Garry's going to be on CNN in a few hours as their "Connector of the Day" (?!) -- at 4:40pm NY time -- and will talk about the Karpov match, politics, maybe Carlsen. Not going to last long so you never know what they'll have time for. It should be live but not 100% sure. There are comments on their website they pull questions from sometimes so it's worth a shot.
Kasparov on CNN
92 Comments
Even if it's not live, it was worth watching at this time just for the bit on the Iraqi comedy show. Awesome.
Was it live somewhere?! I'm watching two different live CNN channels and didn't see it. Recorded, I guess. But I wasn't schlepping on hour on the subway to Midtown to go to the studio with Garry for 10 minutes.
Apparently only live on CNN International. Will be used in clips on US CNN during Wolf Blitzer and later in the week.
Heh.. Bet you wouldnt have said this some years ago, even if it 1 hour was replaced by a couple of days flight! Familiarity breeds .. boredom!
Keep fighting the good fight Garry, you are in our hearts forever. God bless
There's a piece on Chessbase.
"CNN: Kasparov on chess and politics
07.10.2009 – When he is not watching his charge, chess prodigy Magnus Carlsen, on Playchess, Garry Kasparov is talking to the media giants. CNN's Becky Anderson discussed chess, computers and Russian politics with him. "I started from a totally lost position," says Kasparov on the last subject, "but the Russian opposition still exists." Four and a half minute interview."
I can't believe the host was stupid enough to ask Kasparov as a first question if a computer could ever be built strong enough to beat him in his prime. She could just as well have shouted "I am am an idiot and too lazy and proud to read the great research my underlings have prepared so I can learn the basics about the guy I am interviewing".
"I can't believe the host was stupid enough to ask Kasparov as a first question if a computer could ever be built strong enough to beat him in his prime."
Haha, great stuff.
I can't believe that you are so stupid as to not realize that her first question was "would you rather be remembered as a great chess player or a political campaigner?"
Try watching the interview again.
Fair enough, but is it any smarter as a second or third question?
It is commercial television, played to a much wider audience than mere chessplayers like you and me. Give them a break.
"Garry Kasparov is talking to the media giants. CNN's Becky Anderson..."
CNN's Becky Anderson? Media giants?! Heh heh. That one is funny.
.
http://video.nytimes.com/video/2009/10/06/sports/baseball/1247464960472/the-thinking-man-s-games.html?hp
.
Another Video, FWIWorth....
NY Times web site has chess related video with Yankee manager Joe Giradi.
Target level: non-chess playing audience.
.
Maybe it can also be considered a trick question to see how Kasparov would react. Didn't he accuse the other side of cheating at the time?
Whether this is "good" or "acceptable" journalistic style might be another story ... .
Egregiously stupid interviewer. Garry acquits himself well.
Kasparov's answer to the second question is not very truthful; he essentially denies what we all know: computers (Rybka, $60) have reached the point where they can consistently beat the best human players.
i wonder where they are compared to correspondence players.
He said the computers anyone can buy today are good enough to beat "almost every grandmaster."
Almost every, but not quite all of them. So, in Kasparov's view, there must be some GMs who the computers are still not good enough to beat.
The question was along the lines of: "Do you think a computer will ever be built that could play better than Kasparov at his peak?"
The obvious answer is "yes, they are around - Rybka is the name". Or even: "Rybka is very close, but based on what has been achieved, there is no doubt that computers will soon play much better than any player in history, including me".
He side-stepped the question and flat out lied about the state of computer-vs-humans affairs as of today.
"He side-stepped the question and flat out lied..." (Irv)
What did Kasparov do to you to cause you to call him a liar?
CNN has gone downhill.
"He side-stepped the question and flat out lied..." (Irv)
What did Kasparov do to you to cause you to call him a liar?
Perhaps the PCA breakaway...trying to claim the world title that was the rightful property of the national federations (thru FIDE)? Just a guess.
If he helps Magnus Carlsen win the title..under FIDE...that might help mitigate the damage.
Luke asks:
"What did Kasparov do to you to cause you to call him a liar?"
Well, he has done nothing to me, Luke :-)
The reason I say he lied is because he gave an intentionally misleading answer to a very simple question. He knows perfectly well that computers have been playing better than the best humans - including Kasparov at his peak - for a few years already. Instead, he came up with the nonsense about chess having more possible moves than there are atoms in the solar system (btw, does anyone have an idea of where the "proof" of this often-repeated myth lies or who the author is?)
However, it would not be the first time Kaspy lies about chess-related matters: there is the now famous incident where he claimed not to have released the knight against Judith Polgar, only to be exposed by a video of the game in question.
