Final Jeopardy (8 page)

Jennings, with his choirboy face and awkward grin, was a far cry from the tough guys on
CSI
. But he was proving to be a cognitive mauler. Some of his fallen opponents (who eventually numbered 148) took to calling themselves Road Kill and produced T-shirts for the growing club. Yet even while Jennings racked up wins he flashed humor, some of it even mischievous. One $200 clue in the category Tool Time read: “This term for a long-handled gardening tool can also mean an immoral pleasure-seeker.” Jennings, his knowledge clearly extending into gangsta rap, responded: “What is a ‘ho'?” That produced laughter and oohs and aahs from the crowd. A surprised Trebek struggled briefly for words, finally asking Jennings: “Is that what they teach you in school, in Utah?” His response was ruled incorrect. In fact, it could be argued that Jennings's gaffe was right—and far more clever than the intended answer (“What is a rake?”). He could have challenged the call, but he was so far ahead it was barely worth the bother.

What was so special about Ken Jennings? First, he knew a lot. A practicing Mormon who had spent his childhood in Korea and had done missionary work in Spain, he knew the Bible and international affairs. He'd devoted himself to quiz bowls much of his life, the way others honed their skills in ice hockey or ballet, and he had a fabulous memory. Still, his peers considered him only an excellent player, not a once-in-a-generation phenomenon. “None of us who knew Ken saw this coming,” said Greg Lindsay, a two-time
Jeopardy
champ who had crossed paths with Jennings in college quiz bowl tournaments.

Two things, according to his competitors, distinguished Jennings. First, he had an uncanny feel for the buzzer. This wasn't a mechanical ability but a uniquely human one. Sitting at the production table by the
Jeopardy
set, a game official waited for Trebek to finish reading the clue, then turned on the panel of lights on the big
Jeopardy
board. This signaled the opportunity to buzz. Players who buzzed too early got penalized: Their buzzers were locked out for a crucial quarter of a second, opening the door for others to buzz in. Jennings, said his competitors, had an almost magical feel for the rhythm of the buzzmeister. He anticipated the moment almost the way jazz musicians sense a downbeat. “Ken knew the buzzer,” said Deirdre Basile, one of his early victims. “He had that down to a science.”

His second attribute was a preternatural calm under pressure. Like other players, Jennings had a clear sense of what he knew. (This is known as “metacognition.”) But knowing a fact is one thing, locating it quite another. People routinely botch the retrieval process, sometimes hunting for the name of a person standing right in front of them. This problem, known as “tip of the tongue syndrome,” occurs more often when people are stressed—such as when they have less than four seconds to come up with an answer, thousands of dollars are at stake, and they're standing in front of a television audience of millions.

Bennett L. Schwartz, a psychologist at Florida International University, has studied the effects of emotion on tip of the tongue syndrome. He came up with questions designed to make people anxious, such as, “What was the name of the tool that executed people in the French Revolution?” With beheadings on their mind, he found, people were more likely to freeze up on the answer. Memory works on clues—words, images, or ideas that lead to the area where the potential answer resides. People suffering from tip of the tongue syndrome struggle to find those clues. For some people, Schwartz said, the concern that they might experience difficulty becomes a self-fulfilling prophecy. “I know the answer and I can't retrieve it,” he said. “That's a conflict.” And the brain appears to busy itself with this internal dispute instead of systematically trawling for the most promising clues and pathways. Researchers at Harvard, studying the brain scans of people suffering from tip of the tongue syndrome, have noted increased activity in the anterior cingulate—a part of the brain behind the frontal lobe, devoted to conflict resolution and detecting surprise.

Few of these conflicts appeared to interfere with Jennings's information retrieval. During his unprecedented seventy-four-game streak, he routinely won the buzz on more than half the clues. And his snap judgments that the answers were on call in his head somewhere led him to a remarkable 92 percent precision rate, according to statistics compiled by the quiz show's fans. This topped the average champion by 10 percent.

