The Imagineers of War

by Sharon Weinberger

  The danger facing the agency today is irrelevance to national security. In 2003, when the American military in Iraq found that its greatest threat was not from tanks and missiles but from roadside bombs, the Pentagon did not, as it once did, turn to DARPA, an agency stacked with top-notch technical personnel and decades of experience in bomb detection. Instead, it created an entirely new organization that was largely bereft of the type of science and technology expertise long resident in DARPA. What followed is hardly surprising: billions of dollars were spent, and yet casualties from bombs continued to increase.

  Today, the agency’s past investments populate the battlefield: The Predator, the descendant of Amber, has enabled the United States to conduct push-button warfare from afar, killing enemies from the comfort of air-conditioned trailers in the United States. Stealth aircraft, another DARPA innovation, are used to slip across borders to conduct precision strikes and covert operations. Networked computers have shortened the “kill chain” to just seconds, and precision weapons allow the United States to conduct strikes anywhere, even in heavily populated urban areas.

  The question, however, is whether those novelties have successfully achieved what DARPA at one time intended: to create technologies that ensure the United States would not have to go to war and, if it did go to war, that it would achieve a swift victory. Nothing illustrates that disparity better perhaps than the 1960s-era investment in counterinsurgency research, intended to prevent large-scale conventional engagements in so-called limited wars. By 2006, counterinsurgency theory was resurrected as a tool to help conventional forces wage war against insurgents, the exact thing the original DARPA program was meant to prevent. The allure of applying the wizardry of science and technology to warfare seems only to have made the temptation to engage in armed conflict more inviting and to have entangled the United States in a “forever war.”

  —

  If one glances only at the headlines, any concerns about DARPA appear at face value unwarranted, and the agency would seem to be at its peak. Its projects, like the driverless car, are splashed across magazine covers and tech websites, which run breathless stories about the agency’s plans for brain implants to cure mental illness. Lawmakers over the past decade have rarely done more in congressional hearings than laud the agency. Republicans and Democrats praise DARPA, calling it a model for government-sponsored innovation. Politicians, economists, and techies regularly exalt the “DARPA model,” even though it is unclear what the model is.

  Former directors do worry about DARPA and its bureaucratic growth. When DARPA was born in 1958, it had no building. Its senior officials were allotted a few offices in the Pentagon, and members of its technical staff were given windowless offices in the interior rings of the building. For the first few years, the staff directory fit on a standard index card. Today, the directory approaches the size of a small phone book. DARPA may tout a technical staff of only 140 scientists, but they are aided by an army of contract personnel, many who serve as the near equivalent of permanent employees, the very thing it is supposed to avoid. Even its new headquarters seems to run counter to the agency’s once ad hoc, minimalist existence.

  Victor Reis, a former DARPA director, cited one of Parkinson’s laws, not the one on work expanding to fit available time, but the one on the correlation between the decline of organizations and the construction of a “perfect” headquarters. The Pentagon was not completed until World War II was almost over. St. Peter’s Basilica in Vatican City took more than a hundred years to build and was finished well after the papacy’s influence was waning. “By the time you get the real building done, you’re finished as an organization,” Reis said.

  More than the building, Reis expressed concerns about DARPA’s ceaseless touting of its accomplishments. “I’m just a little uneasy that they’re starting to say how wonderful they are, you know what I mean?” Reis told me, when I met with him in his modest government office in downtown Washington, D.C.

  Yet DARPA’s reputation is now so entrenched that the government in recent years has made a series of flawed attempts to “replicate” the DARPA model in other government agencies. The Department of Homeland Security bungled its own version to the point that it exists in name only. The Department of Energy’s ARPA, or ARPA-E, has a budget merely a fraction of what DARPA receives. The intelligence ARPA, called IARPA, has been constrained by bureaucracy. None of them have approached the scope or ambition of their namesake.

