess is that visual artists learn by studying the elements of art, with reference to various artistic works and then practicing a hell of a lot. I totally understand that art is a skill that requires study and practice. But it’s entirely possible I misunderstand how artists work.

My take is:

  1. Technology is always disruptive. Farmers were displaced by agricultural machines. Farriers were displaced by the advent of the automobile.

  2. This technology is not (yet) at the point where it is going to displace any working artists. It doesn’t yet produce reasonable images to an exact specification and functions more as a novelty than anything else. I think at this moment in time, engineers and programmers are more threatened by the current state of AI than creatives are, but I also (as an engineer) don’t feel particularly threatened. And actually If I were a working composer or technical writer I’d be more worried than if I was a visual artist.

53

So a way more accurate way to implement the AI would be to give it a bit of basic art training and then tell it to go look at the real world and represent it as art.

I practice everyday, but I don’t practice by going online and stealing from other artists, I do it by studying real life. For convenience I have bought a lot of model packs of photographs, but I also draw from live models and take a lot of pictures on walks.

I use a lot of my pictures taken on walks for color studies as I am working on my painting.

I would say approximately 10% of my serious adult learning has involved going on the internet to reference other artists and that was very early in my training. I avoid it like the plague now since I want my art to evolve into it’s own thing.

If you want your AI, teach it to draw give it a bit of a frame of reference and then send it out into the world to draw from the real world. That would actually be really interesting to see what it does.

There is such a massive misconception in all of this as to what artists do and what resources they use. I have literally written this same thing so many times yet you fucking idiots can’t fucking read it I guess???

What the AI is doing is called swiping and it is really pretty amateur bullshit. It happens a lot in non commercial fan art. The AI does it better than Human artists, but I’m not sure being better at swiping is actually a triumph. I’ve said this a million fucking times but I think the AI bros have robot dick stuffed so far down there throats they can’t read it. The AI is playing by fan art rules which is fine for non commercial work. not so much for commercial work.

This is the third time I have basically written this out in this thread alone and I’ve written in dozens of time in the last few days and every god damned time fucking AI bros just go “Duh, artist learn by going and copying other artist online. Machine do same”.

54

You learn things the same way that humans learn everything else. (I’m not trivializing all the practice and work that goes into what you do. For background, my sister is actually a professional artist, so I know all the work that she put into developing her portfolio)

Take the dragon example I presented above. How do you draw a dragon? Where did that image come from?

It came from you observing dragons that other people drew.

55

(post deleted by author)

56

Ah, cool. You do you then, dude.

58

The change is not always positive though. British cloth weavers, who were solidly middle class artisans, were replaced with a factory system that was so brutal and immiserating that it is still famous 200 years later. Technological disruption can easily mean that you are now among the superfluous class that is free to starve underneath a bridge.

begin rant
Right now it is the artists who are getting their face eaten by the leopard. I think the correct response is to have solidarity with them, since our own face-eating moment will arrive in due time. We are all on course to have a morning where we wake up, check the news, and realize that the skills and knowledge that we traded for money are no longer of value to society. This is why I’m in favor of an automation dividend/UBI, since otherwise we’ll all just end up living under a bridge/being liquidated by some billionaire’s robot dog.
end rant

59

100% agree

Are they? Are artists losing their jobs to AIs? This isn’t a story I’ve seen reported.

Yuval Harrari talks about exactly this problem in his book Homo Deus and how inadequate our political/economic system is to address it.

60

Short answer:
I’m not sure

Long rambling answer:
Well, when I want to show my Friday gaming group a picture of a new monster, I use Craiyon to generate the image. And if I was doing something slightly more serious & bundle their adventures together to publish for free on itch.io, I would try using MidJourney before trying to hire an artist. So that’s 1 job they missed out on. That is just me, I have no taste and am unable to tell the difference between real art and AI-generated goop. But I think there’s probably a lot of philistines like me out there who would turn to AI or proc-gen art these days rather than going to a professional. I notice it too in the indie 4X games I play, where you are much more likely to see an proc-gen/AI-generated portrait than one drawn by an artist.

I would posit that the situation only gets darker for real artists as the AI models and tooling improves, and that if a dire prediction is not true now it will be true N months or years from now.

I would also theorize that it’s the low-end and beginning artists that are going to be hurt the most the soonest. There will probably always be a place for the best-of-the-best artists, and the AAA projects will shell out money for them. But I wonder what it would be like if the next generation’s Vic Davis was making Solium Infernum in 2025. Would he go with AI art rather than contract artists? On the one hand, some of the custom art for Solium is pretty baller. On the other hand, it seems like AI could do chaotic hell-scapes pretty well. And (if I’m remembering correctly) Vic has mentioned that one of the reasons he quit gamedev was the amount of effort needed to coordinate with the independent contractors. Maybe things go easier for him if he is just typing prompts into a model while he waits for his ActionScript to compile? Actually wait wait, Solium Infernum is a bad example, look at some of his other games like Occult Chronicles where the art is just serviceable rather than stellar, and you could see how small teams like him could make their games with much less demand for artists. Basically for anything on the indie/low side of gaming, where margins are tight and managers are cheap, it is going to be very appealing use a solution that lets them cut out ~40% of their workforce.

61

While I think the AI generated art is interesting, it has a look to it which feels heavy on abstraction and filters. It’s like a blurry attempt to get to the stylistic end result of the product without wasting time on the middle. And while I have an intellectual curiosity for the subject, there’s no way I’ll pore over any of the images I’ve seen, compared to the hours I can spend with artists’ sketches and paintings.

People have been trying to escape paying for artist fees forever. Look at children’s books… I feel like more than half of them these days use photoshop pattern filled collage-like shapes to describe their characters and stories, rather than traditional pen, ink, and watercolor, and they’re much poorer for it; I actively steer away from any of them when choosing to read to my kids.

