When Surveillance Valley by Yasha Levine was published in 2018 I took a rather perfunctory look at it and figured I wasn’t that interested in reading it. The little bit I could read on Amazon didn’t impress me and so I ignored it. Since then I’ve encountered people relying on it to make silly arguments and telling me I should read it.
More than once I’ve had conversations with people that ended with them telling me that I should read Surveillance Valley. I’ve found these conversations frustrating and unfortunate because otherwise we agreed in our disgust with the surveillance state. They were as fed up and frustrated with our dystopic surveilllance based society as I was, but Surveillance Valley was not providing them with the proper historical background to interpret the present correctly.
This blog post is the result of me finally finding the time to sit down to read Surveillance Valley, and analyze it for historical accuracy. The book’s general premise that there is an unbroken lineage of malevolent, counterinsurgent, and militaristic intent at the heart of the Internet is incorrect. But it does have some interesting tidbits in it and I did learn a couple interesting things from reading it.
I’m going to walk through the book chapter by chapter pointing out my issues with it. But before we get to that I want to be clear that this is not an ad hominem attack on Surveillance Valley’s author, Yasha Levine. I’ve got nothing against him personally and the personal attacks he received because of Surveillance Valley are pretty gross. I was particularly disgusted by The Guardian’s opening sentence to their review of the book. Pointing out that Levine is “of Russian extraction” is a disgusting, bigoted slant. The Guardian’s editors should take their job more seriously.
This article is not a personal attack on Yasha Levine and I’m mostly uninterested in Levine’s politics. This article is concerned with the historical narrative of the Internet presented in Surveillance Valley. I also criticize the book for some of its analysis on contemporary technologies, but I’m mostly interested in showing inaccuracies in its description of Internet history.
For the rest of this article I will use the shorthand [SV page #] when quoting directly from Surveillance Valley.
Prologue
In the Prologue Levine lays out his thesis for the book.
My research reveals a third historical strand in the creation of the early Internet – a strand that has all but disappeared from the history books. Here, the impetus was rooted not so much in the need to survive a nuclear attack but in the dark military arts of counterinsurgency and America’s fight against the perceived global spread of communism. [SV page 13]
The idea that the early Internet (i.e., ARPANET) was developed to survive a nuclear attack has persisted throughout the years. Everything I’ve read and heard would suggest that this is not true and that this was not a design goal of the early ARPANET. I suspect the culprit is Paul Baran’s early paper on packet switching. Baran’s paper argues that distributed networks provide better redundancy, and in a nuclear age packet switched networks are good at surviving nuclear attacks. It’s important to note that while Baran played an important role in conceptualizing packet switched networks, there is no evidence to suggest that early Internet pioneers chose packet switching over circuit switching for reasons of survivability in the event of a nuclear war.
Instead of rehashing too much history, I’ll quote Ashwin Mathew’s succinct coverage of this topic.
While the early development of Internet technologies may have their origins in the Cold War era, they were not entirely shaped by the political imperatives of that time. Packet-switched networking may have been developed with the military problem of survivability in mind, but it was eventually put to use for the fairly utilitarian end of access to time-sharing computer resources. Accordingly, the social context within which these technologies were developed was driven by academic researchers, rather than military priorities.
Levine has a lot to say about the war in Vietnam and the US government’s undemocratic suppression of protestors in the 1960s and 1970s. But he fails to really connect any of it to the ARPANET. More specifically, he fails to connect US government suppression of anti-war protestors to the intent of the ARPANET's designers. This connection is key, because if Levine can’t connect the Intent of the Internet’s progenitors to events happening in meatspace at the same time, his primary argument falls apart. Merely juxtaposing events happening at the same time does not show intent or causation.
Chapter 2
The ARPANET first gets discussed in Chapter 2. This is Levine’s first attempt at proving his thesis that the Internet emerged from DoD counterinsurgency programs in use at the time in Vietnam. However, besides retelling some stories about Students for a Democratic Society (SDS) breakins and protests against JC Licklider, nothing is presented that really supports his thesis. The fact that some student protestors believed that a project to network computers was connected to the Vietnam War and suppressing student dissent does not really prove anything about the intent of that project. Levine needs more evidence here to reach his conclusions.
