My mind went directly to the Music Genome Project. The article seems a little like a subset of 'Can machines duplicate arbitrary human categories'. Well, yea,

My mind went directly to the Music Genome Project. The article seems a little like a subset of 'Can machines duplicate arbitrary human categories'. Well, yea, if we program them to. But why would you want to? It's all already very well labeled. I think a more interesting question would be "How would a naive ML network classify various clouds of music types across time?"