[wataruspeedo]

The problem—in my opinion—with arguing "an AI could never do X" is that your argument generally ends up working against you, by creating a well-bounded problem space, which is the first step to solving any problem (in this case, with AI.) I find it hard to create an argument against AI that isn't self defeating.

I personally don't see why a group of smart and soulful people couldn't assemble a database of cultural input data and train an AI to create some meaningful expressions that fundamentally change our human perspective, or change culture, or whatever heuristic you'd like to set for what makes an artist. Particularly as humanity comes to grips with what it means to be human in a post-AI world, this maybe the exact thing that becomes interesting from an artistic perspective.

I guess you could debate about attribution in that case, and a whole bunch about semantics in any case.