The Imagineers of War

by Sharon Weinberger

  The future, in Fields’s view, was not in warships or military planes but in electronics and computers. While DARPA in the 1960s and 1970s had laid the foundation for personal computing and the modern Internet, by the early 1980s the agency’s computer science work had atrophied. The ARPANET had been transferred to the Defense Communications Agency, and DARPA’s subsequent directors pushed the computer science office to work on immediate military technologies. Fields, who had been a close colleague of J. C. R. Licklider’s, helped pioneer the agency’s renewed push into computer science in the 1980s, including a billion-dollar initiative to create artificial intelligence. To justify the massive spending increase, which lacked an immediate Cold War justification, DARPA pointed to Japan. In 1981, Japan had announced plans for the Fifth Generation Computer project, aimed at creating artificial intelligence, and the country’s economy, guided by the powerful Ministry of International Trade and Industry, was the new bogeyman. “We trundled out the Japanese as the archenemies and said we have to leap beyond the Japanese, you know, and whatever,” recalled Bob Cooper, who was then the director.

  “Whatever” was exactly right. The DARPA director privately acknowledged it was a ruse and he had selected the Japanese for convenience. DARPA formulated a ten-year, $1 billion effort to create artificial intelligence called the Strategic Computing Initiative; it would be DARPA’s biggest and most ambitious investment in computing since the ARPANET. The plan would appeal to the “Atari Democrats,” or the young tech-savvy Democrats who believed computing would save the American economy, and it would also pass muster with Republicans, who fretted over perceived threats to American hegemony. The director even went to Japan and traveled around the country, essentially gathering ammunition. “I came back and in private conversations with Congress and senators, I used it,” Cooper bragged. “I mean unabashedly.”

  Despite the influx of money, the ambitious road to artificial intelligence would soon be lined with failed technology projects. DARPA funded Thinking Machines Corporation, a company building supercomputers based on massively parallel processing, only to see the company go bankrupt when its government contracts dried up. The Pilot’s Associate, a talking, thinking computer program that would assist an aircraft pilot—as R2-D2 had served Luke Skywalker in the X-Wing fighter—ended in failure. A “smart truck” meant to navigate autonomously had problems differentiating between rocks and shadows. The billion-dollar artificial intelligence initiative failed to achieve any of its original objectives. By 1989, The New York Times was already writing the program’s postmortem. Showing just how far afield the work had gone from the original vision of artificial intelligence, the article declared that DARPA was “giving up its work on its autonomous land vehicle,” described as one of the “most publicized goals” of the Strategic Computing Initiative. Cooper, who had started and championed the program, was crushed. “It’s over in my mind,” he said.

  —

  DARPA had given up on creating artificial intelligence, but Japan was still a convenient bogeyman. With the collapse of the Warsaw Pact, Tokyo by 1989 had nearly eclipsed Moscow as the focus of Washington’s policy wonks. Only rather than the nuclear balance, the pundits were warning of the United States’ ballooning trade deficit with Japan, which had reached nearly $50 billion in 1989. Fears of an ascendant Japan (and a descendant United States) helped boost the Yale professor Paul Kennedy’s book The Rise and Fall of Great Powers onto the best-seller list, based on the prediction that the United States was heading toward a downward spiral spearheaded by deficit spending and economic stagnation. “There is a basic conflict between Japanese and American interests notwithstanding that the two countries need each other as friends—and it would be better to face it directly than to pretend that it doesn’t exist,” the journalist James Fallows argued in The Atlantic Monthly. “That conflict arises from Japan’s inability or unwillingness to restrain the one-sided and destructive expansion of its economic power.”

  The 1980s deregulation under President Ronald Reagan sparked a national debate over the role of government in managing industries. A group of vocal Democrats came out in favor of helping key industries. By the late 1980s, Congress and the White House, now under the administration of President George H. W. Bush, were sparring over the loaded term of “industrial policy,” or the role of government in boosting the private sector through targeted investments. Senator Al Gore, a Democrat and a technology enthusiast, was a forceful advocate for having the government invest in key areas, like supercomputers, while Republicans blasted attempts to “pick winners and losers.” The free market, the Republicans argued, would ultimately prevail.

