“Michael Levin suggests that the most interesting behaviors of AI systems may not be their explicitly programmed functions but are instead "side quests" or intrinsic motivations that emerge from their interaction with a "Platonic space".”