Future of Go Summit

The Future of Go Summit (Chinese: 中国乌镇围棋峰会) was held by Chinese Go Association, Sport Bureau of Zhejiang Province and Google in Wuzhen, Zhejiang, the permanent host place of the World Internet Conference, in May 2017, featuring five Go games involving AlphaGo and top Chinese Go players,^[1] as well as a forum on the future of AI.^[2][3] It was Google’s biggest public event in partnership with the Chinese government since Google China's search engine was moved out of mainland China to Hong Kong due to the government censorship in 2010. It was seen as a charm offensive launched by Google toward Chinese officials, being part of effort to reopen China's market.^[4][5]

The version of AlphaGo used in this Summit was AlphaGo Master.^[6][7] It used one TPU on a single machine.^[8] DeepMind claimed that this version was 3-stones stronger in games of self-play against itself than the version used in AlphaGo v. Lee Sedol, with Elo rating higher than 4,500.^[9]

After winning its three-game match against Chinese grandmaster Ke Jie, the world's top Go player, AlphaGo was awarded professional 9-dan by Chinese Weiqi Association.^[10] DeepMind announced that AlphaGo would retire, and DeepMind would disband the team that worked on Go and spend their time exploring new AI in other areas instead of Go. After the Summit, DeepMind released 50 games AlphaGo played against itself.^[11]

Schedule

• Ke Jie vs. AlphaGo. 23, 25 and 27 May, all of them from 10:30 – 17:30 UTC+8.^[12]
• Pair Go: Gu Li and AlphaGo vs. Lian Xiao and AlphaGo. Friday 26 May, 8:30 – 12:30, UTC+8.^[12][13]
• Team Tournament: AlphaGo vs. Chinese team (Chen Yaoye, Mi Yuting, Shi Yue, Tang Weixing, Zhou Ruiyang). Friday 26 May, 13:30 – 18:30, UTC+8.^[12][14]

Games

Ke Jie vs AlphaGo

Google DeepMind offered 1.5 million dollars winner prizes for this match while the losing side took 300,000 dollars for participating in the three games.^[15][16][17] AlphaGo won all three games.^[18][19]

Pair Go

The pair Go game was played on 26 May morning. Gu Li, an eight-time world champion, and Lian Xiao, the current Chinese Mingren and Tianyuan titles holder, each had their own AlphaGo teammate, alternating moves in tag team style.^[20] The move sequence was Gu Li (black), Lian Xiao (white), Alphago (black), and AlphaGo (white).^[21]

First 99 moves

Moves 100–199

Moves 200–220

After move 198, the black AlphaGo proposed resignation, but rejected by its teammate Gu Li.^[22] The white AlphaGo and Lian Xiao finally won by resignation after move 220.

This pair Go game provided a glimpse of how human players might be able to use AI tools in the future, benefiting from the computer's insights while also relying on their own intuition.^[20] After the match, David Silver, the lead researcher for AlphaGo said: "Pair Go was a beautiful game. All four players made some brilliant and creative moves. It felt like four painters working together on a shared canvas, all with different styles, all combining together to make something truly beautiful."^[23]

Team Go

The team Go match was held on 26 May afternoon. In this five-on-one match, five world champion players, Chen Yaoye, Zhou Ruiyang, Mi Yuting, Shi Yue, and Tang Weixing (black), joined forces to play against AlphaGo (white).^[24] There was a two-and-a-half-hour set time limit for each side followed by three 60-second byo-yomi overtime periods.^[25] Consulting each other on every move, the human team approached the challenge in a light-hearted manner, obviously enjoying the experience of playing together.^[20] Chosen as the team leader, Zhou Ruiyang positioned the stones on the board during regular time.^[25] Tang Weixing put down the stones in overtime, almost working on his own,^[25] and finally resigned in the face of AlphaGo’s certain victory.^[26] Eric Schmidt, the Executive Chairman of DeepMind's parent company Alphabet, Inc. tweeted: "This speaks volumes about where AI is headed - human players are teaming up with AlphaGo to have even more fun with the game!".^[27]

First 99 moves (87/93/99 at 65, 90/96 at 66)

Moves 100–199 (105/111/119/125/131 at , 108/114/122/128 at 102, 172 at 115)

Moves 200–254 (219 at 213)