Although his prose is more serviceable than sparkling, Stephen Baker chronicles what happens when IBM’s serious researchers confront a high-risk/high-stakes challenge at the intersection of humiliation and breakthrough. Given the immature mix of artificial intelligence techniques and technologies, the Jeopardy challenge was far more difficult than that presented by chess.
After all, chess — the royal game — had been the lab rat of artificial intelligence research for decades. Jeopardy — the game show of the upper middlebrow — sometimes involved more competitive, interpretive, and open-ended interplay than chess. The task of recognizing, evaluating, and processing puns, pop culture references, and subtle wordplay in less than two seconds is a nightmarish programming proposition.
But David Ferrucci, the stressed-out researcher tasked with bringing Watson to life and Baker’s chosen hero, is fully committed. Money plays only a minimal role in this narrative. IBM supported the Jeopardy challenge both as a publicity stunt and as a forcing mechanism to integrate nonaligned strands of its artificial intelligence and analytics research efforts.
I’m comfortable arguing, as Baker is not, that a decade hence, Watson’s triumph in Jeopardy will be regarded as a far more technically and economically significant event in computing history than Deep Blue’s victory. Why? Because the way people interact with machines around seemingly simple questions and answers represents a profound shift in the coevolution of technology. It’s not an accident that one of IBM’s most important prototyping tools in Watson development was Google.
Just observing how IBM modeled, simulated, and evaluated what it takes to win at Jeopardy is an anecdotal treat. Knowledge is not the same as understanding. “This led to an early conclusion about a Jeopardy machine,” Baker writes. “It didn’t need to know books, plays, symphonies, or TV sitcoms in great depth. It only needed to know about them.... Ken Jennings, Ferrucci’s team learned, didn’t prepare for Jeopardy by plowing through books. In Brainiac [Jennings’s pop autobiography], he described endless practice with flash cards. The conclusion was clear: The IBM team didn’t need a genius. They had to build the world’s most impressive dilettante.”
Designing for dilettantism across the breadth and range of Jeopardy categories was enormously difficult. But Google- and Wikipedia-type technologies — combined with computationally intensive statistical learning algorithms — ultimately gave Watson the power to win.
When Ken Jennings lost to Watson, he noted on his (correct) final Jeopardy answer the mock ironic line from a famous Simpsons cartoon: “I for one welcome our new computer overlords.” This was a passing of the pop trivia torch from the most successful human player to his silicon successor. When Jennings completed his run of Jeopardy wins in 2004, no one in computer science — including the Googlers —would have predicted a Watson-like triumph within a decade.
Francis Bacon, the founding philosopher of science to whom the famous phrase “Knowledge is power” is attributed, also observed in 1620 that “we cannot command nature except by obeying her.” In this later observation, he anticipated Carver Mead’s aphorism by roughly 375 years. The essential truth of that prescient insight hasn’t changed a bit. But the technologies have evolved, in every meaning of the word. Their ongoing evolution, these three books agree, is also our own.
- Michael Schrage is a contributing editor to strategy+business and holds appointments at MIT’s Sloan School of Management and London’s Imperial College. He was previously a Washington Post reporter and a columnist for Fortune and the Los Angeles Times.