As IBM's scientists contemplated building a machine that could compete with the likes of Ken Jennings, they understood their constraints. Their computer, for all its power and speed, would be a first cousin of the laptops they carried around the Hawthorne lab. That was the technology at hand for a challenge in 2011. No neocortex, no neurons, no anterior cingulate, just a mountain of transistors etched into silicon processing ones and zeros. Any
Jeopardy
machine they built would struggle mightily to master language and common sense—areas that come as naturally to humans as breathing. Their machine would be an outsider. On occasion it would be clueless, even laughable. On the positive side, it wouldn't suffer from nerves. On certain clues it would surely piece together its statistical analysis and summon the most obscure answers with sufficient speed to match that of Ken Jennings. But could they ensure enough of these successes to win?

Ken Jennings's remarkable streak came to an end in a game televised in November 2004. Following a rare lackluster performance, he was only $4,400 ahead of Nancy Zerg, a real estate agent from Ventura, California. It came down to the Final Jeopardy clue: “Most of this firm's 70,000 seasonal white-collar employees work only four months a year.”

The
Jeopardy
jingle came on, and Jennings put his brain into drive. But the answer, he said, just wasn't there. He didn't read the business pages of newspapers. Companies were one of his few weak spots. He guessed, “What is FedEx?” When Zerg responded correctly, “What is H&R Block?” Jennings knew his reign was over. During his streak, he had amassed more than $2.5 million in earnings and became perhaps the first national brand for general braininess since the disgraced Charles Van Doren.

Harry Friedman, of course, was far too smart a producer to let such an asset walk away. A year later, he featured Jennings in a wildly promoted Ultimate Tour of Champions. This eventually brought Jennings into a threesome featuring the two leading money winners from before 2003, when winners were limited to five matches. Both Jerome Vered and Brad Rutter had retired as undefeated champions under the rules at the time. Rutter, who had dropped out of Johns Hopkins University and worked for a time at a music store in Lancaster, Pennsylvania, had never lost a
Jeopardy
match.

In the 2005 showdown, Rutter handled both Jennings and Vered with relative ease. He was so fast to the buzzer, Jennings later said, that sometimes the light to open the buzzing didn't appear to turn on. “It was off before it was on,” he said. “I don't know if the filaments got warmed up.” In the three days of competition, Rutter piled up 62,000, compared to 34,599 for Jennings and 20,600 for Vered. (These weren't dollars but points, since they were playing for a far larger purse.) Rutter won another $2 million, catapulting him past Jennings as the biggest money winner in
Jeopardy
history.

These two, Rutter and Jennings, were the natural competitors for an IBM machine. To establish itself as the
Jeopardy
king, the computer had to vanquish the best. These two players fit the bill. And they promised to be formidable opponents. They had human qualities a
Jeopardy
computer could never approach: fluency in language, an intuitive feel for hints and suggestion, and a mastery of ideas and concepts. Beyond that, they appeared to boast computer-like qualities: vast memories, fast processors, and nerves of steel. No tip-of-the-tongue glitches for Jennings or Rutter. But would a much-ballyhooed match against a machine awaken their human failings? Ferrucci and his team could always hope.

IN THOSE EARLY DAYS
of 2007, when Blue J was no more than a conditional promise given to Paul Horn, David Ferrucci harbored two conflicting fears. By nature he was given to worrying, and the first of his nightmare scenarios was perfectly natural: A
Jeopardy
computer would fail, embarrassing the company and his team.

But his second concern, failure's diabolical twin, was perhaps even more terrifying. What if IBM spent tens of millions of dollars and devoted centuries of researcher years to this project, played it up in the press, and then, perhaps on the eve of the nationally televised
Jeopardy
showdown, someone beat them to it? Ferrucci pictured a solitary hacker in a garage, cobbling together free software from the Web and maybe hitching it to Wikipedia and other online sites. What if the
Jeopardy
challenge turned out to be not too hard but too easy?

That would be worse, far worse, than failure. IBM would become the laughingstock of the tech world, an old-line company completely out of touch with the technology revolution—precisely what its corporate customers paid it billions of dollars to track. Ferrucci's first order of business was to make sure that this could never happen. “It was due diligence,” he later said.

