Andrej Karpathy - Reinforcement Learning: A Flawed Approach (via searchlore.ai)