Defending the brain: Chess champion Garry Kasparov versus the machines

0
98

When World Chess Champion Garry Kasparov stepped up to the podium to play his first game against the computer Deep Blue he is famously quoted as having commented, “Why is there an American flag and a Russian flag? This is not Russia against America, it is man versus machine. I represent humanity here.”

Five times in all Kasparov has taken up the mantle of defender of humanity, against various computer opponents, and it’s a role which he takes very seriously. For Kasparov believes, as do his legions of fans, that chess is the ultimate expression of the human mind, a supreme mix of intuition and intelligence which no computer should rightly be allowed to conquer. For Kasparov, chess is “life in miniature” and “an art”, a realm a computer should never be able to comprehend. According to Kasparov’s tutor Mikhail Botvinnik however, chess is not just an art. It is “the art which expresses the science of logic”, the “art of analysis”, and what is a computer but an analysing machine?

Kasparov vs Karpov. Credit: Instagram

So when Kasparov moved the first pawn in his first game against his first computer opponent “Deep Thought” in 1989, he wasn’t only defending the human brain against the calculating power of computers. He was also defending his image of a game which he has made his life’s work and passion, against the cold logic which his mentor had ascribed to it. The stage was set for a series of emotionally charged confrontations between man and machine, which would, at their peak, see over two million hits a day being recorded on official competition websites.

Garry Kasparov was born in Baku, the capital of the Russian republic Azerbaidzhan on the 13th of April 1963. He became a grandmaster aged just 17 and at 22 the youngest world champion up to that date. He defended his title three times against arch-rival Anatoly Karpov before splitting from the Federation Internationale des Echecs (FIDE) and playing Nigel Short of Britain for the title under the auspices of the Professional Chess Association (PCA). He again defended his title in 1995 against Indian Grandmaster Viswanathan Anand. In 1990 The Brain Club’s journal Synapsia elected Kasparov as its first “Brain of the Year” and described him as “The World Chess Champion, athlete and humanitarian both, and a cultivated and curious man who closely follows literature, films and politics”.

He retired from professional chess in March 2005 after a series of collapsed deals saw him unsuccessful in his bid to reunify the world championship. He is currently dedicating himself to Russian politics which he says is headed down the wrong path under Vladimir Putin.

Deep Thought: 1989

Kasparov often remarked that his love for chess was started when he solved a chess puzzle “very young” that his parents were struggling with.

Estimated by Kasparov to play at a FIDE rating of “about 2500”, Deep Thought was an early precursor to the now-legendary Deep Blue. The match between Kasparov and Deep Thought was organised by television chess commentator Shelby Lyman and sponsored by AGS Information Services in New Jersey. It was billed as a test of the shrinking “intelligence” gap between man and machine and the match was accompanied by a whirlwind series of media and television appearances across the United States. Media attention was focused on the match for weeks in advance of the actual games and many say that the Kasparov easily achieved his goal of revealing to the American media that chess was appealing to that nation’s public.

Appearances on David Letterman’s show, where he lightly teased the host’s playing abilities, Good Morning America and PBS’s The Eleventh Hour hyped up the match to a point which would only be exceeded years later by the Kasparov vs Deep Blue matches.

As for the games themselves, they were over pretty quickly. The computer technology being what it was at the time, Deep Thought was not able to play at even close to the level of the world champion and it took only a few hours for Kasparov to dismantle his opposition two games to none. The first game, in which Kasparov perhaps showed his opponent a little too much respect, was only resigned by Deep Thought after 52nd move. The second game was another matter entirely, with Kasparov forcing a convincing victory after move 37. Commentating Grandmaster Edmar Mednis had however described the game as over by move 15 with Deep Thought clinging to a hopeless position for long after it had already lost the game. Kasparov stood to rapturous applause after which he commented, not entirely in jest, that the computer still had a thing or two to learn – principally how to resign.

Deep Blue: The Original Match 1996

IBM’s Deep Blue.

While experts agreed that past computer chess games were not an accurate test of the abilities of both man and modern technology, they all admitted that the Kasparov Deep Blue game would accurately present the skills of both sides.

