In fact, DARPA’s funding of natural language processing in the first decade of the twenty-first century did have one major success. DARPA funded wide-ranging artificial intelligence research under a program called Personalized Assistant That Learns, which sponsored work at SRI International. The military was not interested in the work, and the DARPA program was terminated, but SRI International spun off the technology as a company called Siri, which was eventually bought by Apple and incorporated into the iPhone. DARPA’s work on natural language processing was not necessarily going to help soldiers talk to Afghans, but it could help Americans find the closest Starbucks.
Zemach did not object to DARPA’s supporting research into something like the Phraselator; he just did not think it belonged on the battlefield. “When the war kicked off, DARPA got a lot of pressure to be relevant, which is a problem, because DARPA is not good at being relevant,” Zemach concluded.
That view reflected DARPA’s image in the early twenty-first century: a great science fiction agency, but not a place that the Pentagon turned to during wartime. It was a view at odds with the first two decades of DARPA’s existence and a reflection of how much had changed in the ensuing years. When it came to modern threats central to American forces fighting in wars, like the proliferation of roadside bombs, DARPA’s role was tangential or aimed at developments, like driverless cars, that would only help years in the future.
—
In April 2008, Vice President Richard Cheney, one of the chief architects of the Iraq War, feted DARPA at a black-tie dinner for sixteen hundred guests at the Washington Hilton to celebrate the agency’s fiftieth anniversary. At that point, Tether was the longest-serving director, thanks to the political support he had from Cheney and Rumsfeld. Cheney, in fact, gave the keynote speech at the dinner, praising the agency’s work in areas, such as drones, that had become a hallmark of the administration’s ongoing war on terror. “One thing we didn’t have a lot of in Desert Storm was the unmanned aerial vehicle,” Cheney said. “But thanks to DARPA, that technology was advancing rapidly in the early ’90s. And we’ve been able to use it all the time in both Afghanistan and Iraq—for reconnaissance, for remote sensing, and to strike the enemy.”
Outside the Washington Beltway, what had cemented DARPA’s reputation for innovation was not necessarily drones or stealth aircraft but the Internet. The agency’s most important creation had ensured DARPA’s place in history, even if it had emerged from a tiny effort four decades prior. Whether the DARPA of 2008 was capable of producing the types of innovation that had emerged in 1968—when Robert Taylor had published plans for the ARPANET—was not something that was widely debated. Introspection was not a characteristic of the modern DARPA. When Tony Tether wanted to kick off a celebration for the agency’s fiftieth anniversary, he commissioned a series of articles describing DARPA’s successes. The final product, published by a private company, was interspersed with paid advertisements from defense manufacturers. DARPA also hired a video production company to interview all of the living former directors for a brief promotional video as part of the anniversary celebration. The unedited interviews, which were only released after a Freedom of Information Act lawsuit, offered insights into how much the agency had changed over the past few decades.
When asked about DARPA’s role in the war on terror, which was in full swing in 2008, Tether seemed slightly stumped. “While we’re containing them, you do need an ability to go in and really start with the three- and four-year-olds to try to get into their minds to basically teach them, ‘Hey, it isn’t a bad thing. It isn’t a bad thing to deal with non-Muslims,’ ” he said. “And in this country, basically programs like Sesame Street really went a long ways toward integrating this country, toward our kids growing up, saying, and ‘Hey, blacks and whites and pinks and everybody—it’s okay! It’s okay to play with them!’ ”
In vastly simplified form, Tether had hit on the fundamental problem of the war on terror: there was no way to win the war with technology. DARPA’s sensors and drones could find and kill terrorists, but it was not stemming the growth of support for extremism. And yet DARPA could not seem to find any other course to pursue other than creating weapons and technology. DARPA in the post-9/11 era was praised for its gizmos and gadgets and portrayed in the press as a supercool science-fiction-inspired agency that was building universal translators, driverless cars, and brain-controlled prosthetics. But DARPA was no longer, as it had once been, the go-to place for national-level problems. That DARPA’s leaders had once advised presidents and defense chiefs, sent hundreds of employees to work in field offices in Thailand, Vietnam, Lebanon, and Iran, or helped armor the president’s vehicle in the wake of an assassination, was all but forgotten.
