Here’s a nice introduction to some of the evolutionary algorithmic approaches to producing machine-based creativity.
In this case, the focus is on creating music by starting with a selection of musical phrases, then weeding those out by comparing them to music samples selected because of their appeal to the human ear. This fitness test helps the algorithm to hone in on those of its musical phrases that will likely please humans. It then randomly switches, combines, adds and deletes notes and then runs the fitness test against that iteration. This is done again and again, with the resulting music moving more and more towards something that should maintain the appeal to humans, while still be something quite different, and new.
These approaches to artificial creativity raise all kinds of interesting questions about the nature of creativity and about humanity’s role in creative expression in the future.