Google’s artificial intelligence (AI) programme wins its third straight match of the complicated board game Go, giving it victory in a five-match series against one of the world’s top players.
SEOUL, SOUTH KOREA (MARCH 12, 2016) (GOOGLE) – Google’s artificial intelligence (AI) program won a third consecutive game of the complicated board game Go on Saturday (March 12), claiming victory over one of the world’s top players in a five-match series.
The victory for the AlphaGo program, designed by Google subsidiary DeepMind, over South Korean professional player Lee Sedol, the holder of 18 international titles, has surprised many, including its designers and has underscored advances in AI.
“Though I have a lot of experience in Go I have never felt before such severe pressure as I do now, and I suppose my abilities were a bit lacking to overcome that,” a dejected Lee told reporters in the South Korean capital, Seoul, after his defeat.
He said AlphaGo was not perfect and had weaknesses, although he did not elaborate on them, holding out hope that someone might beat it.
“AlphaGo shows the part of its weaknesses, so I doubt whether it has skills that can actually deliver a message to human. Therefore, I think Lee Sedol is the one who lost today, not humanity,” Lee said.
Go, most popular in countries such as China, South Korea and Japan, involves two contestants moving black and white stones on a square grid, with the aim of seizing the most territory.
Experts did not expect an artificial intelligence program to beat a human professional for at least a decade, until AlphaGo beat a player last year.
But Lee, 33, had been expected to put up stiffer competition than the player defeated then.
“To be honest, we are a bit stunned and speechless. Lee Sedol put up an incredible fight again, AlphaGo made a large territory at the bottom of the board, but Mr. Lee found some really amazing tactics to play in the territory and create a really huge co-fight,” said DeepMind founder Demis Hassibis, who earlier tweeted that AlphaGo’s victory was an “historic moment.”
Google executives say Go offers too many possible moves for a machine to win simply through brute-force calculations, unlike chess, in which IBM’s Deep Blue famously beat former world champion Garry Kasparov in 1997.
Instead, they said, AlphaGo has sought to approximate human intuition, by studying old matches and using simulated games to hone itself independently.
Lee had initially expected to win the series, even if he were to lose a match or two. Two more matches will be played – on Sunday (March 13) and Tuesday (March 15).