Image generators are designed to mimic their training data, so where does their apparent creativity come from? A recent study suggests that it’s an inevitable by-product of their architecture.
It's easy to harrumph at this article if you hate AI and all that, but I think it's interesting to try to come up with a somewhat objective definition of creativity. I do think it gets at an important part of the creative process, "Necessity is the mother of all invention". When you're working locally and stuff starts getting weird because of nonlocal constraints, then you have to start getting creative to make it all work coherently as best you can.