if anything, one of the major flaws of tech hucksters' pitch for AI art is that they don't go far enough. prompt engineering is fine and well, but still extremely imprecise in many ways. this could be fixed by integrating AI generators as brushes and other kinds of precision tool that artists can use on a digital canvas / 3d modeling program / whatever. but that approach isn't favored at present, presumably because that wouldn't whet investors' appetites for juicing profit margins by slashing labor costs. DraGAN is a good example, and i'm sure many of you probably remember a bunch of tech demos like this from the years before prompt-based image generators reached their current watershed of quality.
then there's stuff like those neural nets that can generate 3D models from 2D data and i mean, imagine what technologies like these in tandem with a proper art studio workspace could do for artists! especially if it could take your own pre-existing artwork and manipulate it in such a way, being able to take a raster image and treat it something like a 3D model. to say nothing of machine learning integrated elsewhere in the pipeline like the comic book outline generators used on Spider-Verse. immense potential swept under the rug by tech capitalists, and we're just rolling over and letting them.
