The challenge in solving the problem of artificial intelligence is to find a single clear problem definition that can be formalized, for which a solution would mean AI has been cracked. Reinforcement learning can be used to decompose this problem into smaller pieces that work together to solve the larger problem of AI.