Posted on Leave a comment

My Position on generative AI

Spread the love

So it seems like all anyone can talk about right now is Generative AI and how its going to replace anyone who thinks or shapes things for a living. And the first thing I have to say about that is that I’m old. Which might not seem relevant, but give me a little leeway here while I explain myself.

—————————————————
The TLDR is: People also copy other peoples art, and use other peoples art to learn. People forge other peoples art in both legitimate and illegitimate ways. I assert we’re really just upset that people copy art. And I think that copying art is just using the technology wrong. And I lay out the method I use, so that I can demonstrate why its not copying.
—————————————————

When I was born there were no computer graphics, no digital painting, there were an incredible variety of practical effects and ways to achieve computer graphics like effects though. By the time I was sketching my first random cartoon faces instead of paying attention to class graphics were limited to either lines and blocks on television, Pong Space Invader etc. Model makers thought it wouldn’t amount to much… and it all got better and more specialized over time. Model makers still exist by the way, although there are more of them using computers now.

It was probably pretty world ending to people that made their living making animatics for television with cardboard boxes though, they all learned to animate with computers or retired. After pong (or qix which I like a lot too) video games got color and solidity, but behind the scenes they were essentially very specialized pixel art, held together and moved around with sprites, the people that only made black and white video game monitors were pissed (they tried to hold on for a sec with these colored overlays that you put in between the screen and the mirror on the console) its hard to imagine a video game of today on a black and white CRT. I have one. It goes to a space invaders game that a friend gave us, (because it no longer works and needs to be fixed). It also does not make a black and white monitor a color monitor.

The guys that made the lines and blocks didn’t really stop with black and white though they made monitors and games in color, then they made the same things with sprites. So the guys that made monochromatic crts retooled their factories. A real 3d game at the time (they did exist) was still in one color and lines and blocks. I’m thinking of a tank arcade game that you put your eyes up to this periscope looking thing and you played a tank commander that shot other tanks stereoscopic 3d with really limited physics. It was fun, also not something that caught on.

3d movies at the time were the red and blue lines on top of a black and white movie that you watched with the colored glasses. Somewhere from the back people that made plays for the masses just gave up the last ghost they were holding on to.

These were all the things that people older than me were objecting to, I saw their point, but I also didn’t want to play pong (there was a driving game too also with lines). The point though is that to really animate something that was more complicated technologically than south park you had to draw it, on cellophane, make layers, take pictures, You actually can’t get the required equipment to do this any more, and the output would look really bad in comparison to even “hand drawn cartoons” which given the way people make them these days are also very digital when you get right to it.

The guys that animated at the time didn’t really think the lines and blocks were going to amount to much. Simple 3d models finally came for all television graphics, they got better, It reduced the number of animators. Which the world had already reduced the number of animators from when they were all hand painted (my great aunt worked for Disney and Walter Lantz so I’ve seen these types of cells framed as art, they’re beautiful, collectible and not how anyone does it even by the seventies.

And really most of the non “hand drawn” animations are 3d models with surfaces and effects, thereby removing _a_lot_ of the drawing. So, my position is that objecting to only the gen AI part of the current technological environment around art is kind of missing the point. We’re all trying to participate in creative culture, and that’s not a bad thing.

I mean, Technology always changes how people do things… Change is hard… Its also always happening, While there is a difference between fewer people touching hand to paper and nobody touching hand to paper – there’s really not… The hill I wanted to hold when I was a kid was when they got rid of the paper, and the canvas. Photoshopping a picture was possibly the tackiest thing you could do at the time. The method to do this is to take someone else’s picture and use a series of effects and modifications on it to obscure their original photograph possibly lift elements out of multiple pictures and recomposite them as a new image and post it as your own, picture or digital painting, – (at the start of digital painting you moved a mouse over a scan and traced the borders of the existing image and then filled the areas in, it was utter trash).

So I ignored that, and did not try to develop the skills required to do those things, I still can’t really use a drawing tablet very well. But the aspects of it that upset me were all eventually fixed. I just wasn’t around for all that because I was still dreaming of singlehandedly making a traditionally animated movie in my apartment. Which is crazy, you need a whole group of people to do that, even today with digital tools. Now with a lot of fully automatic tools you’re about at the level of being able to do a YouTube short fully animated. A full movie would be a lifetime of work though (look up Yuri Norstein and his wife Francheska Yarbusova, ‘The Overcoat’). Its a really romantic idea, to just do it all yourself by hand, but if you go back far enough you will always be engaged in corner cutting. Do it all with pencils and someone from the stone age is mad you didn’t make your own charcoal and mine your own pigments.

Particularly today with digital painting, the outgrowth of the aforementioned tracing, photoshopping, and line and block making. Someone was legitimately mad about all those things, but as the technology zeroes in on what really needs to be done there are still no free rides.

