中美关系：历史、文化与 AI 竞争——Noah Smith，来自 Econ 102

We have to get out of the mentality of thinking that America makes all the choices, that we have all the power, that we have all the agency, that we're the big one who makes all the choices. We're not the big one anymore. They're bigger than us in most of the important ways. We're big—we're still big—but they're bigger than us in most important ways.

Today I'm beginning what I expect will be an ongoing series of episodes exploring one of the most consequential and fraught issues of our time: U.S.–China relations. My guide today is economist Noah Smith, author of the Noahpinion blog and co-host of the Econ 102 podcast, also on the Turpentine Network.

There's an old joke about a visitor to China: after a week, they think they know something about China; after a month, they believe they know a lot about China; but after a year, they realize they know nothing about China. I mention this because I approach this topic with genuine humility, considering the role that fear of China and Chinese AI plays in American AI discourse today, and considering the many foreign policy mistakes the U.S. government has made in recent decades.

I feel this topic is too important to ignore, and also too complicated to uncritically accept the Washington consensus. But, to be very clear, and to channel the language of a historical Red Scare, I do not now, nor do I ever expect to, have all of the answers.

With that in mind, I consider this a sort of warm-up episode, which I hope will serve as useful context for a number of additional conversations going forward. I think Noah is a really good person to provide such perspective, as he's a super well-read, well-traveled—having lived in Asia for a few years—and extremely open-minded thinker who's willing to challenge prevailing orthodoxies.

I started off by asking Noah to sketch out the historical arc of U.S.–China relations, and I appreciated how his analysis balanced short-term realism—noting that China, and arguably Xi Jinping specifically, was first to defect from a more positive relationship—with a longer-term perspective, observing not only that the U.S. and China have been allies in the past, but that there are no major fundamental reasons the 2 countries couldn't conceivably repair their relationship in the future.

In fact, contrary to anyone who would frame the U.S.–China rivalry as some sort of inevitable or eternal clash of civilizations, Noah also notes that despite our obviously different histories and political systems, there are a number of surprising cultural similarities shared between our 2 peoples.

Along the way, we also get Noah's takes on some of the most important AI-related questions being debated today. What are the U.S. export controls on semiconductor manufacturing equipment and chips themselves ultimately meant to accomplish? How much are they changing China's calculus when it comes to potentially acting on Taiwan, and what will their practical effects be?

Can we expect to deny China the ability to militarize AI technology or meaningfully slow its progress toward AGI? Or will we perhaps simply deny the Chinese public the benefit of AI productivity improvements and economic growth?

I appreciate that Noah answered as many of these questions as he felt confident answering, and also gave his standard fallback “no opinion” answer to a few as well. This is, of course, far from the final word on these topics, and we should all be striving to learn and understand both the other side and the overall situation better as we go forward through time.

What's up, dude?

Hey, guys. How's it going?

Great. Great to be with you again.

Good to see you.

Here's kind of what I thought I would do: a quick preface for you, and then we can jump into it. Basically, I have never quite been able to figure out why the United States, the West, whatever, is so bent on this sort of path of escalation and keeping China down. I'm not even quite sure what exactly it is we think we're trying to do with our export controls.

I've been voicing doubt about this in various places, and somebody just recently tagged me online and said that Robert Wright and I are the only 2 China doves in the world, which is not literally true. But I feel like I'm way out of my depth here in terms of analyzing China, and I also feel like I haven't really been able to get a good, from-the-ground-up, explain-it-like-I'm-5 understanding of the most basic question: Why are we here?

Are there any alternatives that we have? What I would love to hear is sort of the steelman, mainstream case for why we need to do what we're currently doing.

The reason I care, in addition to the fact that I would like the Chinese public to have economic growth just like I would like us to have economic growth, is that it feels to me like a pattern of escalation, weaponization, whatever, is not good for the overall big-picture AI safety picture that I also worry about.

Well, okay, a couple of questions before I do that. Number 1: Are we talking mostly about AI, or are we talking just sort of about the general geopolitical thing? How focused are you asking about here?

I'm very focused on AI, but I do think the broader context is important. For example, I don't personally see DeepSeek as a threat to America at all. It just doesn't seem like a threatening thing to me. I don't understand why people are saying, “China just defeated America,” and so on. I don't get it.

