My preference is pixel art, I love pixel art but I am no good at it and getting AI to generate pixel it is a complete none starter for me.
I agree, AI models are trained on images created by humans so they are in effect imitating thousands of different artists which is why you can end up with images with two ears, actually I think its meant to be some kind of elf ear but just doesn't look good. This is another thing, I regenerated that image yesterday using the same parameters and its changed, the original has a normal human ear. I think its because the software itself has changed over the last few months.
I think there will be jobs for humans that are specifically involved in generating AI content because its so difficult to use, especially if you are anyone artists trying to generate art content.