Most alarmingly, PlanetRisk began seeing evidence of the US military’s own missions in the Locomotive data. Phones would appear at American military installations such as Fort Bragg in North Carolina and MacDill Air Force Base in Tampa, Florida—home of some of the most skilled US special operators with the Joint Special Operations Command and other US Special Operations Command units. They would then transit through third-party countries like Turkey and Canada before eventually arriving in northern Syria, where they were clustering at the abandoned Lafarge cement factory outside the town of Kobane.
It dawned on the PlanetRisk team that these were US special operators converging at an unannounced military facility. Months later, their suspicions would be publicly confirmed; eventually the US government would acknowledge the facility was a forward operating base for personnel deployed in the anti-ISIS campaign.
Even worse, through Locomotive, they were getting data in pretty close to real time. UberMedia’s data was usually updated every 24 hours or so. But sometimes, they saw movement that had occurred as recently as 15 or 30 minutes earlier. Here were some of the best trained special operations units in the world, operating at an unannounced base. Yet their precise, shifting coordinates were showing up in UberMedia’s advertising data. While Locomotive was a closely held project meant for government use, UberMedia’s data was available for purchase by anyone who could come up with a plausible excuse. It wouldn’t be difficult for the Chinese or Russian government to get this kind of data by setting up a shell company with a cover story, just as Mike Yeagley had done.
Initially, PlanetRisk was sampling data country by country, but it didn’t take long for the team to wonder what it would cost to buy the entire world. The sales rep at UberMedia provided the answer: For a few hundred thousand dollars a month, the company would provide a global feed of every phone on earth that the company could collect on. The economics were impressive. For the military and intelligence community, a few hundred thousand a month was essentially a rounding error—in 2020, the intelligence budget was $62.7 billion. Here was a powerful intelligence tool for peanuts.
Locomotive, the first version of which was coded in 2016, blew away Pentagon brass. One government official demanded midway through the demo that the rest of it be conducted inside a SCIF, a secure government facility where classified information could be discussed. The official didn’t understand how or what PlanetRisk was doing but assumed it must be a secret. A PlanetRisk employee at the briefing was mystified. “We were like, well, this is just stuff we’ve seen commercially,” they recall. “We just licensed the data.” After all, how could marketing data be classified?
Government officials were so enthralled by the capability that PlanetRisk was asked to keep Locomotive quiet. It wouldn’t be classified, but the company would be asked to tightly control word of the capability to give the military time to take advantage of public ignorance of this kind of data and turn it into an operational surveillance program.
And the same executive remembered leaving another meeting with a different government official. They were on the elevator together when one official asked, could you figure out who is cheating on their spouse?
Yeah, I guess you could, the PlanetRisk executive answered.
But Mike Yeagley wouldn’t last at PlanetRisk.
As the company looked to turn Locomotive from a demo into a live product, Yeagley started to believe that his employer was taking the wrong approach. It was looking to build a data visualization platform for the government. Yet again, Yeagley thought it would be better to provide the raw data to the government and let them visualize it in any way they choose. Rather than make money off of the number of users inside government that buy a software license, Mike Yeagley wanted to just sell the government the data for a flat fee.
How big is your concern that these constraints will spur China to spin up competitive AI chips?
China has things that are competitive.
Right. This isn’t data-center scale, but the Huawei Mate 60 smartphone that came out last year got some attention for its homegrown 7-nanometer chip.
Really, really good company. They’re limited by whatever semiconductor processing technology they have, but they’ll still be able to build very large systems by aggregating many of those chips together.
How concerned are you in general, though, that China will be able to match the US in generative AI?
The regulation will limit China’s ability to access state-of-the-art technology, which means the Western world, the countries not limited by the export control, will have access to much better technology, which is moving fairly fast. So I think the limitation puts a lot of cost burden on China. You can always, technically, aggregate more of the chipmaking systems to do the job. But it just increases the cost per unit on those. That’s probably the easiest way to think about it.
Does the fact that you’re building compliant chips to keep selling in China affect your relationship with TSMC, Taiwan’s semiconductor pride and joy?
No. A regulation is specific. It’s no different than a speed limit.
You’ve said quite a few times that of the 35,000 components that are in your supercomputer, eight are from TSMC. When I hear that, I think that must be a tiny fraction. Are you downplaying your reliance on TSMC?
No, not at all. Not at all.
So what point are you trying to make with that?
I’m simply emphasizing that in order to build an AI supercomputer, a whole lot of other components are involved. In fact, in our AI supercomputers, just about the entire semiconductor industry partners with us. We already partner very closely with Samsung, SK Hynix, Intel, AMD, Broadcom, Marvell, and so on and so forth. In our AI supercomputers, when we succeed, a whole bunch of companies succeed with us, and we’re delighted by that.
How often do you talk to Morris Chang or Mark Liu at TSMC?
All the time. Continuously. Yeah. Continuously.
