Can A Computer Finally Beat A Human in Go?

by Ying Ying Yang

Recently, Google announced that they developed a computer go program which has the potential to beat professional go players. (And in fact, Alphago did beat world’s second best player a few days ago). Go is a traditional Chinese board game and for decades computer programs can only beat amateur go players. Alphago, therefore, might be a huge breakthrough in the world of go.

What is Go?

Go is a traditional Chinese board game developed in Zhou Dynasty (1046–256 BCE). It is played by two players, who each holds black or white pieces. On a board with 19*19 grid of lines and 361 crosses, players put pieces on the crosses. The goal of this game is to encircle as many areas as possible, and the player who has more areas wins.



Why is Alphago so important?

AI program has been developed and defeated human in many board games. For example, a perfect algorithm for checker was developed through AI. But in the field of go, AI could never beat professional players. There are hundreds of crosses on the chess board, therefore hundreds of possibilities to place the piece. After one step, there are also another hundreds of possibilities. This calculation would be too large for computers. Human beings, however, don’t calculate the possibility of winning for every cross. Instead, we would just focus on a few crosses and decide on the best one. What is important in Alphago is that it tends to think and play more like human. First it analyzes the game strategically to decide on a few spots, thus avoiding mass calculation. Then, since the choices are narrowed down, it can calculate more precisely than human players and outcompete them.

However, many go players would not enjoy Alphago. If Alphago continues to be undefeated, is it still worthy to be a professional go player? What if Alphago finds a perfect algorithm of go? On the other side, Alphago can be a huge advantage. It can be used in go education, and it can be a good practicing go partner for human players.