Slashdot Mirror
World Computer Chess Championships Underway
azaris writes "While the FIDE World Championships for human players in Tripoli, Libya are down to the last two contestants, the computers are playing their own 12th World Computer Chess Championship in Ramat-Gan, Israel. How will the open source chess engine Crafty do against the proprietary closed engines? Will the computers play more interesting chess than their human counterparts?"
www.chessvariants.com is a fun site.
(I'd make it a clicky but I don't know how to do so)
Actually, I think the real reason the computer drew in the latest matches against Kasparov and Kramnik was psychology more than anything else. The computer does not get stressed or fatigued when it is under pressure, nor does it lose morale after a blunder (like Kramnik) or have any fears of losing (like Kasparov).
English is easier said than done.
no, definately not, having identical engines on the same computer can result in completely different games. It comes down to processing power, hash table size, and the actual structure of the engine. In this tournament they put all the engines on identical comptuers, so its the architecture of the engine that is tested.
If you are about to mod me down, keep in mind that this post was most likely sarcastic.
Computers do not settle for draws like humans do in face of complications. This will guarentee some extremely interesting endings.
Also, since Ken Thompson is making great progress on building endgame databases, the games might be all played to end.
New year Resolution: Don't change sig this year
.
Max Froumentin of W3C shows how to animate chess games by converting ChessGML to SVG with XSLT.But why would a "true AI" be susceptible to bluffs? Would a veneer of humanity laid atop an overgrown calculator somehow limit its ability to perform said calculations?
Or, to put it differently, why would I include sufficient ego in an AI to cause it to be able to underestimate an opponent?
No, I for one welcome our new AI overlords.
Why no GNU Chess ?
If the computer considers the library of Emmanuel Lasker, then it could be "more interesting". He was (in)famous for making -ahem- startling moves for someone of his caliber. Traditional chess theory would call them blunders, but they would serve to complicate the board to a degree that his opponents could not as easily determine the best moves before he could. He would thus confuse, recover, and gain advantage before they could adequately respond. A chess program designed to confuse more rigid chess programs could serve to benefit in the same way Lasker did.
blarg.
>Will the computers play more interesting chess than their human counterparts?"
#define PACMAN "ProgrammerAlgoristChessmasterMAN"
I think it becomes a game of PACMAN against other PACMAN, so I always see this as human vs human.
The games are interesting, not because they are "played" by the machines, but because they are indirectly played by the programmers.
Except a human has this neat thing called intuition.
Your intuition can tell you things that will take you hours and hours to prove on paper. Or even in your head, following logic.
Intuition is not much more than having a large sample set from which to draw and using that sample set to infer generalities. Those generalities allow you to recognize certain patterns and also reject other patterns outright so that you don't consciously consider them. When you think about it, intuition really means "I have a hunch", and those hunches are formed on the basis of past experience. There's nothing that prevents a computer from building up a sufficiently large body of samples and, with the proper programming of course, inferring patterns from it as well.
Want to improve your Karma? Instead of "Post Anonymously", try the "Post Humously" option.
... keeping Kasparov to a 3:3 draw in one challenge. Interesting bishop sacrifice it used in one of the games - one of the better AI moves I have seen I must admit :)
http://www.chessbase.com/shop/product.asp?pid=170& user=&coin=
Well, computers have played some pretty damn interesting chess in the past. Let's take the Deep Blue-Kasparov matches as a much overused and overhyped example:
1996 game 1: 23. d5 was a brilliant move on deep blue's part. I would hope that I would find such a good move there. Yes, the computer found that move through calculation, but the move itself shows a great understanding of spatial and pawn structure elements of the position.
1997 game 2: this game was riddled with awesome moves, but 23. Rec1 33. Nf5 and 24. Ra3 are the cream of the crop.
23 Rec1 is annoying and almost a human move. The computer is playing almost perfectly here. I think Kasparov has very little counterplay here. Kasparov's queenside is UGLY and he gets little to nothing in compensation.
24. Ra3 just rocks here. Deep blue is playing the Ruy Loppez like he means it. That move made me really wonder about who was behind the computer. The Ruy Lopez is a rich opening with lots of crazy details regarding strategies in each variation, but deep blue nailed them like any world level player who playes the lopez should. Basically, deep blue couldn't have forseen a Lopez variation, but found the correct strategies all the same. Also in this game, Deep Blue psyched Kasparov out of a draw. In his top form, Kasparov WOULD HAVE SEEN the draw. 45... Qe3 does it. At the least, it is a draw by perpetual check, and if Deep Blue tried to stop it he gets crushed.
33. Nf5 is a very "computer" move, but really blows away Kaspy. It doesn't make sense muc really, but brings the bishop into play fast and kind of psyched out Kasparov.
