I believe you have to think about AI music in the context that it's written for, and the environment you're in. We, as musicians, listen to music in a different way to other people do, so - think about a crowded restaurant where there is lively conversation going on, and some background music ... you're not really thinking about the music, but it registers in the background and gives you the right vibes for eating and talking ... in those situations AI will do well, the music is not a focal point of the night - but, if you're going out to hear a band play, then you don't want to sit in front of a PC with other people and watch it make it's 0's and 1's into music ....
I remember back in the 80's or 90's when I first got into multi-tracking and a friend quipped some remark about how using the multi-track was 'cheating' as we couldn't reproduce the sounds live ... every new technology is a tool that we can use - the problem is the same with cameras. I studied photography in school with analog/film cameras - learning about composition and the rule of thirds etc, film wasn't cheap, so you had to use a roll of film to the best of your ability, now with a digital camera attached to the phone, anybody can take 100 different photos of the same thing and pick out the best of the 100 and throw out the rest. Law of averages says you will get one good photo from those 100 ... the same with AI music, there will be one good song in 100, but it's wading through all the 99 other songs to find the good one.
BTW I tried some AI "song -> video" website and they produced these two videos for my music:
https://www.youtube.com/watch?v=YTECwT8uP4Y
https://www.youtube.com/watch?v=ZuIGwP-1tVk
Cheers
Richard
-[ Musician, writer, guitarist, singer ]-
Bandcamp https://richardmortimer.bandcamp.com/follow_me
Discogs https://www.discogs.com/release/29065579
YouTube https://www.youtube.com/@RichardMortimerMusic