August 16, 1992, was a fateful day in the history of artificial
intelligence: It marked the first time a computer program
had ever challenged a human being for a world championship.
The challenger was a 400-pound slab of silicon named Chinook.
The defending champion was a 65-year-old college professor
named Dr. Marion Tinsley.
The game was checkers.
In recent years, the ancient art of checkers has gained
a reputation as the Rodney Dangerfield of board games. Yet
beneath its surface simplicity lies a fierce struggle every
bit as challenging as more intellectually respectable pursuits
such as chess. The game's deceptive charm--plus the fact
that other researchers had largely overlooked it--persuaded
Jonathan Schaeffer, a computer science professor at
the University of Alberta, to abandon his fledgling chess
program and turn to checkers instead. One Jump Ahead
is his fascinating account of the project, which began
as a mere academic exercise but became an obsession to win
the world championship--and, more specifically, to defeat
Tinsley.
Tinsley is such an intriguing character he merits a biography
of his own. The son of a Kentucky sheriff, he earned a mathematics
degree and became a professor at Florida State University
in Tallahassee. Following a religious experience, he moved
across town to teach at Florida A&M, a primarily black
college, and spent his summers canvassing door-to-door to
win souls to his Baptist faith.
But beneath this gracious Southern exterior lurked a colossus
of the checkerboard. Since clinching the world championship
in 1952, Tinsley won every tournament he ever entered and
every match he ever played. He retired from checkers several
times, only to return to the game stronger than ever. Incredibly,
during his 40-year reign, he lost only five games. In Schaeffer's
words, Tinsley was "as close to perfection as humanly possible."
Tinsley was up against a formidable opponent, however:
Chinook boasted eight brains, a database of 85 billion endgame
positions and the ability to peer as many as 27 moves into
the future. A single keystroke could transform its personality
from cautious to swashbuckling, making it a difficult player
to prepare against.
To non-combatants, Chinook may seem like a boardgame Terminator,
an indestructible automaton. But as Schaeffer's book shows,
its development was surprisingly haphazard. In fact, Chinook
was plagued by bugs: Its vaunted databases were riddled
with holes; its knowledge of checkers was surprisingly unsophisticated,
at times almost laughably naive; and when Schaeffer made
last-minute corrections to the program, he often accidentally
introduced new errors. Far from being an infallible juggernaut,
Chinook comes across as a digital enfant terrible,
capable of sparkling brilliance and maddening blunders in
the same game.
Held in London, the clash of the checkerboard titans generated
intense media coverage and was portrayed as a battle between
man and machine. Tinsley even cast the match in spiritual
terms: "I have a better programmer than Chinook," he told
the Daily Telegraph. "God gave me a logical mind."
In the end, Tinsley vanquished Chinook, thanks to flashes
of brilliance and a mysterious computer malfunction.
Schaeffer learned from his mistakes and challenged Tinsley
to a rematch in 1994. But the return bout was a puzzling
anti-climax: After six drawn games, Tinsley complained of
an upset stomach, withdrew from the match and forfeited
the title. A few days later, doctors discovered a malignant
tumor in his pancreas. Tinsley died the next year, leaving
the cancerproof computer as the world's reigning champ.
One Jump Ahead is not only gripping, it also raises
many provocative questions. Could Tinsley have beaten Chinook
in the rematch? Are computer programs destined to outplay
humans in chess, go, bridge, and even poker? Perhaps most
important, what does Chinook's triumph imply for the future
of machine--and human--intelligence? After all, Chinook's
strength lay in its gigantic databases, which effectively
reduced the game's creative subtleties to a mechanical exercise
in searching a catalog.
But Schaeffer makes a persuasive case that we should not
define computer intelligence by its similarity to the human
brain. Rather, he says, we should look at the results: Can
a robotic program excel at a task which, in humans, obviously
requires imagination and intelligence? The answer, at least
in the case of Chinook, is a resounding yes--and the implications
must be correspondingly profound.
- - - - - - - - - - - -
- - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Willamette Week | originally
published April 5,
2000
|
