It is easy to confuse recordings (or generations) of music for music in its totality, because we use the term "music" to mean both, as a convenience
It is also easy to confuse a digital picture of a painting (or generation of a painting-looking image) with a painting, because we use the term "painting" to mean both, as a convenience
This is a problem of language, and silly because media depicting the art is just one dimension of the actual art
A picture of a painting is one narrow representation of a painting, and a recording of a song is just one snapshot of a song.
AI presents a real challenge, but currently only in those narrow dimensions