« Headaches | Main | Should old people move to Panama? »
Kramnik vs. the computer
The match started today, six games, here is a good overview article. Kramnik gets a million dollars if he wins, $500,000 otherwise.
Here are the rather complicated match rules; Kramnik has unparalleled access to the opening book and workings of the machine. Here are numerous expert opinions, many favor Kramnik. Sorry guys, but I predict one computer victory and the rest draws. Here is commentary by Kramnik. You can watch the games live here.
One commentator put it well: "The last match was drawn - against a weaker version of Fritz on lesser hardware. And there's no reason to think that since that match, Kramnik has learnt to calculate an extra 6 billion positions per second."
But so far, in game one, the outcome was a draw and Kramnik had a slight edge throughout...
Addendum: Kramnik missed a win.
Posted by Tyler Cowen on November 25, 2006 at 03:48 PM in Sports | Permalink
Comments
Of course, I love the chess posts by Dr. Cowen. Further down in the Chessninja comments there is a line about Fischer Random chess. I agree that top GM's only hold their own now because of openning knowledge; eliminate a standard starting position and computers destroy our feable attempts.
Posted by: Richard Pointer at Nov 25, 2006 6:09:04 PM
I predict a computer win by at least 4-2.
Posted by: John T. Kennedy at Nov 25, 2006 7:16:16 PM
Huh. When I (1 dan amateur) won't have to give the best computer programs 9 stones handicap at Go, I'll take notice, perhaps.
Posted by: A Tykhyy at Nov 25, 2006 8:21:36 PM
I'd think the randomized chess would be better for people due to the extensive use of databases of masters' chess games by programs.
Posted by: bhauth at Nov 26, 2006 1:09:57 AM
Another interesting point was the comment by Kramnik:(4th link)
"The day will come when we will no longer have a chance against computers.If I indeed manage to beat Fritz in this match it will probably be the last time that a human being wins against a computer."
He has already hedged himself!(So that means he will be able to ask for greater prize money in the future irrespective of the current outcome just to make the effort worth it.)
Posted by: ajoy at Nov 26, 2006 3:40:14 AM
While computers do use openning books to get to the middle games, the amount of tactical problems presented by a fischer random position would be greater for the human than the computer. At least that is my observation.
Posted by: Richard Pointer at Nov 26, 2006 4:36:29 AM
wow~~~~!!!
not thinking people's lost, i wish krmnik wins computer
Posted by: astraea at Nov 26, 2006 7:35:50 AM
bhauth: I must agree with RP. Computers need extensive databases of openings in order to compete with humans who understand strategic principles on a very deep level. Because computer programs (as written) do not understand those principles except as they are emergent from the tactical considerations within their seen search tree.
Computers don't need databases of "tactics" because their simple ability to search faster, wider and deeper overwhelms human tactical ability in situations where the tactics are not fairly well circumscribed by important and well understood strategic considerations.
Basically, get a human out of a position s/he understands very well, and Fritz will destroy it completely. Get Fritz out of it's book, and the human (if it's Kramnik or similar) will have a slight advantage that *might* lead to a win.
Posted by: Michael Sullivan at Nov 26, 2006 8:29:41 AM
Well it has been years since I played competive chess, and I was not particularly good then but from the description of the rules Kramnik has a much better chance than people are concluding. It really depends on how deep Fritz's opening book is.
If I am playing an inferior player I really don't spend a lot of effort worrying about what he is thinking, I develop my positon using the basic principles of space, time and piece defence and count on solid play carrying me through until he makes the first of a series of mistakes to let me enter the middle game with an advantage.
If I am playing a player of similar talents I had two strategies back in the day. One was to work from a memory of an opening book, which if you have a good memory and a lot of time to study could carry you a long way. Well I had a good memory but was fundamentally lazy so except on the real standbys like Ruy Lopez rarely went more than three moves in. Instead I just used the same principles outlined above but spent some time thinking about what my opponent was doing. But if I had open access to his opening book I would have a pretty good idea of what he was doing and so spend my time on the counter, itself to a degree guided by the opening book. It is only when the opening book stops or the computer makes a move outside of it does Kramnik have to start actually playing chess, and under the rules the arbiter will have informed him that the computer is now calculating variations, or in real terms actually playing chess rather than relying on the memory of a table.
