It is interesting, but their approach is not what I would use. They seem to use sheer computing power to model sounds, rather than modelling music. But certainly that does allow them to model any sounds whatsoever, which in itself would open up interesting possibilities (modelling the sounds of nature, for example).
