Go to Contents Go to Navigation

(LEAD) S. Korean Go player celebrates his first win over Google AI

All News 19:24 March 13, 2016

(ATTN: UPDATES throughout, ADDS more comments, photo)
By Joo Kyung-don

SEOUL, March 13 (Yonhap) -- After proving that the human-like algorithm is not invincible, South Korean Go player Lee Se-dol said Sunday he is very happy and described his victory as "priceless."

"It's my first time getting this kind of congratulations from people by just winning a single game," Lee said after the match.

At the Four Seasons Hotel in Seoul on Sunday, Lee beat AlphaGo, the artificial intelligence (AI) program designed by Google's London-based firm DeepMind, for the first time at the special man-versus-computer Go tournament called Google DeepMind Challenge Match. The self-learning program dropped its black stones in 180 moves. The five-round Go tournament is now 3-1.

"If I was leading 3-0 and lost one, it would have been really painful to me," he said. "But after suffering three straight losses and then winning one, I feel so happy about it."

Lee, who has won 47 Go events including 18 international trophies, said that his victory on Sunday could be his most precious one.

"This victory I'm not going to change it with other ones," he said. "It's a priceless victory."

Throughout Sunday's match, Lee said he found two of AlphaGo's weaknesses.

"First, AlphaGo seems to have difficulty when it is playing with black stones," he said. "Also, AlphaGo makes mistakes like it has a program bug when it sees a move that it didn't expected."

The 33-year-old South Korean had already lost the series win and US$1 million to AlphaGo after suffering three straight defeats earlier. When asked about whether he was in an unfavorable condition from the start because the Google AI already had learned from his thousands of game records, Lee said that wasn't a problem.

"Of course, I would be able to play games well if I had some information (on AlphaGo) from the start," he said. "But basically, I would say my ability wasn't enough (to challenge AlphaGo)."

Lee, a ninth-dan player who went pro at the age of 12, said that three straight losses against the AI did impact him, but wasn't immense for him to give up his tournament.

"I do have stress because the result wasn't good, but since I enjoyed playing these games, it wasn't that painful," he said. "After this victory, much of those stresses are gone."

The win has also brought him up to his usual confidence level before playing the final match against AlphaGo at the same venue on Tuesday.

"Since I won the game with white stones, I want to win with black stones next time," he said. "I think winning with the black stone is more valuable, so I want to do it."

Fellow Go players also praised Lee for displaying his style of the game and believe that he will spice up the last match.

"Lee had good decisive moves in the center," said Song Tae-gon, a ninth-dan player who commentated the match. "I now have a feeling that Lee is getting to know AlphaGo's thoughts."

DeepMind CEO Demis Hassabis congratulated Lee for his victory.

"Lee has proved why he is such a legend," the Briton said. "Because of Lee's fantastic plays, it (AlphaGo) was pressurized to make some mistakes."

Hassabis said he was also happy, but in different way than Lee. The 39-year-old emphasized that AlphaGo wasn't specifically trained to face Lee and the way that its AI gets stronger is by playing more games against an "incredible player" like Lee.

"We are also very happy to test AlphaGo, push it to its limit and find out its weaknesses," he said. "We will go back to U.K. to figure out what has happened."

Google co-founder and Alphabet president Sergey Brin said he is really looking forward to watching the next game.

"We are very impressed and excited to have this match. And huge congratulations to Lee Se-dol," he said. "I will be excited to watch the last game on Tuesday although I will be in California."


Send Feedback
How can we improve?
Thanks for your feedback!