At the same time, I think the threat of a war over Taiwan is big and real. I think a lot of people think of U.S.–China competition as some sort of vibe—who's winning? Some people in China release an LLM that's pretty good and cheap, and suddenly it's, “Wow, they're winning.”

I've never seen things that way. In fact, I'm going to write a post on DeepSeek later today, against my better judgment. Basically, I don't understand why that LLM is a threat to us, and so I don't think our policy should be focused around trying to stop China from developing things like DeepSeek.

In other words, I don't have a good explanation for why we should worry about that, be freaked out, and try to kneecap China's AI ability in that way. I don't put you on the hook for representing the establishment too much, but at the same time, I worry about Taiwan and other wars quite a lot, as well as things that fall short of war, like cyberattacks. There are a lot of concrete things to worry about that are not an LLM being cheap.

Can I start way back at the very beginning?

Go ahead. We should definitely get to the Chinese—or the Taiwan—issue. That's kind of where I came into it, because it's like, “Chip ban—wait, doesn't this make it easier for them to justify an attack on Taiwan?”

But let's go back to the start. Start at the very beginning, if you would. How do you—and maybe abstracting away a little bit from current personalities, because I think we've got singular personalities at the top of both countries right now—but if we forget about personalities, egos, and ideologies, what would you say is the natural strategic relationship between the United States, this giant landmass on one side of the world, and China, this giant landmass and population on the opposite side of the world? Is there any reason for us to be rivals?

There's a question of whether I think we should be rivals in the ideal world, and whether I think there are natural forces pushing us toward rivalry that are unfortunately there. The answer to the first question is no. We shouldn't be.

The United States and China were de facto allies for most of the 20th century, with the exception of a brief period in which they fought a war against each other. Other than the Korean War—which was stupid on both sides and entailed big mistakes on both sides—that relationship was rectified when Nixon, Kissinger, and these people made the move to de facto ally with China against the Soviet Union.

China helped us win the Cold War, or we helped China win the Cold War. We helped China economically integrate into the world. We helped China throw off Japan. We helped China avoid conquest by European powers in the early 20th century, which is not something a lot of people know.

With the Open Door policy, America essentially stretched the Monroe Doctrine and applied it to China. We said that none of you Europeans should try to conquer China. Don't do that.

So, with the exception of the time that we fought a war, America and China were in close cooperation throughout the 20th century, if not in actual alliance—which we were several times. There's no reason we can't be again. That was good. That was fine. The 20th century went well. A lot of people died, but I think overall it turned out well, and things went in a good direction, with some bumps in which a few tens of millions of people died.

Mao was really terrible. He was sort of a bump in the America–China relationship. Culturally, I think America and China are very similar to each other. If you look at how Chinese people live and how they think about the world, it's not very different from America—or at least not from what America classically has been.

American culture has changed a lot recently due to social media and things, but that classic American culture of “I'm going to get rich, but I don't know how,” being obsessed with real estate, driving to the mall, or liking hip-hop but also being slightly racist—I could go on and on about the ways in which Chinese culture and American culture are fundamentally similar to each other.

They're both large countries. China is bigger than us, but we're both pretty big, and we're bigger than the surrounding countries. These things create natural similarities between us and China.

I think there's a popular myth that countries and civilizations that are similar to each other don't fight, or have less reason to fight, because they're similar. Well, no. As my little sister likes to put it, “Fight for niches.”

The fact that China and the United States are so similar means that, in some ways, we occupy the same niche, both in our own minds—“I'm the big country. I should be in charge”—and in the world. CCTV versus Fox News, I guess.

Even China's economy looks a lot like our economy did 100 years ago: massive manufacturing overcapacity, sort of a fast follower in everything, and innovation on a small scale rather than at the basic research scale. That was us. That was early-20th-century America.

We had some big inventions, like the airplane, obviously, but in cars, for example, we were fast followers. In machinery, we were fast followers. Those were the guys inventing the stuff—Britain, Germany, and France—and we followed them.

China is following a similar pattern to us. We dipped our toes in the waters of imperialism with the Philippines and the Spanish-American War, and China looks like it may do that in Taiwan, while trying to grab some territory from India and throwing its weight around.

It's a big, rich country. “Guess who's rich now? Ha. We're going to throw our weight around. You can't boss us around anymore, foreign powers.” That was us.

I think our similarity naturally pushes us toward rivalry. That doesn't mean we have to fight a World War III, but I think this is something we have to deal with.

