Schatz listens patiently as I repackage Raghavan’s concerns in a question. Are there times, I ask, when you just have too much data? When it gets in the way and confuses things? He seems taken aback by this line of questioning. “More data is always better,” he says.
He explains that some organizations have trouble managing loads of data. Some ask the wrong questions. Of course, he won’t tell me about the nature of data that his team is sifting through across the street. He’ll only say that the “statisticians are enjoying a field day” and that “the information age has given math a whole new life.” But clearly, while politicians and civil libertarians wrangle over how many of our personal details should be thrown into this national security analysis, problem-solving mathematicians at the NSA are always happy to get more.
“The information age that we’re in is going to be an entirely new era of what would be called applied mathematics,” Schatz says. His Numerati use every statistical and mathematical tool in their arsenal—“topology, abstract algebra, differential equations, number theory”—to piece together networks, predict migrations, analyze voices, and match photographed faces to others in a database. Schatz says the agency has seen “an explosion of mathematics into new areas.” He describes multidisciplinary teams, where numbers people work closely with engineers and computer scientists. The “customers” for these teams are the intelligence agents—often liberal arts grads, he says (as I nod approvingly). Presumably, they have the on-the-ground knowledge to tweak the data miners’ algorithms and steer investigators toward known hotbeds of terror and menacing clans. If the agents have leads or, better, concrete intelligence, they can turn a foraging expedition into a more targeted hunt.
This small cryptology museum is a monument to the code-breaking heritage of the NSA. Governments and armies throughout history have relied on their cleverest geeks to devise secret codes to protect their vital messages. They’ve also counted on them to hack the secrets coming from the other side. In one of the glass cases here is Nazi Germany’s famous Enigma machine, whose code was broken by ingenious British mathematicians. This was a key to victory in World War II. With the founding of the NSA, in 1952, the U.S. government built code breaking into an entire bureaucracy. It quickly grew into the largest math shop in the world (which it remains to this day, though the agency never divulges the numbers). These code breakers battled on a key front in the Cold War. While the CIA operatives met secretly with sources, from safe houses in Berlin and Moscow to thatched huts along the Mekong Delta, their counterparts at the NSA led quieter lives. They commuted to offices, first in Washington, then later to these glass cubes in suburban Maryland. Their job, quite simply, was to match mathematical wits with their counterparts in the Soviet Union.
When James Schatz signed on at the NSA in 1979, after getting his doctorate in math from Syracuse University, he stepped right into this tradition. For his first 15 years, he tells me, he worked on cryptological math, diving into some of the deepest and as yet unsolved conundrums in mathematics, and fashioning them into a sort of numeric armor to wrap around secret communications. To pierce this armor, the other side had to master some extremely heady math. Throughout the Cold War, cryptology was at the center of a mathematical arms race.
Back then, the NSA team didn’t need to delve into the human psyche. Their more sociable colleagues, the spies and the diplomats, handled that murky domain. However, by the time Schatz was promoted to head the math department, in 1994, changes were afoot. The Berlin Wall was in pieces, and the United States’ new enemies, whether warlords, terrorists, or international money launderers, were scattered all over the world. The challenge at the NSA was less to crack their communications than to find them. How were they organized? Where did they get their financing? What were their plans? This information didn’t travel highly encrypted on secure networks. Much of it was mingling freely with the rest of the globe’s chatter. Just like the rest of us, many of these villains were migrating the details of their lives and their missions onto mobile phones and the Internet. They were camouflaged by networked humanity. This meant that the mathematicians at the NSA faced a new and growing challenge. Many had to shift their focus from pure math to the hurly-burly bundles of words and pictures and smiley faces and mouse clicks that were pouring through the networks. Somewhere in that growing mass of unstructured data, they had to find the bad guys and piece together their networks. “Look at the whole telecommunications industry, all the information that’s flying around on the Internet,” Schatz says. “How are we actually going to tap all that for the good of mankind?” To carry out their central mission—protecting us—the mathematicians at the NSA, like the Numerati elsewhere, had to figure out humans.