The deeper Fritz's opening book goes the more advantage there is for Kramnik. As long as he has equal access he can almost effortless maximize his position and by noting those moves not covered by the opening book can choose his time and place to force Fritz to actually start playing (i.e. calculating) chess.
Like I said my days of playing competitive chess are decades back. But memorizing an opening book and specializing in the more exotic openings could carry you a long way in a high school chess match. A little King's Indian goes a long way. But try some of those offside openings against a chess player that really understands the game and you better have a damn good memory indeed. Because he or she doesn't need to memorize every variation, only to understand the fundamental weakness of the position.
A lot of computer intelligence people rely too much on the iron wall around their rule book. If I know what you are doing, if I can see you mechanically collecting a move from a table of moves, if I can see your Turing Test set of rules for responses, then you are going to have a lot harder time convincing me your machine is "thinking". Consulting a table of integrals is not doing math.
In effect the rules are set up to let Karmnik look over Fritz's shoulder at his cheat sheet. And the deeper and more extensive that cheat sheet turns out to be the easier for Karmnik.
Fritz's team can equalize this to some degree by arbitraily cutting short the opening book at some specified level, that is force Fritz to start calculating moves at an earlier point. Which hands the advantage back to the grandmaster who didn't learn his openings by memorizing a book but by playing and observing games. It will be interesting but if I understand the rules right I predict 3 1/2 to 1 1/2 for Kramnik.
Posted by: Bruce Webb at Nov 26, 2006 9:37:26 AM
Chessbase is a couple million games, not an opening book. IIRC Fritz even sort of interpolates.
Posted by: bhauth at Nov 26, 2006 1:07:54 PM
the grandmaster who didn't learn his openings by memorizing a book but by playing and observing games.
My understanding is that computers use opening books they wrote themselves, just as you claim about humans.
Posted by: Douglas Knight at Nov 26, 2006 9:20:50 PM
Chessbase is a company that produces many products.
1. Fritz chess program
2. Big Database and Mega database.
Fritz does not rely on a database of masters' games to come up with it's opening moves. It has a selected opening book of moves considered to be sound by practice. A large number of masters' games have novelties that are often untested and possibly unsound. Just work with Fritz and a database for a bit and you will find the tactical flaws quickly. Opening books, just like MCO (Manual of Chess Openings), are sort of agreed upon knowledge. Fire up Fritz and you can click on the lines you want the computer to play up to the book where opening theory is exhausted.
http://www.chessbase.com/newsdetail.asp?newsid=1558
The link above will dispell all your thoughts about masters games being the core of Fritz's ability to play chess. It is Chessbase's 2004 April fools joke.
Regards
RP
Posted by: Richard Pointer at Nov 26, 2006 10:48:14 PM
Thanks for giving Chess some press!! Kramnik could use openings "out of book", and hope for the best. Unusual openings, where the database is not as deep. I have never used Fritz, but computers have problems with close positions also. Deep Blue showed us that computers don't always reply as expected. Kasparov tried a pawn sacrifice in Game Two of the rematch, Blue declined and threw Kasparov into a tailspin. It should be interesting to say the least.
Posted by: pawntorook4 at Nov 27, 2006 7:56:03 AM
"Sports"?
Look, chess is a competitive, demanding, grueling, game.
But if Stephen Hawking can play it, it's not a sport.
Posted by: Joe Grossberg at Nov 27, 2006 11:51:20 AM
kramnik missed a mate in one!! he could have won todays game and missed a simple opportunity..
Posted by: pawntorook4 at Nov 27, 2006 1:08:28 PM
let me clarify, he missed the computer move, ie: his sloppy defense led to the computer win.. he was even, and could have eeked out a win, at the very least a draw. talk about brainfart
Posted by: pawntorook4 at Nov 27, 2006 1:47:22 PM
Posted by: at Oct 14, 2008 12:05:00 AM
Mabinogi Gold
Mabinogi online gold
Mabinogi money
cheap Mabinogi gold
buy Mabinogi gold
Posted by: aion kina at Mar 20, 2009 10:47:18 PM