It's easy to say that we should just be nice, back down, and not do it. But all your life, you grew up in a world where America was the dominant power and could basically choose international relations. We could have chosen not to do the Iraq War if we wanted. We could have chosen not to do the War on Terror. We could have chosen all these things.

China gets a vote. If you watch Dwarkesh Patel's podcast with Sarah Paine, they're just really excellent, and I recommend watching it. She's said that we can talk about what our approach to China should be all day long, but at the end of the day, China has an approach to us, and we have very limited ability to control that.

Being nice to China and trying to do everything we can to facilitate its power and economic strength is something we did for decades upon decades. We did that. That's a thing we already did. There's a general consensus now that it was a failure.

I think China turned around and bit us—you know, bit the hand that fed them—in some way. I think that consensus is a little overstated. You did see some real liberalization in China during the Jiang Zemin and Hu Jintao years of the 1990s and 2000s. You went from Mao having a personality cult and slaughtering everybody to Mao impersonators being normalized during the Hu Jintao era, and you had years of peace and prosperity.

I don't think that was as bad as people say. Chinese people deserve not to be poor, and they're not poor anymore. There's still some poverty in China, obviously, but overall, they're not a poor country anymore at all. They developed. They're a developed country, and I'm not going to say that was bad.

I don't think the consensus in America now is that this was bad, that we shouldn't have done it, and that we should have avoided it, let them stay poor, shut them out of the international trading system, and so on. I don't think so.

I think it's better for 1 billion people not to be poor. At the end of the day, poverty is the ultimate enemy of the human race. We've got to conquer it. Then we can solve all our other problems. Absolute poverty—the kind where people don't have enough to eat—is the ultimate enemy of humans. I wrote an essay about this called “The Elemental Foe,” which I recommend—my own thing.

However, I think China's leadership has transitioned. China's system for selecting leaders is bad. Deng Xiaoping made a huge mistake when he crushed the Tiananmen Square protests instead of accommodating them, because he cemented a leadership system that selects leaders through essentially backroom power dealings.

What that means is that eventually you'll get a leader who is extremely good at backroom power dealings, to the detriment of other things. That's what you've gotten with Xi Jinping, and you got that pretty fast, actually.

This is not popularly known, but Deng Xiaoping handpicked Jiang Zemin and Hu Jintao to be the next 2 leaders of China. After that, he said, “You all are on your own, because I can't pick forever.”

As soon as they were on their own, there was a big political knife fight. A guy named Xi Jinping, who was extremely good at getting his cronies in everywhere and dominating the Chinese Communist Party, knifed Bo Xilai and Zhou Yongkang—and I'm probably mispronouncing his name.

Xi was this great political infighter who basically dominated China. He's not the most hawkish guy in China, but he's pretty hawkish, and he definitely sees America as a foe, as an enemy, despite engagement and despite our decade-long policy of engagement.

It was China that defected from that cooperation first, when Xi Jinping came to power in the early 2010s. Before that, we had a cooperative relationship with China under Hu Jintao. We had differences over Taiwan, and they had an authoritarian government, but we had a fundamentally cooperative relationship.

Xi Jinping decided to change this. He decided—we didn't make that decision—that America is China's rival. China has to topple the American-led global order, as he saw it.

Rush Doshi is a very good analyst, one of the most preternaturally calm human beings you'll ever see. He has attention surplus disorder, and he went through decades of Xi Jinping's communiqués, speeches, and so on, summarizing them in a book called The Long Game: China's Grand Strategy to Displace American Order.

It's one of the driest books you'll ever read. It's very hard to get through. But if you think that's hard to get through, try getting through reams and oceans of Chinese Communist Party documents.

Doshi basically said, “Look, Xi Jinping has a paradigm.” There was a paradigm shift in Chinese thinking under Xi Jinping, which was not entirely caused by Xi Jinping himself. There are more hawkish people than him somewhere in the Chinese government, but Xi Jinping is the guy who cemented this into a new idea and created Chinese Communist Party slogans for it, which they love.

His slogan is “Great changes unseen in a century,” which basically means that the current global order is an American imperium, and this will be overthrown by China. China will overthrow that imperium and replace America.

That's how Xi Jinping thinks. That's not how Noah Smith thinks. That's not how people should think. But it's how Xi Jinping thinks, and it's pretty clear that he thinks that.