THE YEAR WAS 2002. NATO troops had stormed through Afghanistan. The U.S. government was threatening to attack Iraq. And Jeff Jonas, like many others, was still obsessed with the attacks that triggered these wars. Jonas, a software entrepreneur in Las Vegas, couldn’t stop thinking about 9/11. Given the information that the government possessed during the weeks and months before the tragedy, could an extremely smart data sleuth with the right tools possibly have unraveled the brewing plot and foiled it? Jonas was no expert on international terrorism, or on Islamist jihad. At this point in his life, he had never even traveled outside the United States. But he was a leading expert on finding people who want to stay hidden. He thought his approach was worth considering.
Jonas, who’s now a chief scientist at IBM, is telling me this story over Chinese food at a strip mall near the Las Vegas airport. He’s a catlike figure dressed in black. He leans across the table toward me as he talks, his neatly trimmed goatee hovering over my fried fish. Following the attacks, he says, he pored over public records, from newspaper articles to grand jury testimony. He was looking for paths that could (and should) have drawn investigators to the bombers. He found that two of the terrorists, Nawaf Alhazmi and Khalid Almihdhar, had been placed on the State Department Watch List two weeks after President Bush got word about planned Al Qaeda attacks. It seems easy, with hindsight, to say that investigators should have tracked them down. But Jonas notes that these two men were linked to previous attacks on the USS Cole and the U.S. Embassy in Nairobi. They were already targets of the highest priority. “We’re not talking about people on illegal visas,” he says. “That number is in the millions. We’re talking about known terrorist killers in the United States. That’s a small list.”
If investigators had looked for them, Jonas discovered, they could have found them in the San Diego phone book. Days after landing on the watch list, the two men reserved plane tickets in their own names. Even without knowing that those aircraft would turn into bombs, investigators should have seen those names pop up. “These guys were hiding in plain sight,” Jonas says. He walks through the evidence, step by step. Roommates shared phone numbers and other connections linking them to the other participants in the plot. It’s true that investigators, given these details, would still have remained clueless about what this network was ultimately up to. They would have seen only that a group of people with links to past acts of terror were busy renting hotel rooms, making phone calls, and buying plane tickets. The arrests would be based on the people’s records and contacts—not what they were planning to do. But detaining them may have foiled their plot. The subtext of Jonas’s argument, of course, is that investigators could have located these killers by making better use of the data and the tools they had at hand.
Why have I flown all the way out to Las Vegas to spend time with Jeff Jonas? I want to find out how a society can monitor itself yet remain free and uninhibited, even sinful, by using the tools of the Numerati. Jonas is the ideal guide for this. He’s vehemently opposed to the use of statistical data mining to predict the next terrorist attacks. He fears its intrusions and false alerts. Yet he trusts that data and surveillance can protect our freedoms without sacrificing our privacy. His method isn’t so different from that of an old-fashioned detective. It starts with
a lead: a suspect, a door to knock on, a sign of suspicious behavior. And it follows the trails of data from there. This is what I call the gumshoe approach, a focused alternative to predictive data mining. Jonas told me months ago, over lunch in New York, that Las Vegas was a perfect test case for gumshoes. That’s what I’ve come to see.
Jonas has built his business and his fortune on following threads of data. He began developing his targeted approach by tracking a group of aquatic killers and their victims. It was early 1995 when the young software whiz first arrived to work at the Mirage Hotel in Las Vegas. The fish swimming in the massive aquarium at the Mirage were worth $1 million. But there was a problem. Fancy fish were disappearing—presumably into the mouths of their tank mates. Jonas’s job was to build a tracking system for fish so that the casino could calculate the survival rate for each type—and avoid investing in Darwinian losers. Jonas, a supremely social animal, learned a lot about the casino business while acquainting himself with the movements of tropical fish. Business was booming, he heard, and this was creating vulnerabilities. As thousands of people streamed through their doors, the casinos were finding it harder to keep their eye out for thieves and cheaters. They needed a system more advanced than the one Jonas was building for the aquarium. In the hunt for humans, they would be looking for specific predators. For decades, the casino had entrusted this work to people. But things were getting out of hand. The numbers were too big.