  DARPA, with Democratic congressional support, had quietly become a part of this debate. The agency was funding Sematech, a consortium of chip manufacturers formed in the 1980s to help advance research on semiconductors. A New York Times article from March 1989 compared DARPA to Japan’s Ministry of International Trade and Industry, even though the latter had nothing to do with the military. “At a time when more industries are seeking Government help to hold their own against Asian and European competitors, DARPA is stepping into the void, becoming the closest thing this nation has to Japan’s Ministry of International Trade and Industry, the agency that organizes the industrial programs that are credited with making Japan so competitive,” the article said. DARPA was “being propelled, only partly by its own choosing, into the role of venture capitalist for America’s high-technology industry.”

  Yet Fields, as the new DARPA director in 1989, embraced this venture capitalist role. He focused on the high-definition television market, arguing it would allow American industry to maintain its lead in superconductors. The new director wanted to place DARPA at the vanguard of industrial policy, pursuing dual-use technology to give the United States an edge in the global economy, even if it was a vision that ran counter to the White House. Congress, at DARPA’s request, approved plans to spend $20 million to help the domestic high-definition television industry. Consumer electronics, Fields insisted, were critical for maintaining the American lead in semiconductors, which were also a key component in military systems. Fields was passionate about investing in consumer electronics but also naive, recalled Ray Colladay, a later DARPA director. Fields “had never run a private sector company or business, or product line, or anything that would give him experience that would help him appreciate how difficult it is to take a piece of technology and incorporate it into a product that people would want to buy.”

  When George H. W. Bush became president in 1989, he fashioned himself as a Republican idealist who opposed any government interference in the marketplace. Oblivious to the political mood in Washington, Fields became a vocal proponent of industrial policy. For those within DARPA who knew Fields, his determination to carve out a position in direct opposition to the presiding administration was not surprising. His undoing was a technology known as gallium arsenide, a potential replacement for silicon chips. Gallium arsenide chips were costly to make and the manufacturing base was still nascent, but compared with silicon, these new chips would be faster and more efficient and have properties, such as hardening against radiation, that made them particularly attractive to the military. DARPA had funded gallium arsenide in the past, but Fields wanted to use the technology as a test case for his ideas about supporting industry: DARPA, he decided, should invest in a gallium arsenide firm.

  Around the same time that Fields was getting involved in industrial policy, Richard Dunn, DARPA’s chief legal counsel, was exploring ways to enable the agency to cut through red tape. Dunn was in touch with a group of retired senior military officials who had been pushing the Pentagon to reform its contracting practices, and Dunn saw it as an opportunity to push through some of his own ideas. A few of those former officials, including the retired four-star general Bernard Schriever, regarded as the father of the American missile program, went to see Senator Sam Nunn, a powerful member of the Senate’s military oversight committee. Soon, Congress gave
DARPA legal authority to enter into something called “other transactions,” which was, in the simplest sense, a way for the Pentagon to fund research companies, skipping the volumes of government regulations that accompany typical military contracts.

  Fields saw the new legal authority as an opportunity for DARPA to act like a venture capital firm, and he turned his attention to Gazelle Microcircuits, a company working on gallium arsenide. Dunn argued it would be best to test a few small agreements—perhaps a few hundred thousand dollars—because DARPA had never used this new authority. But Fields wanted to make Gazelle the test case and took the idea to Charles Herzfeld, who by then was the director of defense research and engineering and DARPA’s overseer. Herzfeld supported it enthusiastically.

