A Day That Will Live In Infamy

The chess-playing supercomputer Deep Blue beat chess champ Gary Kasparov on this date in 1997. The two adversaries had faced each other the year before, and although Deep Blue did manage to win one game, Kasparov won the five-game match. IBM returned in 1997 with a heavily upgraded Deep Blue, and this time, the computer won two, Kasparov won one, and they played to a draw three times. Kasparov demanded a rematch because he believed the computer had cheated, but IBM refused and took the computer apart. Since that time, computers have become more and more successful at beating humans on the chessboard.
writersalmanac.publicradio.org

Low class move by IBM in my opinion. I thought so at the time and I think so now.

These engineers (or more likely, their managers) were not entitled to act as if they had just beaten Gary Kasparov. A computer is a research machine that is the product of mankind’s advancement in science and technology, and IBM was just the organization that ran the test against the best human opponent to measure our progress.

But the measurement was right. Computers passed humans at that point and the gap continues to widen. Now the GMs learn as much as they can from computers and then go do their best in primitive hand-to-hand combat. Chess stimulated a lot of work in the early development of computer science, and for that we can feel some comfort.

The Go players are still beating the computers though, aren’t they? If the early computer researchers had used Go as a paradigm, various things might be different now.

For the astounding progress made by Go programs recently do a search for, Computers are very good at the game of Go, by Chris Ball.
blog.printf.net/articles/2012/02 … game-of-go
Also see, Man vs Machine match: Final results and game commentary at: gogameguru.com/man-machine-m … ommentary/

“Deep Blue, in short, made computers personable.”
From: techcrunch.com/2012/05/11/ibms-c … -birthday/

Did it really?

Too Easy to be cavelier with other people’s money. IBM spent huge sums on Deep Blue beyond tangible returns, but still you criticize their personal character?
And I gather you similarly want to bring back the Botvinnik rule that the champion gets a rematch if he loses; = challenger must win two consecutive matches to settle in as the new champ.

OLD QUESTION: Can chess engine defeat the best human? Answer: No, not really.

NEW QUESTION: Can a human grandmaster plus the best chess engine --defeat-- a copy of the chess engine alone? Answer: Probably Yes; and probably will remain Yes for perhaps another decade?
.

On the contrary, it was Kasparov who exhibited real low-class behavior after that match, with all his ridiculous accusations, his accusations towards Joel Benjamin and others, so much cry-babying that I’ve long since forgotten the details. Good thing Kaspy seems to have grown up since then.

Has he?

How much were the old computer parts worth once they took Deep Blue apart? Old computer parts aren’t usually worth very much. I guess they took the computer apart so that a rematch would be impossible. Really chicken behavior, and the rematch would have been great publicity too.

Besides, I think Kasparov would have lost, esp. if IBM could make any changes to the computer in the meantime. Kasparov was past his prime in age and getting older all the time. Silicon doesn’t get worse with age, doesn’t get tired, etc. You could just see that his head was being fried, trying to play against the silicon monster.

I think the rematch provision is reasonable. It’s OK to have it, or not to have it. Then there’s the intrigue of champions who avoid giving the rematch (Alekhine, to Capablanca) and the public can express their opinions about that – the whole thing is not only reasonable, but great for publicity.

I can identify with his bad feelings after losing. Really caring is part of what made him such a champion, but to face a virtually perfect opponent who will detect any gap in your calculations, exposing your error for all the world to see – he did not deal well with that experience.

So Kasparov reacted under immense pressure, a great champion defeated by a superhuman opponent. The IBM side reacted under – no particular pressure that I can see. They chose to spend the money, then took the box apart to prevent a rematch just because they could. They turned on the electricity and beat the world champion, and they didn’t even have the decency to turn on the electricity again for a rematch.

By the way I’m no fan of Kasparov’s politics or a lot of things he’s done (so I’m not inclined to favor Kasparov) and I am an electrical engineer (so I am inclined to understand and respect the achievement of IBM). Yet, I have no respect for what IBM did to Kasparov after the match.

It wasn’t ‘old computer parts’. It was a fairly top-of-the-line general-purpose mainframe, with 32 Power2 cpus and 512 special-purpose chess processors. After the specific chess equipment was removed, it was still a working and valuable mainframe machine.

