i think ai-generated images are valuable for analysis not in the traditional way, where you figure out what the artist's decisions meant to them and for the meaning of the piece, but in a strange new way where you have to take computer-think into account and can reverse-engineer insights about what went into the database. this works in an 'artistic' way obvs: "when i ask something trained on [x] to give me [y], [z] shows up a lot. let's look into [z] as a theme among [x] about [y], and see if we can gain a new understanding of the subconscious or conscious associations between the two"
but it also works in a technical way: "when i ask this face generator to give us women, it only gives us white women. let's talk to the database people about bringing on some more diverse voices" or if you want to take it to another level, "when i ask this face generator trained off images we use in conjunction with our brand, it only gives us white women. let's talk to the marketing people about bringing on some more diverse voices and see how expanding our representation could potentially polish up our public image among groups we hadn't been targeting as well before" where the images generated are not the product and the point but rather an easily-presented medium through which we can do pattern recognition at a speed and scale unfeasible by analysts whose time is better spent further down the problem-solving process
thinking about the images we got when BMA was running is always interesting because for every cool art thing we got we'd get ten pictures seemed designed to make you think in the second way, from something as explicit as "when I put 'house' in, it unfailingly gives me something that looks straight out of a wealthy american suburb. an american kid would see nothing wrong but if a japanese child or even a british child tried that, they'd think it was broken" all the way to "whenever we ask for two robots standing together, we usually get a big blocky one and a small curvy one in a very limited set of poses......... hm". there were persistent design motifs you could only catch sight of through repeat searches for the same thing, something the BMA game cycle was specifically designed to encourage
and i dont know where the bma bot got its image sources obviously but we had rounds where people were using the phrase 'mcmansion' or misspelling famous painters names and getting pretty accurate results so it probably just scraped the internet? and there's a pretty valuable conversation to be had there about the classification and demographics of internet images, specifically how in most search engines you have to explicitly clarify your query in order to get something that isn't american, that isn't heterosexual, that isn't white. (and on a much less important note about the evolution standardization and arguable sanitization of various internet-specific image categories)
i agree wholeheartedly that any given piece of ai "art" has little to no artistic value because it was created largely at random, designed to mimic and nothing more. it's not art because it wasn't made by a human who can make intentional aesthetic decisions to convey meaning. i'm not an artist if i scroll google long enough to find a picture of an anime girl i like. but if you generate enough images and put them alongside one another, you can make an argument that the ais themselves are a form of art- after all, they were made by people, they're the product of hard work and intentional decisions. and thanks to the database model, they reflect the psyches of the people who created them. consciously and, often damningly, subconsciously