The match, which took place in Philadelphia, was held under the auspices of the International Computer Chess Association (ICCA) and was structured in a standard tournament style with a maximum of eight hours per game. It was the first real test of man versus machine under tournament conditions and carried with it a winner’s prize of US$400,000.

Deep Blue was developed as an improvement on the Deep Thought chip at IBM’s TJ Watson Research Centre in New York by the original chip designers, chess expert Murray Campbell and technical wizard Feng-Hsiung Hsu, as well as a team of scientists including Joe Hoane, Jerry Brody and Chung-Jen Tan. The original two had begun working on their project while still doctoral candidates at Carnegie Mellon University in Pittsburgh, and Deep Blue was just the latest in a long line of prototypes.

Described as playing through a “Brute Force Method”, Deep Blue was capable of analysing up to 20 billion different positions in the time allocated for each move.

But Deep Blue did not simply evaluate every possible board position. It employed a system called “selective extensions” to examine the board. “Selective extensions” allowed the computer to more efficiently search for critical board arrangements. This means that Deep Blue selectively chose distinct paths to follow, eliminating irrelevant searches in the process and thereby saving time.

The core of Deep Blue’s skill, however, lay in its ability to evaluate the billions of possible board arrangements and determine which move resulted in the best position. This was achieved through the computer’s “evaluation function”.

The “evaluation function” is an algorithm that measures the “goodness” of a given chess position. Positions with positive values are good for White, and conversely, positions with negative values are good for Black. If the overall score is negative, for example, this means that Black has the advantage.

Deep Blue’s evaluation function looks at four basic chess values: material, position, King safety and tempo. Material is based on the “worth” of particular chess pieces. For example, if a pawn is valued at 1, then the rook is worth 5 and the Queen is valued at 9. The King, of course, is beyond value because his loss means the loss of the game.

The simplest way to understand position is by looking at your pieces and counting the number of safe squares they can attack. King safety is a defensive aspect of position. It is determined by assigning a value to the safety of the King’s position in order to know how to make a purely defensive move. Tempo is related to position but focuses on the race to develop control of the board. A player is said to “lose a tempo” if he moves unnecessarily while the opponent is making more productive advances.

The software begins this process by taking a strategic look at the board. It then computes everything it knows about the current position, integrates the chess information pre-programmed by the development team, and then generates a multitude of new possible arrangements. From these, it then chooses its best possible next move.

Kasparov had nowhere near the same ability to foresee the future but he did have his famed human intuition and an ability to play against the computer’s weaknesses.

Credit: Youtube

At the time, computers notoriously preferred an open position as this meant that there were more options for future moves. Additionally, the computer is usually less capable of factoring in material swaps for position, regarding the material gained as more important than the position achieved. The ability to play against the computer’s weaknesses proved decisive.

In game one Kasparov played his usual attacking style of game. He opened up the board and left his king exposed allowing Deep Blue to take advantage and stun the chess community by forcing Kasparov to resign in the 36th move. Stunningly, the computer had earlier seemingly sacrificed a pawn for position, a move which Kasparov admitted afterwards had thrown him. He remarked that he, “could smell a new kind of intelligence across the table.”

After this game Kasparov talked to the machine’s designers and verified that it had simply calculated every possible move and deduced that six moves ahead it would take a pawn, recovering the loss. He regained his confidence, because, “the computer hadn’t viewed the pawn sacrifice as a sacrifice at all.”

This defeat led to a change in Kasparov’s tactics and he began playing to the computer’s weaknesses.

In game two, Kasparov, now playing white, elected to use the positional Catalan Opening, which is a hybrid of the Queen’s Gambit and the King’s Fianchetto. This is a conservative, closed opening used specifically to limit the computer’s options. In what was one of the most tense games of the competition, Deep Blue eventually resigned on move 74 when it realised that nothing could stop Kasparov’s three connected passed pawns. The next two games were draws with neither side sacrificing an inch, but an error in game five by Deep Blue saw Kasparov take a one game lead into the decisive game six.

If Deep Blue was human, journalists would have described its loss to Kasparov in game five as being caused by a loss in concentration. As it is, Deep Blue’s play in game five is inexplicable in that after refusing Kasparov’s offer of a stalemate on move 23 the computer seemed to lose its way, moving in an uncoordinated and confused manner. Kasparov quickly set up his pieces into the ideal attack formation and by move 31 was very much in the ascendancy. Deep Blue programmer Murray Campbell conceded on Deep Blue’s behalf on move 51.

