i think that some people are under the impression that "AI art always looks the same" because the big generation-as-a-service models tend to push for a very "same-y" look and people who aren't super involved in the process just going for something quick and easy. which is something like saying "digital art always looks the same" because of the top 10 anime of the season or whatever. if you look at e.g. reachartwork's work there's a ton of variation within pieces, and none of these really resemble the "midjourney house style" people tend to associate with it. and this is all one person. rob sheridan has a different style entirely.
this is why it's more useful to think of AI as a tool than as the "creator". if you want to make lots of interesting weird things, you can make of lots of interesting weird things. if you just want to make waifus, you can make waifus (not that there's anything wrong with that).