DARPA had its high points: its work on the monkey with the mind-controlled cursor grew into a program to develop a prosthetic controlled by the human mind. While a true neuroprosthetic is still many years away, DARPA-funded researchers by 2008 were at least demonstrating a very rudimentary arm that could be controlled by a person twitching a chest muscle. At that point, DARPA had spent more than $150 million, and the product was nowhere near a Luke Skywalker–inspired prosthetic. Still, given the thousands of amputees coming back from the wars in Iraq and Afghanistan, the DARPA program made sense.
More immediately successful was the Grand Challenge. In 2012, Google debuted its driverless car, based on the work by Sebastian Thrun, the Stanford professor who led the winning team in the 2005 competition. The unmanned car competition had done exactly what DARPA had hoped it would do: take a bold technical goal, and prove it was possible. If the Orteig Prize ushered in the modern era of transatlantic aviation, then the Grand Challenge can rightfully take credit for the dawn of autonomous cars. It was Tether’s greatest legacy, even if it did nothing for Iraq and Afghanistan.
The Disneyfication of DARPA was hard to reconcile with the realities of the modern battlefield. DARPA had reimagined itself as an agency of future war, producing technologies that might be used decades later. It was still successful, but it was no longer relevant to war, or at least not the type of war the United States was facing in Iraq and Afghanistan. In his final year of congressional testimony, Tether’s only direct mention of the agency’s current contributions to counterinsurgency was a reference to translation technology. The larger issues facing the United States were beyond the modern DARPA’s mandate. “The counter to this problem is a long, long war,” Tether said in the internally commissioned interview. “I’m not sure whose job this is, but, basically, what we’re trying to do is develop technology to contain them, to contain this threat, which is worldwide.”
He offered one additional thought: “We can’t go and kill them all, you know.”
CHAPTER 19
Return of Voldemort
The only tiki bar in eastern Afghanistan had an unusual payment program. A sign inside the establishment in Jalalabad read simply, “If you supply data, you will get beer.” The idea was that anyone—or any foreigner, because Afghans were not allowed—could upload data on a one-terabyte hard drive kept at the bar, located in the Taj Mahal Guest House. In exchange, they would get free beer courtesy of the Synergy Strike Force, the informal name of the group of American civilians who ran the establishment.
Patrons could contribute any sort of data—maps, PowerPoint slides, videos, or photographs. They could also copy data from the drive. The “Beer for Data” program, as the exchange was called, was about merging data from humanitarian workers, private security contractors, the military, and anyone else willing to contribute. The Synergy Strike Force was not a military unit, a government division, or even a private company; it was just the self-chosen name of the odd assortment of Westerners who worked—or in some cases volunteered—on the development projects run out of the hotel where the tiki bar was located.
The Synergy Strike Force’s Beer for Data exchange was a pure embodiment of the techno-utopian dream of free information and citizen empowerment that had emer
ged in recent years from the hacker community. Only no one would have guessed that this utopia was being created in the chaos of Afghanistan, let alone in Jalalabad, a city that had once been home to Osama bin Laden. Or even more unlikely, that the Synergy Strike Force would soon attract the attention of DARPA, which was reaching back in its history to resurrect elements of AGILE, its most ambitious, and controversial, wartime research effort.
DARPA’s interest in the potential of open-source information came at a critical juncture in Afghanistan. In January 2009, Barack Obama was sworn in as the forty-fourth president of the United States, the war in Afghanistan was in its eighth year, and a resurgent Taliban, whose regime had rapidly collapsed after the American invasion in 2001, was challenging the central government’s tenuous authority in the provinces. Counterinsurgency, the doctrine that had been promoted in Vietnam decades earlier, was back in vogue.
