Friday, June 29, 2018
Surveillance Valley The Secret Military History of the Internet By Yasha Levine Published 02.06.2018 (PublicAffairs)
We face a crisis of computing. The very devices that were supposed to augment our minds now harvest them for profit. How did we get here?
Most of us only know the oft-told mythology featuring industrious nerds who sparked a revolution in the garages of California. The heroes of the epic: Jobs, Gates, Musk, and the rest of the cast. Earlier this year, Mark Zuckerberg, hawker of neo-Esperantist bromides about “connectivity as panacea” and leader of one of the largest media distribution channels on the planet, excused himself by recounting to senators an “aw shucks” tale of building Facebook in his dorm room. Silicon Valley myths aren’t just used to rationalize bad behavior. These business school tales end up restricting how we imagine our future, limiting it to the caprices of eccentric billionaires and market forces.
What we need instead of myths are engaging, popular histories of computing and the internet, lest we remain blind to the long view.
At first blush Yasha Levine’s Surveillance Valley: The Secret Military History of the Internet (2018) seems to fit the bill. A former editor of The eXile, a Moscow-based tabloid newspaper, and investigative reporter for PandoDaily, Levine has made a career out of writing about the dark side of tech. In this book, he traces the intellectual and institutional origins of the internet. He then focuses on the privatization of the network, the creation of Google, and revelations of NSA surveillance. And, in the final part of his book, he turns his attention to Tor and the crypto community.
He remains unremittingly dark, however, claiming that these technologies were developed from the beginning with surveillance in mind, and that their origins are tangled up with counterinsurgency research in the Third World. This leads him to a damning conclusion: “The Internet was developed as a weapon and remains a weapon today.”
To be sure, these constitute provocative theses, ones that attempt to confront not only the standard Silicon Valley story, but also established lore among the small group of scholars who study the history of computing. He falls short, however, of backing up his claims with sufficient evidence. Indeed, he flirts with creating a mythology of his own — one that I believe risks marginalizing the most relevant lessons from the history of computing.
The scholarly history is not widely known and worth relaying here in brief. The internet and what today we consider personal computing came out of a unique, government-funded research community that took off in the early 1960s. Keep in mind that, in the preceding decade, “computers” were radically different from what we know today. Hulking machines, they existed to crunch numbers for scientists, researchers, and civil servants. “Programs” consisted of punched cards fed into room-sized devices that would process them one at a time. Computer time was tedious and riddled with frustration. A researcher working with census data might have to queue up behind dozens of other users, book time to run her cards through, and would only know about a mistake when the whole process was over.
Users, along with IBM, remained steadfast in believing that these so-called “batch processing” systems were really what computers were for. Any progress, they believed, would entail building bigger, faster, better versions of the same thing.
But that’s obviously not what we have today. From a small research community emerged an entirely different set of goals, loosely described as “interactive computing.” As the term suggests, using computers would no longer be restricted to a static one-way process but would be dynamically interactive. According to the standard histories, the man most responsible for defining these new goals was J. C. R. Licklider. A psychologist specializing in psychoacoustics, he had worked on early computing research, becoming a vocal proponent for interactive computing. His 1960 essay “Man-Computer Symbiosis” outlined how computers might even go so far as to augment the human mind.
It just so happened that funding was available. Three years earlier in 1957, the Soviet launch of Sputnik had sent the US military into a panic. Partially in response, the Department of Defense (DoD) created a new agency for basic and applied technological research called the Advanced Research Projects Administration (ARPA, today known as DARPA). The agency threw large sums of money at all sorts of possible — and dubious — research avenues, from psychological operations to weather control. Licklider was appointed to head the Command and Control and Behavioral Sciences divisions, presumably because of his background in both psychology and computing.
At ARPA, he enjoyed relative freedom in addition to plenty of cash, which enabled him to fund projects in computing whose military relevance was decidedly tenuous. He established a nationwide, multi-generational network of researchers who shared his vision. As a result, almost every significant advance in the field from the 1960s through the early 1970s was, in some form or another, funded or influenced by the community he helped establish.
Its members realized that the big computers scattered around university campuses needed to communicate with one another, much as Licklider had discussed in his 1960 paper. In 1967, one of his successors at ARPA, Robert Taylor, formally funded the development of a research network called the ARPANET. At first the network spanned only a handful of universities across the country. By the early 1980s, it had grown to include hundreds of nodes. Finally, through a rather convoluted trajectory involving international organizations, standards committees, national politics, and technological adoption, the ARPANET evolved in the early 1990s into the internet as we know it.
Levine believes that he has unearthed several new pieces of evidence that undercut parts of this early history, leading him to conclude that the internet has been a surveillance platform from its inception.