It’s a fact that the ARPANET was funded from the same purse as counterinsurgency programs happening in Vietnam at the time. They were both funded by DARPA. And let’s be honest, DARPA and the RAND Corporation did some pretty horrible shit in Vietnam. The Phoenix Program and other assassination programs were terrible. Yet Levine’s self-assigned task is to connect this stuff to the ARPANET and eventually the Internet. That’s what he promised in the Prologue and that’s what he fails to do in this chapter.
Levine’s method is to point out that DARPA is doing bad stuff in Vietnam and then start talking about the ARPANET. The reader is left to form the connection themselves. Without explicitly stating it, his method boils down to a kind of follow the money thinking. There’s nothing inherently wrong with this method and it is used in journalism often, especially journalism that investigates corruption and governmental malfeasance. Typically, following the money in investigative journalism means investigating who controls business entities, or investigating capital flows through networks of actors engaged in a conspiracy. There’s nothing fundamentally wrong with journalists using a follow the money method in their investigative journalism, but it has its limits. It assumes that intent and causation follow the money.
Levine’s deployment of this method is facile, however. His essential argument that the ARPANET was developed for military and counterinsurgency purposes rests entirely on the fact that it was funded by DARPA and that it was protested by anti-war activists. He offers no other evidence to advance his argument.
The argument that the ARPANET and the early Internet was funded for domestic counterinsurgency and military purposes is absurd once one investigates the available historical record. A historical record that Levine selectively engages with throughout Surveillance Valley.
Chapter 3
Chapter 3 opens with discussion of an NBC evening news report that aired on June 2, 1975. In it, correspondent Ford Rowan presents evidence that a new computer network, the ARPANET, is being used by law enforcement, the DoD, the CIA, and the FBI to access information about American citizens. Levine then argues that this usage of the network presages Edward Snowden’s leaks from 2013.
There’s just one glaring problem with this narrative. The NBC report is talking about law enforcement and spy agencies accessing information over a network of computers, while Snowden’s leaks were talking about law enforcement and spy agencies surveilling users of the network. These are two different things. Surveillance Valley falsely equivocates between these two different things.
Should we be surprised that government agencies were exchanging information over a general purpose computer network? Probably not, since they were some of its earliest users. But this has absolutely nothing to do with Snowden’s leaks in 2013. There’s no connection between the two.
Levine spends a fair amount of time talking about databases in Chapter 3 and he uses terms like “big data” that have grown popular in the past decade. Disregarding that Surveillance Valley is meant to be a book about the Internet, it never talks about how terrible databases were in the 1960s and 1970s. They were nothing like what we have today.
I could not find any discussion of CIA data recovery and indexing methods from the 1960s. But here is how former CIA employee Ralph McGehee described accessing the CIA’s database of Chinese communists in the late 1950’s in his book, Deadly Deceits: My 25 Years in the CIA:
The huge index room, commonly referred to as “the snake pit,” was surreal. The small window air conditioners would break down under the strain of the masses of sweating humanity trying to squeeze into the narrow aisles between six-foot-high wooden file cabinets. Earthy aromas greeted those who entered from the relatively fresh air of the corridor. There was a noisy din of shouted instructions, disgruntled conversations, and flirtations. To check a name in the index, you had to use fancy footwork to dance around the opened drawers and mobs of humanity. Inevitably you were shoved, elbowed, kicked, and jostled.
We in the China records unit had our own private hell as romanized Chinese names created a special form of file-tracing torture. The Chinese language had no alphabet. It was composed of about 10,000 commonly used ideographs or symbols representing ideas or words. This posed an extremely difficult problem for the records division and the file tracers, none of whom knew how to read or write Chinese. The most common Chinese surname is Ch’en, but there were more than a dozen ideographs that romanized out as Ch’en. The same with given names such as Cheng. A request for a trace on a Ch’en Cheng presented a near impossible task of sorting through the hundreds or thousands of index cards on that name. How to discriminate between them?
A system of numbering every major Chinese ideograph was devised and was slowly being incorporated into the indexing system. Under this system each ideograph was assigned a four-digit number, which was included on the index card; i.e., Ch’en Cheng (7115/2638). In this way the tracer could look only for those Ch’en Chengs writing their names with the 7115 Ch’en ideograph and the 2638 Cheng ideograph—a vast improvement. But the system was new and it required a staff of Chinese to process the numbers.
