This technology is currently putting people out of work and it won’t be getting better as developers continue to refine these programs. Concept artists are the most endangered by this as the companies that utilize their services have already begun trimming departments down. A couple of artists with a few machines can turn out as many (if not more) works than an entire team could.

There is a bit of solace to be found. Even if you ignore the legal precariousness, these programs are far from perfect. It takes time and effort to prompt the computer to get exactly what you want. Even after the ‘art’ is generated, it still requires a degree of touching up. Not everyone has the skill set to do this, but… the gap between semi-professional artists and everyone with a computer is beginning to close.

The threat of automation is here and it isn’t going away. Professions that people assumed were safe are not. We’ve needed to have a serious conversation about this for years, but we won’t see anything actionable until it’s far too late. Much like a number of the other extremely predictable assumptions that academics have been warning us about for decades.

If you want to do something against it, make certain not to buy into companies that are trying to run a profit on these programs. Not only will you be slowing the progression of digital art automation, but you’ll also be saving yourself cash. The majority of those paid ‘programs’ are really just a branch of Stable Diffusion…which is a program you can download to your machine for free.

But also mixed in there is the desire for it be driven by consent and/or public domain when training. It’s emerging tech, and I’m in tech, so I felt like I had to at least see what the state of it was. A lot of missing fingers. A lot of extra fingers, eyes, limbs, and other horrors, but every so often it would spit something out that I really jived with. And I was having fun with it. But that kinda came with guilt.

Like, who actually was sourced on this picture I liked? It doesn’t tell me. I can’t seek them out and ask for a commission. A few mashups of anime images, probably sourced from scraping the entirety of a website.

So, yeah. It’s cool, and I get excited about it, but I want a way to engage with it that makes me feel less sleazy.

Otherwise, I don’t think it’s possible to put the AI monster back in the box and I don’t want to ban any kind of technology.

So I wasn’t going to bring that up, but aren’t AI art generators kind of…sketch? Like they’re taking visual cues from someone’s else art?

Or do I have that confused with something else?

Preemptive “this might need to be it’s own thread” because hoo boy is this a contentious and thorny issue.

@Roz said in MU Peeves Thread:

@Testament said in MU Peeves Thread:

@GF I think it’s unenviable position. Either you’re really good at describing how someone looks, or you’re a good enough artists to make your own PBs. Or you just pay someone to do the art for you, which is always a viable option too.

I dunno if there’s a better option for people who lack both of those skills.

Some folks have had good luck with Artbreeder!

So I wasn’t going to bring that up, but aren’t AI art generators kind of…sketch? Like they’re taking visual cues from someone’s else art?

Or do I have that confused with something else?

Basically don’t sell it.

It’s worth noting that profit is not required to successfully prosecute a case of copyright infringement. Not-for-profit activities may make it less likely for the copyright holder to think it’s worth the money to sic their lawyers on you, but they absolutely could if they wanted.

Trademarks also muddy the waters. When Disney comes after creators for Mickey Mouse art, it’s more a trademark thing than a copyright thing.

Also it bears noting the law almost always lags behind technology, especially when it comes to IT.

Absolutely. And also, the entire gist of copyright law enforcement is a giant “it depends”. Bitter legal battles have been fought over guitar riffs and tiktok dances, and I’m sure this will be much of the same. But my feeling as an engineer is that this is 90% derivative and therefore copyright infringing. The folks behind the AI generator app might well argue that it’s not. Who knows.

Well, since $$$ is in play now those waters will get murky, if not bloody.

What’s the difference between me making a drawing or otherwise art based on or inspired by someone else’s work and a pre-programmed AI machine doing what amounts to the same thing? If that’s the case, then isn’t all fan art ever also qualified as violations of copyright?

I think you can look to existing copyright cases for hints of where the line is.

Things like chords of music, specific dance moves, individual words and phrases are not copyrightable, nor are broad “styles” like blues music, or even Faulkner’s way of ditching all punctuation. They do not constitute “a work” under copyright law.

However, a specific work of music, art, writing, etc. is copyrighted, and that gives the artist sole privilege to create “derivative works”.

If you’re making your own art based solely on a style of some famous artist, it’s unlikely you’re creating a derivative work. You’re doing your own thing, just mirroring some basic techniques they use.

But if you’re literally taking their work and then altering it to just add some other face onto it, that’s pretty derivative. And that’s basically what the AI-generated programs are doing.

The grammar training for things like AI dungeon feel like they may be different. You might have fed Jurassic Park into the algorithm, but it’s unlikely you’ll get anything even remotely resembling Jurassic Park from the output.

Having said that, I am not a copyright lawyer, and the courts are still sifting through all this.

I’m not her, but I’d think it has something to do with it doing so because the AI actually using the painting somebody else did to create something. Like, it’s basically fed into the code. It’s a computer program using something somebody else made to make something new DIRECTLY, rather than just influence or whatnot. With a computer you’re actually putting the art into it.

If that’s the case, then isn’t all fan art ever also qualified as violations of copyright?

This is one of the big arguments against it that I’ve encountered, because these generators are fed preexisting works without artist permission.

Indeed. But I personally believe that will eventually be sorted out under copyright laws. There are already cases floating about, though by my understanding they haven’t been definitive as of yet. It could wind up being a boon to artists if such use is said to be protected (which I believe it should be) and the big names end up having to secure permission and pay royalties.

a) It has to be trained on existing works done by existing humans

This is one of the big arguments against it that I’ve encountered, because these generators are fed preexisting works without artist permission.

a) It has to be trained on existing works done by existing humans, and b) It can only do what it’s been programmed to do.

Machine learning isn’t true learning the way humans learn. It’s just super-elaborate pattern matching, and is limited by the gears and knobs the programmers put on the box.

For these reasons, it can never be as creative as actual humans, by design. It can automate some simple tasks; it can be useful and even fun; but I’m not especially worried about AI-generated art making artists obsolete any more than I’m worried about AI Dungeon making writers obsolete.

Words and phrases from recent roleplay turned to AI art turned to ASCII art turned to a piece of art in-game.

I’m digging it.

and so you can pry Midjourney from my cold dead hands you mONSTERS

Because I can’t draw a straight line.

^Sam Gorski from Corridor Crew, as Big Boss and done in the style that Yoji Shinkawa uses for Metal Gear Solid

^Same guy, as an ‘evil wizard’, I’m fairly sure they used the style of Greg Rutkowski for this

AI image generation is getting better fast.

So what had happened was I watched that stupid episode of John Oliver about AI art generators, and I plugged my character concept into one and posted the results to our private group. And our newest member who is so gentle was like— full-blown offended. Mind you, the results were so bad and dumb and not art (sorry to you, art). I viewed it as a silly thing to occupy a few minutes of my time, but I see where people are coming from where it’s a slippery slope.

A few times now we’ve all pitched in to hire an artist to do one or all of our characters in their style to commemorate the end of a campaign. And there’s no replacing that.