Kasparov Versus Deep Blue Summary
Whether by canny publishing design or mere chance, Monty Newborn’s Kasparov Versus Deep Blue benefited from a considerable accident of timing. The book was in stores on the heels of the May, 1997, match in which world chess champion Garry Kasparov suffered a stunning defeat at the silicon “hands” of Deep Blue, the world’s strongest chess-playing computer program. The result of that contest sent shock waves through the chess world, induced euphoria in artificial intelligence aficionados, and brought the principal members of these arcane and typically obscure communities an enormous amount of unaccustomed attention from the mainstream media. The human champion’s defeat was generally discussed with an undercurrent of alarm, as an ominous sign of the inexorable and apparently inevitable ascendancy of technology over purely human ability. Chess is often viewed as the ultimate test of pure intellect; if humanity’s best can be overcome by an assemblage of circuits, countless pundits asked, what hope is there for the rest of us?
Public interest in such questions made the 1997 duel the most widely followed chess match since the 1972 Cold War confrontation between Soviet champion Boris Spassky and American challenger Bobby Fischer. Television analysts, newspaper columnists, science critics, social commentators, and seemingly anyone else with access to a camera or a byline used the event as a pretext for the discussion of man-versus-machine issues, and literally millions of fans followed the games live over the Internet. The author and publishers of Kasparov Versus Deep Blue were no doubt delighted with the attendant interest in their topic, but prospective readers and purchasers should be forewarned: The book was written well in advance of the 1997 match and does not cover it at all; nor, despite some passing attention in brief chapters at the book’s beginning and end, does the author give much consideration to any larger implications of the confrontation of mind and machine.
Instead, Newborn traces the evolution of chess-playing computer programs up through the February, 1996, match between Kasparov and Deep Blue, a contest that Kasparov won handily (a result that may have led him to underestimate the computer’s playing strength in the 1997 rematch). Although Newborn provides a thorough and informative overview of the subject, his account is likely to appeal chiefly to readers with a serious interest in either chess or artificial intelligence. Readers whose primary interest is in the drama of Kasparov’s defeat would do well to wait for the publication of works devoted to the second match.
For those with an abiding interest in either computers or chess, however, Kasparov Versus Deep Blue is a first-rate read. Beginning with the efforts of computer pioneers Claude Shannon and Alan Turing in the post-World War II era, Newborn recounts the steady advances in chess program sophistication and playing strength. In 1950, Shannon, a Bell Telephone researcher and a leader in the early development of information theory, published the first paper outlining the design of a chess-playing program (although no program was built precisely according to his outline, virtually every actual program has built on his ideas). Turing, a British scientist also regarded as one of the giants of computer history, earned celebrity for designing the machine used to crack the German military code during World War II; by 1951, he too had published simple algorithms that could be used as the basis for an artificial chess player.
From these beginnings, progress was slow but steady. By the 1960’s, scientists on both sides of the Iron Curtain were laboring to create programs that could mimic the decision-making processes of expert players, as it was understood that such technology could well have broader—and possibly military—applications. A 1966 long-distance match between computers at the Massachusetts Institute of Technology and the Moscow Institute of Theoretical Physics (won by the Soviet program, 3-1) provided a rare instance of Cold War technological exchange.
No early program, however, showed much evidence of real playing strength. Fischer, one of history’s strongest players, toyed with Mac Hack, one of the world’s best programs, in three games in the late 1960’s (he later asserted that he could have given the computer queen-and-rook odds—a prohibitively steep handicap—and still won). In 1968, David Levy, one of Great Britain’s top players, bet a group of computer scientists $10,000 that no program would be able defeat him in a match within the next ten years. He won the wager with relative ease, downing a series of leading programs throughout the decade. Before an audience at the 1978 Canadian National Exhibition, the tuxedo-clad Levy handily defeated a program known as Chess 4.7 to claim the prize.
Nevertheless, real progress was being made as the result of steady advances in both computer hardware and programming sophistication. Most early programs sought to mimic the mental processes of the best human players, but these attempts were gradually abandoned. Human experts, faced with the overwhelming theoretical complexity of the game, can calculate good moves by relying on a relative handful of general principles that make it possible to ignore the vast majority of potential choices. Were it not for the utility of this commonsense and somewhat instinctual approach, the almost limitless options would make the game impossible. Attempts to instruct computers to focus in the manner of a human expert (or, in programming argot, to “forward prune” branches from the “move tree” according to general principles) met with little success. Humans can think with a flexibility that enables them to deal with the many exceptions to each rule, but computer programs that attempted to allow for exceptions played erratic, unimpressive chess.
In the early development of the field, therefore, a debate raged between adherents of the human-style approach and advocates of the “brute-force” method, which ignored general principles and relied on computer speed to calculate the consequences of every possible move for a certain distance ahead in a given situation. Increasingly powerful computer hardware eventually settled this VHS-or-Beta debate in favor of the brute-force school. In 1989, Levy was decisively defeated in a four-game match by Deep Thought, a Deep Blue prototype capable of evaluating more than 200 million prospective positions when making a move. By the 1996 match with Kasparov, Deep Blue could evaluate 100 billion possible consequences of each move.
Newborn’s book contains substantial technical exposition, and only the hardier readers are likely to wish to delve into discussions of the minimax theorem, the alpha-beta algorithm, transposition tables, search trees, horizon effects, and other programming arcana. Most of this material, however, is confined to the opening chapters—which, the author acknowledges, can be skipped by the general reader without much loss. The chess content is rather more pervasive, as the book contains move-by-move summaries of dozens of games played by computers. The games, though, are well but lightly annotated, and the summaries provided should make it possible for even the nonplayer to follow the storyline.
The volume concludes with an extensive and rather odd assortment of appendices: a tabular history of major computer chess championships, a listing of the rules governing the Kasparov-Deep Blue match, a tabular “diary” of Deep Blue’s competitive results, a similar account of Kasparov’s performances against leading computers, and an explanation of the chess notation used in the book. The first four are informative but unlikely to interest most readers, while the last seems superfluous; any reader with the expertise to follow the games will surely be familiar with standard notation. Although the appendices certainly do not detract from the book’s value, it seems a pity that the space devoted to them (about 10 percent of the volume) could not have been put to better use—say, a longer and more thoughtful discussion of the man-versus-machine issues that drew so much attention to the Kasparov matches. Two other design quibbles are worth noting: The book is profusely illustrated with photographs and chess diagrams, but the quality of both is surprisingly low. The many photos are mostly low-resolution snapshots of programmers and chess players (neither, it must be said, among the most attractive of subjects!) that add little to the volume’s charm. More serious is the matter of the diagrams, which resemble the on-screen graphics of an outmoded home-computer chess program (the queens look something like squashed spiders or alien spaceships, the knights appear to be melting, and the white pieces in general seem to fade into the black squares). For readers who attempt to follow games in their heads, poor diagrams are a real drawback—and a perplexing one, given that high-resolution printer graphics can now be had even for personal computers.
Despite such flaws, Kasparov Versus Deep Blue is an effective, scholarly overview of the field of computer chess. Readers may well hope that Newborn goes on to produce a similarly well-written, and perhaps intrinsically more absorbing, account of the historic 1997 match that may have signalled the end of human chess supremacy.
Sources for Further Study
The Guardian. April 24, 1997, p. O3.
Science. CCLXXVI, June 6, 1997, p. 1518.
Science News. CLI, March 8, 1997, p. 138.
The Washington Post Book World. XXVII, March 9, 1997, p. 3.