So Jonas built software to help casinos pinpoint known cheaters, grifters, and goons—what casino execs refer to as “subjects of interest.” Called NORA, which stands for non-obvious relationship awareness, it specialized in rifling through many different pools of internal casino data, from personnel files to credit applications, looking for common threads. NORA might see, for example, that Krista, who was on the suspect list, had the same home phone number as Tammy, who had just applied for a job as a blackjack dealer. Were they partners in crime? NORA highlighted the correlations. Then it was up to humans to dig for the answers. But piecing together these connections from a sea of data was invaluable. NORA, quite simply, helped straighten out who was who.
NORA, unlike other data-combing setups, doesn’t look just backward in time. It also reaches into the future. Let’s say a casino is on the lookout for the leader of a gang who uses the Internet to put together teams of crooks. (This is a growing problem.) They’ve dug around and have a few facts on him—say, an alias, two phone numbers, and an address. In a traditional backward-looking data hunt, investigators comb through all of their records looking for signs of him. Nothing turns up? Thank goodness . . . But what if the very next day a friendly, big-tipping tourist checks in at the casino. The phone number he scrawls on the registry form is one of the two from the list. In a traditional system, the casino won’t spot him until they hunt again. But with NORA, each new piece of data—each phone number, each name and address—creates a new query. It asks the system, “Hey, anything fishy about this one?” That’s how NORA extends forward in time. It’s constantly at work, prospecting into the future, assembling the bits of evidence as they arrive.
After Jonas unveiled NORA, it was just a matter of time before other companies and government agencies began knocking at his door. The challenges of finding identities and tracing connections within vast databases were hardly unique to Las Vegas. Anybody interested in sorting through data to find and profile shoppers or patients or voters or workers or lovers . . . in short, entire ranks of the Numerati desperately needed NORA, or something like it. Those hungriest for NORA were wrestling with the biggest and messiest sets of data in the world as they searched for the identities and movements of terrorists. In January 2001, In-Q-Tel, the venture-financing arm of the Central Intelligence Agency, bought a share of Jonas’s company, Systems Research and Development (SRD). And following the attacks of 9/11, NORA was enlisted in the war against terrorism. Four years later, IBM bought SRD for undisclosed millions. This made Jonas a rich man and turned the entrepreneur into a distinguished engineer and chief scientist at Big Blue. His start-up morphed into IBM’s Entity Analytics group. In his new role, Jonas has a lot to say about the use of technology for national security. He sits on panels, testifies at inquiries convened by the president, and is a leader in IBM’s efforts on the defensive front of this new war.
Jonas tells me that he was a beach bum and guitar player as a kid. But in the tenth grade he signed up for a computer class and then took another one. When he ran out of computer courses, he told himself, “I’m out of here.” He passed the graduate equivalency exam and dropped out. Within two years, Jonas was running a booming software business, Preferred Programming Services. But he knew a lot more about writing software than running a business. The debt got out of hand, and he went bankrupt. By the time he was 20, Jonas was sleeping in his car.
He clawed his way back into the software business. Even before finding a place to live, he started up his next company, SRD. The business took off. But at age 24, Jonas hit another big turning point. While checking out a new BMW, he had a salesman take him for a ride—and the salesman drove the car off the road. Jonas broke his neck. For a short while, he was paralyzed. After he regained use of his limbs, he worked his way through a long rehab and back to fitness. But the center of his spinal cord, he says, is dead. And to this day he has only slight feeling on the right side of his body. “I can feel the difference between the point of a pencil and the eraser,” he says, poking his fingers with a fork. He grips a glass of ice water. “But I can’t tell the difference between heat and cold. And I don’t feel pain very well.” Ever since the accident, Jonas has been a perpetual motion machine (with a sky-high threshold for pain). He schedules his business meetings and conferences in places like Singapore, Brazil, and New Zealand to coincide with Ironman Triathlons. One day he’s telling a roomful of executives how to locate people with software, and the next day he’s swimming across a bay, biking up and down a mountain, and running for as long as 14 hours straight. He says he thrives on such episodes. Sometimes, he says, he peels off his shoe and sees that he’s ripped off a toenail.
