Slashdot Mirror
A Shogi Champion Turns to Chess
FFriedel writes "Michael Jordan tried it with baseball, and it, like, didn't work out too well for him. But what about a professional Shogi champion switching to chess? Yoshiharu Habu, one of the most gifted players in the history of the ancient Japanese game, has taken a casual interest in chess - and already reached IM strength. He is currently playing in a tournament in Paris, where chess grandmaster Joel Lautier interviewed him." Shogi is a very odd game if you're used to chess. Most of the pieces have biases toward forward motion, and when you capture an enemy piece, you can bring it back into play for your side.
I presume you refer to resigning and agreeing to a draw? Go and Shogi are the same as Chess in this respect. In all three, one player may resign when he thinks he has lost. In Chess and Shogi, the players can agree that the game is drawn. (A draw is not normally possible in competition Go because they normally use fractional komi.)
Note: the rules for Go say that unless one player resigns, both players have to agree that the game is over. If one player thinks game is not over, he/she continues making moves ... which the other player may ignore.
In Japanese rules, you lose points by making
moves that your opponent ignores. In Chinese
rules it make no difference.
The difficulty is not for westerners per se. It is more a problem for novices who don't know how to judge that a position is lost. This applies equally to Go, Shogi, Chess and many other complex games with a binary outcome.
I agree that it is difficult for a game playing program to know when to resign or offer a draw. But it usually doesn't matter. Who cares if the computer is "impolite" for not resigning? :-)
Chess is much the simpler game from this point of view as the loss of pieces in chess is a one way trip towards simplicity. In Shogi the pieces can come back on and so the game does not have the same steady progress to an endgame. In practice, of course, most games do move towards a sparcer state but it is a major difference in the two games which would affect "solving" them.
TWW
"Encyclopedia" is to "Wikipedia" what "Library" is to "Some people at a bus stop"
> After that there has been virtually no progress in game strategy; all the improvement has been in hardware speed.
> Its a shame, considering that game playing is thought to be one of the easiest problem domains for AI.
I'm not very knowledgable about games specifically, but the impression that I got from my undergrad AI class a few semesters ago was that neither of these statements are true. There has been significant improvement in search techniques since 1962, and game playing is hard enough that it hasn't made a dent that is noticable to the general public. The problem of search is general to a lot of AI, not just games, so these things proceed at the same pace. Perhaps it's just that the domains that the public notices are ones where the statespace is small, but the active research problems (at least in multi-agent systems which is what I am familiar with) often seem to involve really massive statespaces, like games such as chess and go.
I also am pretty sure that the really competent game systems do considerably more than brute force. Chinook if it is the checker-player that I am thinking of has a massive endgame and opening game book, and a lot of heuristics for overall strategy. I don't think any really succesful chess player uses simple brute force either.
In short, saying that game playing is simple compared to other AI seems to me like the people at MIT before looking at vision, saying that they could have that problem solved in a few weeks. It's still wide open, and probably will be for an indefinite amount of time, along with games.
I like to give a plug for this book every now and then: Why Michael Couldn't Hit: And Other Tales of the Neurology of Sports. Even if you don't care at all about sports, this book is a fascinating read. It describes how the brain and neurology is linked with being a world-class athlete. What I found especially interesting is that the author makes a good case that there are small windows during growing up where you must play a particular sport in order to be world-class at it. If you miss the window, you miss your chance. After that, your brain does not have plasticity to devote a specialized part of itself to the sport. He also makes the case that being a world-class musician has similar windows.
I would imagine that there are similar cases to be made about being world-class at a particular mental sport such as chess.
Highly recommended.
Sometimes it's best to just let stupid people be stupid.
Chess programs are not simple brute forcers. They could only do that if they could calculate forward to the checkmate and then apply all the movements to it.
This is obviously not possible since chess games use 6-10 times as many moves as the best computer can foresee.
So the computer must rely on an algorithm to:
a) Quickly discard stupid moves. You don't have to compute 12 moves deep if after 3 moves youve already lost the queen and two towers.
b) Anticipate the opponents moves.
c) Judge and compare different positions and take the best one out of them.
All these factors are extremely complex, and there has been a lot of software-based progress in the last years.