[youtube]https://www.youtube.com/watch?v=mzpW10DPHeQ[/youtube]
Welcome to theÌę91avÌęcoverage of a momentous match in Seoul, South Korea. This video is a livestream of the Go match between Lee Sedol, one of the worldâs best players, and AlphaGo, Googleâs artificial intelligence program. The two will be duking it out over the next week. Itâs currently 3-1 to AlphaGo.
Advertisement
Like Gary Kasparov versus IBMâs Deep Blue in the 1990s, Lee Sedol versus DeepMind is a milestone in artificial intelligence. Itâs a test of the power of current AI and how society reacts to it. And itâs a litmus test for the work we can expect AI to start doing in the near future.
Updates will be appearing here regularly from our reporter, Mark Zastrow, on the ground at the Four Seasons hotel and from our team in London who will be analysing the news.
What is Go?
The game of Go involves two players placing black and white counters to conquer territory. It is played on a 19 by 19 board, which allows for 10171 possible layouts, versus roughly 1050 possible configurations on a standard 8 by 8 chess board. To give you a sense of scale, itâs estimated there are 1080 atoms in the universe. âGo is probably the most complex game ever devised by man,â says DeepMind founder Demis Hassabis.
The matches will start at 1 pm local time on the following days (commentated on in Korean and English):
9 March: First match 1-0
10 March: Second match 2-0
12 March: Third match 3-0
13 March: Fourth match 3-1
15 March: Fifth match 4-1
Final update: Itâs over. The historic match between Google bot AlphaGo and human Go grandmaster Lee Sedol drew to a close this morning.
It was a victory for artificial intelligence, with AlphaGo taking the final game at the Four Seasons hotel in Seoul in South Korea to win the series 4-1.
Lee started game five with a similar strategy to the previous game, grabbing territory in the corners in a defensive crouch, inviting AlphaGo to overplay in the middle. âDare I say, this might be its weakness,â said Ben Lockhart, a top US amateur studying Go full-time in Seoul.
But AlphaGo clawed its way back, until the game hung so thinly in the balance, the commentary crews couldnât agree on who was ahead.
In the end, it was AlphaGo who finally won, having played the game out to the slimmest of margins.
âWhat a heartbreaking end to this series,â said Andrew Jackson of the American Go Association.
After the game, Lee said the experience had led him to question his beliefs about Go.
âI donât believe that AlphaGo is superior to me,â he said. âI think there is still more humans can do against artificialÌęintelligence.â
Read theÌęstory of the fourth match, in which Lee finally wins a game to bring the series to 3-1Ìę(a short excerpt is below):
Lee started conservatively, ceding the centre of the board. But on move 78, he turned the game around with an astonishing wedge play in the middle. Gu Li, one of Leeâs rivals, commenting on the game in China, called it the âhand of godâ, the kind of language that was being applied to AlphaGo in previous games.
AlphaGo responded to the unexpected move with a weak counter, which set off a brilliant sequence from Lee to capitalise. According to Demis Hassabis, one of DeepMindâs founders, AlphaGo didnât realise its mistake until eight moves later. âLee Sedol beat AlphaGo at its own game,â said Jackson.
Lee sat up straighter as he closed in on victory. The press room began to buzz as AlphaGo played increasingly bizarre moves, the death throes of an algorithm. Across the table from Lee, Aja Huang, the AlphaGo programmer who acts as its human avatar and places its stones on the board, appeared resigned. And then, a message appeared on his monitor: âThe result âwhiteresignâ was added to the game information. AlphaGo resigns.â
In his moment of victory, if Lee cracked a smile, I didnât see it. His focus was back on the board, replaying the game.
Read theÌęstory of the second match, in which the AI went 2-0 upÌęin the five-match seriesÌę(a short excerpt is below):
In the minutes after todayâs match ended, Jackson said AlphaGoâs flawless play left him in shock. âThings that looked questionable in hindsight turned out to be correct. Thatâs its hallmark.â
At yesterdayâs post-game press conference, Lee looked shell-shocked. Today he seemed resigned. One reporter asked what AlphaGoâs weaknesses are? Lee laughed: âWell, I guess I lost the game because I wasnât able to find out.â
As Lee filed out of the press conference, his thin frame lit up by flashing cameras, a lone Korean reporter cried out, âLee Sedol, fighting!â
No one responded.
Ìę
Read theÌęstory of the first match to give you a taste of the action so far as the AI went 1-0 up in the five-match seriesÌę(a short excerpt is below):
As the endgame played out, Lee appeared downbeat, his hands shaking at every move. He flashed a wan smile as he resigned.
The question is whether Lee can recover. In AlphaGoâs previous match, experts say Hui Fanâs moves grew erratic after dropping the first game.
âItâs so shocking. I expected AlphaGo to win one game, but I didnât expect it to be the first one,â said Myung-wan Kim, a Korean 9-dan professional living in Los Angeles and commenting with Jackson.
âI am in shock, I admit that,â said Lee. But he said he had no regrets about accepting the challenge and planned to come back strong in the rest of the matches. âI didnât think AlphaGo would play the game in such a perfect manner.â
Read more:ÌęDoes a machine beating a Go master mean human-like AI is close?
For more on machine minds, watch our Explanimator video:
[youtube]https://www.youtube.com/watch?v=bctMvKrB_y0[/youtube]
Ìę
Ìę
