The author’s primary claim seems to be that AI art defaults to what most people prefer (‘accessible’, pleasing to the eye/ear, etc.), which he puts in the bad c

The author’s primary claim seems to be that AI art defaults to what most people prefer (‘accessible’, pleasing to the eye/ear, etc.), which he puts in the bad category even if created by a human. However, I did not notice whether he checked if AI can generate images/poems that aren't so accessible/aesthetically pleasing, by asking AI to do so? It seems like a huge part of his claims about AI generating art (that's similar to bad human art) assumes that accessible/aesthetically pleasing art is all AI can do, and it can’t do otherwise if instructed so. Otherwise, I guess the argument would devolve to: AI defaults to generating art that humans prefer (‘bad’), but AI can also do far more as well (if asked) when generating images/poems? (Although OP article is from '24, so capabilities may have changed.)