Irv can't be cornered on this question. Kasparov started lying around the time the first match with Karpov was halted and he hasn't stopped since. If he ever caught himself telling the truth he'd lie immediately just to keep his hand in. In his chess, where "truth lays bare the presumption of a lie", can be found the greatest truths of the game. I think Lasker said that, but he may well have been predicting the rise of Kasparov, who never lied at the board but was unable to tell the truth away from it.
If you could pick any 2 top 10 players and the variations they'd play against each other, what would it be? I'd pick Vassily Ivanchuk as White and Alexander Morozevich as Black in the French Tarrasch Morozevich Variation C03 with 4.c3 c5 5.dxc5 Bxc5 6.Bd3. They've never played the Tarrasch against each other, but I'd give Vassily 9-5 odds of winning. I'm very interested in other's opinions of this and which players and variations you'd pick. Maybe Mig will even join in. ;)
Computers beating Kasparov at his prime!? Uh, didn't Deep Blue beat him 12 years ago? Today, Rybka running on my home PC could beat him (or Carlsen or Anand or Topalov or Wang Yue or Kramnik or Aronian or Biguier...)
If you took away the opening book and Rybka had to think for itself, it couldn't beat any of the top 100 players. It's a calculator that is incapable of independent thought, not a chess player. It can only follow our commands. Every computer programmer understands this.
Yes, opening books are very important for computers. This reminds me of an episode many years ago when one engine was "told to" play 1.e3 e5 2.e4 against his stronger(?) colleague - surrendering the advantage of the first move but getting him out of book immediately (of course he was provided with the relevant book). The opponent found the best moves - for a long time it was a normal Ruy Lopez with colors reversed - but took plenty of time which he lacked later on.
BTW, when man against machine matches were still interesting [read: when the result wasn't predictable] some GM's (don't remember their names) adapted well playing anti-computer chess: a seemingly calm, non-theoretical opening followed by a slow buildup on the kingside. The danger remained beyond the engine's horizon until it was too late. Could this still work today? But I guess players like Kasparov couldn't force themselves to play something like the Colle system.
Finally on correspondence chess: While some may consider it immoral or unethical, I guess most top players de facto play "Advanced Chess", getting help from computers at least to check tactical lines. In terms introduced here by Manu, this comes down to a combination of chess understanding (strategic play, long-term planning) and calculation power (concrete decisions or variations). If computers had a chance "on their own", this would be the end of correspondence chess - just like game adjournments had to be abandoned quite a while ago.
I disagree. It wouldn't do worse from the opening than equality with White and slight disadvantage with Black. From there, it would go on to beat pretty much everyone.
Irv shows some ignorance again:
Approximate number of variations in chess = 10^120
See reference Shannon's famous paper on chess.
Approximate number of atoms in the universe = 10^80
See the NASA website: http://map.gsfc.nasa.gov/universe/uni_matter.html
The number of variations in chess is MUCH bigger than the number of atoms in the universe. Assuming of course current theories and calculations are correct. But it is certainly not a myth.
Onischuk fan says:
"If you took away the opening book and Rybka had to think for itself, it couldn't beat any of the top 100 players"
ROTFL!
Why don't we just un-plug them? Then, ANYONE can beat them.
It's the old "let's cripple the computer so that we can pretend we can still beat it" argument.
"It's a calculator that is incapable of independent thought, not a chess player."
Yet, it plays chess much better than the best humans. So much for the need to "understand" chess. When a "stupid" machine, incapable of "independent thought", completely lacking in "understanding" of the game still beats the best humans, any smart person can only conclude that the game is purely tactical - a complex, beautiful little game of calculation and nothing more.
Bottom line is: computers play better than humans. Period.
"Bottom line is: computers play better than humans. Period." (Irv)
"If you took away the opening book and Rybka had to think for itself, it couldn't beat any of the top 100 players." (Onischuk fan)
Irv is right, Onischuk fan is wrong. Larry Kaufman has run test matches with Rybka against GMs with Rybka starting the game with less material (already out of book), sometimes a move or two behind, and Rybka crunched the GMs pretty badly. No book, less material to start with. Also, I think there have been chess 960 games between GMs and computers (no book), and the GMs ended up face down on the floor.
"When a "stupid" machine, incapable of "independent thought", completely lacking in "understanding" of the game still beats the best humans, any smart person can only conclude that the game is purely tactical"
But it's not purely tactical for human beings who don't have perfect recall and constantly improving hardware :) The whole history of computers and chess is fundamentally just a misunderstanding. Chess is (rightly or wrongly) associated with human intelligence so there was the theory that if computers could beat humans at chess they'd have proved their intelligence - but of course that's just nonsense. It'd be the same with building an android to run the 100m - it'd be an impressive technical feat, but of course in the end the android would be able to beat any human every time.