“I think I can get you into a crow’s nest,” Jonas tells me one afternoon. He’s referring to the surveillance room high above the casino floor. He’s eager for me to look at the world through the security managers’ eyes, not so much for all the details they can see from up there, but for what they ignore. This is the key to all kinds of surveillance, he says: what to focus on. It’s crucial in Las Vegas. People come here to do all kinds of things they don’t dare try back home. They crave the freedom to blow money, drink way too much, and follow virtually every animal impulse that stirs them, from staring down a bartender’s dress to arranging a ménage à trois—all off the record. In short, they want to sin anonymously, which is another way of saying that they’re looking for freedom and privacy. For this, oddly enough, they come into a dark world bristling with cameras. It may be a visit to the future, in which cameras and other sensors surround us and protect us from harm. We can only pray that the powers that be will stick to that mandate and respect our secrets. Jeff Jonas argues that casinos have come as close as anyone to mastering this balancing act.
“THERE IT IS.”
“What?”
“See what she did with her hand?”
“Play it again.”
Thanks to Jonas’s connections, I’m up in the crow’s nest of a major casino in Las Vegas. It’s dark. Most of the light comes from the dozens of TV monitors blinking from the wall. Four of us are gathered around one of these TVs. We’re looking at a young woman who’s drinking and joking with her friends and having the best of times—while she cheats at the blackjack table. To be fair, we’re studying every hand she has played, and we see her cheat only once. It happens quickly. The cards are dealt. Her bet is placed. She has a blackjack, a winning hand. And with a lightning gesture—which I find especially impressive, considering how heavily she’s drinking—she adds another $5 chip to her bet. That’s illegal, a gamin
g violation, as they call it in Las Vegas.
Granted, she hasn’t sneaked loaded dice onto the craps table or filed down the ball bearings in the roulette wheel. But her move is a crime, and I’m privy to the discussion in this digital crow’s nest about how justice is to be administered. This is life at the heart of a surveillance society, I think, as I look around. It’s an impressive view. The casinos all have fixed cameras staring down at every table. Tracking cameras cover nearly every inch of floor space. All of these send video to the banks of TV screens in front of us. When I arrived, we testdrove these tools by following one customer, a man who had just walked into the casino. He was carrying a backpack over one shoulder. His eyes were probably still adjusting from the desert glare outside to the dusky shades of the gambling den. The surveillance team tracked his movements, switching from camera to camera. They exchanged terse coordinates, like pilots on a bombing run. We watched from screen to screen as the man made his way through the banks of slot machines, past the craps tables and the bar, and finally to the hotel’s front desk. I was hoping, for his sake, that he wouldn’t scratch himself or stick an idle finger into his ear along the way. He was performing for a crowd.
A few minutes later, the surveillance team got word from the floor to study the behavior of a woman at a blackjack table. She’s the one we’re watching now. She’s in her twenties, I’d say. She’s playing with two others, both men. They’re smiling and joking. She’s wearing a low-cut blouse with spaghetti straps. She takes sips from her drink, which she holds with her left hand, and she occasionally reaches up to fix a strap that keeps falling from her shoulder. On one monitor we watch the hands she’s playing. She has two cards totaling 14 and asks for another. She gets a king. Whoops. That puts her over 21, which means she’s lost the hand. It hardly seems to ruin her fun. Meanwhile, at another monitor, someone else on the team has rewound the tape. (Yes, it’s a VCR, which seems surprisingly primitive to me.) He’s watching every hand she has played. He’s the one who sees her cheat. “Got it,” he says, and promptly cues it up for us. We watch the illegal move again and again in slow motion. Isn’t the dealer looking right at her? Didn’t he see it? Did the others do something to distract him? It’s hard to tell.
The Numerati Page 13