Nearly forty years after Vietnam, General David Petraeus and his followers resurrected elements of ARPA’s “glorious failure” as modernized counterinsurgency, or COIN, a “hearts and minds” campaign that emphasized providing security for the local population. The writing that had the “greatest influence on Petraeus’s thinking,” according to the journalist Fred Kaplan, was a counterinsurgency book by David Galula, the French officer sponsored by DARPA in the early 1960s under Project AGILE. Petraeus lifted Galula out of decades of obscurity, dusting off his writing and incorporating elements of it in a new counterinsurgency manual. “The objective is the population” became an often-cited sentence from Galula’s DARPA-funded study of Algeria. In Iraq, COIN was heralded as a success, elevating Petraeus and a new generation of “COINdinistas,” as the newly minted experts were called, to near-rock-star status. “By 2009 [counterinsurgency] was being celebrated as the answer to America’s mounting woes in Afghanistan as well,” wrote Greg Jaffe, a Washington Post reporter.
DARPA in 2009 also returned to counterinsurgency. That year, the agency launched an ambitious initiative in data mining, the very work that had created one of the agency’s most public debacles of the past decade. Only rather than rooting out terrorists in the United States, as Total Information Awareness had attempted, DARPA’s new focus was on Afghanistan. DARPA eventually brought two data-mining programs to Afghanistan: a highly secretive data analysis program meant to predict insurgent attacks based on “big data” science used by companies like Amazon to predict customers’ purchases; and a program based on the emerging science of social networks, attempting under the guise of humanitarian work to enlist an unwitting army of Afghan civilians to spy for the American military. And so DARPA’s first deployment to a war zone since Vietnam began with a group of well-intentioned hacktivists trading beer for data at Afghanistan’s only tiki bar.
—
In February 2009, the month after Obama took office, Tony Tether was ordered to resign to make way for a new DARPA director. Even though Tether had already been at DARPA longer than any other director, he recalled being devastated by the order to leave. Obama’s pick to head DARPA, announced in July 2009, was a public watershed for the agency: Regina Dugan became the agency’s first female director.
As a program manager at DARPA in the 1990s, Dugan had earned a reputation for boldness—and some alleged recklessness—visiting minefields and combat zones to test bomb detection technology. When she started making the rounds in the Washington Beltway as DARPA director, her choice of attire—short skirts, stiletto heels, and leather jackets—generated as much buzz as her credentials. Dugan’s financial ties to a family-owned firm receiving DARPA contracts for bomb detectors proved fodder for critics, though she insisted Pentagon lawyers signed off on her recusal from dealing with the company. A self-styled technology enthusiast, Dugan liked talking about theories of innovation, such as Pasteur’s quadrant, an approach that emphasizes finding an idealized type of research that combines scientific exploration with practical applications. Her lecture style was often better suited to the world of TED, the popular technology conference, than to old-school military briefing rooms. In a marked departure, Dugan canceled DARPATech, the Disney fest that had come to symbolize the Tether era. “There is a time and a place for daydreaming. But it is not at DARPA,” she told Congress. “DARPA is not the place of dreamlike musings or fantasies, not a place for self-indulging in wishes and hopes.”
One of Dugan’s first moves was to hire Peter Lee, a prominent university computer scientist, to run a key DARPA office. Lee, the head of Carnegie Mellon’s computer science department, was reluctant at first, even though he and Dugan were already friends. The Information Processing Techniques Office, which paved the way for the ARPANET, had long since moved away from supporting basic computer science research, focusing instead on more traditional weapons technology like ARGUS-IS, a 1.8-billion-pixel camera that could be flown on a drone and used to keep watch on entire cities. As a member of the Computing Community Consortium, Lee had co-authored a paper titled “Re-envisioning DARPA” that forwarded ideas for how to bring the agency back to its golden age roots of the ARPANET and computer networking. Dugan “started to lean on me” to take the job as head of DARPA’s Information Processing Techniques Office, Lee recalled. He agreed, but a month before he was supposed to take over, Lee had dinner with Dugan to talk about DARPA. “You know, Peter. I don’t think you should take over IPTO,” she said, just as he was dropping her off after dinner. “You should just start a new office.” Dugan did not say what the new office would do, other than it should be a “pure expression of what DARPA could be.” Lee had no idea what that meant.
