AlphaGo

AlphaGo is a computer programme developed by Google DeepMind that uses deep reinforcement learning and Monte Carlo tree search to play the board game Go. In March 2016, AlphaGo defeated the reigning world champion, South Korean professional Lee Sedol, 4–1 in a five-game match held in Seoul. The match was widely covered as a landmark in artificial intelligence: Go had been considered intractable for computers due to its astronomical branching factor (roughly 10^170 possible board positions), and prior estimates had placed human-level Go performance at a decade or more away. The win came 18 months ahead of the most optimistic expert predictions.

AlphaGo was built on two deep neural networks, a policy network to select moves and a value network to evaluate board positions, combined with Monte Carlo tree search. The earlier AlphaGo Fan version defeated European champion Fan Hui in October 2015; the version that played Lee Sedol was AlphaGo Lee; a subsequent AlphaGo Master version defeated 60 online games against top professionals in January 2017. The programme was created by a team led by David Silver and Demis Hassabis at DeepMind. The Nature paper (January 2016) describing the system became one of the most-cited AI papers of the decade.

AlphaGo was superseded by AlphaZero in 2017, which matched and surpassed AlphaGo from scratch in 24 hours using a single generalised algorithm without opening books or endgame databases. AlphaGo represents the moment reinforcement learning moved from academic interest to mainstream awareness; the Lee Sedol match was watched by an estimated 200 million viewers worldwide. Its creator, David Silver, later Left DeepMind to found Ineffable Intelligence in 2025, raising the question of what a fully autonomous, label-free system might achieve where AlphaGo's human-curated training data once set the ceiling.