The first piece of evidence he cites comes by way of ARPA’s Project Agile. A counterinsurgency research effort in Southeast Asia during the Vietnam War, it was notorious for its defoliation program that developed chemicals like Agent Orange. It also involved social science research and data collection under the guidance of an intelligence operative named William Godel, head of ARPA’s classified efforts under the Office of Foreign Developments. On more than one occasion, Levine asserts or at least suggests that Licklider and Godel’s efforts were somehow insidiously intertwined, and that Licklider’s computing research in his division of ARPA had something to do with Project Agile. Despite arguing that this is clear from “pages and pages of released and declassified government files,” Levine cites only one such document as supporting evidence for this claim. It shows how Godel, who at one point had surplus funds, transferred money from his group to Licklider’s department when the latter was over budget.
This doesn’t pass the sniff test. Given the freewheeling nature of ARPA’s funding and management in the early days, such a transfer should come as no surprise. On its own, it doesn’t suggest a direct link in terms of research efforts. Years later, Taylor asked his boss at ARPA to fund the ARPANET — and, after a 20-minute conversation, he received $1 million in funds transferred from ballistic missile research. No one would seriously suggest that ARPANET and ballistic missile research were somehow closely “intertwined” because of this.
Sharon Weinberger’s recent history of ARPA, The Imagineers of War: The Untold Story of DARPA, The Pentagon Agency that Changed the World(2017), which Levine cites, makes clear what is already known from the established history. “Newcomers like Licklider were essentially making up the rules as they went along,” and were “given broad berth to establish research programs that might be tied only tangentially to a larger Pentagon goal.” Licklider took nearly every chance he could to transform his ostensible behavioral science group into an interactive computing research group. Most people in wider ARPA, let alone the DoD, had no idea what Licklider’s researchers were up to. His Command and Control division was even renamed the more descriptive Information Processing Techniques Office (IPTO).
Licklider was certainly involved in several aspects of counterinsurgency research. Annie Jacobsen, in her book The Pentagon’s Brain: An Uncensored History of DARPA, America’s Top-Secret Military Research Agency (2015), describes how he attended meetings discussing strategic hamlets in Southeast Asia and collaborated on proposals with others who conducted Cold War social science research. And Levine mentions Licklider’s involvement with a symposium that addressed how computers might be useful in conducting counterinsurgency work.
But Levine only points to one specific ARPA-funded computing research project that might have had something to do with counterinsurgency. In 1969, Licklider — no longer at ARPA — championed a proposal for a constellation of research efforts to develop statistical analysis and database software for social scientists. The Cambridge Project, as it was called, was a joint effort between Harvard and MIT. Formed at the height of the antiwar movement, when all DoD funding was viewed as suspicious, it was greeted with outrage by student demonstrators. As Levine mentions, students on campuses across the country viewed computers as large, bureaucratic, war-making machines that supported the military industrial complex.
Levine makes a big deal of the Cambridge Project, but is there really a concrete connection between surveillance, counterinsurgency, computer networking, and this research effort? If there is, he doesn’t present it in the book. Instead he relies heavily on an article in the Harvard Crimson by a student activist. He doesn’t even directly quote from the project proposal itself, which should contain at least one or two damning lines. Instead he lists types of “data banks” the project would build, including ones on youth movements, minority integration in multicultural societies, and public opinion polls, among others. The project ran for five years but Levine never tells us what it was actually used for.
It’s worth pointing out that the DoD was the only organization that was funding computing research in a manner that could lead to real breakthroughs. Licklider and others needed to present military justification for their work, no matter how thin. In addition, as the 1960s came to a close, Congress was tightening its purse strings, which was another reason to trump up their relevance. It’s odd that an investigative reporter like Levine, ever suspicious of the standard line, should take the claims of these proposals at face value.
I spoke with John Klensin, a member of the Cambridge Project steering committee who was involved from the beginning. He has no memory of such data banks. “There was never any central archive or effort to build one,” he told me. He worked closely with Licklider and other key members of the project, and he distinctly recalls the tense atmosphere on campuses at the time, even down to the smell of tear gas. Oddly enough, he says some people worked for him by day and protested the project by night, believing that others elsewhere must be doing unethical work. According to Klensin, the Cambridge Project conducted “zero classified research.” It produced general purpose software and published its reports publicly. Some of them are available online, but Levine doesn’t cite them at all. An ARPA commissioned study of its own funding history even concluded that, while the project had been a “technical success” whose systems were “applicable to a wide variety of disciplines,” behavioral scientists hadn’t benefited much from it. Until Levine or someone else can produce documents demonstrating that the project was designed for, or even used in, counterinsurgency or surveillance efforts, we’ll have to take Klensin at his word.
