From May 23, Google’s artificial intelligence program AlphaGo is taking on the world’s No. 1 player Ke Jie in three games in a five-day festival of Go in the game’s birthplace, China.
The millennia-old board game of Go had long been thought of being beyond the reach of AI algorithms thanks to its notorious complexity, with each move involving a much higher number of possible plays than, for example, chess. But AlphaGo’s previous sweeping wins against a long list of top human players have proved that AI can play the game at a level no human likely ever could attain, something thought impossible as recently as 2014.
That’s why 19-year-old Ke stands as the last hope against complete machine domination over one of the world’s most difficult board games.
The backstory
Just over a year ago, AlphaGo set a milestone in the history of AI by taking down South Korea’s Lee Sedol, one of the world’s best Go players. But Ke didn’t want to play AlphaGo at the time, instead agreeing to take on the AI in the future.
This January, Google’s DeepMind machine-learning team secretly put an upgraded AlphaGo onto two online board-game platforms to test its skill against top human players. During the test, AlphaGo racked up 60 wins and zero losses. Ke lost all the three games he played with the AI.
Ke revealed that he was taken to hospital after the three losses, but didn’t provide details. If that were not the case, Ke wrote on China’s Twitter-like Weibo, he would have used the “one last move” he had prepared.
Last month, DeepMind announced that the “Future of Go Summit,” co-hosted by the Chinese government, would be held May 23-27 in the eastern Chinese coastal town of Wuzhen, and that the centerpiece of the event would be a three-game match between Ke and AlphaGo.
The odds
AlphaGo can continue to improve, playing against itself and humans, learning as it simulates, which makes it harder and harder to conquer as each day passes.
Fellow player Gu Li, a former world champion, gives Ke only 10% chance of winning just one game against AlphaGo; AI expert Lee Kai-Fu, a former Google and Microsoft executive, says Ke’s chance of winning even one game is almost zero (link in Chinese). Still, Korea’s Lee did win one game against AlphaGo last year—apparently unpredictability works.
Deepmind has said that AlphaGo isn’t meant to diminish Go, but to make human players “stronger and more creative,” as they test out new moves and strategies after being inspired by the AI.
Team play
Other matches will feature professional Go players teaming up with and against AlphaGo. In the “Pair Go” format, a Chinese player will play against another, but both will have their own AlphaGo teammates that take over every other turn. In the “Team Go” format, five Chinese players will try to defeat AlphaGo together. These matches are scheduled in between the three games between Ke and AlphaGo.
How to watch live
The first game between Ke and AlphaGo is scheduled May 23 10.30am Beijing time (May 22 10.30pm EST); and the second and final game will take place at the same time on May 25 and May 27, respectively.
DeepMind says it will livestream the matches and post updates and expert comments on this page and on its Twitter account.