Some artists do use randomness as part of generative music techniques, and it sparks a lot of controversy. In actual fact, randomness is not very useful for composition simply because even if you get something really nice out, you aren't going to get it again. So you either have to record everything, which becomes impractical, or you head into John Cage territory where you embrace the non-reproducibility and it becomes part of the art.
Far more effective to simply wire your studio up in such a complex manner that it behaves chaotically, which leaves you with all the serendipitous, 'surrendering to the machine' type benefits of randomness, but with the added benefit that if you hit rewind and play you will hear the same thing again.
The almighty Autechre, known for their generative music, are quizzed on this in their very interesting and geeky Sound on Sound interview: http://www.soundonsound.com/sos/apr04/articles/autechre.htm (the bit on randomness is halfway down in the "talking about our generation" box.)
Warning: UX people may have a slight breakdown when confronted with the screenshots of their custom software: http://media.soundonsound.com/sos/apr04/images/autechremax2.... (the color spectrum in the bottom right is what fascinates me, talk about indirection...)
You should read the article before commenting.
(Incidentally that Bernstein lecture is great, thanks for linking it)
If you really want to learn about music, I think it's better to study good music. If you thought adding rhythm had surprising effects, this video will blow your mind: http://www.youtube.com/watch?v=g0ZE38BQmvQ It's an analysis of Mozart in Leonard Bernstein's Norton lectures at Harvard.