< Earlier Kibitzing · PAGE 5 OF 6 ·
|Jan-06-05|| ||OJC: Good point acirce. I didn't know the match trends but they are not surprising when one's opponent doesn't tire. Kasparov versus a computer is somewhat analogous to Karpov versus Kasparov (since Karpov tired more quickly than Kasparov and Karpov fared less well later in matches than early) |
|Jan-06-05|| ||OJC: I see the point that knight13 and whitehat are making too. It would be great to see such a match take place so that each game would matter less to the human and in that way at least it would be less stressful. |
Botvinnik said a match for the world championship takes a year off your life. A super GM might lose a decade in a 24 game computer match ;). Just ask Garry.
|Jan-07-05|| ||Knight13: Deep Junior 9 is rated "3048." A GM is rated from 2500 to 2800. I don't believe humans can be rated over 2900. 3048 and 2813(Garry Kasparov's current rating) 200+ points a way. Did anybody tell you a player lower than his/her opponent's rating by 200 and still win the math? I mean Deep Junior is a computer, not a human. Computers make almost the perfect moves but humans don't. |
|Jan-08-05|| ||Hidden Skillz: knight13..from where r ugettin these rates for deep junior?? |
|Jan-08-05|| ||Everett: Knight13
Humans still make moves that computers don't see.
|Jan-08-05|| ||acirce: http://web.telia.com/~u85924109/ssd...
This is the latest SSDF rating list, the one that I think is the most recognized when it comes to computers. Deep Junior 8 is 2749, Deep Junior 9 is not yet in.
|Jan-09-05|| ||lostemperor: I was sure Knight13 is talking about USCF computer ratinglist which is higher than Swedish computer ratinglist in which computers play other computers. |
|Jan-09-05|| ||BradMajors: <Deep Junior 9 is rated "3048.> No. Computers can never be "rated" because they don't play in the same rating system as humans. Even on the computer rating lists, which don't mean anything, they don't have ratings over 2800. The matches they play against Kramnik and Kasparov shows the humans played far better chess. Perhaps you refer to ICC or playchess rating, which is entirely a different thing. |
|Jan-09-05|| ||lostemperor: <BradMajors> computers do play in human tournaments although not often. The Fidelity Mach IV the best from the late 80's/ early 90's has an official 2325 USCF rating against humans and got a national master's title. |
|Jan-11-05|| ||Whitehat1963: What's the best program to buy these days? And how much does it run? |
|Jan-11-05|| ||Skylark: The difference between human players and computer players is simple. Humans pick up tactical nuances and see particular lines, while not even looking at ridiculous lines, because of a little thing called intuition. Computers have a high calculative ability, but because even the most efficient programmed computer in the world has to calculate EVERY varitation before picking its move, the computers are inferior to humans at chess. I'm not saying that I could go and kill Deep Junior 9 right now; hell, I can't even beat Deep Junior 7. What I'm saying is that people like Anand and Kramnik and Kasparov are able to calculate faster more accurately than computer programs because they don't have to look at every move. Intuition also allows some of the fine combinations in history too; Tal wrote himself that many of his combinations were too deep even for him to calculate all the specifics and variations. But his intuition and experience told him that it was the right thing to do. Computers don't learn from their mistakes; and they don't see patterns. All they do is crunch numbers. Sure, for analysis and practise, they're great. But in my opinion, tournaments and matches should be left to people. |
|Jan-12-05|| ||Everett: <Skylark> well put |
|Jan-13-05|| ||Whitehat1963: <Skylark> and <Everett>, you're both under a common misconception about the nature of computer chess software and how it works and how computers play. So was I until I read this article. I highly recommend reading it: http://www.kurzweilai.net/meme/fram... |
|Jan-15-05|| ||Everett: Computer programs lose to computer programs, evidence enough that they are not perfect.|
Further, humans are capable of playing a perfect game, albeit rarely.
Combine these two, and humans (as chessplayers) will never be completely defeated by humans (as tool makers.)
Also, do computers learn? Can a program be taught to not repeat a variation, or see its mistake? If not, Any program can misread merely one position and lose every game with that color to a human with a memory of a 10 year old. All he would have to do is repeat the moves.
He states computers have taken a "modest" step toward pattern recognition, and fails to describe how this "neural net" works. Perhaps I didn't see another research page. I don't see how a computer can assess any position without attaching "values" so whatever patterns are put into the computer, it still has to value them (ie. which pattern is better/safer/winning compared to another, conflicting pattern arising in the same game)
My complete amateur opinion believes that there is no software to be created that will assess every position perfectly, (perhaps because it is designed by humans) and because of this, a human will be able to triumph...