China gets a vote. We don't get to decide unilaterally the relationship between the United States and China. We just don't. People who think we do get to unilaterally decide that are coping. They're clutching for the feeling of control, this illusion of control that they always felt like they had before.

In the 1990s, the U.S.–China relationship was entirely determined by the United States. In the 2000s, it was 85% determined by the United States. Now it's less than half. China is leading the dance. They're in the driver's seat. That's a fundamental fact, and I think we have to acknowledge it.

Xi Jinping's attitude shift—his huge attitude shift—toward the United States predates our attitude shift. It predates Trump, predates Biden, predates everything. Back in 2013, the only person saying that China was our rival was Chuck Schumer, this old dude nobody was listening to.

That consensus shifted very quickly, and Xi Jinping was the first to shift it. He was the first.

Economic decoupling between the United States and China has been proceeding since the beginning of the Xi Jinping era. He's tried to remove all Chinese dependence on U.S. economic inputs—chips, machinery, airplanes, everything—making China totally economically self-sufficient.

That was something he did before the United States implemented export controls, tariffs, or any of this stuff. The narrative that the United States just woke up one day and decided to kneecap China and keep it down is wrong.

Xi Jinping woke up one day and decided that China was going to overthrow the U.S.-led global order. We're basically reacting to that now. I say “basically” because Trump watched a lot of CNN back in the 1990s and decided that America was losing the industrial race against Japan. Then he transferred that anxiety—which was mostly misplaced at the time—onto China mentally.

He said, “How come we have this big trade deficit with them? We've got to fix it.” Trump had some agency here, and some of this was already going on. It wasn't entirely Xi Jinping, but it was mostly him who shifted this relationship.

I think we have to respond to that. We're in a position we're not used to being in. We're used to being the big one—the one who acts while everyone else reacts. Now we're reacting. We're not the big one anymore, and I think that's a mentality, a mindset, we've got to get used to.

So what do you think is the threat in concrete terms? Because the word “overthrow”—what could actually happen? What’s the bad thing? Do they want to be part of some international reserve-currency bundle, or what do they want?

No, that’s not important. They actually don’t want to be the reserve currency. China should have one of the reserve currencies at this point. The RMB should be one of the dominant reserve currencies on the planet, but they don’t want that.

What China wants is, number 1, control of the seas. Trade goes over the ocean in physical goods, and trade in digital goods goes over internet cables. China wants to control those things. They would like to control the lanes of global trade.

You see this in the South China Sea, where they’ve basically said, “This is our maritime territory. Your ships may not pass here until we say so.” That’s an area through which much of the world’s trade flows.

That’s different from the U.S.-led global order. We said, “Here’s the United Nations Convention on the Law of the Sea. Anyone can send their ships through these waters. It’s international waters.” We have rules about how you own the waters next to your coast, and then the international waters are free.

The United States Navy, which was overwhelmingly the most powerful navy until recently, was the policeman of the seas. We made sure that there was freedom of the seas, that global trade could continue, and that the internet could continue. Much of the internet was basically United States-controlled, and we enforced freedom of the internet. We said, “You can send your data over these cables. We built these cables. You can send your data over them. Go for it.”

Of course, China built the Great Firewall, but the international internet cables through which digital trade happens—all services trade and so forth—were free because the United States said so. Freedom of the seas happened because the United States said so.

That created the trading regime that, ironically, produced the big American trade deficits that Trump dislikes. But it also created the trading regime that allowed the world to enjoy an unprecedented period of growth. It tied China to the United States in important ways and made us mutually interdependent, and that was why we did it.

We believed that if we bound the world together in the mutual interdependence of commerce, that would prevent conflicts and wars. After World War II, our priority was preventing conflicts and wars.

China would like to have sovereign control over extended areas of the ocean, where the Chinese Navy decides which ships may pass and which may not. Much of the United States’ economy depends on trade that passes through those waters. Much of the United States’ economy similarly depends on digital trade that passes over the internet cables that China would also like to control. They’re going around cutting some internet cables, anyway. That’s a preview of what’s to come, I think.

When you talk about the U.S.-led global order, I don’t think you’re talking about the reserve currency. I don’t think you’re talking about the IMF, the World Bank, or things like that. I think you’re talking about trade, and you’re talking about the lanes of trade: the sea and the internet. Those are the things that China would like to control.

