In the Review’s September 21, 2023, issue, Ben Tarnoff reviews two recent books on the history of Silicon Valley, where “money begets money with an ease that would make Andrew Carnegie weep.” His essay ends with a reflection on the “neurotic character” of most of the region’s capitalists, who champion liberal parties and politics with one hand and preserve “a profoundly unequal social structure” with the other.
Tarnoff’s first two books were about, respectively, pre–twentieth century American counterfeiters and Civil War–era San Francisco writers. For the past seven years, his preoccupation with money and power in northern California has been channeled into writing about the business of tech. His most recent book, Internet for the People (2022), is a history of how the Internet was privatized and a manifesto for turning it into a public good. Meaningful “democratic decision-making can’t take place,” he argued in its opening pages, “so long as the Internet is controlled by firms that are compelled to prioritize profit-making as a matter of survival.”
We e-mailed this past weekend about ChatGPT, recent advances and setbacks in tech worker organizing, and Marxism’s “large and fractious family.”
Max Nelson: How did you start writing about the tech industry?
Ben Tarnoff: The short answer is I needed a job, so I started working in tech. I’ve always had an interest, and some proficiency, in technology, so it wasn’t too big of a stretch. By then—around 2015—I had spent many years working as a writer and an editor, including a stint as an editorial intern at The New York Review. I had published two books about American history. But I had begun to feel aimless, isolated. I felt like I was just drifting from one thing to the next.
Tech gave me a place to write from. I found that I liked the work. But, just as importantly, it rooted me in a world. And that rootedness helped me develop something like an intellectual project, or at least a set of themes that made my writing feel more lively and coherent.
In your latest book, you argue that many of the problems critics identify on the Internet today—algorithmic bias, privacy encroachments, digital redlining—can be traced back to the moment in the 1990s when government policymakers decided to turn the Internet over to private corporations. How did that moment make it possible for Silicon Valley to become what in your essay you call “the most efficient capital-accumulation machine in history”?
Silicon Valley is a creature of public investment. Without federal funding, the technologies that form the basis of what we now call the tech industry wouldn’t exist. This is not a new or controversial point, but it bears repeating, because the industry—or, to be more precise, those authorized to speak on its behalf—tends to have a short and selective memory.
The Internet is an especially stark example of the industry’s reliance on government-led innovation. It took billions of dollars of public money and decades of public management to create. Only after the foundations of the technology were in place and its viability was thoroughly demonstrated did industry step in, and the government stepped back. Even then, much more work had to be done to make the Internet we know today. In the mid-1990s, it was a specialized tool used mostly by researchers. Turning it into a mass medium was going to take effort and imagination.
There were many open questions about how precisely to popularize the technology. Unfortunately, the lopsided terms of privatization, which empowered industry to unilaterally determine the future of the Internet and retained no role for the public sector that had bankrolled the invention of the technology, ensured that the transition would be guided wholly by commercial imperatives. We are now dealing with the legacy of that decision. An Internet dominated by the profit motive is also an Internet that violates our privacy, amplifies right-wing propaganda, and intensifies various kinds of social inequalities.
In your essay last year about Twitter, you noted that “the financial floodplains of Silicon Valley” have been “drying up” since the Federal Reserve started raising interest rates. What longer-term effects do you think these shifting economic conditions will have on an industry so thoroughly organized around risky investment—“a gambler’s faith,” as you put it in your new review? How could these changes affect the terrain of struggle for tech workers?
It’s a complicated picture. On the one hand, the higher cost of capital and the post-Covid slowdown in revenue growth have taken a toll. Valuations, both public and private, are down. Hundreds of thousands of employees have been laid off. On the other hand, we’re seeing a mini-boom around generative AI, prompted by the release of ChatGPT. AI-related startups are getting funded at a fearsome rate. Nvidia, which makes so-called AI chips, is doing insane business. And the AI frenzy has helped push the Nasdaq up 32 percent so far this year. (By comparison, it was down 25 percent this time last year.)
The tech industry is structured for groupthink. It needs a perpetual supply of shiny new paradigms to impel new waves of investment. Generative AI seems to be the latest one. Still, there is more substance to generative AI than to cryptocurrency or the metaverse, which were the most recent candidates for the next new thing. Crypto never managed to become anything more than a casino, and the metaverse barely exists. By contrast, generative AI has all sorts of obvious use cases. It also has a “wow factor” that helps get the blood and capital flowing. ChatGPT, Midjourney, and the like may not signal the imminent arrival of an artificial superintelligence—I have my doubts—but they do demonstrate real technical sophistication of the kind that we didn’t see in, say, NFT apes.