If you read Hsu’s book on the match (“Behind Deep Blue: Building the Computer that Defeated the World Chess Champion”), he states that the reasons there was no rematch boiled down to
Kasparov wanted IBM to pay him for a third match; IBM felt that if he wanted a rematch, he could raise the funds
After Kasparov lost, he insulted and attacked everyone involved in the project, accusing them of cheating, and after that, IBM was no longer willing to do Kasparov any favors
IBM felt that it had already achieved maximum publicity from the match for the millions of dollars already spent

There were already significant changes between Deep Blue and Deep Blue II. IF there was a third match, Deep Blue III would have been larger.

Heck, I thought you were referring to the date of my divorce in 1989.

I guess they took the computer apart because the IBM team had fulfilled their objective.

IIRC some time after the match there was an item in the press that IBM had sold about $800 million dollars worth of computers with a similar matrix of components and computing power. That was about a 40 to 1 return on investment, a great deal for any company. Having achieved the goal the company turned to more practical real world applications. Kasparov’s complaints and accusations guaranteed that IBM would not play any sort of rematch against him.

From what I could see, GK had too much respect for Deep Blue because of his own use of computers to prepare. He overestimated its ability to evaluate and conduct positions. As a result he played a series of anti-computer style games to avoid its strengths. By doing so he also limited his own chances and played about 300 points or so below his own rating IMO. Had he tried to play what he calls “real chess”, the machine at that time probably would have not been capable of keeping up with his opening preparation, calculating ability in dynamic positions, or his positional judgement in closed or semi-closed positions. The program, left to its own evaluating devices in selecting openings, had a curious quirk of selecting some variations that it “thought” were good, but were actually inferior. Thus the need to get GMs to prune the bad lines and make it play other lines as if by rote rather than through playing according to its evaluative algorthim. GK never gave himself a chance to catch them out in selecting lines that he knew from previous analysis were slightly worse.

Playing a match with the computer is an unequal affair. You might match up well with the computer in the opening and outplay it tactically or positionally in the middle game. However, allowing the machine to access endgame databases in order to play out games is a tremendous advantage as the computer’s perfect memory is going up against a tired opponent who, even if he is well versed in endgame theory, is still using his own analytical abilities for many moves. For true scientific purposes, the machines should never have been given the ability to access databases but used only its algorthim alone to play.

Human players are used to playing against other players who play for ideas and who strive for the initiative. Computer programs are reactive rather than proactive when they are playing. They are good at refuting ideas and bad tactics, but less good in positions where the landmarks are not so clear and have nothing for them to bite on to calculate. It often seems to wait for you to do something. A good idea generated from a computer is more of an accident than something that is based on patterns, imagination, or the types of principles we use to select moves. Give it something to do, to break down tactically, and the program is in its element. It has difficulty generating something interesting on its own without special human guidance and evaluation.

Good point

Did anyone else play naturally against that computer? Kasparov was the greatest human chess calculating machine on the planet at that time, and if he was scared of the silicon beast, I don’t doubt his judgment.

Ideas win the game for humans, but really it’s tactics that win games. We just think of them, or groups of them, as ideas. The computer will coldly find any little hole in the tactics behind your apparently good idea and blow it up. I vaguely remember, didn’t Kasparov try playing naturally and experience bad things happening?

At some point it’s not a matter of confidence. It’s a matter that the superhuman computer is stronger. I can’t fault Kasparov’s match strategy or play, as the leading representative of my species.

It should be noted that IBM did not set up a match to play lesser rated players than GK. It would have been embarrassing if the machine had lost to a mere 2200+ player or even a 2400+ range player who was conversant with computer chess. A loss to the World Champion is vexing but can be rationalised. Losing to a lesser light, well that was nigh unthinkable as it would lead to curtailment of funding and derailing of careers. They might have learned more by playing a variety of players.

BTW, it was always a little annoying that the computer people blamed losses on hardware problems than on the software algorithm. The enhancement of speed made the machine look more ply ahead, but missed the point IMO. The focus should have been on making the evaluative functions better with as little use of energy or chips as possible to conform with what they claimed the machines would be ultimately used for - deep space missions or smart battlefield devices. For many reasons “wetware” ( brain) still works better than hardware because our internal software, though flawed, is often more flexible and faster in decision-making. We also do better because our emotions spur our thinking in multiple directions. In short, when we are having fun we play better.