Then, the week before Lee was supposed to start at DARPA, Dugan called him on the phone and asked him to come to Washington, D.C., immediately. She wanted him to meet with Secretary of Defense Robert Gates, who was scheduled to visit DARPA’s headquarters. As he drove from Pittsburgh toward Washington, Lee grew nervous. He was an ivory tower academic who had written a paper telling DARPA what he thought it should do, but now that he was actually about to become a senior DARPA official, he realized he had no plan of his own. His apprehension grew worse when he was pulled over for speeding. He had visions of multiple tickets and eventually having his license suspended. “Oh my God, I’ll be driving back and forth to Pittsburgh and D.C., and I won’t have my license,” he thought.
The speeding incident, however, provided Lee with inspiration for the new office. He had recently learned about Trapster, a smart-phone app that allows users to map and share information on speed traps using GPS. Trapster enabled a virtual army of tipsters to create a real-time map warning drivers of areas where the police may be lying in wait. For Lee, who was interested in social networking technologies, and in avoiding speed traps, Trapster finally provided him with an idea that he thought might interest the Pentagon chief. Instead of plotting speed traps, he imagined a Trapster-like application that could track potential bomb attacks in Afghanistan. Crowdsourced data was allowing millions of people to monitor events in real time. Already, communities of people were collaborating online to track nuclear proliferation, spotting potential test sites in North Korea. Humanitarians were using crowdsourcing to monitor elections and respond to natural disasters. If crowdsourcing could plot speed traps and spot election fraud, perhaps it could be used in war zones. When Lee presented the idea, the defense secretary seemed to like it. So did Dugan, who encouraged Lee to pursue it. That was the beginning of Lee’s Transformation Convergence Technology Office, a name so incredibly awkward it seemed tailor made for hiding secret programs.
—
Dugan greeted Lee’s idea with enthusiasm for a specific reason. Crowdsourcing was part of the booming field of “big data,” an area that DARPA had avoided for the better part of a decade following the public brouhaha over John Poindexter’s Total Information Awareness. Poindexter’s work had become the agency’s Voldemort, the “program that shall not be named,” as Dugan put it. “It was an area that the agency wasn’t doing a lot of work in. It didn’t make any sense,” she said.
“This field was explosive.”
Big data was indeed exploding, not just in the private sector, where it was being used to predict consumers’ movie rentals and book purchases, but also with the military, which was parsing reams of data being sucked up by sensors monitoring Iraq and Afghanistan. One classified program, described opaquely by Bob Woodward in his book The War Within, was used to comb through data to “locate, target, and kill key individuals” in Iraq. Woodward touted the technology in interviews as “very top secret” and “one of the true breakthroughs” of the war in Iraq. He later revealed in a subsequent book more details, including the name, the Real Time Regional Gateway, an NSA computer program better known by its acronym, RTRG. It was designed to pull together many feeds of information—everything from intercepted phone calls to information on bomb attacks—and analyze that data to identify insurgent networks and predict attacks.
Once Woodward went public with RTRG, senior intelligence officials began to drop more details about this top secret program. According to the retired air force colonel Pedro “Pete” Rustan, who helped develop the technology, RTRG started as an intelligence program in Iraq that tracked insurgents by intercepting phone calls and triangulating the location of insurgents. The system worked by collecting and analyzing streams of data in real time and then pinpointing insurgents by location. “If you’re smart enough to combine all that data in real time, you can determine where Dick is out there,” said Rustan, then at the National Reconnaissance Office. “He’s in block 23 down there, and he just said he’s going to place a bomb.”
The Imagineers of War Page 40