But so what? Would that mean that humans would no longer run the 100m? Chess will remain a game where humans continue to use intuition and understanding as well as pure calculation of variations. The very fact that by nature we're not perfectly equipped for it is what makes it worth the effort.
Henry,
That's the "observable" (visible) universe. Things are not as simple as you think...including the way at which Shannon arrived at his number of possible chess moves.
Check it out - it's bound to open your eyes quite a bit.
I basically agree vwith you, Mishanp. One little clarification, though: the game is purely tactical, even for humans. We have limited tactical skills compared to computers, so we do the best with whatever we have, but we are still working on purely tactical terms: "if I do this, he may do that or that and that position I evaluate as ---- for me...etc".
If you observe carefully, the main difference between players of different classes is that as you go up the ratings, players make tactical mistakes of increasing "subtlety" - in other words: the stronger the player, the more difficult it is to spot the flaws in his game. This is true of computers. They can and do blunder, but it is awfully difficult for humans to spot the computer's mistake bacause it is often way beyond the human's tactical vision.
"One little clarification, though: the game is purely tactical, even for humans."
Apparently every GM is wrong and the opening is "purely tactical" and not strategic.
Onischuk fan writes:
>>> Apparently every GM is wrong and the opening is "purely tactical" and not strategic.>>>>
Computers have confirmed a long-standing, not-very-popular theory: what we call "strategy" is just long-term tactics.
That's why "stupid" machines, without an "understanding" of the game, without a "brain" or an ounce of "intelligence" can beat the sh*t out of the best human players.
Even more paradoxical (and something worth thinking about for those who glamorize the game): did you know that a very weak player ( say, 1800-rated) who could never beat a 2400 IM can create, without outside assistance, a chess program that will easily defeat those players he can never beat on his own?
Think about that one for a second...
The 9 round 2009 U.S. Women's Championship is under way in St Louis. IM Anna Zatonskih defeated IM Irina Krush in Round 3 with the Lasker Defense and has 3 and 1/2 points after 4 rounds. Krush has 2 points. Only 5 years apart in age, their intense closely matched rivalry includes not only being the top 2 women in the US, but also being numbers 32-33 in the women's world!
"a very weak player ( say, 1800-rated) who could never beat a 2400 IM can create, without outside assistance, a chess program that will easily defeat those players"
Maybe this also means that the "weak" player does not or cannot realize his full potential over the board? Computers never are and never get nervous, tired, sick or jet-lagged ... not even after several hours of play. And they never make obvious blunders such as missing a forced mate or the only defence available. IMO this is also implied by comments such as "the piece sacrifice looks dangerous, but Rybka is not impressed".
As far as strategy vs. tactics is concerned: the best strategy doesn't help if you then commit a tactical oversight. Regarding Irv's "if I do this, he may do that or that and that position I evaluate as ---- for me...etc" - at the end of this thought and calculation process, a human player still relies on his "chess understanding" (unless the variation is obviously winning or drawing).
Concerning comments by acirce and Luke: If computers do not need an opening library to beat the strongest human players, then why do they have one? The Chess960 example is not too relevant IMO: Here humans are deprived of their usual "pattern recognition" capabilities, one of their strengths in conventional chess.
"And they never make obvious blunders such as missing a forced mate or the only defence available."
Well, it happens :) Example:
http://www.chessgames.com/perl/chessgame?gid=1352348
(19...Rfd8)
>>
That's why "stupid" machines, without an "understanding" of the game, without a "brain" or an ounce of "intelligence" can beat the sh*t out of the best human players.
something worth thinking about for those who glamorize the game>>
You lack a fundamental understanding of what computers/programs really are. The computer and its associated software is really a triumph for humanity - humans programmed it.
Just because someone builds a tool that can calculate using complex algorithms the best possible moves (most of the time) does not take away the beauty of the game.
If however, we were wired like computers or even interface our brains very easily to a chess program then the game would cease to be interesting.
The fact that humans found the game interesting for hundreds of years is a testament to its greatness.
If computers actually had intelligence in the future and actually become "beings", then too this discussion may be relevant as the android beings will mock us for finding this game fascinating. Of course they will assert they are superior beings as well in many other areas ...
"One thing I don't understand is why they didn't just change it to Qb7 since it's such an obvious move."
