In March 2016, DeepMind’s AlphaGo defeated Lee Sedol 4–1. The system used neural nets trained on human games and self-play to guide MCTS.
Move 37 in game 2 became a symbol of machine creativity. The world saw AI could shine in domains of intuition, not only pure tactical calculation.
AlphaZero (2017) removed initial human learning in chess, shogi, and Go. The self-play + search thread remains central in agents and AI for science.
