This is an excellent summary of the problems with AI Art beyond even the legal issues of copyright- though it indirectly addresses them as well- and especially points out why the comparisons between human learning and the algorithm “learning” only show a deep ignorance of the human process— and why relying on the algorithm will create nothing but stagnation.
From Christopher Doehling:
“I’m often accused of not understanding how gen-AI works. Nahh. I understand. Counter: a lot of tech people don’t understand human creativity or learning. I have backgrounds in all three. But mostly in art and creativity. Check my LinkedIn profile.
Before you come at me with “computers learn/create like peoples do” consider: it may be you who are in a strange land. Not me. Before you go running into the jungle, you might want a guide.
For example, ever wonder why it took us so long to learn how to draw and paint "realistically"(to make images that look like what we see)? It's because by default, our brains learn concepts, not visuals. For most, our eyes are used to recognize, not replicate.
It's like our mind throws away the visual information that is explicit, and exchanges it for understanding. I quickly know a cat from any angle I see it. I know that they are furry quadrupeds that purr when you pet them unless they scratch you. I know a cat when I see one.
My eyes also help me quickly understand the 3d space the cat occupies, so that I know where it is in relationship to my body, so that I can pet it or avoid it. My eyes help me understand "cats". Can I draw one? Not easily, because my brain, by default, doesn't care about that.
Unless... I want to care. I want to make images that represent or communicate my understanding of "Cat". Early art is more abstracted/symbolic because we expressed concepts first before explicit visuals. 4 lines= 4 legs. Shape language tells the story, with an arresting style...
Even if the exact visual (what my eyes saw) is not transmitted. Its not that we didn't want to. We didn't know how, any beginning artist experiences the same problem. Your brain wants to express what it knows conceptually, not what the eye sees.
But over time, our concepts and understanding grew to include things like optics, math, color theory. Tools we could (with great effort) apply to our artistic expressions as well.
Filippo Brunelleschi (re?)discovered linear perspective not just by looking at the world around him or at other art, but by application of those concepts. and then, finally, we could (again?)draw and paint what we saw. We could also make others see what we had only imagined.
So, we draw what we know, about what we see. Even if what we see is other art, even if I do a master copy, It comes from a place of concept. We are seeking to understand technique, another's experience, another's knowledge, not just absorb a visual for later source material.
evidence: If you have not gained at least some of the same conceptual understanding that the master did, you probably will not be able to copy their work, at least not convincingly.
"Generative" Ai (as it is) is not only unlike humans in the way that it learns, it is the polar opposite. It can copy what it sees. it can combine what it sees. But it does so without any understanding at all. About anything. At all.
A computer does not know what "cat" is. It may have some pixels->patterns that are keyworded "cat" but that is all it has. It can denoise from those latent images/parts of images, but it will only do so as instructed by our keyword requests and/or randomly seeded math, etc.
The only concepts delivered into an AI gen image are those given by the original artists or photographers. If you see a cat, its because someone else (or many someones) gave you cats to see. All the Ai did was serve it up, blindly combined.
The uniqueness of each of us, our experiences, and the concepts we learn and teach are what makes art evolve. If Ai had "taken over" for us at, say, the medieval period (in Europe), Art movements would have ended there too. Renniassance, Baroque, Impressionism, Cubism,...
Etc. they never would have happened. because no matter what prompt you gave, all you would get would be remixes of Medieval paintings, or anything previous to that time period. Ai doesn't make anything really new. not the way we do. It only (blindly) combines what's already made.
That's what's at stake here. We are on the brink of handing our creativity over to something that isn't creative. Why would we limit ourselves like that? If you think it makes art easy, it doesn't. It's an illusion. All you have is the art made up to this point. and no more.
If you want to be an artist, be one. No matter what your skill level, it's better than this. You are contributing of yourself to the world. You are contributing. period.
P.S. it’s not that Ai doesn’t have valid uses as a real tool. But when we get the idea that it’s a pet pro artist that “does the dirty work for me”, that’s a dark path. The dirty work is what moves us forward. it’s also the fun part, and we are the only ones who can really do it.”
https://twitter.com/dolimac/status/1635286958330224641?s=46&t=MInooHF4e3-CHmlyx2cj8w
