Google's Alpha Go defeats world No. 1 Go player

The reddened eyes of world No. 1 Go player Ke Jie of nine dan rank were in tears. When the cameras took a close up of his face, he covered his face with the hands. He was obviously sobbing inside. At around 1:00 p.m. on Saturday, Ke Jie had his third match against AlphaGo, AI-powerd Go program developed by Google’s DeepMind. While he has displayed different facial expressions and gestures during his go matches in the past, this time his face hardened as a statue when his defeat was almost decided.

Before Ke showing tears, he suddenly disappeared from the closed-door go playroom in Jiaxing, Zhejiang, China for about 10 minutes. Some 2,300 audiences including reporters home and abroad who had been watching the match with a large-sized screen installed in a conference room downstairs were stirred up saying, “Where did he go?” and “Did he give up the match?” The commentator on the scene did not make specific explanation, either. “He seemed to have cried in a men’s room. He didn’t sleep well last night and the game was not easy. It must have been too much for him deal with,” said Ke Guofan, Ke Jie’s father.

Ke of nine dan rank has a winning rate of 80 percent or higher when he begins the game with white stone. Although he made a prior request to grab a white stone this time, he gave up the match at the 209th move. At around 5:00 p.m. after the match, he came down to some 1,000 audiences who had been watching his game on the first floor and confessed, “It was painful to have a go match against AlphaGo. I was completely hopeless that I could win the game against AlphaGo.”

“I’m very sorry that I lost. I wish I could have done better,” said Ke, accusing himself. Having shown relatively bright expression‎ after the 1st and 2nd matches, his dismal look and choked voice this time elicited rounds of applaud and words of encouragement. “I admit that I had made bad moves from the very beginning when opening the game. I couldn’t sleep last night, agonizing over how to deal with AlphaGo.”

To help go lovers improve their go capabilities, DeepMind announced its plan to develop a tool that can work as a go map based on the moves AlphaGo has made so far and to publish a paper that shows how the AI-powered go player has evolved since its match against Lee Se-dol of nine dan.

Ja-Ryong Koo bonhong@donga.com