1 goofy poster at chessgames said that. The human player is supposed to play Shredder and the operator at the same time. I thought this was a 1 on 1 player versus computer match. lol
Thomas wrote:
"Maybe this also means that the "weak" player does not or cannot realize his full potential over the board?"
Yes, Thomas. I'm pretty sure there is a very strong player lurking inside every weak player.
Irv:
Haha still trying to argue? Showing more ignorance?
First, yes it is the visible universe. But the invisible parts are not "atoms". As far as counting "atoms" the calculation is correct. Remember, the statement was about the number of "atoms", not "all kinds of particles".
Second, okay suppose you also want to count the invisible part, assuming they are also atoms. To get the total atoms, including visible and invisible, you just need to multiply it with approximately 25 (why, you can read yourself in the NASA article, assuming you can read). Even then the result is less than 10^82 which is still FAR smaller than the number of chess variations.
Shannon's calculation of chess variations is simple. And notice that the number he gave is a lower bound. Which means it is AT LEAST that many. It could even be more.
Do you even realize the magnitude of the numbers we are talking about?
No arguments from me, Henry.
You're absolutely right in anything you say or believe.
:-)
The opening book is a huge advantage, as has been said earlier. For the playing field to be more even, both sides (computer and human) should be given access to the same information. If Kasparov or Carlsen were to sit down and play Rybka, but have at their fingertips a current database of 3 million + games (the same way the computer engine does) ... I dunno if the computer would be able to win. I doubt the human would win either.
"I dunno if the computer would be able to win. I doubt the human would win either."
Ah, so it'll be a draw I guess.
Besides having an opening book, usually the computer also uses endgame tables. Which means it doesn't have to calculate moves by itself at all in the endgame, and still knows the perfect move. To make it fair, give the human also access to endgame tables.
Miktal wrote:
"For the playing field to be more even, both sides (computer and human) should be given access to the same information."
But the computer doesn't have access to the information Botvinnik gave to a nubile 13-year old Kaspy during those lazy summer afternoons in Moskow...
Rybka and supercomputers still don't come anywhere near calculating all the way to the end. If they could, the Holy Grail (aka 32-piece tablebase) would exist and the byteheads here could legitimately claim that computers play "perfect" chess, rather than merely "better-than-the-best-humans."
Therefore, it follows with perfect logic that Rybka and her sisters, at the end of many (most?) of the lines they calculate, must apply an evaluation algorithm. By definition, that evaluation is NOT based on "calculation" (in the sense of looking still further ahead)... but rather on applying rules and principles to weigh and compare various features of each position.
And the latter - the rules and principles that allow computers to evaluate the hypothetical future positions that are the end result of all their calculation - are born out of what? STRATEGY! In other words, they're the product of the very "human intelligence" that Irv denies has any role in computer chess!
Someone please show me where I've erred in this analysis.
IMHO there is a misconception about how humans -- especially amateurs -- play chess and what drives their move selection.
One of the advantages that higher-rated amateurs have is they know certain opening schemes fairly well. Thus, they can use their knowledge to build fairly safe positions where they are unlikely to fall to tactical problems.
They don't crunch every possible move during a game.
Nor do they have perfect recall of their favorite opening lines...so occasionally they make errors.
This translates into those "subtle tactical errors" that one of the posters mentioned....as you go up the rating chart, players don't drop pieces (at least not frequently) or pawns or weaken squares, etc.
But they are still tactically vulnerable.
Contrast this with computers with an opening book and/or endgame tablebase. They will reproduce the opening perfectly. No error in move choice (i.e. no faulty memory). That doesn't mean the opening book is perfect, but there will be no memory errors. This is a HUGE edge over humans...so that the computer can have a safe edge by move 15 or 20 or 25 or whatever the limit of its book might be (vs. the human).
This is similar to what higher-rated humans do to lower-rated humans -- they use their opening knoweldge to get safely superior positions by move 15 or 20.
Now take away that perfect opening book -- machines are much more vulnerable.
The point about Kaufmann's odds games should also include certain possibilities:
1. Some odds positions may actually favor the odds-giver position (such as removing the f-pawn).
2. Odds positions are much less familiar to humans -- thus negating their knowledge of the openings as well. See the point about tactical errors in novel situations vs. fewer errors in familiar positions.
I'm not sure that odds games show nearly as much...as just turning off the book and making the program "solve" from the starting position. And of course no move sequences can be hardwired into the programming (that would be a source of "opening book cheating" by this argument).
IMHO the sheer move-crunching isn't nearly as important as the perfect recall at the beginning and end of games...that is what makes today's programs so hard to defeat.