There are some other harms, some other bad things, that Xi Jinping would like to do to the United States—ways that he would like to hurt the United States in order to make us less of a potential threat to Chinese power in the world. One of those is obviously to cut us off from trade and from control of our own trade.

A second thing is the destabilization of our society. The more we end up fighting over pronouns, AI safety, police defunding, or whatever we’re fighting about this week, the more our society becomes disunited and unable to oppose whatever China wants to do.

Chinese power is optionality. Power’s option value. Power doesn’t mean, “I do a thing.” Power means, “I can do a thing if I want.” That’s what power means. It’s option value.

China wants the option value of being able to do whatever it wants. As long as the United States is a big, rich, powerful country, we get a vote. They get a vote, but we get a vote too. China would like to change that.

When I say China would like this stuff, I don’t mean that your average Chinese person on the street would like it. I mean Xi Jinping and his regime, who now dominate the Chinese Communist Party, would like us to be weaker simply so that we don’t get a vote in the future.

If you think that this is not realistic, we were generally a much nicer country than China, and yet we also did the same thing to all our rivals. We weakened Germany and Japan after World War II, and we weakened Russia after the Cold War, so that they didn’t get a vote. China wants to do this to us a bit more because they’re a bit meaner than we are.

China would like to destabilize our society. I don’t like that because I like living in a stable society. I grew up in a more stable society than the one I live in now, and I kind of liked that.

Whatever you get on the internet and argue about, whatever flame wars you have, whatever cultural thing you think is a threat to you—China would like you to have more of those flame wars. Xi Jinping would like you to feel more of that threat from your fellow Americans and from your neighbors, because the more we fight each other, the less option value we have to stop China from doing whatever it feels like doing, today or tomorrow.

I think China has made concerted attempts to destabilize our society.

Can you be more concrete on that? Are we talking about TikTok on the scale of its algorithm, or what?

Yes, TikTok’s algorithm is absolutely part of it. Influence operations are another part of it. But the biggest part of it is actually called sharp power, which is that the Chinese party-state has basically gone around using economic leverage to force people to say and do various different things.

So far, it has used that pretty narrowly to protect its perceived interests over Taiwan and Xinjiang, and to promote its own stuff. It used its sharp power to stop the NBA from standing with the Hong Kong protesters, and all that stuff. But sharp power can also be used to favor actors that destabilize American society—partisan shouters and culture warriors—and to weaken actors that want to unite and oppose anything China wants to do.

If there’s someone in America who says, “No, guys, we’ve got to unite to stop China from closing off freedom of the seas, or taking Taiwan, or taking part of Japan or India. We’ve got to stop them from controlling our supply of critical minerals. We’ve got to stop them from doing these other things,” China can use sharp power and control of online algorithms to silence those people—the people who would have us unite—so that the voices that would divide us become the only voices left.

That’s a threat.

I hear that, and I’m still not super convinced. It doesn’t sound like my grandchildren are going to be speaking Chinese.

They’re not, nor does China want them to.

So it seems like the threat is sort of marginal.

No. No one wants you to speak Chinese, because that would mean that you could pollute China’s own thought ecosystem with your—

I was wondering how long I was going to be welcome on RedNote, where I just popped in for a minute.

They don’t want you. They want you to not matter, to be at the periphery of the world, where you don’t matter. They want you to go fight your little battles over whatever parochial thing you fight over, and not bother them. They want you to have no power to stop them from doing anything they want.

That’s what they want. They don’t want you to speak Chinese. When I say “they,” I don’t mean your average Chinese person. I mean Xi Jinping and his cronies. That’s their image of power.

Their image of power is that you become Argentina. To America, Argentina is not even on the map. They don’t matter. Malaysia is fine; I like it. Argentina occasionally provides us with entertainment and occasionally provides us with supplies of raw minerals, but they don’t matter. They’re a nonentity to us geopolitically. We don’t have to think about Argentina. Argentina doesn’t get a vote on anything.

They want you to be Argentina to them. For the way you’ve grown up and the things you’re used to, becoming Argentina is bad. If you lived in Argentina, it would be worse than where you live now. That’s a threat, whether you choose to accept, believe in, and resist that threat or not. The threat exists.

I mean, you pick a country. I happen to know Argentina better than most countries, and when I think about Argentina, I feel like their problems are largely the result of mismanagement over many successive generations of leadership. We may have a similar problem, but I don’t think Argentina as a nation can point to somebody else and say, “You did this to us.” I don’t feel like if we end up entering hyperinflation or whatever, we’ll be able to point to China and say, “You did this to us.”