Mr Quinn wrote, “The computer will coldly find any little hole in the tactics behind your apparently good idea and blow it up. I vaguely remember, didn’t Kasparov try playing naturally and experience bad things happening?
At some point it’s not a matter of confidence. It’s a matter that the superhuman computer is stronger. I can’t fault Kasparov’s match strategy or play, as the leading representative of my species.”

This is the last game of the last match:
This game was played on May 11, 1997.

Deep Blue–Kasparov: 1.e4 c6 2.d4 d5 3.Nc3 dxe4 4.Nxe4 Nd7 5.Ng5 Ngf6 6.Bd3 e6 7.N1f3 h6 8.Nxe6 Qe7 9.0-0 fxe6 10.Bg6+ Kd8 11.Bf4 b5 12.a4 Bb7 13.Re1 Nd5 14.Bg3 Kc8 15.axb5 cxb5 16.Qd3 Bc6 17.Bf5 exf5 18.Rxe7 Bxe7 19.c4 1–0

Much has been written about the match and this particular game and I will not rehash it now. I do, though, question how anyone could write that he, “…can’t fault Kasparov’s match strategy or play…” after playing over this game. I put the game into 365chess.com and was unable to find it in the Masters database, instead finding it in the Big database, which oughta tell you something. Mr Quinn calls Mr Kasparov “…the leading representative of my species.” I will point out that Kasparov’s opening preparation, with his use of the ideas of leading GMs who were paid by him for their ideas, and of computer programs, is legendary. Although many different excuses have been offered as to why “the leading representative of my species” would have chosen a discredited line in a variation championed by the World Champion he vanquished, none have satisfactorily answered the question. I will not accuse the former player of taking a dive like others, including Dr Elwyn Berlkamp, a professor at the University of California at Berkeley, did in a lecture on the game of Go he gave at that school some years ago. I purchased the CD at the National Go Congress in Black Mountain,NC, in 2006. I will say that it would seem to be the only plausible reason for the last game disaster by “the leading representative of my species.”
Mr Quinn writes, “The computer will coldly find any little hole in the tactics behind your apparently good idea and blow it up.” Mr Kasparov’s idea was bad and the program, not computer, found the “hole” that was large enough to drive a Mack truck through.
If, in the near future, like, say, 12/21/12, “the leading representative of my species” must face-off with aliens for the fate of humanity, I can only hope our chosen representive is much better than the one chosen for humans in the battle with the machine in this case.
Armchair Warrior

Many people kept trying to make computers play like humans, and to find a way to use evaluative functions to improve playing strength without increasing the number of plies, or to cut the tree.

It didn’t work very well. This was a big debate but the debate has been long settled. Brute force is the way to win with a computer.

I think Larry Kaufman was recently working on things like evaluative functions for Rybka, so they are trying and may have some success. But I believe it’s only at the margins. Brute force is what counts for a computer.

And also, brute force is extremely important for humans, though coupled with tree-pruning. One can look at all sorts of chess strategy books about the styles of grandmasters and how a class player can play more like a grandmaster, but really GMs are extremely good, extremely consistent calculating machines. That’s the difference between an IM and a GM in my experience: they both have a very broad and deep understanding of chess strategy, but the GM is almost always better at calculation and will almost always win a game that depends on concrete calculation. The GM can implement his vision on the board, because he makes the details work for him reliably. I think it’s partly that the GM can prune the tree better and end calculation faster (this exchange sac is clearly OK, no more calculation, on to the next variation) and partly clearer sight of the board, almost total lack of mistakes like thinking pieces are still where they currently are when calculating a variation in which they’ve moved, and that sort of thing. The second half of that is what I would call “computer-like calculating ability.”

And I bet that 2700 GMs are a bit better calculators than 2600 GMs. But I’ve seen Aronian lose by tactical mistakes, so 2800 GMs may not be necessarily better than 2700s at calculating.

Yes this was a mistake. One of those things humans do. As you say you are not asserting that he took a dive.

People used to say that when they played Bobby Fischer, they felt he pushed them off the board. I’ve had similar experiences myself against strong opponents; the pressure is somewhat physical. And that tends to produce blunders. As Fischer said, he liked to destroy the opponent’s ego.

Does this answer your concerns? Haven’t you ever played worse because you lost confidence against a stronger player? Haven’t you ever pushed a weaker player to the point they lost all their ability to fight against you? For the first time perhaps, Kasparov was on the receiving end of this, and he did not take it well. That’s how I interpret the events.

Until there is a quantum computer. :smiley: