 |
| Korean professional go player Lee Se-dol, right, plays against the artificial intelligence system AlphaGo during the Google DeepMind Challenge Match at the Four Seasons Hotel in central Seoul, Wednesday. On the left is DeepMind's lead programmer Aja Huang putting stones on behalf of AlphaGo. / Courtesy of Google Koea |
Google's AlphaGo scores surprise win against world go champion in first game
By Yoon Sung-won
The artificial intelligence (AI) machine AlphaGo astounded the world by beating world go champion Lee Se-dol in the first round of a historic five-game match, Wednesday.
AlphaGo's early dominance also contributed to its win despite mistakes at the end of the game, making it neck-and-neck before Lee resigned.
"I was so surprised because I did not think that I could lose," Lee said after the match. "In terms of the match, I think my mistakes in the early stages led to defeat in the end. I did not imagine that AlphaGo would play such a perfect game."
Lee said he does not regret accepting the challenge and expects to win the remaining games.
"I really enjoyed the game and do not regret accepting the challenge," he said. "I have won several world championships and will not be shaken after defeat in the first game."
Lee's loss came as a shock to the international go world as many assumed that machines still have a long way to go to outperform experts like him in go, which requires not only accurate calculations but also intuition.
"AlphaGo did not play like a machine, although it excluded the emotions of a human professional go player," said Kim Seong-ryong, a professional 9-dan go player who was a commentator on the match. "AlphaGo remained coolheaded even after some obvious mistakes that many top-class professionals said were critical."
DeepMind's lead programmer Aja Huang placed the stones on behalf of AlphaGo and Lee played the game with black stones.
The game, which was the first out of five that will run through March 15, was highly contested from the beginning with the AI system playing better than expected.
Lee, who is said to have a complicated and thrilling playing style, tried irregular moves to shake the AI. But AlphaGo did not stagger and made aggressive moves in turn as the two players proceeded to the central part of the board after one hour into the game.
At 4 p.m., three hours into the match, AlphaGo made several mistakes as the game drew near to an end and Lee almost seemed to seize the chance to win.
But in the end, he failed to secure enough points to overcome the 7.5 komi points he yielded in starting the game.
"I am glad that we could complete the match without any technical glitches," said David Silver, Google DeepMind's go project team leader.
More than 250 reporters from around the world gathered at the match arena in the Four Seasons hotel in central Seoul.
Political figures including Seoul Mayor Park Won-soon, Minjoo Party of Korea interim leader Kim Jong-in, Saenuri Party floor leader Won Yoo-chul and Science, ICT and Future Planning Minister Choi Yang-hee visited the venue.
The Google DeepMind Challenge Match between Lee and AlphaGo was hosted by Google, as a way to demonstrate the AI system developed by its London-based subsidiary DeepMind, based on its advanced machine learning technology. A $1 million prize is up for grabs.
In October last year, AlphaGo beat the three-time European go champion Fan Hui 5-0. This came as a surprise to the go world as it was the first victory of a machine against a professional human player in the game.
Lee is the world's top go player with 47 wins and 18 world championships. He started his professional career in 1995 when he was 12 years old, and clinched his first international title at the Fujitsu Cup International Go Championship in 2002.
Originating in China about 3,000 years ago, go is the oldest board game played at a professional level. In the game, two players compete to surround more territory as they alternately place black and white stones on a board with 19 by 19 horizontal and vertical lines.
Humans have already been defeated by AI in other games such as chess.
![[INTERVIEW] Hip-hop group Uptown returns after 13 years with new lineup](https://img.koreatimes.co.kr/upload/thumbnailV2/0eae35da5e9c40c38cf439e550a2eb3e.jpg/dims/resize/178/optimize)
![[INTERVIEW] ASTRO members aim to shine in musical theater](https://img.koreatimes.co.kr/upload/thumbnailV2/df16760dc50f469d93434f66a069465c.jpg/dims/resize/178/optimize)
