the only attraction that image-generating neural networks hold for me is exploring the unique qualities of the network. They can't replace humans, obviously. I don't want to use them as an artistic tool, 95% of the people who do are hacks with no vision, and I'm much more fascinated by drawing out the weirdness embedded in the embedding.
Exploring the phase space of images and generating truly bizarre shit is where they shine. I've enjoyed experimenting with this (at least before the FUCKERS over at OpenAI took away my free daily tokens), but I know a lot of folks would rather not see it at all, so when I post the weird shit I'll hide it behind appropriate CWs, as above.