How do we salvage generative art and its long history from the LLM-based "AI art" clogging up the internet at the moment?

I spoke with @aparrish for my latest in Filmmaker magazine.

https://filmmakermagazine.com/121867-joanne-mcneil-large-language-models-allison-parrish/

Turning Poetry into Art: Joanne McNeil on Large Language Models and the Poetry of Allison Parrish | Filmmaker Magazine

Publication with a focus on independent film, offering articles, links, and resources.

Filmmaker Magazine | Publication with a focus on independent film, offering articles, links, and resources.
@jomc @aparrish I'm so glad to see someone finally write on this topic without reducing the whole problem of "AI" art to questions of authenticity and the agency of a human author/artist when there's lots of legitimately interesting ways to use things like text generation. this post mentions horse_ebooks and throughout this whole stupid chatgpt fad I've thought a lot about that and early deep learning/LLMs where it was precisely the production of randomized and somewhat coherent but highly uncanny outputs that made them interesting, compared to what is now popular which is to produce something optimized to reach a statistical lowest common denominator and basically invariably end up sounding like a sitcom script
@nyx @aparrish @jomc I was also a fan of the way some people used Deep Dream, it could create some original, nightmarish pieces, many of which I still like, while Dall-E only shows me things that are similar to what I have already seen
@nyx @aparrish @jomc I think it could end up being similar to how things played out in music - electronic or electronically processed music that tried to sound like live music invariably sounded like shit, but electronic music that embraced the artifice and sense of regularity is great
@julieofthespirits @aparrish @jomc for sure, the thing I've noticed in these discussions is that people seem to always lean towards wanting to make these pseudo-philosophical statements about art that are all to varying degrees downstream of the conservative PragerU-tier idea that "art" is either like Western and European rennaissance stuff or it's not "real art" (whatever tf that is). it misses the point, one which as you said we have really good established examples of really with electronic music, of how something that is fundamentally *in*human, artificial, uncanny, etc. can have artistic merit (again, whatever that is).

idk to me it feels very obvious and simple that you know it when you see it when something produced by technological means has some interesting quality to it. I've only ever seen that though with the recent "AI" fad when it involves it not functioning as intended, which is itself kind of interesting to explore really but there's always a downward pressure from these "AI" companies ofc precise to *prevent* people from "mis"using their products to do anything interesting.
@nyx @aparrish @jomc yeah, I think that if there's a reason Dall-E is proving the most popular model of AI image generation it's precisely because it's good at replicating the kind of art that Prager U people like, rather than things that could actually be used by contemporary artists interested in AI's uncanny possibilities
@julieofthespirits @aparrish @jomc that's exactly my thoughts too. been wanting to write something on this at some point but I also don't want to unintentionally end up feeding into a fad that I've been sick of seeing people talk about for awhile already lol
@nyx @jomc @aparrish There's something about early A.I. art before it learned to mimic us that I find interesting (and artistic). It was always psychedelic and geometric and makes me feel as if I'm peering into a very different reality, or peering beyond reality itself. It makes me think of the story "The Dreams in the Witch House," where the main character stays in a haunted house and reality starts to fall apart and the protagonist peers into many other dimensions at once.
@piceaglauca @aparrish @jomc a few years ago I was really interested in the possibilities that GANs/LLMs had because of these qualities of producing things that felt extremely inhuman and only vaguely resembled anything that actually exists, but I should have known that corporate "AI" technologies would reach equilibrium in the most boring ways possible