what's the AlphaGo

author:frank  counts:321  time:2017-01-06


From Wikipedia, the free encyclopedia
AlphaGo logo
AlphaGo logo

AlphaGo is a computer program developed by Google DeepMind in London to play the board game Go.[1] In October 2015, it became the first Computer Go program to beat a professional human Go player without handicaps on a full-sized 19×19 board.[2][3] In March 2016, it beat Lee Sedol in a five-game match, the first time a computer Go program has beaten a 9-dan professional without handicaps.[4] Although it lost to Lee Sedol in the fourth game, Lee resigned the final game, giving a final score of 4 games to 1 in favour of AlphaGo. In recognition of beating Lee Sedol, AlphaGo was awarded an honorary 9-dan by the Korea Baduk Association. It was chosen by Science as one of the Breakthrough of the Year runners-up on 22 December 2016.[5]

AlphaGo's algorithm uses a Monte Carlo tree search to find its moves based on knowledge previously "learned" by machine learning, specifically by an artificial neural network (a deep learning method) by extensive training, both from human and computer play.

History and competitions[edit]

Go is considered much more difficult for computers to win than other games such as chess, because its much larger branching factor makes it prohibitively difficult to use traditional AI methods such as alpha–beta pruningtree traversal and heuristic search.[2][6]

Almost two decades after IBM's computer Deep Blue beat world chess champion Garry Kasparov in the 1997 match, the strongest Go programs using artificial intelligence techniques only reached about amateur 5-dan level,[7] and still could not beat a professional Go player without handicaps.[2][3][8] In 2012, the software program Zen, running on a four PC cluster, beat Masaki Takemiya (9p) two times at five and four stones handicap.[9] In 2013, Crazy Stone beat Yoshio Ishida (9p) at four-stones handicap.[10]

According to AlphaGo's David Silver, the AlphaGo research project was formed around 2014 to test how well a neural network using deep learning can compete at Go.[11] AlphaGo represents a significant improvement over previous Go programs. In 500 games against other available Go programs, including Crazy Stone and Zen,[12] AlphaGo running on a single computer won all but one.[13] In a similar matchup, AlphaGo running on multiple computers won all 500 games played against other Go programs, and 77% of games played against AlphaGo running on a single computer. The distributed version in October 2015 was using 1,202 CPUs and 176 GPUs.[7]

Match against Fan Hui[edit]

In October 2015, the distributed version of AlphaGo defeated the European Go champion Fan Hui,[14] a 2-dan (out of 9 dan possible) professional, five to zero.[3][15] This was the first time a computer Go program had beaten a professional human player on a full-sized board without handicap.[16] The announcement of the news was delayed until 27 January 2016 to coincide with the publication of a paper in the journal Nature[7] describing the algorithms used.[3]

Match against Lee Sedol[edit]

Main article: AlphaGo versus Lee Sedol

AlphaGo played South Korean professional Go player Lee Sedol, ranked 9-dan, one of the best players at Go,[8][needs update] with five games taking place at the Four Seasons Hotel in Seoul, South Korea on 9, 10, 12, 13, and 15 March 2016,[17][18] which were video-streamed live.[19] Aja Huang, a DeepMind team member and amateur 6-dan Go player, placed stones on the Go board for AlphaGo, which ran through Google's cloud computing with its servers located in the United States.[20]The match used Chinese rules with a 7.5-point komi, and each side had two hours of thinking time plus three 60-second byoyomi periods.[21] The version of AlphaGo playing against Lee used a similar amount of computing power as was used in the Fan Hui match.[22] The Economist reported that it used 1,920 CPUs and 280 GPUs.[23]

At the time of play, Lee Sedol had the second-highest number of Go international championship victories in the world.[24] While there is no single official method of ranking in international Go, some sources ranked Lee Sedol as the fourth-best player in the world at the time.[25][26] AlphaGo was not specifically trained to face Lee.[27]

The first three games were won by AlphaGo following resignations by Lee Sedol.[28][29] However, Lee Sedol beat AlphaGo in the fourth game, winning by resignation at move 180. AlphaGo then continued to achieve a fourth win, winning the fifth game by resignation.[30]

The prize was $1 million USD. Since AlphaGo won four out of five and thus the series, the prize will be donated to charities, including UNICEF.[31] Lee Sedol received $150,000 for participating in all five games and an additional $20,000 for his win.[21]

On June 29th, at a presentation held at a University in the Netherlands, Aja Huang, one of the Deep Mind team, revealed that it had rectified the problem that occurred during the 4th game of the match between AlphaGo and Lee Sedol, and that after move 78 (which was dubbed the "hand of God" by many professionals), it would play accurately and maintain Black's advantage, since before the error which resulted in the loss, AlphaGo was leading throughout the game and Lee's move was not credited as the one which won the game, but caused the program's computing powers to be diverted and confused. Aja Huang explained that AlphaGo's policy network of finding the most accurate move order and continuation did not precisely guide AlphaGo to make the correct continuation after move 78, since its value network did not determine Lee Sedol's 78th move as being the most likely, and therefore when the move was made AlphaGo could not make the right adjustment to the logical continuation.[32]

Unofficial online matches in late 2016 to early 2017[edit]

On December 29, 2016, a new account named "Magist" from South Korea began to accept matches from other professional players on the Tygem server, and changed its account name as "Master" shortly on 30 December 2016, then moved over to the FoxGo server on 1 January 2017. On January 4, 2017, DeepMind confirmed that the "Magister" and the "Master" were both played by an updated version of AlphaGo.[33][34] AlphaGo won 60 out of 60 games in total (30 on each server).[35] Master played at the pace of 10 games per day. Master quickly gained notice by many players on the Tygem server due to being exceptionally skilled. Many quickly suspected it to be an AI player due to little or no resting between games. Its adversaries included many world champions such as, Ke JiePark Jeong-hwanYuta IyamaTuo Jiaxi, Mi Yuting, Shi Yue, Chen Yaoye, Li Qincheng, Gu LiChang Hao, Tang Weixing, Fan TingyuZhou RuiyangJiang WeijieChou Chun-hsunKim Ji-seokKang Dong-yunPark Yeong-hun, and Won Seong-jin; National champions or world championship second place winners such as, Lian Xiao, Tan Xiao, Meng Tailing, Dang Yifei, Huang Yunsong, Yang Dingxin, Gu Zihao, Shin Jinseo, Cho Han-seung, and An Sungjoon. All 60 games except one were fast paced games with three 20 or 30 seconds byo-yomi. Master offered to extend the byo-yomi to one minute when playing with Nie Weiping due to his old age. After winning its 59th game Master revealed itself in the chatroom to be control by Dr. Aja Huang of the DeepMind team,[36] then changed its nationality to United Kingdom. After these games were complete, the co-founder of Google DeepMind,