I agree with most of what you say, Jon. this is where we 'part ways":
I think that both computers and humans go through the "tree" of variations, stopping at various points to assess the position at that point. The "position at that point" is what the human and the computer think it will be based on what they have calculated as possible replies by the opponent. The assessment is just the final calculation, where a number of factors are weighted: development, piece activity, space, threats/king safety, number of available moves, etc. The usual stuff.
So, why is it that the computer and the very strong (GM) humans often reach different results (in the sense that computers regularly beat humans these days)? Because humans make errors in their calculations and often end up assessing/expecting the wrong position!
But the process is the same.
"But the computer doesn't have access to the information Botvinnik gave to a nubile 13-year old Kaspy during those lazy summer afternoons in Moskow..."
Haha! Well, if anyone has something concrete that Botvinnik gave (information-wise) to Kasparov, I would be more than happy to allow it in the computer's database (if it wasn't there already). The thing is, Kasparov's cumulative experience of the game is not the same as the computer's openings and endgames database ... in the case of the computer, they are completely distinct from one another. Whereas the aspects of the game that Kasparov has learned over time are an integral part of his playing ability ... i.e., they cannot be turned "off" like the various databases computers use. It just seems to me that the computer has access to more information at the start than the human, and that information is not the result of the playing experience of the program, but rather the injection of huge libraries of positions that are simply recalled, not calculated.
"...at the end of many (most?) of the lines they calculate, must apply an evaluation algorithm."
Very good point! I honestly had not thought of that before... So, if all the computer had was calculation, then all possibilities that ended with equal material within its horizon would have equal value and the computer would not be able to select a single move among many. Hmmm...
irv wrote:
"But the computer doesn't have access to the information Botvinnik gave to a nubile 13-year old Kaspy during those lazy summer afternoons in Moskow..."
Kindly keep your thoughts on young boys to yourself.
Standec wrote:
" if all the computer had was calculation, then all possibilities that ended with equal material within its horizon would have equal value and the computer would not be able to select a single move among many."
Have you thought that many positions can be out-right winning even if the material is equal? Or drawn even if one side has substantially more material - like the unique case of the rook +2 pawns vs. rook drawn position?
The computer (like all humans), while evaluating any given position, takes into account - calculates - many factors besides material.
That's exactly the point. If the computer used NO heuristics, all positional elements would be lost and nothing beyond its horizon could be seen. Without a positional algorithm, all final positions in a tree would only be calculated by material balance (which is obviously a poor understanding of chess). flyonthewall is correct that the computer clearly has to use something besides just tactical calculation.
The problem with your position that computers should be deprived of their opening book is that it becomes a slippery slope with no end in sight.
Today, it is the opening book.
Tomorrow, it might be the endgame tables.
Next time, it will be about the microprocessor being too fast. Then, the prgram might be accessing too much RAM...
How much do you think we need to cripple the machines before we accept the computer's superiority in chess?
i think a far more realistic and honest approach is to admit that the computers are too strong for humans. We can always continue enjoying the game as much as we want against other humans. We don't race against cars - we don't insist on taking the wheels out in order to make a human vs. car race competitive.
It may surprise you, but I don't enjoy playing against computers. I don't even enjoy using databases or training with chess software. I greatly enjoy the game against other humans. I don't really see much point in playing against machines. But I will not pretend that they don't play stronger than the best humans. And I would certainly not feel proud to beat a crippled computer program.
>> The computer (like all humans), while evaluating any given position, takes into account - calculates - many factors besides material. >.
Humans who always "calculate" the next move are normally not good players. Good players normally "judge" their move to be optimal based on strategical consideration and pattern recognition.
Any human trying to mimic the way a computer "calculates" will be an inferior player.
And computers dont really play chess -- Its not like someone programmed the computer the rules and then make it LEARN to improve its game - that would be a breakthrough in AI.
Standec wrote:
"That's exactly the point. If the computer used NO heuristics, all positional elements would be lost and nothing beyond its horizon could be seen. "
You're confused, Standec.
1. Nothing beyond the computer's horizon can be seen. That's why they use the expression: horizon. The computer can NOT "see" beyond its horizon.
2. The computer - once again, like all humans - does not "calculate" several moves and then "evaluates". That doesn't happen. The computer , like all humans, evaluates ALL THE TIME, not just at the end of variations. Every half move involves calculation/evaluation. Without that constant evaluation the computer - or the human - would not know when to stop wasting time on a clearly losing variation. They would have to play everything out to the end in their "mind". That is obviously not the case.
