The Google AI that defeated a Go champion in 2016 was AlphaGo.
Developed by Google DeepMind, AlphaGo defeated South Korean professional player Lee Sedol in a five-game match held in Seoul from March 9 to 15, 2016. AlphaGo won four games to one, a landmark result because Lee was a world-class 9-dan professional. The match demonstrated that machine-learning systems could handle Go's enormous number of possible positions.
AlphaGo combined deep neural networks with Monte Carlo tree search. One network evaluated board positions, while another estimated promising moves; the search system then used those predictions to choose its play. Its training included human games and extensive self-play, allowing the program to develop strategies that sometimes surprised experienced Go players.
The 2016 match was not AlphaGo's first professional victory. In October 2015, an earlier version defeated European champion Fan Hui 5–0, but the win over Lee attracted global attention because of Lee's stature. AlphaGo is also distinct from AlphaZero, a later system trained through self-play across Go, chess, and shogi.