I guess it is supposed to help us find tumors and whatnot, synthesize images rather than rely on an artist to do it, etc. But if you listen to a lot of ML talks they sometimes say, "we trained this on images but the technique works on any 2D or 1D signal" so it's applicable to signal problems in general.
Currently both Big Sleep and Deep Doze are generating 512x512. These ones are representative: https://postimg.cc/HVspWgPn
Few are "collages", most images have full-area coherence.
For some reason Colab let me run this all last night. Wow just wow. Eventually I got the feeling that it matters if I'm watching.
Each of these links has about 9 images and the prompts that made them. Sometimes the image does not look like an animal right off the bat, then it seems like you asked for something the network had to say.
What those are, I have no idea.