Game six was one of the shortest in the tournament. Just before move 40, the more than 600 spectators applauded and laughed as they saw the champion pick up his watch from beside the board and put it back on his wrist. This familiar gesture showed that Kasparov believed the game was over, and he was proved correct in just three moves when Deep Blue resigned for the third time in the match.

Kasparov had won a decisive victory but the computer had managed to prove itself far stronger than pre-match predictions had suggested it would be.

After the match, Kasparov spoke of his victory saying, “It is weakest in a position where it doesn’t have a plan. You have to limit its unlimited potential. You have to be careful not to create weaknesses in your own position, not to leave hanging pieces, not to leave a king threat. You have to play solid, positional chess because any mistake will be punished by the machine more severely than by a human player.

“My overall thrust in the last five games was to avoid giving the computer any concrete goal to calculate toward… In the end, that may have been my biggest advantage: I could figure out its priorities and adjust my play. It couldn’t do the same to me. So although I think I did see some signs of intelligence, it’s a weird kind – an inefficient, inflexible kind that makes me think I have a few years left,” he added.

How wrong he was. Just over one year later Deep Blue was to come back after a programming overhaul to give Kasparov the biggest surprise of his career.

Deep Blue: The Rematch 1997

Legends of the game: Kasparov vs Karpov

The fifteen months between the first match between Deep Blue and Garry Kasparov saw the computer’s programmers make a number of significant changes. It was in early March 1996, just a couple of weeks after the first match, that CJ Tan, the Deep Blue project manager for IBM, made a list of ways in which the computer could be improved.

The list, of little under one and a half pages, included an examination of the computer’s hardware, a suggestion that the computer’s opening repertoire be augmented, an increase in databases dealing with chess endgames, the creation of a more powerful and sophisticated evaluation function for chess positions, the hiring of additional grandmasters to work with the team, and the development of a strategy to exploit an opponent’s time trouble and to disguise the computer’s own strategy on the board.

The computer’s understanding of opening became a particular focus as with so many undeveloped pieces, the computer’s understanding was bound to be limited. Kasparov had foreseen this during the original matches and had developed openings which specifically deviated from standard openings to force Deep Blue into unfamiliar positions as early as possible. To work on this, and other problems, chess Grandmaster Joel Benjamin was brought in to help the computer work through its weaknesses and to spot and pinpoint holes in its knowledge.

“For example,” Benjamin said, “I noticed that whenever Deep Blue had a pawn exchange that could open a file for a rook, it would always make the exchange earlier than was strategically correct. Deep Blue had to understand that the rook was already well placed and that it didn’t have to make the exchange right away. It should award evaluation points for the rook being there even before the file is opened.”

A few weeks before Deep Blue was to meet the champion for the second time, another anomaly was discovered. For some reason, as the computer began searching the lines of possibility, it began seeing a queen where there wasn’t one when examining certain moves. This problem was eventually traced to a newly designed chip linking Deep Blue’s powerful multiprocessors with its chess software and was corrected over two weeks. According to the New York Times, mentioning the anomaly of the imaginary queen to designer Feng-Hsiung Hsu, even many years later, still sent him into a fit of guilt.

In addition to the deeper understanding of chess that was given to Deep Blue, the computer was also made faster. According to the development team, the new Deep Blue was able to explore 200,000,000 positions per second, compared to an estimated three positions a second which Kasparov could examine. The new Deep Blue would, therefore, be likely to outperform the machine of just fifteen months previously without a problem, which explains why many refer to these matches as being between Kasparov and Deeper Blue.

The rematch saw Kasparov and the newly refurbished Deep Blue again play each other over six games under tournament conditions at the Equitable Center in New York City from May 3 to May 10, 1997. At stake was a US $700 000 prize purse to the winner and $400,000 to the loser.

The match took place on a set that looked like a study, complete with bookshelves, a potted plant and duck decoys. Behind the set was a dressing room where Kasparov was allowed to rest between moves. This was made difficult, however, by the fact that the computer responded so quickly. Even going to the toilet usually cost Kasparov time on his clock.