  Gazelle Microcircuits had already raised $10 million, and Fields wanted DARPA to be treated like an investor. Fields ordered Dunn not to write a statement of work. “If it was a venture-capital-supported firm, then he wanted this to be like a venture capital investment,” Dunn said. Arati Prabhakar, a DARPA scientist in charge of the contract, sat in on company board meetings. “She operated in a completely different manner than a normal program manager. She was on the inside of that company,” Dunn recalled. He was concerned about the vagueness, and after he consulted Prabhakar, they decided it would be best to specify the money was for research and development, distancing DARPA from the notion of investment. It was too late. On April 9, 1990, DARPA issued a press release announcing the Gazelle agreement, which gave the company $4 million over twelve months. “The agreement between DARPA and Gazelle was the first of its kind, based on DARPA’s recently granted authority to support advanced research and development by innovative means other than traditional contracts and grants,” the press release stated.

  Gazelle would design a one-gigabit-per-second and faster gallium arsenide chip, and DARPA, for its investment, would get access to Gazelle’s research and patents. Prabhakar hailed the award as a new way of doing business for DARPA. “Gazelle is typical of the companies we couldn’t work fruitfully with in the past,” she said. When The New York Times picked up the story, however, the deal was cast as an investment, as Fields had been advocating. “The Defense Department, through its advanced research arm, has for the first time made what is essentially a venture capital investment in a young Silicon Valley company,” the paper reported. Suddenly calls started coming in from the White House, and Fields was in a panic. “Bring me the agreement quick,” he ordered Dunn. Fields paged through the agreement and saw the statement of work that Dunn and Prabhakar had added. “Oh, thank goodness,” Fields said.

  Fields hoped that the statement of work—the very language he had originally not wanted included—would help distance him from the idea that DARPA was taking an equity stake in the company. At that point, it was too late. According to an unpublished DARPA history written by Dunn, Secretary of Defense Richard Cheney “was somewhat displeased by Dr. Fields’ high-profile public support” for dual-use technologies, like gallium arsenide. More displeased, however, were the White House and the president.

  Dennis McBride, a DARPA program manager, was waiting to see Fields when White House lawyers started to file into the director’s office. “Dennis, you’re postponed because these White House lawyers need to see Craig,” the secretary told him. As McBride prepared to leave, he could hear Fields being fired. “Craig, you should not have had this press conference,” one of the lawyers said. “President Bush loves you but you just spit right in his face.” Pentagon officials initially claimed that Fields had been “offered an opportunity” for another job in the Pentagon. It was a face-saving measure; a few weeks after being removed from his DARPA position, Fields quietly left the Pentagon for good. He would never comment on the episode, and even when asked during an interview commissioned by DARPA nearly twenty years later, he tersely responded, “Well, I think you should check the public record.”

  As Fields was getting fired, Herzfeld was scuba diving in the Caribbean with his wife. Herzfeld had recently returned to the Pentagon, full of ambition. Once one of the most powerful positions in the Pentagon, the director of defense research and engineering was now just a high-level bureaucrat with a nice office. In the fall of 1986, the president had signed the Goldwater-Nichols Department of Defense Reorganization Act, the largest restructuring of the military since the National Security Act of 1947. The legislation was the culmination of several years of debate over the failure of the military services to work together cohesively. Among other changes, a newly created undersecretary of defense for acquisition would become the “weapons czar,” while the director of defense research and engineering was downgraded to a second-tier job. Donald Hicks, who held the position at the time, resigned in protest. The change was not cosmetic; it meant DARPA reported to a lower level of the Pentagon.

  Herzfeld was supposed to be DARPA’s direct link to the Defense Department’s leadership, but he found himself frozen out of major decisions and with no access to the defense secretary. “Did you hear Craig got fired?” Herzfeld’s military assistant asked him when he returned from vacation, walking into his E Ring Pentagon office, several miles away from DARPA. Herzfeld had not, because no one had told him. It was a telling statement on the role of DARPA—and military science and technology generally—at the sunset of the Cold War. The agency once again had no agreed-upon mission, no political support, and, for the moment, no director. Even as its Cold War innovations were making their way to the battlefield, DARPA was edging away from war, or perhaps being edged out of it.