As for the ARPANET, Levine only provides one source of evidence for his claim that, from its earliest days, the experimental computer network was involved in some kind of surveillance activity. He has dug up an NBC News report from the 1970s that describes how intelligence gathered in previous years (as part of an effort to create dossiers of domestic protestors) had been transferred across a new network of computer systems within the Department of Defense.
This report was read into the Congressional record during joint hearings on Surveillance Technology in 1975. But what’s clear from the subsequent testimony of Assistant Deputy Secretary of Defense David Cooke, the NBC reporter had likely confused several computer systems and networks across various government agencies. The story’s lone named source claims to have seen the data structure used for the files when they arrived at MIT. It is indeed an interesting account, but it remains unclear what was transferred, across which system, and what he saw. This incident hardly shows “how military and intelligence agencies used the network technology to spy on Americans in the first version of the Internet,” as Levine claims.
The ARPANET was not a classified system — anyone with an appropriately funded research project could use it. “ARPANET was a general purpose communication network. It is a distortion to conflate this communication system’s development with the various projects that made use of its facilities,” Vint Cerf, creator of the internet protocol, told me. Cerf concedes, however, that a “secured capability” was created early on, “presumably used to communicate classified information across the network.” That should not be surprising, as the government ran the project. But Levine’s evidence merely shows that surveillance information gathered elsewhere might have been transferred across the network. Does that count as having surveillance “baked in,” as he says, to the early internet?
Levine’s early history suffers most from viewing ARPA or even the military as a single monolithic entity. In the absence of hard evidence, he employs a jackhammer of willful insinuations as described above, pounding toward a questionable conclusion. Others have noted this tendency. He disingenuously writes that, four years ago, a review of Julian Assange’s book in this very publication accused him of being funded by the CIA, when in fact its author had merely suggested that Levine was prone to conspiracy theories. It’s a shame, because today’s internet is undoubtedly a surveillance platform, both for governments and the companies whose cash crop is our collective mind. To suggest this was always the case means ignoring the effects of the hysterical national response to 9/11, which granted unprecedented funding and power to private intelligence contractors. Such dependence on private companies was itself part of a broader free market turn in national politics from the 1970s onward, which tightened funds for basic research in computing and other technical fields — and cemented the idea that private companies, rather than government-funded research, would take charge of inventing the future. Today’s comparatively incremental technical progress is the result. In The Utopia of Rules (2015), anthropologist David Graeber describes this phenomenon as a turn away from investment in technologies promoting “the possibility of alternative futures” to investment in those that “furthered labor discipline and social control.” As a result, instead of mind-enhancing devices that might have the same sort of effect as, say, mass literacy, we have a precarious gig economy and a convenience-addled relationship with reality.
Levine recognizes a tinge of this in his account of the rise of Google, the first large tech company to build a business model for profiting from user data. “Something in technology pushed other companies in the same direction. It happened just about everywhere,” he writes, though he doesn’t say what the “something” is. But the lesson to remember from history is that companies on their own are incapable of big inventions like personal computing or the internet. The quarterly pressure for earnings and “innovations” leads them toward unimaginative profit-driven developments, some of them harmful.
This is why Levine’s unsupported suspicion of government-funded computing research, regardless of the context, is counterproductive. The lessons of ARPA prove inconvenient for mythologizing Silicon Valley. They show a simple truth: in order to achieve serious invention and progress — in computers or any other advanced technology — you have to pay intelligent people to screw around with minimal interference, accept that most ideas won’t pan out, and extend this play period to longer stretches of time than the pressures of corporate finance allow. As science historian Mitchell Waldrop once wrote, the polio vaccine might never have existed otherwise; it was “discovered only after years of failure, frustration, and blind alleys, none of which could have been justified by cost/benefit analysis.” Left to corporate interests, the world would instead “have gotten the best iron lungs you ever saw.”
Computing for the benefit of the public is a more important concept now than ever. In fact, Levine agrees, writing, “The more we understand and democratize the Internet, the more we can deploy its power in the service of democratic and humanistic values.” Power in the computing world is wildly unbalanced — each of us mediated by and dependent on, indeed addicted to, invasive systems whose functionality we barely understand. Silicon Valley only exacerbates this imbalance, in the same manner that oil companies exacerbate climate change or financialization of the economy exacerbates inequality. Today’s technology is flashy, sexy, and downright irresistible. But, while we need a cure for the ills of late-stage capitalism, our gadgets are merely “the best iron lungs you ever saw.”