Perhaps the most difficult of all problems was the records division’s habit of lending out its record copies of documents. Someone conducting a trace on Ch’en Cheng would attempt to gather all documents mentioning that name. If there were 1,000 references to documents, then the tracer theoretically would attempt to gather every one of those documents. But as frequently was the case, other people outside of the records division had some of them. To do a thorough job, the name tracer would have to go to the office of each person and ask to see the document. The main corridor in the I-L complex was almost a half-mile long, and it was not infrequently that one traveled from one end of the building to the other and back again. In many cases the person who had borrowed the document had been transferred, and the document was lost forever.
Databases were only introduced in the mid 1970’s and it’s no secret that the CIA was Oracle’s first customer. The first relational database was developed in 1977 and it really wasn’t until the early 2000’s that the kind of massive databases that we now refer to as “big data” could be used effectively to engage in mass surveillance. Of course, none of this has anything to do with the ARPANET, or surveilling Internet users, but even Levine’s discussion of database technology is biased and facile without some discussion of how immature databases were in the 1960s and 1970s.
In the late 1960’s, as engineers at MIT, UCLA, and Stanford diligently worked to build a unified military computer network, the country convulsed with violence and radical politics – much of it directed at the militarization of American society, the very thing that the ARPANET represented. [SV page 114]
The ARPANET was not a representation of a militarizing American society. Even the current Internet doesn’t represent a militarization of society. I protested the Reserve Officer’s Training Corps (ROTC) at my undergraduate alma mater James Madision University (JMU) because the ROTC promotes the militarization of American society. I agree that the USA has a problem with its culture being militarized. But this has nothing to do with the ARPANET or later Internet.
The early history of the Internet/ARPANET is well documented. Not only do we have a plethora of available documentation that has been digitized and placed online, but most of the early Internet pioneers are still alive and available to talk with people. It’s quite easy to look at this record and determine that from its earliest inception the ARPANET was intended to be a general purpose network.
For example, the Department of Defense’s Report, A History of the ARPANET: The First Decade, written in 1981 plainly states the objectives of the ARPANET on page 2.
(1) To develop techniques and obtain experience on interconnecting computers in such a way that a very broad class of interactions are possible, and (2) To improve and increase computer research productivity through research sharing.
The Request for Comment (RFC) document series is another historical trove of documents that Surveillance Valley completely ignores.
"II. Some Requirements Upon the Host-to-Host Software
Simple Use As with any new facility, there will be a period of very light usage until the community of users experiments with the network and begins to depend upon it. One of our goals must be to stimulate the immediate and easy use by a wide class of users. With this goal, it seems natural to provide the ability to use any remote HOST as if it had been dialed up from a TTY (teletype) terminal. Additionally, we would like some ability to transmit a file in a somewhat different manner perhaps than simulating a teletype. [RFC 1: Host Software]
The key innovation of the ARPANET was inter-process communication between different types of computers. By 1969 computer users were already able to communicate with computers at distance. Long distance teletypes for programming computers were already in use. What the ARPANET enabled was the ability for computers to talk with other computers over a network. This lead to the invention of sockets to enable cross network inter-process communication. Sockets allow a process on one computer to exchange data with a process on another computer. Even after 50 years the Internet still essentially works like this.
Because a process may create another process, and because in general the two processes are indistinguishable when viewed externally, I know of no reasonable way for two processes to request connection directly with each other. The function of sockets is to provide a standard interface between processes. [RFC 49: Conversations with Steve Crocker (UCLA)]
Sockets are general purpose channels for transferring data between processes on different computers. There’s nothing inherently militaristic in their design or implementation.
The ARPANET scenarios presented at the 1972 International Conference on Computer Communication are diverse, as one would expect from a network that was designed for general purpose computational applications. They include; games, English language conversational tools, mathematical applications, database queries, and file transfer applications. More detailed discussion of the conference and the general purpose nature of sockets can be found here.
I have no reason to doubt NBC’s reporting that computers connected to the ARPANET were storing and transmitting data on American citizens, possibly illegally. The NBC report mentioned at the beginning of Chapter 3 would seem to confirm this. But there’s a big difference between a general purpose network constructed for lots of different applications and a network built for the sole purpose of spying on Americans. Not every use of the early ARPANET was morally justifiable, just like not every use of today's Internet is morally justifiable.
Chapter 4
I don’t have many quibbles with chapter 4. I do find it odd that the privatization of the Internet often gets mixed in with hippie counterculture narratives, LSD, and libertarian pseudo-revolutionaries. The privatization of the Internet is mainly a story of lucky government contractors being in the right place at the right time, and government bureaucrats not understanding or not caring about the money-making potential of the Internet. In a more serious book about Internet history I would expect more discussion of boring contracts and agreements than of counter culture. Instead, in chapter four it feels like Levine is responding to a common cultural narrative about the Internet’s early history instead of creating his own fact based historical narrative.