  The attempts to replicate DARPA belie the temptation to draw some fantastical lessons about management science. Should organizations get rid of all their employees every three to five years, as DARPA does? Should science agencies do away with peer review, as DARPA often does, in order to pursue revolutionary ideas? In a culture that reduces analysis to bullet points, celebrates the intellectual reductionism of TED talks, and prays at the altar of PowerPoint, it is important to remember that not everything can be reduced to an organizational chart. DARPA should not be treated as a black box management tool that can be dropped on top of any organization to make it more innovative.

  It is tempting to reduce DARPA to a caricature. The truth is that DARPA’s legacy cannot be easily packaged as “innovation in a box.” Its successes—and failures—have always been a function of its unique bureaucratic form, which arose from its historical role as a problem-solving agency for national security. Rearranging boxes on an org chart, or cubicles in an office, will not produce another ARPANET. With the exception of having technical staff managing research, and a director, the agency has never had a fixed organizational structure.

  In fact, DARPA’s style often runs counter to fuzzy management theories of collaboration. So-called kumbaya moments at DARPA are few and far between. With some notable exceptions, the program managers often know little of what their colleagues in other offices are doing. In one case, a program manager described to me being shocked to meet a stranger on a Pentagon shuttle bus, only to learn that both men had worked for several years in DARPA. In a cast of thousands, this would not be surprising, but in a relatively small agency it is more unusual. DARPA, as one former director called it, is “140 program managers all bound together by a common travel agent.”

  For every management lesson that DARPA might hold, there is a counter lesson: the agency pioneered a novel system for detecting nuclear tests that simultaneously modernized the field of seismology. It was a huge program, with high-level White House attention. Yet the ARPANET was started by a psychologist hired to run research programs that the agency’s leadership did not particularly care about. He isolated himself, pursuing his own grand plan for an “intergalactic computer network” that led to the ARPANET, the foundations of the modern Internet. That both programs could exist in a single agency should give pause to anyone looking for easy answers about management and innovation.

  —

  Just a few miles away from DARPA, Stephen Lukasik’s house in Northern Virginia looks out over Lake Barcroft, the same body of water where, some fifty years earlier, William Godel’s young daughters had tried to navigate in the “Jesus shoes” intended for Southeast Asia’s waterways. Tony Tether is one of Lukasik’s Lake Barcroft neighbors. Though cordial, the two former directors are not quite friends.

  On one of their encounters, Lukasik asked Tether, “What are you trying to do in DARPA?”

  “We want to fix it so that if we can find them, we can kill them,” Tether replied.

  Tether was likely being facetious; even his science fiction version of DARPA was about much more than creating killing machines. Yet his answer struck at the heart of Lukasik’s concern: there did not seem to be any thought given to the overarching problems that DARPA was supposed to solve; it was just generating technology.

  Lukasik has spent the past four decades contemplating DARPA’s legacy in national security. The walls of his basement are neatly lined with books that span topics ranging from Stalinism to cyber warfare. Notably absent are any books on management theory, a topic that
Lukasik openly mocks, even though it is often what people want to discuss with a former DARPA director. Now in his eighties, Lukasik is sometimes perplexed that his grandchildren’s friends are in awe when they find out he was once the head of DARPA, the agency whose inventions regularly appear in whiz-bang television shows featuring futuristic weapons. Lukasik is deeply proud of DARPA’s legacy but openly disappointed that the agency he helped mold is regarded as a science-fiction-inspired gadget lab. His vision for DARPA is still as an agency that solves important national security problems. “That is my growing discomfort with DARPA,” Lukasik told me. “It is not because DARPA is irrelevant as an institution, but is DARPA doing what it ought to be doing for the security of this country?”

  While some officials argue that DARPA is as good now as it has ever been—and that could be true in terms of the quality of science and technology—there is no denying that the agency has largely been absent from the past ten years of national security debates, which have centered on terrorism and insurgency. In late 2014, Lukasik advised the Defense Department to have DARPA start a new long-range strategic planning study to look at future technology for national defense, as the agency had done during his tenure. The Pentagon agreed, even co-opting the old name the Long Range R&D Planning Program (it was shortened to “Long Range Research Development Plan”). Rather tellingly, this new study was launched without DARPA.