It is asinine to call Kasparov a "liar" based on his failure to give a categorical statement, in an off-the-cuff statement, about the strength of computers relative to humans. Indeed, a noncategorical statement in response to the question was appropriate. Under what conditions … at what time controls? I would not be prepared to categorically state that a computer could beat Topolov (or Kasparov in his prime) consistently at a particularly long time control. Perhaps it could, perhaps not. Speculating one way or the other is hardly a lie.
Well, I don't think I am trying to gimp the computers to the point where humans can beat them consistently. I guess I am just a bit irked by the "man vs. machine" mentality this issue leads to ... we don't race cars on foot, or try and calculate pi with a pencil and paper past what the world's super-computers can do. Some would say that in chess, it is computational force vs. computational force, but it really isn't. The computer comes with these huge databases, opening books, etc. It has the efforts of dozens of computer experts behind it, chess theoreticians, etc. It truly is like trying to outrace a car on foot ... the human invention in both cases is given a technological advantage. I think that computer's are clearly "stronger" than human beings at chess, but it is the human beings efforts that make it so (i.e., giving it access to nearly every recorded game in history, multiple-piece tablebases, and extensive opening books).
Danz wrote:
>>> "And computers dont really play chess" >>>>
You're absolutely right, Danz. My apologies. I was under the impression there is chess-playing software being sold - including Rybka.
;-)
Mr irv,
You seem not to understand what a computer is. It runs programs.
ANd programs so far have not matched what humans can do - otherwise we would be living in a world of Asimov with the 3 laws of robotics.
I have not heard of any software self-learning from the rules and improving..
Rybka does not play chess in that sense. I give Rybka 2 a mate position from a game in 1900 and it cannot find the best continuation. After I "teach" it by showing it.. I come back and it repeats the same thing...It learns nothing and knows nothing..
Maybe you should try to grasp this concept first.
Playing chess includes the process of learning the rules and applying them. The improvement in skill is part of chess.
So far the computer/program are just chess tools not actual beings that beat us in the game.
As far as the chess strength of this "chess tool"
goes, its not that impressive since:
GM + chess program > chess program on its own.
How come?
Chess programming software still has a long way to go.
Danz's comment is most interesting to me, because it zeroes in on an issue I view as far more important than whether computers can beat top humans, or even whether computers "understand" the game as well or better than top humans.
The issue is: what does computers' superiority mean for us, our understanding and appreciation of the game, and its theory?
The answer is "Very little." Computers are clearly better than the top humans... and for us (even for Kasparov or Anand!), that doesn't and shouldn't matter.
That's because computer chess inherently differs from chess as played by humans. So, the amount that we can learn from them is far more limited than computer acolytes seem to feel. We can't copy their methods, as danz points out. We are just biologically incapable of emulating computer style to any significant degree.
So, the decisive qualities in human-vs-human chess will remain tactical vision GUIDED BY INTUITION, JUDGMENT AND STRATEGIC UNDERSTANDING. Or: strategic understanding, intuition and judgment, augmented by tactical vision.
One often hears chess pundits blithely assert that computers transformed the game by turning young kids into tactical powerhouses like Radjabov and Carlsen. I think that's a misconception. Even for the young generation of GMs who got so strong so fast with the aid of computers... computers functioned as study aides - NOT mentors, coaches or tutors. Carlsen didn't LEARN anything about chess that's directly attributable to his software, hardware and data. Having access to modern chess-study tools simply allowed him to look at more games and more variations more quickly than he could have done using pre-digital technology (printed books). Technology did not affect the content of what he learned.
And if you really want to get down to it, the computer can't do anything whatsoever if the human unplugs it ... good game, computers. :-p
Computer chess is as interesting as listening to a Beethoven piano sonata played via MIDI playback. It's flawless and faster than any human can ever perform it, but it lacks all of those things that make the music compelling to listen to.
Regondi, as a guitar maestro you should know that speed is not at all necessarily synonymous with good music.
touche!
So, I'm wrong in saying that nothing beyond the horizon can be seen, because nothing beyond the horizon can be seen?? You're right, I am confused.
A positional evaluation of a position means nothing if it doesn't predict the future value of a position. When a computer uses positional algorithms, it is not "seeing" beyond its horizon, but judging how valuable an end position will be. Without that non-material judgement, it would have no way of seeing which options are better than another given material equality.
One often hears chess pundits blithely assert that computers transformed the game by turning young kids into tactical powerhouses like Radjabov and Carlsen. I think that's a misconception.
- - - - - - -
You're right that the *main* benefit of working with technology is simply the speed and ease with which you can access large amounts of data.