I remember reading Wired, Mondo 2000, and R U Sirius as a kid and really grooving with the potential of a tech utopian future. It was my 1990's teenage tech-utopian phase that I'm happy I grew out of.
In 2023 I read True Names and the Opening of the Cyberspace Frontier. It’s an anthology of collected essays from different authors published in 2001, but most of the articles were published in the 1990’s. It’s eponymous, essay True Names was originally published in 1981 by Vernor Vinge. True Names contains the first use of the term cyberspace in fiction.
Reading True Names and the Opening of the Cyberspace Frontier in 2023 reminded me of the things I read as a kid about cyberspace and utopian futures. It also felt somewhat quaint and silly. I was reminded of my childhood fascination with computers, but also of American society’s childish fascination with the Internet. It brought me back to a place and time when the dream of the Internet and technology was much more pleasant than it is today.
Is documenting the zeitgeist of 1990’s cyberspace utopian thinking necessary for a history of the Internet? Levine seems to think so, but I’m not convinced. Surveillance Valley would be a better book if it focused less on past zeitgeists and cultural reminiscence, and more on historical facts.
Chapter 5
Chapter 5 permits Surveillance Valley to finally arrive at Levine’s recentist comfort zone. The chapter starts off by talking about Google and search, having forsaken the book’s original goal of discussing the Internet. Many people get confused between the World Wide Web and the Internet, but these things are not the same. Google is not the Internet, Google is not the World Wide Web, and Silicon Valley is not at the center of the Internet, etc.
It sounds silly to state these obvious facts, but I sometimes wonder if they don’t need to be restated more often. When I talk about the Internet with different audiences I never know what level of understanding each audience brings to the table. Chapter 5 causes me wonder how many of my misgivings about Surveillance Valley lie in Levine’s basic misunderstanding of how technology works. Many of the distinctions that are obvious to computer scientists seem not so obvious to Levine. Thus, he repeatedly fails to draw meaningful distinctions between different technologies.
The Internet has come a long way since Larry Page and Sergey Brin converted Google from a Stanford PhD project to a multi-billion-dollar company. But in a lot of ways it hasn’t changed much from its ARPANET days. It’s just gotten more powerful.
[..]
The Internet is like a giant, unseen blob that engulfs the modern world. There is no escape, and, as Page and Brin so astutely understood when they launched Google, everything that people do online leaves a trail of data. [SV page 246]
Either Levine does not understand the distinction between the Web and the Internet, or he is being willfully ignorant of this distinction for the sake of his argument. I understand that for many Internet users the Web is the Internet. But the Web is not actually the Internet. A serious historian would clarify this distinction and connect what’s happening on the Web back to their original thesis. At the very least Levine should provide some understanding of this distinction to his readers. Instead we get a seamless merge from one to the next. For Levine, the ARPANET, the Internet, Google, Silicon Valley, and many other entities are seemingly all part of the same surveillance behemoth, unchanging throughout history with the perennial goal of spying on you.
Chapter 5 is strong when it discusses the horribly invasive practices of large web services corporations, and conversely, it is weak when it attempts to incorporate their behavior into a comprehensive historical narrative. This observation can be made about Surveillance Valley more generally. It’s much better at discussing recent events than it is about discussing history.
Chapters 6 and 7
Chapters 6 and 7 discuss the Edward Snowden revelations and Tor. I’m thankful to Edward Snowden for taking the risks he did and for making the disclosures he made. It’s unfortunate that there was no significant political fallout from Snowden’s leaks. We should probably all assume that whatever the NSA is doing now is much worse than what Snowden revealed. We don’t know their capabilities, but at least now we understand that the law does not constrain them.
The discussion of the Tor Project was interesting. I’ve played around a bit with Tor and while I’m not an expert in it I understand basically how it works.