  —

  On August 2, 1990, some eighty-eight thousand Iraqi troops invaded oil-rich Kuwait. Less than six months later, after sanctions and international condemnation failed to persuade the Iraqi strongman, Saddam Hussein, to back down, the United States led an offensive that began in the predawn hours of January 17, 1991. Eight AH-64 army Apache helicopters slipped across the border from Saudi Arabia and into Iraq on a carefully rehearsed mission to destroy key Iraqi radar sites and open up a safe air corridor for air force aircraft. “There is no way we could build a corridor to get through safely with these jets without these surface-to-air missiles taking us down,” explained Dennis McBride, who was in charge of DARPA’s simulation work at the time. “We had to take out the surface-to-air sites.”

  In the days leading up to those attacks, DARPA-built simulation systems at Fort Rucker, in Alabama, were linked up to counterparts at Central Command, reviewing the tactics for taking out Iraq’s air defense systems. It was, as McBride explained, a “nap of the earth” mission, meaning the helicopters had to fly very low and close to the ground to avoid detection by radar. Simulation was the only way to see where the mission might fail. At U.S. Central Command headquarters, General Norman Schwarzkopf, who was directing the war, reviewed the simulations. “We were instantiating it in the simulation, saying, ‘This is a good idea but this one is not,’ and ‘Here’s why,’ and he would readjust,” McBride recalled. “He personally planned that very first mission with us, with the simulation capability.”

  Shortly after 2:00 a.m. on January 17, the Apaches approached the Iraqi air defense sites—flying an approach first mapped out in DARPA’s simulators and rehearsed in the Saudi desert—and destroyed the Iraqi air defense sites. A few hours later, an air force F-117 Nighthawk—a stealth aircraft derived from DARPA’s Have Blue prototype—flew safely through that air corridor. The F-117’s first bomb destroyed an Iraqi air force site, and the second leveled a telecommunications hub in the center of Baghdad.

  Less than six weeks later, the first Joint STARS aircraft, a DARPA-sponsored airborne tracking radar, spotted a massive convoy of Iraqi vehicles fleeing Kuwait and passed the data directly to strike aircraft. The resulting destruction of some two thousand Iraqi vehicles earned the escape route the name “highway of death,” and Air Force Magazine praised DARPA’s airborne radar as “one of the more unlikely heroes of Operation Desert Storm.” The Gulf War ended up as the proving ground for DARPA techn
ologies that were developed to fight the massive military forces of the Warsaw Pact. Its simulators were helping plan the war. The F-117 Nighthawk, the air force’s stealth attack jet and successor to DARPA’s Have Blue, was conducting air strikes. Joint STARS, which emerged from DARPA’s Assault Breaker program, was still a prototype but already changing how the military fought. Yet in the Pentagon, officials were not interested in any ambitious new DARPA projects.

  “Look, we don’t need now any new magic inventions,” George Lee Butler, a senior official in the Joint Chiefs of Staff, told Victor Reis, who took over as DARPA director after Fields was fired. “What we need is something to basically drive costs down. We know our budgets are going to decrease.” The Gulf War had just ended, and Reis admitted this view was a bit of a “culture shock” for an agency that had staked its reputation on developing revolutionary technologies, like stealth aircraft and advanced radar, but it was also the new reality. The Gulf War might have demonstrated DARPA’s innovations of the past two decades, and bolstered its reputation, but it also came at a time when the Pentagon was mainly looking to save money.

  Within days of the Gulf War’s ending, Reis got a call from Jack Thorpe, the DARPA program manager who had pioneered the agency’s simulation work. By 1990, the SIMNET program was done, and though Thorpe was long regarded as one of DARPA’s most imaginative program managers, he had largely been sidelined. Even for an organization known for cutting through red tape, Thorpe had a reputation for bending the rules. When DARPA was told its simulator building at Fort Knox, Kentucky, was possibly in violation of the law because all new military construction required prior congressional authorization, he had a trailer hitch installed and claimed it was just a temporary structure. By 1991, Thorpe, who had managed to stay on as a DARPA employee for a decade, was working from a small office in Europe. “He had sort of been put there to get out of the way,” Reis recalled.