The first victory of AlphaGo against Sedol, which took place last week, was considered a very important milestone for the field of artificial intelligence, it is an unprecedented event, taking into account the ranking of the human player. The victory achieved by conquering the other three match amplifies the milestone.
The path to win the last match was not without obstacles for AlphaGo, as confirmed by Demis Hassabis, founder DeepMind (the company controlled by Google that developed the software). The game, in fact, opened with an error on the part of your computer called the next comeback. The time world champion Go Lee Sedol succumbs with the attestation of his undisputed skills: "Lee Sedol was amazing," admits Hassabis.
As agreed earlier, the machine learning algorithms and system similar to the human neural networks used by AlphaGo have managed to cut a milestone that experts have deemed extremely difficult, because of the nature of the game. The amount of moves and combinations that can fulfill in the Go game is in the first place very superior to that present in the game of chess.
To win a Go, therefore, you can not resort only to brute force by analyzing all possible combinations - activities in which they excel and computers that do not require the application of artificial intelligence techniques - it is necessary that the computer is able to develop its own approach and its own strategy by analyzing a large database of professional players moves. The prize of one million dollars that Seedol failed to grab a winning challenge destined to go down in history will be donated to charity by Google.
No comments:
Post a Comment