I think one major question that arises is the extent to which brains, or AI neural networks, can explore the unknown in a meaningful way.
There is some question about whether the networks are able to extrapolate outside of the space of the (embedded capabilities and training distribution). Some think that what looks like true creativity and inventiveness is actually just interpolation in a very high dimensional space.
https://www.biorxiv.org/content/10.1101/764258v3
If all of this is true, then it means that randomness is necessarily the real source of creativity in AI and possibly also in people. It would mean that you would create new works of art by taking interpolating inside the convex hull of the high dimensional space you're drawing from, and then you would mutate it randomly, and the result will then potentially grow the convex hull of that space (if the new generated point is outside of it).
In the case of humans, we have some advantage, because when we generate a new point outside of the convex hull, we are immediately able to test whether we like it or not. AI might one day be able to guess that however. Ultimately, the reason we like it may be due to the type of challenges it affords us when listening to it, or some mutual information it has without emotions, etc.
And, people might be able to draw from this space of not just music to get some kind of interpolated result, and then translate it into music.
Alternatively, you could just randomly generate music without any inspiration, and you might get something outside of that convex hull, but it most likely will be terrible music by human ears. There might be a small possibility it will be great. But if there are such great pieces of music which are outside the convex hull of existing music, then we probably won't find it quickly, we have to likely wait until enough mutations have occurred to expand the convex hull out into that unexplored space (unless the assumption isn't true and there is some kind of divine source of creativity).
Also, music that is outside of the convex hull might sound terrible to people just because it's too strange to us. It might not be until the convex hull reaches it that we have the perquisites to understand and appreciate it. In other words we will may evolve to like that completely new music that we wouldn't like it if we heard it suddenly without treading the path to get to it.