However, I don't think it's the only factor. I believe working with computers *does* fundamentally affect the way you play; computers have a very materialistic, concrete "style" and after you work with an analaysis engine for tens/hundreds of hours, surely your play assumes some of the same characteristics. I've noticed this pattern in my own play, and can't imagine that it doesn't also apply to top players.
Standec wrote:
"So, I'm wrong in saying that nothing beyond the horizon can be seen, because nothing beyond the horizon can be seen?? You're right, I am confused."
Yes you are very confused, but not because of the reason you offer. I'll show you. Look at what you wrote:
"If the computer used NO heuristics, all positional elements would be lost and nothing beyond its horizon could be seen."
The point is that WITH or WITHOUT heuristics, the computer can't "see" beyond its "horizon". Your statement implies that there are certain conditions ("using heuristics") that allow the computer to see beyond its horizon. That's false, so I'm assuming that you are not lying - just confused.
*******
That said, I'll give the thread a rest. Thanks everyone for the exchange of ideas!
I'm pretty sure that I made clear that a positional evaluation of a position is assigning a future or potential value. Strict calculation can't give you that. Not sure how that can be considered lying; it is just obvious fact. No reason to make things personal.
[sorry for the full quote, but it's been a while, at least many comments down the line ... :]
"Thomas wrote:
"Maybe this also means that the "weak" player does not or cannot realize his full potential over the board?"
Yes, Thomas. I'm pretty sure there is a very strong player lurking inside every weak player."
First of all, I didn't say that EVERY "weak player" could create a computer program, did you?
Little do I know in detail, but I guess creating a chess computer program requires
1) at least some chess understanding
2) programming skills
[in the old days of Kasparov vs. Deep Blue, hardware was also an issue - no longer as currently damn strong commercially available engines run on ordinary PCs]
I think the strongest engines are team efforts where several people provide complementary skills. "Chess understanding" would, at the very least, include the relative value of pieces. Further steps may include 'general' answers to questions such as
When should one go for the bishop pair?
When are doubled pawns acceptable?
.....
In any case, I don't think a complete outsider (someone who has never heard of chess before or just knows the rules) could create a strong chess program on his own ... .
To put my remark about "a weak player not realizing his full potential" in a wider context: A "mediocre" OTB player could also be a comparatively better book author, coach or blogger. I don't want to (re)start discussions on Mig's ratings, let's just say it is <2600 - same for Dennis Monokroussos and Peter Doggers + colleagues from Chessvibes.
Thomas just slyly wrote:
"A "mediocre" OTB player could also be a comparatively better book author, coach or blogger. I don't want to (re)start discussions on Mig's ratings..."
Anything else that you don't want to do?
Oh yes! That's my point exactly. A MIDI playback might be twice as fast as any human can play but who cares? Who would want to listen to it (other than a curiosity)? Having said that, however, a certain minimal speed is required for a good number of pieces to be musical. "Austurias" for example will never be a great musical performance played at 60 quarter notes to the minute.
Hey Clubfoot - What's with your axe-grinding on Garry Kimovich?
Irv seems to be stuck in a 1989 version of chess computers!
1. Yes, some programs can go beyond their "horizon" in the sense that if a line appears to be promising...the ply depth can be altered to allow additional examination.
That is essentially what humans do when they think there is a mate or combination to win material...vs the usual low ply depth search for positional moves.
The days of "computer only goes X play and no more..." are over -- the horizon effect (computer pushing off small material loss by choosing lines that push the loss past the "horizon" of its search so that it loses larger material eventually) is not the computer-busting tactic that it used to be.
2. Humans don't evaluate positions each ply -- or at least not consciously.
But they *do* apply their evaluations to specific positions or lines of play.
They don't evaluate everything....or even most lines.
However, computers *must* apply a numerical evaluation to everything.
3. The "perfect opening book/perfect endgame table" is the crucial element IMHO
An analogy.
You are on Jeopardy. You have your memory.
Your opponent has access to Wikipedia and a fast search function. You do not.
Answer this question:
Which European power joined France, Turkey and England against Russia in the Crimean War?
If you *know* the answer is "the Kingdom of Sardinia" then you can buzz in.
If you don't know the answer...then the fellow with error-free access to Wikipedia will beat you.
So it goes with man vs. computer at chess.
If you know that 9. a3! is the right answer, you can play it. If you have to search for it....then your computer opponent with the error-free memory record of 9. a3 will beat you.
That's not chess...anymore than accessing Wikipedia is playing Jeopardy.
4. Yes, you can build computers that "learn" from experience at chess positions.
However it is slow, tedious and requires plenty of data input (this position favors White, this one is equal, Rxc2 is the right move in this position, etc.).