[..] I understood why Tor had backing from Silicon Valley: it offered a false sense of privacy, while not posing a threat to the industry’s underlying surveillance business model. [SV page 326]
Maybe, but I doubt it’s as simple as that. I suspect different individuals and organizations are interested in the Tor Project for different reasons. If the US State Department wants to foment numerous regime change operations, which I think they do, then Tor is probably more useful if it’s actually secure. That security requires funding. Large web services companies probably want to keep an eye on an evolving technology space that’s incredibly important to their business. People and organizations participate in open source projects for many different reasons. Sometimes it’s because they use the software in-house, sometimes it’s because they like their employees to experience career growth. Many organizations that employ engineers who actively participate in open source projects are not even aware that their employees develop open source software.
Levine consistently asks the wrong questions. He seems to think there’s some contradiction between the interests of privacy advocates and US empire. But once one understands that they both desire secure communications, it is clear how their interests align. He also misinterprets the takedown of Ross Ulbricht (a.k.a Dread Pirate Roberts) of Silk Road fame, claiming that: "The fall of Silk Road pricked Tor’s invincibility."[SV page 398], after explaining that Silk Road had been infiltrated by the FBI and DHS, "almost from the very beginning."[SV page 398] It’s as if Levine doesn’t understand that Tor does not make a website immune to hacking.
I’ll believe anything if given enough evidence of it. I’ll believe that Tor is backdoored by the government if someone can prove it. I’ll also believe that Tor is not backdoored if someone can prove that. These are technical questions and the former is much easier to prove than the latter. But Levine’s methods and general lack of understanding prevent him from engaging in effective analysis.
The Tor Project is not the Internet. While it’s interesting to learn about the personalities in the Internet Freedom space I’m consistently left with the question: “What does this have to do with the Internet?” I think Tor is kinda neat. I’ve played with it a few times and it looks like interesting technology. Perhaps because I never bought into the idea that it was privacy’s savior I’m not all that bothered by it being funded by the US Government. The vast majority of Internet users will never hear of Tor and will never use it, so why make such a big deal about it?
Epilogue
More false equivocation. In the Epilogue Levine travels to a Holocaust concentration camp and talks about how IBM sold tabulation machines to the Nazis. He also waxes a bit about the meaning of all this technology stuff. The Epilogue fails at tying up any of the book’s loose ends and provides no relevance to his arguments about the Internet.
My Conclusion
Surveillance Valley presents a false narrative of Internet history. Some of the more recent topics it covers are factually covered well, but even those suffer from poor analysis. The ARPANET and later the Internet were not created with counterinsurgency in mind. DARPA was engaged in counterinsurgency operations and research while it funded the ARPANET, but Levine presents no evidence that shows how the creators of the ARPANET were concerned with counterinsurgency or intended the militarization of American society.
Throughout, Surveillance Valley suffers from projected causation and false equivalence. Its primary choice of method, follow the money, is unsuitable for resolving its thesis, which causes its analysis to arrive at incorrect conclusions.
Additionally, the book suffers from recentism. Levine’s attempts at drawing an unbroken surveillance lineage between using the ARPANET for information retrieval and using the Internet to surveil its users are falsely equivocated. The NBC news report from 1975 claiming that personal information of US citizens was transferred over the ARPANET bears no relation to Snowden’s revelations of 2013. Instead, it only demonstrates another example of Levine’s deployment of false equivalence to dupe non-technical readers into Surveillance Valley’s false narrative of Internet history.
Levine is most comfortable and most accurate when he is talking about the present. I appreciate his muckraking into the Tor Project even if the conclusions he draws are often wrong. I hope he continues to call out Silicon Valley companies and the US government for their bullshit, and I also hope that he gets better at analyzing the information he uncovers. I recommend he finds others in the Internet and technical communities that share his frustrations with privacy discourse and learns from them. He should spend less time listening to the noise of social media discourse and more time talking with experts. Too much of Surveillance Valley feels like it’s responding to narratives constructed by influencers instead of people who actually know what they’re talking about.
I hope Levine continues to write about technology and that he continues to rake muck about the large corporations and governments that lie to us. I appreciate good investigative journalism when it’s accurate. Like Levine, I’m disgusted with the ubiquity and sophistication of the surveillance we experience using the Internet. We’re constantly being tracked and profiled. Still, neither the ARPANET nor the Internet began with that intent in mind.
This may be difficult to believe from our current perspective. We have a tendency to be very recentist in our analysis of technological trends and that is not helped by the media’s obsession with tech trends and business fashions. But we need to come to grips with this history if we want to have any chance at rescuing ourselves from our current surveillance based dystopia.
If I made any mistakes in this article I would appreciate being informed of them. To contact me please mail me at andrew [at] depht.com.