"The notoriously intricate strategy game Go has a new champion.
AlphaGo, Google’s Go-playing computer program, has topped Lee Sedol, the world’s reigning player, in a five-game [4-1] match played over a week in South Korea.
The program, a breakthrough in artificial intelligence, learned to play Go by watching human experts and by playing millions of games against itself."
According to Wiki:
The objective of the game—as the translation of its name implies [encircling game]—is to have surrounded a larger total area of the board with one's stones than the opponent by the end of the game.[8]
Once placed on the board, stones may not be moved, but stones may be removed from the board if captured. This is done by surrounding an opposing stone or group of stones by occupying all orthogonally-adjacent points.[9] The two players place stones alternately until they reach a point at which neither player wishes to make another move; the game has no set ending conditions beyond this. When a game concludes, the territory is counted along with captured stones and komi (points added to the score of the player with the white stones as compensation for playing second) to determine the winner.[10] Games may also be won by resignation.
...In combinatorial game theory terms, Go is a zero-sum, perfect-information, partisan, deterministic strategy game, putting it in the same class as chess, checkers(draughts) and Reversi (Othello); however it differs from these in its game play. Although the rules are simple, the practical strategy is extremely complex.
The game emphasizes the importance of balance on multiple levels and has internal tensions. To secure an area of the board, it is good to play moves close together; however, to cover the largest area, one needs to spread out, perhaps leaving weaknesses that can be exploited. Playing too low (close to the edge) secures insufficient territory and influence, yet playing too high (far from the edge) allows the opponent to invade.
It has been claimed that Go is the most complex game in the world due to its vast number of variations in individual games.[112] Its large board and lack of restrictions allow great scope in strategy and expression of players' individuality. Decisions in one part of the board may be influenced by an apparently unrelated situation in a distant part of the board. Plays made early in the game can shape the nature of conflict a hundred moves later.
...Go begins with an empty board. It is focused on building from the ground up (nothing to something) with multiple, simultaneous battles leading to a point-based win. Chess is tactical rather than strategic, as the predetermined strategy is to trap one individual piece (the king).
4 comments:
There's no intelligence, just programmers discovering a mechanical way to do something.
The article says the computer watched...hmmm.. human players, and played a million games against itself. I guess you are saying the computer would be able to do nothing without the initial programming.
Everything is mechanical. Usually a position can be evaluated roughly for goodness, and that's used to evaluate a possible series of a few moves ahead of time. Then it picks one that forces the best result in say five moves.
Programmers change what the rough evaluation is based on.
Onoe way to do that is to play against itself with one side looking one move further ahead, and using it to tune the goodness criterion for the one move less side.
Then the programmer is writing the code that weights and unweights various measures of goodness.
But there's nothing intelligent going on. Just add up the scores and pick the best move according to the scoring.
Thanks rh :)
Post a Comment