  The current DARPA is so narrowly focused on technical problems it is hard to see how its mandate could have allowed it to come up with anything more creative in Afghanistan than computer algorithms. Nexus 7 was significant as an attempt to address current problems of warfare and insurgency by exploiting cutting-edge science and technology. And it was the first time since the Vietnam War that DARPA deployed personnel to a war zone. Yet its narrow scope also highlights how much had changed. In the Vietnam War, DARPA had sought to understand the fundamentals of society and the causes of insurgency; by 2011, DARPA in Afghanistan was seeking simply to predict the next IED attack. The wide-ranging exploration of human behavior that led to the hiring of J. C. R. Licklider seems unlikely today in an agency whose notion of social science is limited to computer programs that spit out predictions like a Zoltar Fortune Teller machine. “This may be more like an entropy process,” Lukasik said. “Once you move in that direction, you move in the direction of more detail, and if that’s the case, you run the risk of becoming irrelevant, because your measure of survival is political adroitness rather than technical excellence and solving important problems.”

  Evidence of this entropy process was highlighted in June 2013 when The Guardian and The Washington Post published reports about the NSA based on leaked documents provided by Edward Snowden. The documents revealed the depth and scope of the agency’s post-9/11 mass surveillance. Lukasik blamed DARPA’s loss of Total Information Awareness, John Poindexter’s data-mining program, as a contributing factor to the fiasco. Research that could have and should have been conducted by DARPA in the open was instead “transformed into intrusive government policies,” Lukasik wrote in his personal memoir.

  One can only guess what William Godel, who launched DARPA’s original counterinsurgency program, would have made of the current agency, whose press releases tout devices that can help soldiers scale glass skyscrapers, while American forces fight in a country dominated by mud houses. For Godel, technology was part of a larger strategy, not a narrow operational tactic. Project AGILE failed, but it was, as Charles Herzfeld proclaimed it, a “glorious failure.” By comparison, Nexus 7 was a failure regardless, not even because the technology was faulty, but because the national security problem it was trying to address—insurgency—could not possibly have been solved by any algorithm, no matter how elegantly designed. Even if it worked, it would be, at best, an inglorious success.

  —

  More than fifteen years after the 9/11 attacks, and over two decades since the end of the Cold War, the dilemma for DARPA is finding a new mission worthy of its past accomplishments and cognizant of its darker failures. In 2014, the agency announced the creation of the Biological Technologies Office, with a heavy focus on neuroscience, building on work DARPA has sponsored since the 1960s. Its new research, part of a larger White House initiative in brain science, has received wide attention. Helping soldiers recover from the devastating effects of traumatic brain injury, one of the more noble goals of the new office, is worthy of DARPA’s attention and one of the most exciting areas of research that it is currently pursuing.

  When I interviewed Justin Sanchez, the acting deputy director for the office, in 2016, he was refreshingly aware of DARPA’s earlier work in this field. The science and technology, he noted, had evolved in the forty years since the days of biocybernetics, as had the agency’s focus. In past years, DARPA had openly said that its goal was to develop ways to have the brain directly control weapons. DARPA officials are careful now to couch the agency’s work in terms of medical applications. The focus is “on restoring the injured war fighter,” Sanchez told me. “That has been the recent motivation for trying to understand the kinds of brain function research that we’re doing.”

  One of the agency’s programs, called Restoring Active Memory, is developing neuroprosthetics, essentially neural implants, which can help repair injured brains. After just two years, DARPA had developed prototype medical devices, and the work has moved to studies involving human subjects. “We already have some initial tests to see how interfacing with a neuroprosthetic device will affect the ability to form and recall memories,” Sanchez said. Similarly, another program, called SUBNETS, short for Systems-Based Neurotechnology for Emerging Therapies, is building implantable medical devices for people suffering from a variety of neuropsychiatric conditions, ranging from post-traumatic stress disorder to depression. There, again, Sanchez cited progress. “We’re also already in human subjects and have some preliminary evidence that is showing that we can understand some neurosignatures related to anxiety and we can modulate the brain with respect to anxiety,” he said.

