AlphaGo was one of the key moments of modern AI, when a computer Go player defeated the Go Grandmaster Lee Se-dol. It had previosuly been believed that the complexity of Go, in particular the large branching factor of its search space made it intractable with even near-future technology. AlphaGo was initially taught using records of large numbers of human games, but then adopted adversarial techniques to improve its performance. AlphaGo uses traditional computer game techniques, such as Monte Carlo tree search, guided by heuristics from deep neural networks. AlphaZero, the successor of AlphaGo, dispenses with the need for human training.
Used on pages 3, 10, 125, 149, 155, 220, 232, 234, 572