If you’re going to shape a Gen AI image to what you actually want, you generate perhaps 10 or 15 images with Gen AI refining the prompt so its closer to what you actually want, then take the ones that are closest and cut them apart and use a combination of all of the normal digital techniques to combine elements and overlay that into a consistent image. And truth be told since the prompt is never 100% what was in your head it is still even more of a collaboration with the AI at that point.

One objection, (and the objection I hear most) is that AI was trained on a lot of images that were publicly available, including images that have use restrictions on them. Well, so were human artists, we looked at other peoples art, and tried to make things somewhat similar. And, if you spent any time in art school you probably copied your fair share of old masters. And, if you make fan art you put your own spin on what a team of other people have created. But really the animators animating all those cartoons did all that already. If you ask a professional artist to forge a painting (and if they’re willing to do it) you can reproduce just about anything, by hand, in a way that will pass at least basic scrutiny. If they are really practiced at recreating things, and research and put effort into it they can probably make something that would pass museum level inspections. Its still a copy by that artist of something that someone else did. The problem starts, when you try selling that as if it was the original piece of art. The REAL distinction, is that the thing you are selling was represented as one thing, but is in fact another. Now, its perfectly all right to sell my copy of the Mona Lisa as a “Jason copy of Leonardo Davinci’s Mona Lisa,” if its good maybe its worth a few hundred dollars, if I try to say its the original and sell it for millions I’ve committed a crime… if I sell my copy I’ve cleaned out the debris from art school.

If I try to ape another persons style, I’ve just committed the sin of bad manners, which really just means the art is unoriginal and therefore worth less, it is not without value (but maybe its value is just that its a prestretched canvas). And typically you will hear an artist say something to the tune of “Yeah, I was trying a Mark Rothko thing,” and if they did it well and they’re going to try to adapt that then they say “I think I’ve really made it my own here, I’m probably going to revisit that a lot.” And then people say “yeah he was really influenced by Mark Rothko, look at these two paintings together,” The people that do that are called art history buffs. So, is it legitimate to reproduce someone else’s style with AI? My position is that if you accidently reproduce someone else you’re just doing it wrong, and its tacky. But if you’re reproducing yourself that’s ok.

And, really if you reproduce something that exists by telling the AI to reproduce all the influences that made up the original and you somehow arrive at sailor moon, we should all bear in mind that sailor moon is just drawings of schoolchildren in traditional Japanese sailor outfits with ribbon wands and a few minor mythological references thrown in. Possibly The person making the AI piece is not reinventing the wheel, but the actual cartoon didn’t reinvent the wheel. Neither did mickey mouse, star wars, battlestar galactica, or any number of other “Intellectual Properties” of media companies out there, star wars took a world war 2 movie and a western, mashed it into a heroes journey tale and added fairly standard elements from sci-fi and fantasy books. Is it a new thing? yes, is it totally original? no. Can you reproduce it by identifying the same design references they used? yes, Is it legitimate to closely reproduce it? not if I say I’m Disney or actually use stills or something. People still do that, when you get anywhere near the elements that go into major movies then the overwhelming amount of images drawings, reproductions, and cosplays out there mean that you get ‘really close.’ Humans were doing that already, really we’re just mad we can now get a machine to do that basic part of the process.

Sailor moon for instance, has magic, the magic of a story that draws people in and makes them care about it. The journey of the characters and the narrative that created the experience of that story is key. Other parts may be key to other stories, like Simon Stalenhag for instance where the story developed in response to the art rather than starting there, or a pretty huge list of indy graphic novels. And, I have a set of sailor moon t-shirts, which were generated by telling the ai to pose a beautiful Korean woman cosplaying in a traditional Japanese sailor schoolgirl outfit, then removing large parts of what it created, coloring the rest, and creating reusable 3-D elements that were added to the background and generating the background in a separate way, then researching the character and translating things into Japanese out of what the characters say. And, really, I think its legally original enough that I could argue copyright in court and win – I also conceptually am not really representing the character more than a person cosplaying the character. I also did a lot of work beyond just waving a wand and saying “AI make me a t-shirt.” I also think I’ve done enough transformative work to the output of the AI that I can represent it as “my work.” Anything I make in this manner shares almost none of the pixels with the first pass of what came out of the AI. I’m also not going to deny that AI was part of the process. Generative fill in photoshop is AI. Digital artwork’s jitter control means that the strokes you “see” aren’t actually the artists real handiwork. Creating files does not create tangible objects… and AI is just another step in that technological procession that I was talking about at the top of this.

So, yeah I’m using it (I meaning Jason), later down the road we’ll be objecting to some other change… possibly once robots gain real consciousness and art is actually generated by non-human hands. Its a way to make something look nice, and as an artist my goal is to make nice looking things. The penstrokes on the Orca t-shirts also required a lot of work to make them look good in another media rather than on the page I made the pen strokes on. And as an artist I say there is humanity in it.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.