I heard about C2PA and I don’t believe for a second that it’s not going to be used for surveillance and all that other fun stuff. What’s worse is that they’re apparently trying to make it legally required. It also really annoys me when I see headlines along the lines of “Is AI the end of creativity?!1!” or “AI will help artists, not hurt them!1!!” or something to that effect. So, it got me thinking and I tried to come up with some answers that actually benefit artists and their audience rather that just you know who.
Unfortunately my train of thought keeps barreling out of control to things like, “AI should do the boring stuff, not the fun stuff” and “if people didn’t risk starvation in the first place…” So I thought I’d find out what other people think (search engines have become borderline useless haven’t they).
So what do you think would be the best way to satisfy everyone?
The same debate happened when photography was invented. Photography led to some of the most iconic art movements of the 20th century and the deliberate departure from more realistic paintings, because nobody needed to hire a painter or sketch artist just to remember what someone looked like anymore.
My expectation is that AI will continue to be used to generate what has already been identified as particular styles. And these styles will go out of fashion with overuse, much like the 50’s was oversaturated with mass produced designs. Which led to pop art.
Tl;dr It’s all just part of the creative development of humans and our adaptation to automation. If we want to address why artists starve, AI in art is only a symptom of a much larger issue around human worth.
I strongly agree with the comparison to photography. People like Man Ray and David Hockney opened doors to new art impossible before photography. And it did not eliminate drawing, painting and similar arts. It did force some change upon them, though. It seems that we have discovered something like a new art form. That should be exciting. It will certainly be uncomfortable
The AI has particular styles because the current AIs are trained in a particular way. As we go further, there is no reason to assume that AI would not be able to copy/mimic any stile. While photography is distinguishable from painting, the AI painting can be indistinguishable. That’s, I think, is fundamental difference.
A painting can be distinguished from a photo from its physical dimensions and attributes like texture. If I take a photo of a painting, almost nobody is going to file a court case against me for putting it in a book or website or tv. We’re collectively ok with that level of reproduction.
An AI work can currently be determined by looking for ‘errors’ that aren’t likely to be drawn in by artists or found in reality. Inconsistent shadows, exact repetition of organic pattern details, extra limb bits, doesn’t construct language fluently, etc. We’re not necessarily ok with that level of reproduction though because it is slightly less obvious at first glance.
Just like photos are a lossy format for paintings, AI is a lossy reproduction tool. While it will definitely be harder to distinguish with time, we’re sort of still left with the core problem. What fidelity reproduction is necessary for something to no longer be original? And what does the originality actually matter when people don’t need to rely on producing it to eat?
Hell, people even bought NFT apes with cut and pasted Mr. Potato Head parts for thousands. People have photocopied photos of the Sistine Chapel in books for close to free. Maybe it’s just cool that the imagery is resonating with people regardless of the medium.
Also, we should feed people without the expectation of financial return, thanks for coming to my tired tedx talk.