Google’s DeepMind revealed AlphaZero, an AI that taught itself chess from scratch in just four hours. It defeated the world-champion program Stockfish, proving that "reinforcement learning" could master any rules-based system.