Jaime Brooks, goes into an in-depth look at the impact of large language models on music and recording artists:

It continues to be the case that in streaming era, recording income is much lower than in the physical music era. Many big artists can scarcely be bothered to chase after it anymore, preferring to refrain from releasing new music until it can be incorporated into a larger plan to promote tours or clothing lines. The platform model would need to do an enormous amount of heavy lifting just to make recordings appealing as a revenue stream again, and that’s assuming it wouldn’t also do irreperable harm to the other pillars of a traditional recording artist’s business in the process. That’s not an assumption I’d personally be comfortable making, because “artist careers” have been driven by scarcity since the dawn of the recorded music era itself. As a society, we simply did not care so much about individual performers until after recordings became the primary way we all engage with music.

The artist model and the platform model are not just incompatible, but actively corrosive to one another. They cannot co-exist. I’m not arguing that one is better than the other, I’m saying that no living person will be able to do both effectively at the same time. The artist-as-platform model isn’t an evolution of the recording artist concept as we currently understand it, but a completely different proposition that would change the sound and character of popular music just as much as recordings themselves did during the “great music shift” of the fifties and sixties if it ever becomes dominant. For Vocaloid Drake to thrive, Drake the “recording artist” would almost certainly need to be destroyed. This is what advocates of “AI” are pushing for when they call for established artists to “open-source” their names, voices, and likenesses. They don’t understand how any of this works, and they don’t want to. They’re just here to break things. It’s all they know how to do.

The whole piece is worth your time.