Final Jeopardy (6 page)

Still, Ferrucci believed his team had a fighting chance, though he wasn't quite ready to commit. He code-named the project Blue J—Blue for Big Blue, J for
Jeopardy
—and right before the holidays, in late 2006, he asked Horn to give him six months to see if it was possible.

ON A LATE SUMMER
day in 2004, a twenty-nine-year-old software engineer named Ken Jennings placed a mammoth $12,000 bet on a Daily Double in
Jeopardy
. The category was Literary Pairs. Jennings, who by that point had won a record fifty straight games, was initially flummoxed by the clue: “The film title ‘Eternal Sunshine of the Spotless Mind' comes from a poem about these ill-fated medieval lovers.” As the seconds passed, Jennings flipped through every literary medieval couple he could conjure up—Romeo and Juliet, Dante and Beatrice, Petrarch and Laura—but he found reasons to disqualify each one. Time was running out. A difference of $24,000 was at stake, enough for a new car. Jennings quickly reviewed the clue. On his second reading, something about the wording suggested to him that the medieval lovers were historical figures, not literary characters or their creators. He said he couldn't put his finger on it, but it had “a flavor of history.” At that point, the names of the French philosopher Peter Abelard and his student and lover, Heloise, popped into Jennings's mind. It was the answer. He just knew it. It was their correspondence, the hundreds of letters they exchanged after their tragic separation, that qualified them as a literary pair. He pronounced the names as time ran out, pocketed the $12,000, and moved on to the next clue.

In answering that single clue, Jennings displayed several peerless qualities of the human mind, ones that IBM's computer engineers would be hard-pressed to instill in a machine. First, he immediately understood the complex clue. Unlike even the most sophisticated computers, he was a master of human language. Far beyond basic comprehension, he picked up nuance in the wording so very subtle that even he failed to decode it. Yet it pushed him toward the answer. Once Abelard and Heloise surfaced, more human magic kicked in: He knew he was right. While a
Jeopardy
computer would no doubt weigh thousands, even millions, of possibilities, humans could look at a mere handful and often pick the right one with utter confidence. Humans just know things. And good
Jeopardy
players often sense that they'll find the answer, even before it comes to mind. “It's an odd feeling,” Jennings wrote in his 2005 memoir,
Brainiac.
“The answer's not on the tip of your tongue yet, but a light flashes in the recesses of your brain. A connection has been made, and you find your thumb pressing the buzzer while the brain races to catch up.”

Perhaps the greatest advantage humans would enjoy over a
Jeopardy
machine was kinship with the fellow humans who had written the clues. With each clue, participants attempt to read the mind of the writers. What response could they be looking for? In an easy $200 category, would the writers expect players to recognize a Caribbean nation as small as Santa Lucia? With that offhand reference to “candid,” could they be pointing toward Voltaire's
Candide
? Would they ever stack the European Capitals category with two clues featuring Dublin? When playing
Jeopardy,
Jennings said, “You're not just parsing the question, you're getting into the head of the writer.” In this psychological aspect of the game, a computer would be out of its league.

Computers, of course, can rummage through mountains of data millions of times faster than humans. But humans compensate with mental shortcuts, many of them honed over millions of years of evolution. Instead of plowing through copious evidence, humans instinctively read signals and draw quick conclusions, whether they involve trusting a stranger or deciding where to pitch a tent. “Mortals cannot know the world, but must rely on uncertain inferences, on bets rather than demonstrable proof,” wrote the German psychologist Gert Gigerenzer. In recent decades, psychologists have unearthed dozens of these rules, known as heuristics. Many of them would guide humans in a
Jeopardy
match against a much faster computer.

