AI “art” is prone to nonsensical hallucinations with people gaining extra fingers, background elements bleeding together, and everything just looking uncanny. It’s technically possible that you could train yourself to try and imitate this “style”, but it would be both incredibly difficult to learn, and also be incredibly unrewarding- even putting aside having to prove you did, in fact, make said piece of art, it would still be as ugly as the AI stuff no one likes anyway.
Are you familiar with “analogue horror”? It’s a genre of (mostly) video horror that uses video and audio filters that make it look as if it’s shot on a 1980s or 1990s video camera (or even a 1980s/1990s home camcorder.) That sort of style evokes a lot of nostalgia in 80s/90s kids.
It obviously doesn’t add fidelity to the image to use such filters that blur, add white noise, and add video artifacts like you might see from a malfunctioning VHS player. It removes fidelity.
But it adds a “quaint” and sometimes surreal feeling to the media, particularly for folks who have been exposed to a fair amount of that medium of video. Or even folks who have only been exposed to retro recreations of elements of that medium.
I’m sure something similar is already happening to the brains of the younger generation. They’re forming connections to AI slop. And some day, I have to imagine elements of AI-generated video content will be used – on purpose – in new media to evoke a sense of quaintness, nostalgia, and otherwise “mid-2020’s-ness”.
I don’t think they’ll exclusively use “GenAI” technologies (stable diffusion, Dall-E, etc) to give media that feeling either. It’s weird to think about now, but they’ll probably be make ways to add AI-slop elements in novel ways that the actual GenAI technologies aren’t capable of. (Again, making a connection with VHS-looking filters, I don’t imagine most people making things like analogue horror content today are using actual VHSs and vintage camcorders.)
So, I can agree with your premise, OP, but only with the slight addendum of “for now”.
I hate that this is what the mid-2020s is going to be remembered for, but I guess every decade has something to be embarrassed about. NFTs/blockchain/cryptocurrencies, Beanie Babies, The Macarena, gestures broadly at the 80s, etc. But great things have come out of all of those decades as well.
add AI-slop elements in novel ways
Datamoshing can have an AI-generated quality.
I think the shrooms just kicked in.
A lot of AI art used e.g. in YouTube thumbnails is uniformly busy, bright and colorful. Which means that nothing stands out, while the eye is immediately tired of the detail. It’s pretty much the opposite of good art practices. That said, it’s quite reminiscent of 80s fantasy art, which also was often overly elaborate and loud.
Counterpoint: AI slop does not have a “style” But a look that is an artifact of all the the material it is trained on and the limitations of the technology. It has a series of visual tropes, for example the odd glossy sheen it tends to have. The oversaturated colors; the tendency to frame centered subjects like social media posts; the rigid composition of clip art and stock images; Even the best image or video generations tend to all have these traits and it’s hard to coerce them to move away from them.
So in a time where even human made media tends to be flattened and averaged by the over influence of pop culture, the biggest disservice you can give yourself as a creative is making your work even more sterile by imitating AI generations.