He had a new researcher on his team, James Fan, a young Chinese American with a fresh doctorate from the University of Texas. As a newcomer, Fan was free of institutional preconceptions about how Q-A systems should work. He had no history with the annual TRec competitions or IBM's Piquant system. Trim and soft-spoken, his new IBM badge hanging around his neck, Fan was an outsider. Unlike most of the team, based in New York or its suburbs, Fan lived with his parents in Parsippany, New Jersey, some seventy miles away.
He was the closest thing Ferrucci had to a solitary hacker in a garage.

Fan, who emigrated as an eighteen-year-old from Shanghai to study at the University of Iowa and later Texas, had focused his graduate work on teaching machines to come to grips with our imprecise language. His system would help them understand, for example, that in certain contexts the symbol H
₂
O might represent a single molecule of water while in others it could refer to the sloshing contents of Lake Michigan. This expertise might eventually help teach a machine to understand
Jeopardy
clues and to hunt down answers. But it hardly prepared him for the job he now faced: building a
Jeopardy
computer all by himself. His system would be known as Basement Baseline.

As Fan undertook his assignment, Ferrucci ordered his small Q-A team to adapt their own system to the challenge, and he would pit the two systems against each other. Ferrucci called this “a bake-off.” The inside team would use the Piquant technology developed at IBM while the outside team, consisting solely of James Fan, would scour the entire world for the data and software to jury-rig a bionic
Jeopardy
player. They each had four weeks and a set of five hundred
Jeopardy
clues to train on. Would either system be able to identify the parent bird of the roasted young squab (What is a pigeon?) or the sausage celebrated every year since 1953 in Sheboygan, Wisconsin (What is bratwurst?)? If so, would either have enough confidence in its answers to bet on them?

Ferrucci suspected at the time that his solitary hacker would come up with ideas that might prove useful. The bake-off, he said, would also send a message to the rest of the team that a
Jeopardy
challenge would require reaching outside the company for new ideas and approaches. He wanted to subject everyone to Darwinian pressures. The point was “to have technologies competing,” he said. “If somebody's not getting it done, if he's stuck, we're going to take code away from him and give it to someone else.” This, he added, was “horrific for researchers.” Those lines of software may have taken months or even years to develop. They contained the researcher's ideas and insights reduced to a mathematical elegance. They were destined for greatness, perhaps coder immortality. And one day they could be ripped away and given to a colleague—a competitor, essentially—who might make better use of them. Not everyone appreciated this. “One guy went to his manager,” Ferrucci said, “and said that the bake-off was ‘bad for morale.' I said, ‘Welcome to the WORLD!'”

So on a February day in 2007, James Fan set out to program a Q-A machine all by himself. He was relatively isolated in a second-floor office while the rest of Ferrucci's team mingled on the first floor. He would continue to run into them in the cafeteria, and they would attend meetings together. After all, they were colleagues, each one of them engaged in a venture that many in the company viewed as hopeless. “I was the most optimistic member of the team,” Fan later said, “and I was thinking, ‘We can make a decent showing.'” As he saw it, “decent” meant losing to human champions but nailing a few questions and ending up with a positive score.

Fan started by drawing up an inventory of the software tools and reference documents he thought he'd need for his machine. First would be a so-called type system. This would help the computer figure out if it was looking for a person, place, animal, or thing. After all, if it didn't know what it was looking for, finding an answer was little more than a crapshoot; generating enough “confidence” to bet on that answer would be impossible. The computer would be lost.

For humans, distinguishing President George Washington from the bridge named after him wasn't much of a challenge. Context made it clear. Bridges didn't deliver inaugural addresses; presidents were rarely jammed at rush hour, with half-hour delays from New Jersey. What's more, when placed in sentences, people usually behaved differently than roads or bridges.

But what was simple for us involved hard work for a Q-A computer. It had to comb through the structure of the question, picking out the subjects, objects, and prepositions. Then it had to consult exhaustive reference lists that had been built up in the industry over decades, laying out hundreds of thousands of places, things, and actions and the web of relationships among them. These were known as “ontologies.” Think of them as cheat sheets for computers. If a finger was a subject, for example, it fell into human anatomy and was related to the hand and the thumb and to verbs such as “to point” and “to pluck.” (Conversely, when “the finger” turned up as the object of the verb “to give,” a sophisticated ontology might steer the computer toward the neighborhood of insults, gestures, and obscenities.)