What are your conversations like?
These days we talk about advanced packaging, planning for capacity for the coming years, for advanced computing capacity. CoWoS [TSMC’s proprietary method for cramming chip dies and memory modules into a single package] requires new factories, new manufacturing lines, new equipment. So their support is really, really quite important.
I recently had a conversation with a generative-AI-focused CEO. I asked who Nvidia’s competitors might be down the road, and this person suggested Google’s TPU. Other people mention AMD. I imagine it’s not such a binary to you, but who do you see as your biggest competitor? Who keeps you up at night?
Lauren, they all do. The TPU team is extraordinary. The bottom line is, the TPU team is really great, the AWS Trainium team and the AWS Inferentia team are really extraordinary, really excellent. Microsoft has their internal ASIC development that’s ongoing, called Maia. Every cloud service provider in China is building internal chips, and then there’s a whole bunch of startups that are building great chips, as well as existing semiconductor companies. Everybody’s building chips.
A month later, the world saw images of mass graves in the Kyiv suburb of Bucha, dead limbs sticking out of the sand. Outside our building one morning, on an old brick wall that was previously empty, was a fresh message, the paint still wet: “Russians, go home.” My boyfriend went back to Russia so he could obtain a European visa, promising he would be back in a month, but he never returned.
I spent the rest of the year on the move: Cyprus, Estonia, Norway, France, Austria, Hungary, Sweden. I went where I had friends. The independent Russian media that I’d always consumed went into exile too, setting up operations where they could. TV Rain began broadcasting out of Amsterdam. Meduza moved its Russian branch to Europe. The newspaper Novaya Gazeta, cofounded by the Nobel Peace Prize laureate Dmitry Muratov, reopened in Latvia. Farida Rustamova, a former BBC Russia correspondent, fled and launched a Substack called Faridaily, where she began publishing information from Kremlin insiders. Journalists working for the independent news website Important Stories, which published names and photos of Russian soldiers involved in the murder of civilians in a Ukrainian village, went to Czechia. These, along with 247,000 other websites, were blocked at the behest of the Prosecutor General’s Office but remained accessible in Russia through VPNs.
“During the first days of the war, everything was in a fog,” says Ilya Krasilshchik, the former publisher of Meduza, who went on to found Help Desk, which combines news media and a help hotline for those impacted by war. “We felt it our duty to inform people of what the Russian army was doing in Ukraine, to document the hell that despair and powerlessness leave in their wake. But we also wanted to empathize with all of the people caught up in this meat grinder.” Taisiya Bekbulatova, a former special correspondent for Meduza and the founder of the news outlet Holod, tells me, “In nature you find parasites that can force their host to act in the parasite’s own interest, and propaganda, I believe, works in much the same way. That’s why we felt it was our duty to provide people with more information.”
I wanted to continue my work in journalism, but the publications that had fled Russia weren’t hiring. My application for a Latvian humanitarian visa as an independent journalist was rejected, and I didn’t have the means to pay the fees for US or UK talent visas.
In her new memoir, Burn Book, Kara Swisher cites a 2014 profile that dubbed her “Silicon Valley’s Most Feared and Well-Liked Journalist.” She might prefer to downplay the first and emphasize the second. Some people would switch that around. But there is no dispute about Swisher’s impact: When it comes to tech punditry, she’s at the top of the heap.
No tech journalist has built a bigger brand for herself. Her three-decade career is a study in hard work and uncommon confidence. She rose from being a reporter at The Washington Post to The Wall Street Journal’s internet reporter and then, in her biggest leap, the cofounder of the All Things D Conference and website with her revered mentor, tech reviewer Walt Mossberg. In one of their most famous interviews, she and Mossberg moderated a blissfully convivial joint session with lifetime rivals Bill Gates and Steve Jobs in 2007 that brought many in the audience to tears. Swisher and Mossberg left the Journal in 2013 and started the successful Code conference, with Swisher heading a news site. Her interviews can be tough, the most famous being with Mark Zuckerberg in 2010, when he was so rattled by the way Swisher and Mossberg pressed him on privacy that he literally sweated through his hoodie. In addition to interviewing the entire tech CEO pantheon, Swisher has tossed questions at figures in politics and culture—Hillary Clinton, Kim Kardashian, Maria Ressa, and so on. All the while Swisher has broken plenty of news, fueled by her deep sources. In the past few years, she has mastered the podcast medium with two hits—On With Kara Swisher, an interview show, and Pivot, with business professor Scott Galloway—as well as a coveted stint hosting HBO’s Succession podcast. Swisher also had a short, high-profile run as a New York Times op-ed columnist. She’s played herself on Silicon Valley and The Simpsons. Her current affiliations are with Vox and New York magazine, and she is a permanent panelist on The Chris Wallace Show, a CNN Saturday morning talkfest.