...1 out of every 5,000,000,000 or so... ;)
|Jan-15-05|| ||dbquintillion: i think the obvious point about a man-machine 24 game match is that man gets fatigued and machine doesn't. i always thought the reason the super gms agreed to 6 game matches was because they knew they would tire in a longer match while the computer would experience no such effect. |
|Jan-15-05|| ||pawn52: <...there is no software to be created that will assess every position perfectly...> What is so odd about computers is when they analyze a position and find a brillant Queen sacrifice, for example, that goes into a won endgame, they would put it as #1, however, if they find a variation that would result in the possibility of a draw, it will discard the queen sacrifice variation and play the variation that has drawing chances because it will assume that the human player (or computer) will find the best way to counter that variation. |
|Jan-15-05|| ||Everett: <pawn52>
That's an interesting point. Didn't know it. But it does raise another limiting point of programs; the art of psychology. Just as Shirov's style unnerves Kramnik, yet Kasparov's does not, computers may come up against humans that are too difficult to beat through making the "best" moves, but will not be able to make the adjustments to make the most "challenging" or "uncomfortable" moves.
How many of you out there play variations, moves, combinations, etc., not because it is the best move, but because it will get into your opponents head? How many times has this succeeded?
Chess is a battle of nerves - David Bronstein Sorcerer's Apprentice
|Jan-15-05|| ||pawn52: <Everett> I have done it quite a bit. On playchess a couple nights ago, I played a guy who I had played several times before, all as Black believe it or not. He thought he knew my two Black openings King's Indian Defense and Siciain. Well, the play went like this: 1.d4 Nf6 2.c4 e6 3.Nc3. He expected me to play 3...d5, which I have done in the past, but I decided to throw him off balance and play 3...b6, The Queen's Indian Defense. I got lucky b/c he knew squat about the opening so I was able to gain a huge advantage during the opening, unfortunately, I blew it in the middlegame and was two pawns down in the endgame while dealing with two good knights vs two BAD bishops (mine); by that I mean they could hardly move anywhere, the passed pawns and the knights tied them down. He won it pretty easily form there. |
|Jan-15-05|| ||azaris: <pawn52> Just to note: After 1. d4 Nf6 2. c4 e6 3. Nc3 don't play 3...b6?! as White will play 4. e4! and get a great game. |
|Jan-15-05|| ||Knight13: <Humans still make moves that computers don't see.> Then see Indochess Man vs Machine (2005). If human were making moves that computer didn't see then why the computers just crushed the humans? |
|Jan-15-05|| ||Whitehat1963: <Knight13>, could it be due to the fact that only one of the humans was rated above 2600 -- ironically, the one who performed worst of all in the tournament? I'm not saying that computers will not routinely defeat even the best players soon. No, that's inevitable because of the rate at which computers get faster and faster, and their algorithms improve. Ten years from now, neither Kasparov, nor any of the big guns, will even be able to draw with a computer if the computer holds the white pieces. And the best they will be able to do is draw when the computer plays black. But for now, I think the top five or 10 players in the world can at least compete somewhat evenly with the best programs. But that collection of players in the Indochess Man vs Machine (2005) tournament, were far from the best players available to defend human pride. |
|Jan-15-05|| ||Whitehat1963: On the other hand, I think someone should organize a simul exhibition between the top computer programs and the top humans -- but in Fischer Random Chess. Let's see how the computers perform when their extensive opening books get thrown out the window! Maybe they'll still whip the humans, but I think not. If they did, then it's definitely over for man-machine matches. |
|Jan-15-05|| ||sandyobrien: in my opinion, a program is not better than a human until it cannot be beaten a human.|
my resoning is this: if the difference between computers and humans is that computers always preform at top-shape because they do not experience fatigue and other psychological things that affect humans. then if it is truely better, it should never be beaten by a human at top shape.
however, even the best computers are beaten at times by humans, therefor i think it's fair to say that humans are currently better.
is that fair?
|Jan-16-05|| ||Everett: <Whitehat1963 I think someone should organize a simul exhibition between the top computer programs and the top humans>|
You go ahead and do that. IMHO, it would be more worthwhile if we took the top GM's and put them in one tournament without the help of seconds or computers.
Which makes me believe that Karpov is perhaps the last real chesschampion, since he supposedly does not use them at all for verification or training purposes. No idea if this is true...
Everyone else is sucking at the silicon teet.
|Jan-18-05|| ||popski: <sandyobrien> Yes, I agree on this one. |
< Earlier Kibitzing · PAGE 5 OF 6 ·