  Modulating the human brain with neural implants holds the potential to treat any number of illnesses and injuries, but it also rightfully raises questions with ethicists, who wonder about the potential pitfalls of having the Pentagon involved in an area that touches upon the core of what makes us human. When I asked Sanchez, for example, whether DARPA would ever consider funding classified work in neurotechnology, he provided a careful answer. “I think we can say none of this is classified at this point in time,” he replied. “We always just keep our eyes open and never want to be caught off guard on that front. We’re actively just looking at the space and seeing where the opportunities are. I think we’ll have to make those determinations as we learn more about how neurotechnology plays out.”

  The potential for secret neuroscience work should give pause to anyone familiar with the Pentagon’s history of human testing. Moreover, though Sanchez and other DARPA officials downplay the potential for the agency’s neuroscience research to lead to weapons, it is impossible to ignore the reality that if the current work were successful, it would have applications in those areas. The world is still adapting to the drone revolution; is it really ready for brain-controlled aircraft? Such technology may be decades away, but there are other issues to be considered with this work. If DARPA succeeds, it could indeed revolutionize neuroscience. Yet if it fails, or is involved in scandal, such as a human subjects study gone wrong, the potential backlash against the agency could have repercussions as serious as those that resulted from Total Information Awareness.

  The ultimate question with this research, as with so much of DARPA’s work, is whether the agency will be allowed—or should be allowed—to pursue something with such truly high risk. Like many ambitious areas that DARPA has pursued in the past, from counterinsurgency to computer networking, its neuroscience work could transform the world by revolutionizing medicine, and it could lead to weapons that change the way we fight in future wars. Whether that world will b
e a better place is unclear.

  Acknowledgments

  If I could trace the idea of writing a DARPA history back to any particular moment in time, it would probably be to a conversation in Washington, D.C., in 2004 with my friend Robert Wall. We debated what a careful examination of DARPA might reveal about its legacy. Is it a genius factory? A Pentagon boondoggle? A refuge for crackpots? More than a decade later, I do not have an unequivocal answer, but many of the questions I raise in this book were guided by that initial conversation with Robert.

  Credit for moving that idea from a coffee shop to a formal book proposal goes to my extraordinary agent, Michelle Tessler, who gently nudged me until I committed thoughts to paper. The person most responsible for taking that proposal and turning it into a cohesive book is Andrew Miller, my editor at Alfred A. Knopf, who guided me through several rounds of revisions to help craft a narrative for an agency that has lived its life episodically. I also thank Emma Dries, editorial assistant at Knopf, for providing valuable comments on the draft.

  A number of friends and colleagues, in addition to Robert, read and critiqued versions of the manuscript, shared contacts, and pointed me in helpful directions. My friend Ann Finkbeiner, co-president for life of the Garwin Fan Club, provided critical comments on the rough draft. Steven Lee Myers offered support during some particularly tough times and provided thoughtful edits on the completed manuscript. I am also indebted to Richard Whittle, who has helped me in countless ways, and to Noah Shachtman, who will forever be my “work spouse.”

  The writing of the book spanned several nomadic years living in Washington, D.C.; Kraków, Poland; New York City; and Cambridge, Massachusetts. In Washington, I benefited from conversations with Jonathan Moreno of the University of Pennsylvania. I also learned a great deal from informal exchanges with Mark Lewis, Richard Van Atta, and David Sparrow, all of the Institute for Defense Analyses, and Steven Aftergood of the Federation of American Scientists. Shane Harris, one of the nicest and smartest national security reporters in town, generously shared knowledge and contacts. After I moved from Washington, my friends Askold Krushelnycky and Irena Chalupa hosted me during several return visits while I was conducting archival research. I also thank John Schidlovsky of the International Reporting Project for his support, and I am particularly grateful to the fearless Washington attorney Jeffrey D. Light, who took my Freedom of Information Act case to court and won.