I don't know why you're mentioning art being derivative, but that's not the thing that worries me. What worries me, is that a good enough model will take away

I don't know why you're mentioning art being derivative, but that's not the thing that worries me. What worries me, is that a good enough model will take away the incentive to write music for many, and as a consequence it will also remove performers. This will reduce demand on music teaching and instruments, which will then both become nearly inaccessible. Since learning music isn't a question of following a few youtube videos, this will leave the world with just AI music. Jazz and classical music are probably exempt, since it relies on subsidies, their audiences care about the actual performance, and AI compositions will not draw enough of a crowd to make it financially interesting. But popular music will suffer, and that's what makes development of these models straight evil.