Articles

Interesting, but as an amatour go player (and computer scientist) I would claim that the game of Go has not been really cracked if it is enough to increase the size of the board from 19 to 23 to give back the upper hand to the human player.

In fact I would claim that, if there is an n>=19, such that the human player beats the computer player consistently on a board of n*n, then the game has not really been solved.

Pietro

I may not be a computer scientist but as a european 3 kyu Go Player I am trully looking forward to seeing GO Software playing an even game with the very best pro players. furthermore, this would be a great tool for people who cannot afford Go Tutoring.

I can only say this: see you in 10 years time ;)

:Ahem: This is very interesting, and I feel interested to see how brute force computation could work out. There is, however, one small problem: Even 1kyu-2dans can easily look ahead in a game of go by almost 40 moves. As one gets stronger, say a 1profesional, it is not unusual to be able to look ahead almost 80 moves at a time. As the game progresses, it is easier to look ahead further for humans, while it would be harder to computers. I am 9kyu, and I can all ready look, on average, 18 moves ahead when doing tsumego problems.

While a 12-ply depth would be useful in chess, that is sorely lacking in a professional game, where players easily go beyond 80-ply by the mid-game. I wonder how that would affect the search tree?