This is really interesting, but it also makes the failure points of computer generated music evident. Namely: it works best with simple, "jammy" music; it requi

This is really interesting, but it also makes the failure points of computer generated music evident. Namely: it works best with simple, "jammy" music; it requires a large existing base of music to pull from; and it has no way of using compositional techniques for rhetorical effect. (Increasing/decreasing tension, "going somewhere", expressing emotions, etc.) In other words, it can't create innovative music or music that has something interesting to say. It can only recombine the past. I like the idea of using these tools for educational purposes, however. Can we derive the "rules" for music of different cultures by feeding them through this kind of algorithm? If so, it would give us a fantastic insight into different musical traditions around the world, even if it couldn't write that music for us.