Google invents AlphaGo to dominate the game of Go

Go is a game that has been around for more than 2500 years and originated in China. It's still commonly played today and involves people taking turns placing black or white stones on a board to capture the opponent's stones or surround an empty space capturing territory.

The game is incredibly complex with Google stating that there are over 1,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000,000 possible positions. Google also claims that is more positions than there are atoms in the universe.

The massive number of possible positions means that the game is incredibly complex to play making it hard for computers to play the game with skill, until now. Traditional AI methods with a search tree for possible positions don't work in Go. Google's system is called AlphaGo and combines an advanced search tree with deep neural networks.

The neural network takes a description of the Go board and process it through 12 network layers with millions of neuron-like connections. One neural network is the policy network and selects the next move; the value network predicts the winner of the game. The system is able to predict the move of the human opponent 57% of the time. AlphaGo beat all other computer players 499 out of 500 times. In a match against European Go champion Fan Hui, AlphaGo won five games to zero marking the first time a machine ever beat a professional Go player.

SOURCE: Google