Generative AI is much better at art than coding.
Mostly because humans invented this convenient thing called abstract art - and since then tolerates pretty much everything that looks “strange” as art. Must have been a deep learning advocate with a time machine who came up with abstract art.
calcopiritus@lemmy.world 2 months ago
Don’t need to be abstract art, it manages to make many kinds of art.
The difference between art and coding is that if you pick a slightly different color or make a line with slightly the wrong angle, it doesn’t change much. In code, however, slight mistakes usually result in bugs.
raspberriesareyummy@lemmy.world 2 months ago
I meant that thanks to abstract art we’re willing to forgive “image glitches” in art by deep learning models.
sugar_in_your_tea@sh.itjust.works 2 months ago
I don’t think that’s the case though. Obvious glitches like 6-fingered hands can be avoided by generating a bunch of samples and picking the best one, and less obvious glitches tend to be overlooked, not considered a “feature” due to an appreciation of abstract art.
AI art works best for pieces that need to fade into the background, like stock images and whatnot to accompany more important copy. If it’s taking center stage, it needs a lot more hand-holding that probably makes it about as costly as just having a human create it.