人機大戰，巔峰競技 - Game I: AlphaGo 棋開得勝

In a landmark battle between man and artificial intelligence (AI), the world champion of the game Go was narrowly defeated by his computer opponent. A huge milestone has just been reached in the field of AI: AlphaGo, a program developed by Google's DeepMind unit, has defeated legendary Go player Lee Se-dol in the first of five historic matches being held in Seoul, South Korea. Lee resigned after about three and a half hours, with 28 minutes and 28 seconds remaining on his clock, while AlphaGo had 5 minutes 30 seconds left.. The series is the first time a professional 9-dan Go player has taken on a computer, and Lee is competing for a $1 million prize.

Google's DeepMind AlphaGo program beat South Korea's Lee Se-dol in the first of a series of games in Seoul.

The BBC's Stephen Evans in Seoul said Mr. Lee appeared "nervous, sighing and shaking his head" at the outset of the match. Throughout most of the game he had seemed to have the upper hand but in the last 20 minutes, AlphaGo took an unassailable lead. Mr. Lee then forfeited, handing victory to his opponent.

"I was very surprised, I didn't expect to lose. [But] I didn't think AlphaGo would play the game in such a perfect manner." said Mr. Lee after the match. "A mistake I made at the very beginning lasted until the very last." DeepMind founder Demis Hassabis expressed "huge respect for Lee Se-dol and his amazing skills," calling the game "hugely exciting" and "very tense." Team lead David Silver said it was an "amazing game of Go that really pushed AlphaGo to its limits."

The Go player has won 18 world championships since turning professional at the age of 12. He said AlphaGo's early strategy was "excellent'' and that he was stunned by one unconventional move it had made that a human never would have played.

Despite his initial loss though, he did not regret accepting the challenge.  He said "I am in shock, I admit that, but what's done is done. I enjoyed this game and look forward to the next. I think I failed on the opening layout so if I do a better job on the opening aspect I think I will be able to increase my probability of winning." Lee was surprised both by how strong AlphaGo's opening was, and by some unexpected moves.

Algorithm vs intuition

The five-day battle is being seen as a major test of what scientists and engineers have achieved in the sphere of AI.  Experts say Go presents an entirely different challenge because of the game's incomputable number of move options which means that the computer must be capable of human-like "intuition" to prevail.

"Playing against a machine is very different from an actual human opponent," Mr. Lee told the BBC ahead of the match. "Normally, you can sense your opponent's breathing, their energy. And lots of times you make decisions which are dependent on the physical reactions of the person you're playing against.”

The computer program first studied common patterns that are repeated in past games, Demis Hassabis, DeepMind chief executive explained to the BBC. "After it's learned that, it's got to reasonable standards by looking at professional games. It then played itself, different versions of itself millions and millions of times and each time get incrementally slightly better - it learns from its mistakes"

Lee will face off against AlphaGo again tomorrow (3/10) and on Saturday, Sunday, and Tuesday (11pm ET the previous evening in the US.) Whatever happens in the rest of the series, AlphaGo's victory today is a colossal moment for AI — but Lee is going to be looking for revenge. "It was a very close and tense game today," says Hassabis. "When we came into this match we thought anything was possible and we still think that now — there’s still four games to go."

References:

1. Google’s AI beats world Go champion in first of five matches / 5 hours ago

http://www.bbc.com/news/technology-35761246

2. Google's DeepMind defeats legendary Go player Lee Se-dol in historic victory

By Sam Byford on March 9, 2016 02:32 am

http://www.theverge.com/2016/3/9/11184362/google-alphago-go-deepmind-result