AlphaGo Zero Masters Chess After Just 9 Hours of Self-Play

DeepMind's AlphaGo Zero has made significant strides in the world of chess, achieving superhuman levels after just nine hours of self-play training. This neural network, initially developed for the game of Go, has now been applied to chess with astonishing results.

The Rise of AlphaGo Zero

AlphaGo Zero was developed by DeepMind and learned to play Go by playing against itself, without using any human game data. This approach allowed it to defeat the original AlphaGo 100-0 after just three days of training. The success of AlphaGo Zero in Go laid the groundwork for its application in other games, including chess.

In chess, AlphaGo Zero reached a superhuman level after just nine hours of self-play training. It won 28 games, drew 72, and lost none against Stockfish, a top chess engine. This achievement highlights the power of neural networks in mastering complex games through self-play. DeepMind's Demis Hassabis, a chess master in his youth with a rating of 2249, has been instrumental in the development of AlphaGo Zero.