Ben Morss, human
For my senior thesis, I created a program that did its best to compose original tunes. At the time, neural nets didn't seem real promising. So I created a system that used a Markov chain to generate a random musical motif, that analyzed that motif and determined likely consonances and dissonances, that created a chord progression from a generative grammar, and which then tried to fit that motif and variations over the resulting chords. My idea was that it was good to start with maximum randomness to simulate "creativity", then work hard to fit that into a logical structure. The difficult part was determining heuristics to navigate the search space.
If you're interested in this project, please get in touch!