As for what this shift means for worker struggles, it is important to note that generative AI is a labor-intensive technology. The data on which AI systems are trained needs to be sorted and labeled by humans. The outputs also need to be evaluated, especially to prevent AI interfaces from engaging in racist outbursts or other undesirable behavior. Most of the people who do this work are poorly paid; some are starting to organize. In May, a group of more than 150 Nairobi-based workers who helped to create ChatGPT and who maintain Facebook and TikTok voted to form a union. This is the sort of struggle that deserves the close and careful attention of journalists and organizers. AI tends to be presented in popular discourse as quite abstract, but at root it’s profoundly industrial, and the working conditions of the people turning its cranks can be downright Dickensian.
There’s been something of a boom recently in Marx-influenced histories of the tech industry written for popular readerships, like Shoshana Zuboff’s The Age of Surveillance Capitalism and Malcolm Harris’s Palo Alto—the latter of which is one of the two books under review in your latest piece. Having yourself written a book that roughly fits that description, how has the Marxist tradition helped you think about the recent history of Silicon Valley?
The Marxist tradition is a large and fractious family. Marx was a restless, kaleidoscopic thinker who was forever revising his views, and whose work contains innumerable gaps, ambiguities, and contradictions. This makes him terribly fun to read. It also, I think, makes him more useful. There are infinite Marxes, a Marx for every occasion. So while I welcome the infusion of more Marx into popular writing about technology, and am delighted that a book as unflinchingly Marxist as Palo Alto has found a wide audience, I’m not sure I could neatly summarize his influence. Marxism is best thought of as a set of questions, not answers.
Speaking for myself, what I have probably found most valuable in Marx is the idea that history has a structure. That is, history is not a purely contingent sequence of events that can be expressed as the sum of individual choices. Instead, there are objective constraints on human action, constraints that arise from the particular society into which we are born and that are largely, though not exclusively, determined by its economic practices—which is to say, how that society produces the things that people need.
This materialist way of thinking is a good solvent for stripping away the mystifications of the tech industry, which likes to narrate its success as a story of exceptional individuals. A materialist approach directs our attention away from the personal qualities of a figure like Steve Jobs and toward the broader social forces that helped develop his particular drives and demons, and that ultimately endowed them with such historical importance. But one must be careful: it’s easy to get so intoxicated with structural explanations that human beings vanish from the analysis. And that’s a problem, because people do have agency, and they can act in all sorts of strange and unpredictable ways. Ideas also have a life of their own. The hard part is figuring out how all of these elements come together. (It’s different each time.) To say that history has a structure is not to say that its structure is simple!
Three years ago you wrote a pamphlet about the history, composition, and prospects of the tech worker movement. What major changes in the movement’s direction have you observed since then? Where do you see its most promising expressions now?
I’m not convinced that anything like a “tech worker movement” still exists. During the Trump years there was a series of high-profile collective actions by white-collar tech workers, especially at big firms like Google. A growing number of people in high-status professional jobs were coming to see themselves as workers. They were using the techniques of worker organizing to demand greater control over how their workplaces were run and what kind of work they did. They were also, in small but significant ways, building solidarity with the large pool of subcontracted and contingent workers who do everything from running data centers to cooking food for Silicon Valley office cafeterias.
That’s not happening at a meaningful scale in 2023. My sense is that a shift took place around 2020. Management fired some crucial organizers; several others burned themselves out. But most importantly, Trump left the White House. The surge of white-collar tech worker organizing had a lot to do with the elevated level of social mobilization that we saw during Trump’s presidency. During those years many people became politicized for the first time. They joined left-wing organizations, participated in protests, started reading groups. Biden’s election sapped much of that energy. Without the existential threat of Trump, political passions cooled.
White-collar tech worker organizing hasn’t stopped, but it’s mostly moved to smaller shops. There are now around a dozen white-collar tech worker unions in the US. Organizers have partnered with national unions such as CWA and OPEIU to help them navigate the arduous road to legal recognition. There have been significant victories, such as the unionization of more than six hundred software engineers, designers, product managers, and data analysts at The New York Times—the biggest bargaining unit of white-collar tech workers in the country by a wide margin. But the current organizing environment doesn’t seem capable of sustaining the intensity of the Trump period. When everyone was feeling political, it was easier to start a conversation about how to build a broad movement to remake the industry from below.
In sum, I would say that from 2017 to 2020, white-collar tech worker organizing had the dimensions, and the aspirations, of a movement. In the years since, it has evolved into a (very junior) member of organized labor. There have certainly been advantages to building real unions with real contracts—it marks a level of maturity that would’ve been unthinkable in 2017. It’s also, inevitably, less exciting. The emphasis falls on things you can win at the bargaining table, like salary floors and agreements about remote work. Bigger questions that are hard to pose within the strictures of US labor law—control over work, solidarity among different kinds of tech workers—recede. This is not through any fault of the organizers themselves; it simply reflects the less propitious conditions in which they are now forced to organize. Of course, there’s always next time.