This time it was Kasparov who drew white for game one – an advantage, as it allowed him to direct the game. He opened with a Reti specifically designed to keep the game as closed as possible and limit Deep Blue’s options. Despite this, the momentum of the game swayed back and forth between the players and both had ascendancy at different times during the game. Kasparov’s defensive stance is very blatantly evident in the fact that not once did he move a piece across the fourth rank.

The Deep Blue team tendered their resignation on move 45 after a rook for a bishop and pawn sacrifice by Kasparov ensured he had two passed pawns which could not be stopped.

After the match Kasparov cautioned the spectators, who gave him a two-minute standing ovation, that despite the fact that he had won the first game the tournament was far from over. “If we carry on in the same style, I will need a lot of energy,” he admitted.

That night the champion and his team spent a few hours evaluating the game, trying to determine some lessons they could take away before heading out for a celebratory lobster dinner at the Ocean Club in Manhattan.

The next morning Kasparov took a walk in Central Park before settling down to play a game which he knew would be tactically more difficult than the previous day’s affair for the simple reason that Deep Blue would now be making the first move.

Credit: Youtube

Game two began with Deep Blue initiating the notoriously difficult Ruy Lopez opening – a strategy which, while exceptionally difficult to initiate, is almost impossible to counter if done correctly. The first 17 moves for each player only took 25 minutes whereafter the game settled into a more normal pace. A knight swap almost two hours into the game threw Kasparov, and his demeanour became more and more concerned from that point on. The champion was clearly rattled. He became more and more animated, often revealing his unhappiness and disbelief after every move. It was only a matter of time, and on move 45 Kasparov surrendered. Some of the computer’s moves had clearly unnerved Kasparov and he admitted afterwards that the level of intelligence shown by the machine during this game had been truly startling. Even more shocking for Kasparov was the discovery by the internet community that he had, in fact, resigned a drawn game. After being out-played throughout the game, and with humiliation on the horizon, Kasparov resigned not noticing that Deep Blue had made a critical error which allowed Kasparov a perpetual check. After his team revealed the hidden draw, Kasparov fell silent for a full five minutes before he eventually conceded that he had made an error. He later claimed that this was the first time he had resigned a drawn position.

Games three and four of the series were both draws that saw Kasparov attack viciously only to be repelled time and time again by the computer. At an average of five hours each these games were physically sapping for Kasparov, yet the length and difficulty didn’t affect the computer at all and it returned to the table for game five as strong as it was on day one.

Kasparov’s exhaustion is revealed in his telling comments following game four, “It was a tough game. I think I was winning at some point, but I didn’t manage the position very well.”

“I was tired and I couldn’t figure it out. Obviously the position was winning at some point, but I couldn’t finish it off,” he added.

The Deep Blue development team were, on the other hand, happy to accept a draw. “For the last few hours, we were hoping we’d get the draw,” said a relieved Murray Campbell. “Deep Blue was just hanging on for those last few hours.”

Game five saw Kasparov once again take control of the white pieces. It was to become what Kasparov himself described as the best game of the tournament.

Initially, it seemed as though Kasparov was, at last, going to play the attacking chess which had made him famous because immediately after the opening, he went on the offensive by aggressively seeking to control the middle of the chessboard. However, he soon settled back into the conservative style that has characterized his play during the previous four games. This was because despite the fact that Kasparov held a slight advantage going into the middle game, his overly aggressive play had opened up the board and allowed Deep Blue to neutralise the position. The complicated positions which Deep Blue forced had Kasparov pondering his moves for far longer than is usual, which in turn placed severe time pressure on the Russian. At one point he was forced to make fourteen moves in just twenty-one minutes, ensuring the pressure was never on Deep Blue. Toward the end, Kasparov sensed a chance to queen his pawn. Rushing it to its destination he was surprised to see Deep Blue completely ignore the imminent threat. Commentators and other watchers alike were also astounded when Deep Blue managed to set up a perpetual check, saving itself from the passed pawn and forcing Kasparov to offer a third stalemate in a row.

After the game, Kasparov said he was impressed by Deep Blue’s ability to play a very rational game of chess. “The computer played like a human today. I have to praise the machine for understanding some very deep positions.”