Despite the title, Burn Book is less a scorched-earth exposé than a primer for Swisher newbies and those who want to know the tech world from an insider perspective. On her podcasts she loves to riff on the big trouble she’s courting by revealing the skeletons in tech’s closet, but for her regular listeners there’s little in Burn Book that they won’t have already heard. (She explains that the title is a play on her Mean Girls reputation, a reference to the book of rumors written by the movie’s high school bullies, and that the cover shot of her face with her trademark Ray-Bans, a raging inferno reflected in the lenses, is kind of a joke.) In the memoir, Swisher slashes her way through the tech world like John Wick with a word processor, vanquishing vain CEOs and clueless legacy media bosses and emerging without a scratch. Those humbled bros include Elon Musk, a former pal who’s now a nemesis. But unlike Musk, who Swisher says recently declared her an “asshole,” most of the tech world still, well, likes and fears her. Other journalists dream of interviewing the likes of OpenAI CEO Sam Altman. At one stop on Swisher’s book tour, Altman is slated to interview her.
During my afternoon with Swisher at her house in a tony neighborhood in northwest Washington, DC, she took frequent breaks for fond exchanges with three of her four children, her wife Amanda Katz (an editor for The Washington Post), and her ex-wife, Megan Smith, a former US chief technology officer, who dropped in. Our conversation, though, was feisty, as we talked about her storied career, why she abandoned the conference business and The New York Times, and how she answers to the charge that she’s mean.
Steven Levy: What prompted you to write a memoir?
Kara Swisher: I didn’t want to. Jonathan Karp, the publisher of Simon & Schuster, bugged me for years to write something. I was much more interested in the blogs or the podcasts or whatever. I never really liked writing my books. The process was so slow. And I’d had enough of these [tech] people. I don’t like most of them anymore. I didn’t want to reflect on them. I’m sick of them. They’re sick of me. And Walt Mossberg was supposed to write his memoir, right?
The impact on the public sphere has been, to say the least, substantial. In removing so much liability, Section 230 forced a certain sort of business plan into prominence, one based not on uniquely available information from a given service, but on the paid arbitration of access and influence. Thus, we ended up with the deceptively named “advertising” business model—and a whole society thrust into a 24/7 competition for attention. A polarized social media ecosystem. Recommender algorithms that mediate content and optimize for engagement. We have learned that humans are most engaged, at least from an algorithm’s point of view, by rapid-fire emotions related to fight-or-flight responses and other high-stakes interactions. In enabling the privatization of the public square, Section 230 has inadvertently rendered impossible deliberation between citizens who are supposed to be equal before the law. Perverse incentives promote cranky speech, which effectively suppresses thoughtful speech.
And then there is the economic imbalance. Internet platforms that rely on Section 230 tend to harvest personal data for their business goals without appropriate compensation. Even when data ought to be protected or prohibited by copyright or some other method, Section 230 often effectively places the onus on the violated party through the requirement of takedown notices. That switch in the order of events related to liability is comparable to the difference between opt-in and opt-out in privacy. It might seem like a technicality, but it is actually a massive difference that produces substantial harms. For example, workers in information-related industries such as local news have seen stark declines in economic success and prestige. Section 230 makes a world of data dignity functionally impossible.
To date, content moderation has too often been beholden to the quest for attention and engagement, regularly disregarding the stated corporate terms of service. Rules are often bent to maximize engagement through inflammation, which can mean doing harm to personal and societal well-being. The excuse is that this is not censorship, but is it really not? Arbitrary rules, doxing practices, and cancel culture have led to something hard to distinguish from censorship for the sober and well-meaning. At the same time, the amplification of incendiary free speech for bad actors encourages mob rule. All of this takes place under Section 230’s liability shield, which effectively gives tech companies carte blanche for a short-sighted version of self-serving behavior. Disdain for these companies—which found a way to be more than carriers, and yet not publishers—is the only thing everyone in America seems to agree on now.
Trading a known for an unknown is always terrifying, especially for those with the most to lose. Since at least some of Section 230’s network effects were anticipated at its inception, it should have had a sunset clause. It did not. Rather than focusing exclusively on the disruption that axing 26 words would spawn, it is useful to consider potential positive effects. When we imagine a post-230 world, we discover something surprising: a world of hope and renewal worth inhabiting.
In one sense, it’s already happening. Certain companies are taking steps on their own, right now, toward a post-230 future. YouTube, for instance, is diligently building alternative income streams to advertising, and top creators are getting more options for earning. Together, these voluntary moves suggest a different, more publisher-like self-concept. YouTube is ready for the post-230 era, it would seem. (On the other hand, a company like X, which leans hard into 230, has been destroying its value with astonishing velocity.) Plus, there have always been exceptions to Section 230. For instance, if someone enters private information, there are laws to protect it in some cases. That means dating websites, say, have the option of charging fees instead of relying on a 230-style business model. The existence of these exceptions suggests that more examples would appear in a post-230 world.