This is a very interesting article into training reinforcement learning algorithms on environments that they create themselves.
This is an incredibly detailed and interesting exploration into the world of creative artificial intelligence, by Samim Winiger and Roelof Pieters.