Note to (non-paying) readers:

A while back, I decided I was going to stop using paywalls. That’s been working out exactly as I expected: my stress levels have gone down a bit, my posting has become more erratic and infrequent, and my revenue has been cut almost in half. All of that is fine with me, but it would also be fine with me — in fact, I’d love it — if you chose to become a paid subscriber anyhow. And if enough of you do, then maybe I can keep the content un-paywalled but also keep up a more regular publishing schedule. That’s the dream, anyway.

—LW

Subscribe now

In late October 2001, while the entire country reeled in the shell shock of 9/11, Congress passed the USA PATRIOT Act. For a few weeks after the fall of the Twin Towers, a moment of collective grief, introspection, and solidarity had seized the nation, but by now it had largely passed, and the mood had turned toward jingoism, paranoia, and thirst for vengeance. The legislation passed easily, with only a single dissenting vote in the Senate.

But even in the context of that bipartisan consensus, the legislation was shocking. Under its surveillance powers, the government could mass collect Americans’ phone and internet metadata, access your medical, banking, and phone records without a warrant, and even find out what library books you’d checked out.

The law was sold as an emergency measure to pursue the conspirators behind the 9/11 attacks. Many of its provisions were subject to Congressional reauthorization four years later. But as everyone knew, those sunset provisions meant very little. Congress might tinker with reforms around the edges, but it was never going to repeal the bulk of the law. Today, 25 years later, there’s a member of Congress who was in preschool when 9/11 was perpetrated, but the legal surveillance infrastructure the PATRIOT Act put in place remains.

Passage of the PATRIOT Act birthed the political and legal apparatus that grew into the gargantuan NSA domestic spying machine exposed by Edward Snowden — the real “deep state.” Snowden’s revelations led to some significant reforms, such as the end of the direct collection of telephone metadata by the government. But again, the apparatus remains in place, and an entire new industry has emerged to operationalize it, led by Palantir.

I suspect that very few people seriously believe that there’s meaningful oversight of domestic surveillance by the U.S. government — that, on the whole, our Congressmembers are jealous of guarding our privacy, antagonistic toward the surveillance industry, and sufficiently skeptical of the White House’s routine invocations of national security. For the most part, we’ve just gotten used to it. Yes, of course the government is spying on us. Of course they could freeze our bank accounts to punish us for dissent, as has already happened in Canada. Of course facial recognition could be deployed by federal agents to punish Americans for expressing dissent on social media, like the owner of the New York Knicks is already doing.

I suspect that, on an intuitive level, most Americans understand the way this stuff works: if a technology emerges that the state can deploy to consolidate its power over the public, the state will seek to use it. If the government overreaches, then Congress may make a show of putting some constraints in place while voters are still paying attention, but those constraints will be limited and, with time, they will erode in the face of new technological and legal workarounds. Authoritarianism will continue to expand, sometimes in leaps, but more typically in small increments.

I say all this as a prelude to posing the question that is the title of this post: Why does it matter if China dominates AI?

The idea that if we don’t plow forward on AI advancement as freely and recklessly as possible then China will do it first is by now so hackneyed that our brains barely detect the massive assumptions underlying it. One of those assumptions is that the Chinese government will have some kind of unilateral control over whatever they create, or in any case more control than we do, which is curious since it’s China that is building its AIs open source and American companies that keep their LLMs proprietary. Another assumption is that China’s military power will quickly achieve parity with ours, which is probably true, but a bit puzzling to hear from those who have pushed the administration to relieve the export controls on Nvidia chips to China that are designed to prevent precisely that outcome.

The main thrust of the argument, however, seems to be just the general idea that China is autocratic and we’re democratic, and you really don’t want such dangerous and powerful technologies falling into the hands of dictators. Who knows what they’ll do with such power?

I have a lengthy piece that will come out close to America’s 250th birthday making a detailed historical case that American democracy has always depended on certain material conditions, and that those conditions no longer prevail. But for now, suffice it to say that it’s foolish to assume, even if you believe the questionable proposition that America does in fact remain a democracy, that it will continue to be one in an age of AI. Judging by the recent history of mass government surveillance in the United States, it’s safer to assume that the more AI advances, the less free we will become, whether you live in China, the United States, Switzerland or Brazil. Not only will we be subject to novel, Black Mirror-style regimes of monitoring and control, but as human labor diminishes in value, the public will lose our ability to bargain with the institutions that rule over us over such things as our civil liberties.

It’s funny that there’s a trend in the tech industry to name their surveillance products and companies after made-up words from Lord of the Rings, like Palantir and Anduril. The lesson from that whole series, which is not in the least bit subtle, is that whoever gets the Ring of Power becomes corrupt and evil, no matter how virtuous their initial intentions. People like Alex Karp claiming that the end of the world will come if China beats us to AGI is a lot like Gandalf telling Frodo he needs to wear the ring so that Sauron won’t put it on first. The tech right are a lot of things, but stupid isn’t one of them. My guess is that they’re well aware of this literary irony. Their appropriation of Tolkien is a winky-noddy admission that they literally do want to be Sauron. That’s the whole point. Either the CCP will become Sauron or the American tech oligarchs will, and Silicon Valley is mustering their orcs to make sure it’s their own.