After five games the score was tied at 2.5 to 2.5 and all hinged on the final game, one in which Deep Blue was playing white and therefore carried all the aces.

The deciding game was played under an atmosphere of extreme pressure and, owing to the nature of these encounters, only one player felt it. Many have commented on the series of events which ultimately lead up to Kasparov’s concession just nineteen moves into this final game. They have theorised that the two games he lost playing black were lost due to the fact that he had chosen to play in the style of his arch-nemesis Anatoly Karpov instead of remaining true to himself. They have suggested that maybe he had thought he had a refutation for the opening in game six which did not come off. The truth however probably lies in the unbearable pressure and the fact that Kasparov had tired significantly over the week while his opponent felt no physical ill effects.

Choosing to open with the Caro-Kann variation, Kasparov made a common opening blunder. He allowed Deep Blue to sacrifice a knight for a pawn thereby denying him the right to castle. The look of horror on Kasparov’s face when he realised what he had done is evidence enough that the move was nothing more than a common blunder. Less than an hour into the game Kasparov conceded and a computer had claimed it’s first Grand Master scalp under tournament conditions.

After the match, Kasparov admitted that he had not been in the best frame of mind before the game and that this had probably cost him.

“For me, the match was over yesterday,” he said. “I had no real strength left to fight. And today’s win by Deep Blue was justified.”

He also famously attacked IBM, accusing the company of cheating, saying that he understood how computers played and that there were moves made during the match which it was not possible for a computer to play. In the press he came off as a sore loser but perhaps he had a point. IBM had Deep Blue dismantled after the game and didn’t allow anyone to examine it at all. Additionally, the games were analysed years later by the computer programme Fritz and despite the fact that Fritz was able to, with a few hours thought, come up with 99% of Deep Blue’s moves there was still one percent which it could not replicate even when investigators gave it significant clues.

Kasparov discusses that last game

Deep Junior: 2003

Many thought that Kasparov’s next computer challenge would not come off after it was repeatedly delayed from October 2002 to November, from November to December and finally from December to January 2003. The venue for the match was also repeatedly changed but it was eventually played in New York at the same venue where 6 years earlier Kasparov had lost in the epic match against Deep Blue. The match-up saw Kasparov take on a programme known as Deep Junior. Unlike Deep Blue, which was a computer built solely for the purpose of playing chess, Deep Junior was a software programme which could be run on an ordinary home PC. At the time of playing it was a three-time World Computer Champion and had won the 15th World Microcomputer Chess Championship (WMCCC) in Paris in1997, the 18th WMCCC in Maastricht, the Netherlands in 2001; and the 10th World Computer Chess Championship (WCCC) also held in Maastricht in July 2002. Against Kasparov, the software was run on a machine with four 1.9 GHz Pentium 4 processors and an effective three gigabytes of memory.

The first game of the six-game match saw Kasparov exploit an apparent weakness in Deep Junior’s opening book from which the computer never recovered. Deep Junior’s handlers resigned just forty minutes into the encounter. The programme’s developers took the opportunity, as was allowed to them by the rules, to patch the error and the second game proved a much closer and hard-fought affair. Despite the fact that Kasparov had dominated most of the game, it ended in a draw after a sacrifice allowed the computer to force a perpetual check.


GM Garry Kasparov preparing to face off against Deep Junior in Game 5. Photo by John Henderson on Wikipedia.

Game three saw the match tied after Kasparov refused the chance to force a draw and blundered, thereby losing the game. It was an extraordinary game, with Deep Junior repeatedly making moves that no human player would dare. Among these inexplicable moves was a castle into Kasparov’s strong Kingside attack and the taking of a pawn which no human would ever possibly consider. Due to Kasparov’s blunder, however, in which he thought he had a possible checkmate but ended up costing himself significant material losses, he resigned on move thirty-seven, knowing that the computer would not blunder and allow him back into the match.

The next two games were back and forth draws with neither side really seizing the advantage or playing to their best ability. Kasparov resorted back to the same techniques which he had used against Deep Blue, playing closed games and limiting the computer’s options. Whenever a draw was even remotely possible he seized it, clearly fearful of a repeat of game three.