Artists use computer shortcuts all the time, and not just time saving ones. Procedurally textured brushes to get traditional paint texture, motion capture for animation, image libraries for textures, 3d posed models to rotoscope into comic panels. If anything, it’s lowered the bar and allowed a lot of mediocre, low-fundamental skill, technical artists to get through the door and find paying jobs, and expand the range of what we have out there (which I think is a good thing).

Regardless of the legalities involved, I really doubt this stuff will replace the desire for human made art for most art directed, paid gigs, beyond some initial novelty products. Maybe it’ll wind up cutting some of the time consuming craft, the same way printing 3d models put some sculptors out of work, or forced oil painters digital.

But as an avid consumer and sometimes producer of art, I’m more inclined to think it’s created a new novelty medium or style, rather than replacing traditional old ones. In which case, I wouldn’t be too worried (yet).

But I guess I would be angry to have my stuff ripped off, but then, I’m guilty of taking all kinds of things online and reusing them in presentations etc. Tricky thing to crack down on.

62

Yes, I agree — this would be “intelligence”. What is currently happening is not. (In my opinion, which could change, etc.)

63

https://blog.quintarelli.it/2019/11/lets-forget-the-term-ai-lets-call-them-systematic-approaches-to-learning-algorithms-and-machine-inferences-salami/

Because of this misconception, we proposed we should drop the usage of the term “Artificial Intelligence” and adopt a more appropriate and scoped-limited terminology for these technologies which better describe what these technologies are: Systematic Approaches to Learning Algorithms and Machine Inferences.

Now we have redefined the name, will we still support the idea that SALAMI will develop some form of consciouness ?

64

Yeah, when I think of how digital process has in some ways lowered the bar and almost helped bring about a cultural “house style” at this point it is kinda a bummer. I think as that style becomes devalued by the flood of AI art I’m hoping that traditional media and more bespoke digital art styles gain more value. We’ll see.

I’ve definitely been chatting with my colleagues about that for sure. That and how to build communities around sharing art in ways that can’t be scraped. That way artists can innovate without feeding this machine.

I don’t really think AI art will wholly replace commercial artists (though it may thin their ranks considerably), since in order to get really good results from it you need and artist at the wheel. My real fear is that professional artist will come to mean AI pilot. Blegh. I find that prospect repugnant.

On the other hand, it’s possible that AI will go in another direction and instead of just swiping whole cloth from the internet maybe artist will get personal AI assistants that they train themselves and grow with them over their careers. That could be interesting too.

Who knows. What I do know is that I find what it is doing now pretty distasteful. The total unwillingness of AI proponents to acknowledge that there is a world of nuance between what the AI is doing and what artists do is pretty frustrating as well. It’s like they know there is an issue in there so they refuse to acknowledge the nuance exits. In the end I’m sure it’s about them wanting to have this thing up and running for commercial work as soon as possible.

65

For my current pathfinder game, I use thispersondoesnotexist to get headshots, and then a headshot-to-painting AI program to get drawn character headshots for NPCs.

I guess I really think they are cool tools for throwaway stuff. This isn’t taking an artist’s job away, because the alternative would probably just be google image search and stealing artwork from deviantart. If I ever wanted to publish any of this publicly, I would probably use those generated portraits as something for a real artist to use as inspiration.

But, this isn’t a published work, just a gaming thing for me and my friends.

I like it as a tool for a non-artist person like myself to cover this part of the GM role.

But I also think that AI art generators using real living artists images for creating their algorithm is icky. I guess if they are using dead people’s art that is better? They aren’t really stealing from anyone trying to make a living at that point?

66

Honestly I think this is great. When I think of fan projects and how AI art could be such a positive impact on that I think that’s pretty cool.

Fan art has always been pretty open season and I see absolutely no issue in the way the AI does it’s thing in terms of that.

And yeah, non living artists work and public domain works are certainly fair game for anyone and anything in a commercial setting. If the AI programmers had found a way to limit to that this entire discussion wouldn’t be happening at all, at least my part in it.

67

We already have that, it’s Youtube.

yeah,exactly. Pretty much every film for us begins with the studio bringing us a brief and a mood board with a bunch of stuff downloaded from the internet. These are people who will make tens of millions of dollars on their product, and they’re not paying the artists of those concepts/photos/videos squat.

But what they do then is, using those references, start to hone their vision using paid artists (of whom will also hit the internet to build their concepts… I’ve had a few kickbacks from prod for watermarks on my background textures, just in case they reach marketting). I don’t see this part getting replaced by lumpy AI mishmashes anytime soon.

68

I actually didn’t even think about it, but I also felt bad using real people’s faces with the AI art generator too. Like, I was stealing a real person’s image.

I wanted it to be completely AI working with AI all the way down. (Though thispersondoesnotexist dot com is using real faces to generate fake ones)

69

Do you feel bad using photo reference for a drawing? Or a live model? What’s the difference?

70

Well, I don’t know about how an artist would do things.

But I would be worried that my AI art converter would come up with something that looked like a real famous person, and I would feel that would be derivative.

I am sure a real artist could be given reference photos and come up with something that was not derivative looking.

71

It just depends on what your brief is. If I’m storyboarding a sequence with Robert Downey Jr, the client wants to know who’s who on the frame. I’m not going to squint real hard and try to work off my own memory and impression of what I think Downy Jr looks like, I’m gonna just copy his face from photos or film.

If you’re trying to draw a generic DnD rogue character and copy RDJ’s face super realistically from a photo, well, be prepared to get called out for it.

What’s a ‘real’ artist, anyway.