It's interesting that the same ideas are finding their way into AI lately. The AlphaStar Starcraft 2 agent was trained first using imitation learning to mimic human games, then was evolved using a sort of cultural evolution whereby different agents competed and the best were saved for the next generation.