The final game was also a draw, the declaration of which saw the capacity crowd boo the Russian. At the time Kasparov was believed to be in a slightly stronger position with winning options. His acceptance of the draw offered by the Deep Junior team angered his fans, with many saying the Deep Blue games had so upset him that he wasn’t prepared to risk the ignominy of defeat again. Kasparov admitted that he had been afraid of a loss but said that the decision to accept the stalemate was motivated more by the blunder which he made in game three than by the earlier losses to Deep Blue. The final score of 3 to 3 was generally thought to be more than Deep Junior deserved.

X3D Fritz: 2003

Kasparov took on X3D Fitz with coverage on ESPN. Credit: Youtube

Kasparov’s final computer chess challenge as a professional saw him take on the unique X3D Fritz. The X3D Fritz combined the world’s most dominant chess software at the time with 3D imaging software to create a game wherein there would be no pieces or board. Kasparov played by looking at a 3D image of a board generated by Fritz and seen through a pair of 3D glasses. This was the first time the computer was actually able to make its own moves on the board, rather than have a handler make the moves for it.

Game one was well attended and played out in a scenario Kasparov had seen many times over the years playing against computer opposition. Despite the fact that he dominated for most of the three hours of play Kasparov was just unable to finish his opponent and X3D Fritz managed to hang on for a draw. Kasparov stated that playing with the 3D glasses had proved unusual and difficult initially, as was the need for him to speak his moves, but said he was generally happy with the day’s play.

“There were a few moments in which I could have improved,” he said, adding “I had a solid position and it seems that X3D Fritz was over-estimating its position, which gave me chances.”

As with game one, game two also ran to a scenario with which Kasparov was very familiar, though this time it was a scenario he most certainly did not appreciate. The game had been finely balanced with both sides switching dominance regularly, when on move 32 Kasparov moved one of two rooks which had been protecting each other. The rook’s absence allowed Fritz’s queen to take one of Kasparov’s pawns “for free”, and subsequently move in for the kill.

After the game Kasparov despondently spoke to the press of his disappointment saying, “You work hard for three hours, you get a very promising position, you make a blunder and you go home.”

One game down with only two to play Kasparov needed to somehow find two wins if he was going to defeat Fritz and he must have gone to bed the night after the second game fearing a repeat of the Deep Blue fiasco. It is therefore to his credit that he came back in game three as strongly as he did. Determined to win and therefore hopefully avoid another humiliating defeat to computer opposition, Kasparov produced one of the finest displays of anti-computer chess ever seen. After cramping the computer’s position early, Kasparov quickly went a pawn up. He was therefore able to sacrifice a pawn to break through X3D Fritz’s defensive line on the queen side. Fritz tried to hurry its attacking pieces back from the King side but by then the damage had already been done. Fritz resigned shortly thereafter in lieu of massive material loss. Once again Kasparov and a computer had equal scores going into the final day.

The next day saw a gritty battle with neither side prepared to succumb. Avoiding blunders appeared to be Kasparov’s primary concern as both teams set up a number of tricky traps for the other only to see their effort wasted as the opponent skilfully avoided falling into them. Ultimately French Grandmaster Joel Lautier saw his early prediction proven correct when a draw was declared by threefold repetition.

At the post-match conference, Kasparov indicated that he thought his play against computers had significantly improved over the years as the computer only managed a draw due to “an unfortunate blunder”.

Kasparov’s win against Fitz

After the fourth game against X3D Fritz, Kasparov remarked that the programme he had just played was far stronger than Deep Blue, and that he was surprised at the level of development in the field in so short a time. He added, however, that even as the machines develop, man continues to learn and so the two should provide excellent competition for each other well into the future.

He was wrong. In the years since computers have built on Deep Blue’s 1997 breakthrough to the point where the battle between humans and machines is not even close. It is said that world champion Magnus Carlsen won’t even play his computer.

“He uses it to train, to recommend moves for future competition. But he won’t play it, because he just loses all the time and there’s nothing more depressing than losing without even being in the game.” says chess grandmaster and author Andrew Soltis.

Carlsen will soon take on an as yet, unnamed challenger for the 2021 World Chess Championship and as they sit down, both will know they can be easily beaten by a R1500 computer.