my favourite piece of "AI art" is something that was not intended to be AI art at all, and was made and posted online as a joke. which is "salmon in the river", from 2023.
this is how the AI apparently understood the above prompt. salmon filets jumping in water.
in this case, not only do i think this is interesting art, but i think it is interesting art only because it was not made by a human. like, this reads to me as art about alienation and commodification, right? obviously we all know that under capitalism products become separated from the labour and processes that created them. you buy a shirt and you don't think about the hands that sewed it, and so on. the commodity form hides its own origins. and food in particular hides not just labour, but life. the animal disappears. like, if you think about how meat reaches you at a western supermarket, typically it arrives on a styrofoam tray, wrapped in plastic, cut into shapes that don't resemble the body they came from. a chicken breast doesn't look like a chicken, and likewise a salmon filet does not look like a salmon. many times people actually find it gross or distasteful to see the animal! the filet is literally the shape that says don't think about it.
so the art, then. the filets are appearing in the river, which is where the living salmon would be. the commodity form is occupying the space of the creature. the erasure itself is swimming upstream. that's sick!!! and the wrongness of it, the visual absurdity, is exactly what reveals how much work the commodity form normally does. We're used to seeing filets in kitchens, on plates, in supermarkets. In those contexts, they look right. they look like what salmon IS! it's only when you put the filet in the context of a living animal that you suddenly see how strange it is and how much has been removed. it's good art!!!
AND YET if human artist had made this image, i don't think it would be very good art. filets swimming upstream as a commentary on commodity food culture is fine but it would feel very on the nose in a banksy, makes-u-think kind of way. this would be a human saying "here's what I think about how we relate to what we eat." like imagine this as a political cartoon, right? immediately the exact same image would make me want to fucking roll my eyes. it would be kind of insufferable!! and to me i think that's because it is making an argument. the artist has to be visibly making a point and the finger is always wagging. we live in a society, bottom text. UGH!
whereas, when an AI produces filets in the river, it's not making a claim. it doesn't think and it doesn't care. it is just outputting based on what it's been trained on and based on the prompt. it's saying "here's what salmon actually means in the aggregate of human visual culture." it says something in and of itself that an AI image generator was asked to create salmon swimming in a river, and it produced filets. boneless, skinless, ready-for-the-pan filets, floating serenely through the rapids. and that's because the AI was trained on us. it was clearly not primarily trained on, e.g Coast Salish art like on carved salmon with eyes and spirits or more generally on cultures that really focus on holding the sustenance and the creature together. instead it was trained on an aggregate blob of the internet, including the very massive and alienated western commercial relationship to food. our images, our photographs, our stock photos and food blogs and recipe sites. and in that corpus, salmon is overwhelmingly a filet. when you throw everything into the pile and ask "what is salmon," the commodity form rose to the top.
thus when the model produces filets in the river, it's not really making an error in the same way we would; it's just accurately reflecting what "salmon" means in the aggregate, and putting it in a context that makes it seem incredibly absurd. it literally works as art because it's not A Guy saying eyy, look what you've done, it's just showing what we've done, without commentary or judgment and without even knowing it's showing anything at all.
but then ALSO. was this really an "error" generated by an AI? or was this a human who prompted it to make a picture of salmon filets in a river and posted it as a glitch for internet points? i don't know!!!
and at first this bothered me, because I've been so hard on arguing that the image only works because it wasn't made intentionally. like, that the lack of human intent is what gives it evidentiary weight and what transforms it from trite political cartoon into Good Art.
but i literally think that's still true. if a human made this, if someone deliberately prompted an AI to produce filets in a river and then framed it as an accident for heckin reddit updoots or even legitimately to make a political point, then what they created is a piece of art about the difference between intentional and unintentional meaning. they used the AI as a medium, but more importantly, they used our expectation of AI failure as a medium. The art isn't just the image; it's the image plus the caption plus our willingness to believe it.
either way, the image only works as art if we believe an AI made it, because that belief is what transforms "salmon in the river" from a heavy-handed political cartoon into a piece of evidence about how we see. but only a human could post it and have it work, because the act of posting it as a mistake is also an artistic gesture, regardless of whether it actually was one. the AI can't do that part. it doesn't know what it made or why it's either profound or funny. but a human couldn't do it alone either, because a human making this image deliberately would just be making a statement, and statements are easy to dismiss. the art exists in the gap between the generation and the framing, and that gap is where the human goes.