The most elementary heuristic is based on favoring the first answer to pop into the brain. That one automatically starts in the front of the line; it is more trusted simply by virtue of arriving early. Which ideas pop in first? Following another heuristic, they're often the answers contestants are most familiar with. Given a choice between a well-known place or person or an obscure one, studies show that people opt for what they know. “If you ask people, ‘Which of these two cities has a larger population,' they'll almost always choose the more familiar one,” said Richard Carlson, a professor of cognitive psychology at Penn State. Usually this works. If a
Jeopardy
player has to name the most populous cities in a certain country, the most famous ones—London, Tokyo, Berlin, New York—often fit the bill. This approach can lead to bloopers, of course. But it happens less often in
Jeopardy
than in the outside world. Why? Again, the writers, being human, work from the same rules of thumb, and they're eager to connect with contestants and with the nine million people watching on TV. They want the contestants to succeed and to look smart, and they want people at home to feel smart, too. That's critical to
Jeopardy
's popularity. “You can't forget that it's a TV show,” said Roger Craig, a six-time
Jeopardy
champion. “They're writing for the person in the living room.” And that viewer, like Ken Jennings—and unlike a computer—races along well-worn mental paths to answer questions. These paths are marked with signs and signals that call out to the human brain and help it navigate.

A century ago, the psychologist William James divided human thought into two types, associative and true reasoning. For James, associative thinking worked from historical patterns and rules in the mind. True reasoning, which was necessary for unprecedented problems, demanded deeper analysis. This came to be known as the “dual process” theory. Late in the twentieth century, Daniel Kahneman of Princeton redefined these cognitive processes as System 1 and System 2. The intuitive System 1 appeared to represent a primitive part of the mind, perhaps dating from before the cognitive leap undertaken by our tool-making Cro-Magnon ancestors forty thousand years ago. Its embedded rules, with their biases toward the familiar, steered people toward their most basic goals: survival and reproduction. System 2, which appeared to arrive later, involved conscious and deliberate analysis and was far slower. When it came to intelligence, all humans were more or less on an equal footing in the ancient and intuitive System 1. The rules were easy, and whether they made sense or not, everyone knew them. It was in the slower realm of reasoning, System 2, that intelligent people distinguished themselves from the crowd.

Still, great
Jeopardy
players like Ken Jennings cannot afford to ignore the signals coming from the caveman quarters of their minds. They need speed, and the easy answers pouring in through System I are often correct. But they have to know when to distrust this reflexive thought, when to pursue a longer and more analytical route. In the same game in which Jennings tracked down Abelard and Heloise, this clue popped up in the Tricky Questions category: “Total number of each animal that Moses took on the ark with him during the great flood.” Jennings lost the buzz to Matt Kleinmaier, a medical student from Chicago, who answered, “What is two?” It was wrong. Jennings, aware that it was supposed to be tricky, noticed that it asked for “each animal” instead of “each species.” He buzzed for a second chance at the clue and answered, “What is one?” That was wrong, too. The correct answer, which no one came up with, was “What is zero?”

Jennings and Kleinmaier had fallen for a trick. Each had focused on the gist of the clue—the number of animals boarding the biblical ark—while ignoring one detail: The ark builder was Noah, not Moses. This clue actually came from a decades-old psychological experiment, one that has given a name—the Moses Illusion—to the careless thinking that most humans employ.

It's easy enough to understand. The brain groups information into clusters. (Unlike computers, it doesn't move packets of encoded data this way and that. The data stay put and link up through neural connections.) People tend to notice when one piece of information doesn't jibe with its expected group. It's an anomaly. But Noah and Moses cohabit numerous clusters. Thematically they are both in the Bible, visually, both wear beards. Phonetically, their names almost rhyme. A question about Ezekiel herding animals into the ark might not pass so smoothly. According to a study headed by Lynn Reder, a psychologist at Carnegie Mellon, the Moses Illusion illustrates a facet of human intelligence, one vital for
Jeopardy
.

