AlphaGo was a model developed by DeepMind which achieved computational superiority in the game of Go, which was long considered one of the most challenging games for a computer to master.
AlphaGo has a policy and value head, which predicts potential moves and predicts the likelihood of winning respectively. It was trained with Reinforcement Learning (RL), playing millions of games against itself and learning from wins/losses.
After AlphaGo, DeepMind developed AlphaZero, a neural chess engine which beat Stockfish, the state-of-the-art in superhuman chess play. In just 24 hours of self-play, AlphaZero achieved a superhuman level of play in chess, shogi, and Go by defeating Stockfish, Elmo, and AlphaGo Zero (its predecessor).