I never understood the hype around diffusion models. Same thing for deep fakes. If you look closely, these models make horrendous mistakes such as hair growing out of ear, uneven eyes, etc., which shows that there is no sense of understanding the underlying concepts in these models. I find GPT-3 a great achievement, but when it comes to visual arts, you can't just call whatever SD/DE2 create "art". At best, these models just reflect the patterns and visual cues they've seen in the training data, but they have no idea as to "what" they are painting. It's actually funny that one of the first applications of these models is in arts where there's no real-world hard-rule to quantify just how terrible/great the results are, because at the end of the day, people will claim that art is all subjective.