Most of what humans experience as perception is actually furnished by the memory. This is because the conscious brain can only process a trickle of data. Psychologists agree that only one to four “items,” either thoughts or sensations, can be held in mind, immediately available to consciousness, at the same time. Some have tried to quantify these constraints. According to the work of Manfred Zimmerman of Germany's Heidelberg University, only a woeful fifty bits of information per second make their way into the conscious brain, while an estimated eleven million bits of data flow from the senses every second. Many psychologists object to these attempts to measure thoughts and perceptions as digital bits. But however they're measured, the stark limits of the mind are clear. It's as if each person's senses generated enough data to run a 3D Omnimax movie with Dolby sound—only to funnel it through an antediluvian modem, one better suited to Morse code. So how do humans re-create the Omnimax experience? They focus on the items that appear most relevant and round them out with stored memories, what psychologists call “schemas.”

In the Moses example, people concentrate on the question about animals. The biblical details, which appear to fit into their expected clusters, are ignored. It's only when a wrong name intrudes from outside the expected orbit that alarms go off. In one experiment at Carnegie Mellon, when researchers substituted a former U.S. president for Moses, people noticed right away. Nixon had nothing to do with the ark, they said.

Even after falling victim to the Moses Illusion, Jennings found no fault in his own thinking. “The brain's doing the right thing!” he said. “It's focusing on the right part of the question: How many animals did the biblical figure take onto the ark?” That, he said, is how the brain
should
work. “It's just that the question writer has found a way to work against you.” Those sorts of tricks, he added, are uncommon on
Jeopardy
.

Strangely enough, the cerebral carelessness that leads to the Moses Illusion also serves a useful function for human thought. Filtering out details not only eliminates time-consuming busy work. It also allows people to overlook many variations and to generalize. This is important. If they focus too much on small changes, they might think, for example, that each time a friend gets a haircut or a suntan, she's a different person. Instead, the brain settles on the gist of the person and is ready to look past some details—or, in many cases, to ignore them. This can be embarrassing. (Sometimes it
is
a different person.) Still, by skipping over details, the brain is carrying out a process that is central to human intelligence and one that confounds computers. It's thinking more broadly and focusing on concepts.

The
Jeopardy
studio sits on the sun-drenched Sony lot in Culver City. Seven miles south of Hollywood's Sunset and Vine, this was a suburban hinterland when Metro-Goldwyn-Mayer (MGM) started making movies there in 1915. In later decades it turned out such classics as
The Wizard of Oz
and
Ben Hur
—all of them introduced by the iconic roaring lion. Following years of mergers and acquisitions, the lot became the property of a Japanese industrial giant—a development that likely would have shocked Samuel Goldwyn. Sony later gobbled up Columbia Studios, which had belonged to Coca-Cola for a few years in the eighties. On the Sony lot, the MGM lion gave way to Lady Liberty holding her torch. In the summer of 2007, as IBM considered a
Jeopardy
project, tourists on the Sony lot were filing past the sets of
Spiderman II
and Will Smith's
Happyness
. Others with free passes lined up for
Jeopardy
. If they made their way past the fake Main Street, with its cinema, souvenir shop, and café, they would come across a low-slung office building named for Robert Young, the actor who played the homespun 1970s doctor Marcus Welby, M.D.

This is where Harry Friedman worked. Friedman, then in his late fifties, was the executive producer of both
Wheel of Fortune
and
Jeopardy,
the top- and second-ranked game shows in America.
Wheel,
as it was known, relied on the chance of a spinning wheel and required only the most rudimentary knowledge of common phrases and titles. Its host was a former TV weatherman named Pat Sajak, who had been accompanied since 1983 by the lovely Vanna White. She had showcased more than four thousand dresses through the years while turning the letters on the big board and leading the clapping while the roulette wheel spun. For some
Jeopardy
fans, even mentioning the two games in the same breath was an outrage. It would be like card players comparing the endlessly complex game of Bridge to Go Fish. Nevertheless,
Wheel
attracted some eleven million viewers every weeknight evening, and about nine million tuned in to
Jeopardy
. Harry Friedman's job, while touching on the world of knowledge and facts, was to keep those millions of people watching his two hit shows. In a media world exploding with new choices, it was a challenge.