If you want a pretty reasonable, actually-supported-by-facts reason to dislike AI art, remember that the neural network technology supporting it inherently trends towards the mean.
For example, with automatic translation, when you translate from a language that's more gender neutral to English, which uses "he" or "she", biases sneak in - he treats the patients, she cooks dinner, stuff like that.
(Similar to how a lot of the image generators had a period where it turns out they were adding "racially ambiguous" to all prompts on the back-end, because results trended white-as-default).
Or, for another point of comparison, the whole argument of "are video games art?" turns out to have the boring answer of "yes, but that doesn't mean art is inherently meaningful".
Image generators tend towards average results, (they'd probably be really good at inoffensive hotel lobby art designed to provoke nothing), so on the whole humans will probably produce more interesting results out of an ability to get weird with it - you don't have to believe in the soul to prefer human art!
...But, by the same token, AI outputs curated by humans have, in personal experience, actually gotten pretty interesting.
Duchamp's Prelude to a Broken Arm (shut up about Fountain, it's not even the only readymade he did) is literally just a stock snow-shovel anyone could buy, hung up in a gallery.
But the nature of a snow-shovel titled the way it is inherently implies a story - someone shoveling snow, slipping, and breaking their arm. The title's implication creates that mental image in the viewer, and suddenly the piece is having a conversation with the viewer even though literally anyone could have made it.
Properly curated and guided, AI output can become intriguing. It's just... very easy to use it in a lazy way leading to boring, biased output.