re: #65 The Ghost of a Flea
The prompt “Grateful Dead” creates swirly rainbows and that’s just the most boring possible parsing of that prompt into visuals.
That’s what I get from AI generated visuals in general, they’re boring except for when they’re uncanny but not in a way that’s challenging or seems to communicate something.
It reminds me when when fantasy novels would all have a house style so there was a same-y-ness, but then someone would do a distinct cover design and there would a dozen imitators soon after.
I mean, on some profound level they’re just collages, and mediocre art is often a borrowing of bits to create a collage-like new piece, and that’s why it has that feel.
Which is exactly what I figured an AI would be good at. Collecting a bunch of bits, slewing them around in different patterns (according to a few rules) and then vomiting the results back out fairly rapidly compared to having a number of human artists work on a set of concept items. From there a human eye (the beholder) could select one or two as having some promise and they would be used. (Or made the input for another cycle by the AI - which I expect would quickly go stagnant since the input variety would become sort of ingrown.)