So you see, computers can play interesting chess. These are only 2 great games I have around of a computer-human match. there are others, but theses are the most dramatic.
09 F9 11 02 9D 74 E3 5B D8 41 56 C5 63 56 88 C0 is the magic number.
However I was surprised to learn that the top human checkers players can easily trounce the computer. I would guess that checkers would be orders-of-magnitude a "simpler" problem than chess. Maybe it's that chess gets all the buzz, since it's considered to be the ultimate thinking-man's game.
Checkers has all but been solved. See this Mathworld article for more info. Basically, there's an estimated 10^12 to 10^18 different positions in a game, with a possibility for only having to solve 10^9 of them. With sufficient memory (Beowulf cluster, anyone?) checkers can be completely solved such that you can guarantee either a win or at worst a draw for the first person to move.
Want to improve your Karma? Instead of "Post Anonymously", try the "Post Humously" option.
Go is much harder than Chess for computers because brute force doesn't work even remotely as well. The branching factor is much higher (until the endgame, there are 100-360 possibilities per ply, compared to a dozen or two for chess), and the depth you might need to search is much deeper (consider a ladder starting on one side of the board whose outcome depends on the stones on the other side, 30 ply down the tree, and determines the life or death of a large group).
Note that I'm not saying the good Chess programs are pure brute force. They are basically a combination of brute force and good AI working together, but the brute force is a critical component of their success. With Go, the AI has to pull all the weight, and it isn't nearly good enough.
Crafty may be open source but it looks like the rules won't allow competitors to use substantial parts of another competing program's code. So having the source available to everyone isn't a liability for Crafty.
Mebon
Don't the top-notch computer programs already do this in one form or another? I mean, it seems like such an obvious line of research for pattern matching and pattern recognition people to explore. Or perhaps not -- that may be one of the reasons that computers aren't as good at Go as they are at chess, Go appearing (to a rank beginner) to depend more on pattern recognition and less on straight-forward deductive analysis.
Intuition plays a strong role in the play of human players great and small and it is the basis by which one can understand the differences not only between human and computer players but between interesting and uninteresting chess games.
Having a hunch about the nature of a position and the posibilities therein have allowed some of the greatest tactical games ever played and this is the identifying characteristic of the matter in understanding the nature of the game. Were humans different, the game before computers would have been different: humans who saw every possibility in a continuation leading to a 'decided' position at the end of each line would have simply announced the result or range of results and the nature of chess itself would be unrecognizable to us ('Mate in at least 37 or at most 103!').
Human intuition allows the 'miracles' of chess--the elegance of chess--in those games that make the game breathtaking and that inspire players to play in the hope of generating them (think of classics like Morphy-v-Consultants or Lasker-v-thomas or many of Mikhail Tal's best games). The intuition or, indeed, inspiration, of games like those are more than instances of the inference from generalities; they are instances of a grandiose specific arising from a game's sea of possibilities. It is the elegance of a queen sacrifice leading not to ineluctable mate (a combination like the end of Morphy-v-consultants), but to a powerful attack with a favorable conclusion (say, the end of Lasker-v-Thomas or of Reti-v-Capablanca) which, as an act, is as difficult to quantify as is the word, 'beauty.'
In a broad sense, a machine's ability to process advantage takes the wonder out of the thing because you know that there is nothing going on but the examination of a great number of positions but it is hard to imagine to imagine programmers 'weighting' their programs for positions conducive to the types of continuations that made of chess-players bother with chess in the first place.
The short form of all of the above is: 'computer programs either are or will soon be the strongest players on earth, but their games tend to be dry.'
To mail me, remove the 'mailno' from my email addy.
"Yeah. It smells, too..."
I used to play high school chess (hey, I was the cool one on the team...really!). There was this kid we called "The Zapper." He had cerbreal palsy and most people, the first time they saw him, respected him for working against his physical limitations and competing like anyone else. And he wasn't that bad for his age. But...he was a terrible loser! He would literally throw hissy fits and knock the pieces all over the table, the floor, where ever, when he lost, and yell out obscure threats. He was one of these people who used his condition as an excuse to be a big baby whiner, and I found it quite shocking. It helped me see past handicaps to the people beyond, good and bad. Good people come in all shapes and sizes, and so do bad people.
As a teenage chess player, I had long hair and listened to loud, hard rock and metal on my walkman, but I would play really boring, solid moves. I got a draw off Boris Spassky in an exhibition once playing the Caro-Kann. My friend played a double King Pawn and lost in 5 hours in a wild King's Gambit game, the last game going. I kind of wish I'd played more aggresively now, although I cherished the draw for many years and had a calculus test to study for.
Professor of Astronomy, Author of Spider Star & Star Dragon (Tor)