Review of “Imagineers of War: The Untold Story of DARPA, the Pentagon Agency That Changed the World,” by Sharon Weinberger

DARPA (previously ARPA) has had generally good publicity as far as I have heard about it — progenitor of the ARPA-net, which led to the Internet; and sponsor of forward-looking competitions in autonomous vehicles, crowdsourcing, and bi-ped robots. This book details its more controversial yet unknown history as an agency with an amorphous mission somewhere at the intersection of science and war. 
The fundamental tension investigated by the book is one of mission — is DARPA a research agency with aspirations to apply advances to war and foreign policy, or is it a war department whose expertise lies in science? Consistent with this tension, its first four large projects consisted of space rockets, missile defense, nuclear test detection, and counter-insurgency. The ambiguity, in its early years, contributed to bureaucratic nightmares, expensive boondoggles, and some of the most infamous events of recent American history.

For example, ARPA was heavily involved in developing the theory of counter-insurgency; for Vietnam, it came up with the notorious Agent Orange as a tool to destroy crops and foliage that the Viet Cong was using as cover; it ended up causing health problems for up to a million people. Elsewhere, ARPA plotted counter-insurgency tactics for Iran and even against the Civil Rights movement in the U.S. Even its successful projects were often hampered; it built early space rockets better than those of the Navy or Air Force, before NASA’s inception, but was grounded due to departmental politics until after Sputnik won Russia part of the Space Race. When NASA finally landed a man on the moon, ARPA’s contributions to the rocket that did it were largely forgotten.

Early ARPA history was not entirely a failure, however. The book credits ARPA with essentially rejuvenating the entire academic field of seismology in the process of figuring out how to detect nuclear tests anywhere in the world (with one chief challenge being, even if one has sensors in the right places to detect tremors, how one distinguises a nuclear test from a natural seismic event). It further, in the 70s and 80s, led the development of stealth aircraft and drones that are direct predecessors to the ones in use today.

More recent DARPA history seems to be a mixed bag (in terms of successful technological investments) — in the late 1980s, it spent a billion dollars on failed AI projects, and the 1990s were full of various meanderings as the defense establishment as a whole was struggling to find a mission in a post-Cold War world. Then, after 9/11, DARPA started an eventually controversial (Total Information Awareness) program in mass (American) data collection and processing as a counter-terrorism effort; when news of the program became widespread in 2003, it was supposedly killed. In reality, the NSA absorbed its work and made it even more secretive, and it became a foundation of the work that would be explosively revealed by Snowden a decade later. On the flipside, the agency also sponsored the DARPA Grand challenges of 2004 and 2005, which were races for driverless cars across 150 miles in the desert. The program is widely acknowledged as spurring the autonomous vehicles research that is now bearing fruit: after no car traveled even 8 miles in 2004, five cars completed all 150 miles in 2005; the leaders of the top two teams — from Stanford and CMU — went on define such work for much of the last decade.

On a more wholistic scale, the author Sharon Weinberger argues that the agency today has much smaller ambitions than it used to — from tackling counter-insurgency, missile defense, and nuclear test detection at once with in-house research, to mostly funding external scientists and hosting competitions. This shift, to motivating and conducting longer-term research, she states, is probably accompanied by better overall scientific practices. However, it pales in its vision and potential impact, when compared to the DARPA of old.

The book itself is solid, and I enjoyed it. It does its job as an institutional history, similar in style, scope, and quality to The Idea Factory: Bell Labs and the Great Age of American Innovation. It seems to be fair to the agency, balancing its scientific successes with its (scientific, moral, and political) failures, and it is very much focused on the science and technology rather than bureaucratic squabbles. The author has a fantastic ability to discuss technology at just the right level of detail for her audience. Unlike some, she doesn’t over-use jargon in a manner that serves more to impress (and annoy) than educate the reader, but she also doesn’t shy away from explanations for why an idea did or did not have merit. For example, I learned that one reason stealth aircraft is so difficult to design is that a 2x improvement in detection distance (e.g. detectable from 10 minutes away to 5 minutes) requires a 16x decrease in radar surface. This feature, for example, helps one understand that some of the agency’s supposed failures were brilliant ideas that were simply a bit ahead of their time.

My only (small) quibble with it is that it spent a little too much time making fun of crazy ideas and technology boondoggles dreamt up and pursued by the agency (such as a missile defense force field, a rotating metal net to protect the President during speeches, or a wire strung between 2 helicopters to detect weapon staches by induction), without giving context of how much was actually spent on such ideas as a percent of DARPA’s funding. If DARPA’s mission truly is one of science, it is expected that most ideas fail, even without such bad science; the difficulty is that one doesn’t know which 10% of ideas will succeed before trying them.

On the other hand, the author seems well-aware of this tension, and so the criticism is minor. She seems to argue that many of these concoctions were doomed to fail — not only were the ideas crazy, but they were pursued in bad faith and through poor scientific practices, such as excessive secrecy, lack of peer review, and led by non-domain-experts. out. Further, at one point, she explicitly balances the agency’s due diligence into parapsychology research (including work at Stanford on whether a mother rabbit could over distance detect the death of her baby bunny, and whether the supposed phenomenon could be used as a communication mechanism) with the agency’s very successful research into computer networking and human computer interaction — suggesting that, a priori, it was not clear which agenda would prove more successful.