That is entirely why they programs "cheat" by using memory-perfect opening books and ending tablebases.
The "self-learning" programs are worse than humans :)
I had a dream I was as articulate as you, then I woke up. ;)
Hey noyb - fine, thanks, but that's not a question. I would take an axe to the root but for the historical record, and there's just no grinding away at that.
I think Irv makes many good points. The one question I would re-ask is why GM + computer beats computer alone. Where exactly would the GM contribution be. Not with tactics anyway!
I think too many people have an emotional repulsion to Irv's claims. Chess has undoubtedly been devalued in the public eye and perhaps in the chess community by the obvious strength of computers. Nevertheless I think the Kasparovs, Kramniks, Karpovs etc etc etc do have real understanding see paragraph one.
In the early days of computer chess they played like fools and you would build up an overwhelming position but often screw up in carrying out the attack overlooking some tactical "fluke". I suspect this is partly what happens at the super GM v compuer level.
In Kramnik's last match he lost two games (one overlooking a one move mate and two playing a last game unfamiliar sicilian trying to even the score) but in general play he was at least equal to the computer.
"In the early days of computer chess they played like fools and you would build up an overwhelming position but often screw up in carrying out the attack overlooking some tactical "fluke". I suspect this is partly what happens at the super GM v compuer level."
Really? Got some recent examples of that? I mean, it may happen, but very very rarely against the strongest engines... Kramnik's match is not a good example since that is not what happened there, and besides, Rybka is much stronger than Deep Fritz was.
Feeble, Clubfoot, feeble...
Big feeble gutless words. Prove me wrong or get out of the ring.
From feeble to pathetic. I just asked you why you attacked Kasparov you big baby! If you can't handle the heat, YOU get out of the kitchen! What a whiner...
¨Ah, the inevitable flinging of the monkey-poo.¨
What happened noyb? The organ grinder fired you?
I thought so. noyb brings his usual peashooter to a gunfight, loses his only nerve and runs home crying with desperate parting epithets like "big baby" and "whiner".
Nah, I had better things to do on a nice weekend than joust with mental midgets...
Clubfoot - I'm calling you out. You still haven't explained why you attacked Garry Kasparov. If you are unwilling to explain yourself, I'm "assuming" you are totally full of it. But then given your inability and/or unwillingness to explain yourself, I'd say you've more than proven that.
Later, chumps!
This post may be described as-
monkey poo.
Little tip here: when you call someone out twice for the same thing on the same thread after you've already been answered, it starts looking a little insincere. I responded above and invited you to prove me wrong, yet you continue with the desperate name-calling, concluding with a shot at all posters on this thread. Insults are not as good as evidence.
Really? If I had to judge who has the better grasp of how chess engines work based on the comments in this thread, Irv wouldn't be the loser.
You misunderstand the concept of a horizon. By your logic, we could say that a computer can easily go beyond its horizon by spending 1 hour on a move instead of 5 minutes. That's hogwash. The horizon is simply being extended in this case, which is quite fine seeing as it's not defined as some fixed number of plies. The concept remains fully intact because there is ALWAYS a limit beyond which the computer can't see irrespective (of any practical amount) of time spent, and it is entirely possible in some cases that the final evaluation (good or bad move) could change further on. Actually though, if you know anything about tree searches, the increase in depth is not linear with the amount of time spent on the search, so the horizon tends to increase slowly as a function of search time regardless of how efficient the engine is at identifying and dismissing unfavorable lines at every ply.
I'm not sure the claim that "humans don't evaluate positions each ply, or at least not consciously" is a significant difference in the way humans and computers play chess. The computer also has means for identifying bad moves quite quickly - this probably only takes a few plies in many cases. Who cares whether or not the human was conscious of doing the same thing? This is simply the human maximizing his time efficiency, and there are doubtless some occasional missed tactics as a result.
The cry to remove opening books and endgame tablebases has always been nonsense. Computers have expansive memories with perfect recall, so what? That's just another advantage, no more or less remarkable than their impressive number crunching ability when compared to humans. If you decide to limit the machine's memory or deprive it of opening theory (which all humans have access to) then you might as well slow down its CPU as well. The man vs machine concept becomes meaningless (it it wasn't already) if you're deliberately selecting a substandard opponent.
Also, the human doesn't have hash tables or infalible memory, so to be fair allow the human to write down all positions and lines as they think about them, together with evaluation notes. Also the human doesn't have the circuitry to process millions of positions per second, so we should therefore allow the human, say 1000 x more time than the computer? Where do we stop!