SEOUL, March 10 (Korea Bizwire) – Google’s artificial intelligence (AI) program AlphaGo defeated South Korean Go player Lee Se-dol for the second straight time in the ancient board game Thursday, proving that its earlier victory wasn’t a coincidence.
In their second showdown at the Four Seasons Hotel in Seoul, AlphaGo, the AI designed by Google’s London-based firm DeepMind, collected another win by resignation in 211 moves after playing nearly four and a half hours.
The self-learning machine now leads 2-0 at the five-round tournament named Google DeepMind Challenge Match. AlphaGo already made history Wednesday by forcing Lee, a top class Go player who has 18 international events, to throw his black stones in just 186 moves.
Lee, a ninth-dan player who went pro at the age of 12, will take a full-day rest Friday before he aims to even the score against AlphaGo this weekend in their third and fourth matches. The tournament, which rewards US$1 million to the winner, wraps up next Tuesday.
Switching the stones played on Wednesday, Lee this time played with the white stones, which gives him “komi,” or 7.5 compensation points, at the end of the match under the Chinese rules. In this tournament, scoring calculations are also based on the Chinese rules, also known as “area scoring,” in which a player’s score is determined by the number of stones that the player has on the board and territories earned.
AlphaGo, whose moves are actually placed on the board by Google DeepMind programmer and amateur 6-dan Go player Aja Huang, opened its game by placing the black stone at the “flower point,” or 4-4 point, in the top right corner in just five seconds. This was 1 minute, 25 seconds faster than its first move made in the first match. Lee also placed his stone at the flower point, but in the left bottom corner.
AlphaGo then made a new move that it has never shown before against professional Go players by placing its black stone at the 3-4 point in the top left corner. In its five matches against European champion Fan Hui last October, which AlphaGo won 5-0, and Wednesday’s match against Lee, the human-like algorithm always occupied 4-4 points in its first two moves.
Taking the 4-4 point on the board, also called the “star point,” is considered to be the standard opening move in modern Go games.
As the machine hinted it would play a somewhat different game from the past, Lee went opposite of his usual aggressive and unconventional style of play.
After the two sides engaged briefly in the right bottom corner, AlphaGo placed a surprising 13th move at the top center, which South Korean commentator Kim Seong-ryong, a ninth-dan professional Go player, described as “a move that would not exist in a human Go game.” Lee answered back by placing an unconventional move in the left side center.
But the 13th move wasn’t as confusing as the 37th move of the game, which AlphaGo put right in the center of the board. While commentators were also hammered by the AI’s mysterious move, Lee left his seat, had a sip of coffee and stared at the board to figure out the counter to AlphaGo’s move. It took more than 20 minutes for Lee to finally place his stone.
Two hours later their tug-of-war on territories heated up, breaking out from the left bottom corner. An hour later, the two players fought in the center of the board as Lee jumped into the center to challenge AlphaGo’s black groups. At that point, Lee had 17 minutes in his clock, while AlphaGo had 44 minutes.
While exchanging moves in the top right corner, Lee ran out of his allotted time of two hours, and had to place his stones in 60 seconds. In this tournament organized by Google and the Korea Baduk Association (KBA), each player has three lots of 60 seconds overtime counting after their given time is expired.
Pressured by time and the cold moves of AlphaGo, Lee tried to get territories in the center in the final phase, but the machine’s high-level calculations didn’t allow the human to take the advantage. AlphaGo also entered its overtime use by the end of the game, but the 33-year-old South Korean was too late to reverse the game.
“We really underestimated AlphaGo’s level,” said ninth-dan commentator Kim Seong-ryong. “If all of AlphaGo’s moves were calculated to the end, it’s impossible for a human to win.”
Go, known as “baduk” in Korea, originated in China more than 2,500 years ago. It involves two players alternately putting black and white stones on a checkerboard-like grid of 19 lines by 19 lines. The object is to claim larger territories than one’s opponent by surrounding vacant areas of the board using one’s own stones.
Go has been viewed as one of the hardest games for computers to master. Google said the possible number of board configurations of Go is larger than the number of atoms in the universe.’