Now the pressure on the South Korean top player Lee grows Sedol, on Saturday to win the third of five scheduled games in Seoul. The 33-year-old was on Thursday in the second game in four and a half hours. When opening defeat the day before the playing strength of the program have him in surprise, “today I’m more than that, I’m speechless,” he said afterwards
The very first victory of AlphaGo against one of the world’s top Go player. was seen as a milestone in the development of self-learning machines and artificial intelligence. For the strategy game Go, with its many possible play down was to last as too complex for computers. The match is open until 15 March and will be shown live on the Google Video website YouTube. It is also about a prize of one million dollars – in the victory of the software, the amount will be donated
He had to admit that the new defeat clearly failed, Lee said.. In the first game, the computer did after his feeling still “problematic trains”. “But now played AlphaGo the almost perfect match.” It will be difficult for him in the third game on Saturday, Lee, who had expressed very confident ahead of the match, the forces comparison between humans and computers to decide for themselves said. He would do his best to “win at least one game.”
commentators had until recently spoken of a close game, but the computer conquered some more space on the board. He was very impressed with the ingenious, “dangerous looking moves of AlphaGo”, the professional player Michael Redmond commented in the YouTube broadcast
The rules of derived originally from China Go are simple in principle.: Two players try on a game board – a grid of 19 vertical and 19 horizontal lines – to conquer territories. But they take turns placing black and white stones. Completely encircled opponent’s pieces are removed from the board. but on the board with 361 fields a huge number of trains is possible, which makes it even for a powerful computer more difficult to ascribe the development of the game.
The second game was expected with additional interest because the software the first train had and therefore could make the game. “AlphaGo presented pace and for Lee to win space,” Redmond said. However, the software secured relatively quickly with a loose chain black stones the top of the board – which contributed in the end to decide. In the rest of the board is a dense tile unfolded on stone.
Then the machine made by approximately forty-five minutes a train staggered Redmond in great astonishment. To get something you really rarely seen, he said, as AlphaGo a stone diagonally placed on the still largely free right side of the board in addition to a free-standing white stone from Lee. The train sends Koreans for more than a quarter of an hour to ponder and after that he spent his time much faster than AlphaGo
Lee Se-dol tried until the very end, abzuknapsen by creative positioning of individual stones parts of the occupied AlphaGo area -. But at the end it was not enough. He went as the first of the planned two hours and he had access to the additional quota, in which is provided a maximum of one minute for each train. The YouTube commentators spoke of a “dramatic end of a dramatic game”.
AlphaGo was developed by the British company Deep Mind, Google two years ago bought, according to media reports for $ 500 million. Cofounder Demis Hassabis wrote in Twitter, the second win is hard to grasp for himself. “AlphaGo has not played this game some wonderful creative traits.” In the software Although millions trains of the best human players were initially programmed – but they can even do so. Hassabis often speaks of teaching computers to think.
No comments:
Post a Comment