The researchers at Google DeepMind
announced that AlphaGo, a computer program that uses deep learning methods, has defeated a professional human player in the ancient Chinese game of Go. Go has been called the most complex games ever created by humans.
Go is a two player strategy board game, where the players alternately place black and white pieces on a 19 X 19 size board, the goal is to attain more territory over your opponent. Traditionally, humans have always outperformed machines in Go. Therefore, AlphaGo’s defeat of the European champion, Fan Hui (5-0) is a significant in breakthrough in Artificial Intelligence. The AlphaGo program uses deep neural networks that that learn to mimic professional Go players, the program also learns by playing against itself.
Involved in the development of AlphaGo is computer science PhD student Chris Maddison
, who is an intern at Google DeepMind and is supervised by Geoffrey Hinton
(Hinton known as the Godfather Deep Learning). Also involved in the research is alumnus Ilya Sutskever
– who was a Research Scientist on the Google Brain project and is now Research Director at OpenAI
AlphaGo is set to play a match against the world’s top Go player, Lee Sedol, in March.
Photo: Chris Maddison
U of T News
– Google Blog
– The Globe and Mail