Behind Deep Blue
By Feng-hsiung Hsu
Princeton University Press; 320 pp. $27.95

In May of 1997 an historic event occurred:  the world chess champion, Garry Kasparov, lost a 6-game match to a computer, having never previously lost a professional match to anybody.  The computer was a custom machine, designed and programmed by a team at IBM Research in Armonk, New York.  It went by the name Deep Blue, its design having been developed from a predecessor named Deep Thought, and blue being the stock color of IBM’s mainframe lines.  The defeat was not completely unexpected:  Kasparov had encountered Deep Blue at a similar match in Philadelphia fifteen months previously, and lost one of the six games.  The 1997 event, however, is generally taken to mark the point at which computers became able to play chess as well as any human being at all.

Feng-hsiung Hsu was a member of the team that built Deep Blue, and he played at the board, carrying out the moves the computer generated, for the first of those six games.  Born in Taiwan in 1959, Hsu took a degree in Electrical Engineering there before moving to the U.S. in 1985 to study at Carnegie Mellon.  His passion at the time was computer chip design, much promoted by the Taiwan authorities during the 1980s as a promising  development path for a resource-poor country.

One of the book’s surprises is that one doesn’t have to be very good at chess to design a machine that can defeat grandmasters.  The author knew Western chess when he came to the U.S.A. (there is a Chinese game, with different pieces and different rules) but preferred Go.  “I was a very weak chess player when I moved to the United States,” he admits.  Even at the time of the Philadelphia match in 1996, he is still saying:  “I could barely play chess.”   Hsu had the assistance of grandmasters, of course, but he himself seems never to have learned to think like a chess player.  He takes wry pleasure in recounting an exchange with Kasparov in which the World Champion asked for a view from Deep Blue’s perspective of a game just played.  “Sensing that I was not seeing the game on the same level as he and Deep Blue, Garry gave up after a few minutes...”

I was pleasantly surprised by Hsu’s book.  I know quite a lot of Chinese computer geeks.  Each of them is admirable in his (they are all male) own way, but I wouldn’t want to read a book by any of them.  Furthermore, I am a chess duffer.  Hsu got my attention and kept it, though, bringing this strange story to life with a fluent, modest style, some side excursions into academic politics, a dash of wit, and riveting accounts of the games — and the gamesmanship — that led up to the May, 1997 victory.

That the games are so fascinating is an odd thing in itself.  They are described in detail, and the most important ones are written up using chess jargon in an appendix.  That is all over my head; yet the interactions across the board, especially those with Kasparov, still manage to be gripping.  I found myself thinking of those compulsively watchable game scenes in the old Steve McQueen movie The Cincinnati Kid.  (I don’t play poker, either.)

The fundamental disadvantage that the creators of Deep Blue had to overcome was that human chess players, at any rate those at the Kasparov level, can get an instinctual “feel” for an opponent’s style, and outwit that style by sheer imagination, playing to the computer’s weaknesses and forcing it into errors.  That was what happened in the sixth game in Philadelphia in 1996, the last game of the match Kasparov won.

The first four games of the 1996 match gave computer scientists a glimpse of the holy grail.  We could see the top of Mount Everest ...  In the last two games, Garry forced Deep Blue into unknown territory, probed for its weaknesses, exploited them, and played Deep Blue as if he were a virtuoso conductor.  Mount Everest grew a thousand feet.  

There is not too much technical detail here, though a certain amount is of course unavoidable, and might have come with better explanations.  (A glossary would be a good idea.  What, for example, is a “net list”?)  The business of designing special-purpose computer chips seems to be fearsomely complicated.  Hsu speaks casually of a chip containing 36,000 transistors, each of which had to be correctly placed relatively to the others — and this was in 1985!  A great deal of code has to be written, too:  several hundred thousand lines, for the 1997 victory.  Mercifully, there is even less about that, perhaps because the author is not much interested in software.

Part of this book’s peculiar charm is that Hsu is level-headed about what he and his colleagues accomplished.  The 1997 match against Kasparov was billed rather crudely in parts of the press as “man vs. machine.”  Kasparov himself was prone to this kind of talk, referring to himself before a 1989 exhibition match against Deep Blue’s predecessor, Deep Thought, as: “the defender of the human race.”   The author pooh-poohs all this.  Writing of the 1997 encounter, he says:

[T]he match was never really “man versus machine,” but rather “man as a performer versus man as a toolmaker.”  Whatever the outcome of the match, when we cheered the winner, we were cheering for a unique human achievement.  When Garry [Kasparov] won in 1996, the Deep Blue team cheered and applauded his outstanding personal performance.  When Deep Blue won in 1997, society at large, not including Garry Kasparov, finally recognized Deep Blue for what it was, namely, the advancement of a powerful tool created by human beings.

That “not including” is a reference to the fact — all too plain, though Hsu is gentlemanly and restrained about it — that Kasparov has worked hard to keep alive the tradition of chess superstars being difficult and obnoxious human beings.