There are other things that haven’t been mentioned, right? They steal intellectual property from American companies, and we could go down a reasonably long list of grievances one by one. But I kind of feel like everything that we’re accusing them of doing—or most things—we’re kind of doing too.

We’re bullying Colombia and threatening them. They won’t take some symbolic thing that doesn’t matter, just like Daryl Morey standing up for the Hong Kong protesters or whatever. We’ve got one plane full of dudes going to Colombia, and we’re going to threaten to sever the relationship over it until they back down. China’s doing that to our private actors, and we’re doing it to other states.

Your point about the similarity definitely resonates with me. If I bring this to the AI moment, I’m like, “Okay, they want to be more powerful.” I would also say they are quite powerful. Whether we want them to be or not, as you said, they have a vote. They’re producing a lot of frontier AI research these days. They’re an advanced country capable of frontier-advancing work.

What are we trying to accomplish right now? The main point of escalation seems to be the chip export controls, and now increasingly an AI arms race, or even, as Alexandr Wang from Scale AI recently put it in a paper about the AI war, something going back a few years.

My understanding of why we were trying to deny China these chips was that we thought they were going to militarize them and we wanted to keep them off the frontier. Now we fast-forward to today, and we’re in this DeepSeek-plus-eight-days moment. It seems like they’re staying at the frontier of AI, or very close.

They’re not quite at the same level as the very best American companies, but they’re better than any of the top 3 American companies. DeepSeek is maybe number 4 in the world, maybe number 5, depending on what else you count. You could count Meta, Tesla, or xAI.

We’re not keeping them from reaching the frontier. It seems like they probably have enough chip capacity domestically to manufacture whatever drones they want to make.

I don’t think DeepSeek is a threatening thing. It doesn’t seem that way. Maybe I haven’t thought about it enough, and I’m just not thinking about it correctly, but I don’t see that specifically as a threat.

So what do you think we’re trying to accomplish? What is the most coherent account of what the export controls are supposed to do?

The purpose of export controls is to maintain a military technological edge over China.

First, let’s talk about the military balance. We have nuclear weapons, and China has nuclear weapons. As long as no one figures out how to hack and cancel the nuclear weapons—which we always try to do—or detects the ballistic-missile submarines, which is another way you could potentially eliminate second-strike capability, neither country is under existential threat from a World War II-style war.

But that also means that in order to stop from losing a war, we would have to destroy the world. That’s the kind of escalation you’ll only do when the alternative is that you die anyway.

Nuclear weapons are only a defense against existential destruction. They’re not a tool of crisis management. Vladimir Putin tries to use them that way, but it doesn’t really work. You can’t say, “Let me have that little chunk of territory, or I’ll nuke the world.” No, you won’t destroy the world over that little chunk of land. You won’t destroy the world over a trade treaty, either. You’ll destroy the world if you’re going to die anyway.

That’s the only time you’ll destroy the world, if you’re even in the ballpark of a rational actor. You don’t even have to be rational, but if you have an ounce of self-preservation instinct, you won’t destroy the world unless you’re about to get killed anyway.

Nuclear weapons do not create a balance of power by themselves. You need conventional weapons as well. John Mearsheimer wrote the whole thing about conventional deterrence, and he was right about this. That was back in the days when John Mearsheimer was right about things, before he got sort of old and kooky. Everyone gets old and kooky, and it’s going to happen to you and me as well.

Our conventional deterrence against China—in other words, if China wants to take over a U.S. ally and we want to stop them from doing that—is very weak. If China has this massive production capacity, it can produce a bazillion missiles, a bazillion drones, and so forth. To offset that, we can’t completely rely on technological advantages. We have to produce things ourselves and get our allies to produce more things.

But we can partially offset it by maintaining a military technological advantage. They have more stuff; we have better stuff. That helps us. It doesn’t solve the whole problem, but it’s one thing that helps.

There are many aspects to maintaining a military technological advantage, but one of them is precision weaponry. Have you heard of the Third Offset?

Starting in the 1980s, and really accelerating after that, we got precision weapons—weapons that could find a target and hit the target instead of just throwing something into the general area. Precision missiles, precision drones, precision everything. That’s how you fight now.

The reason this is powerful has to do with inverse-square laws and inverse-cube laws in physics. A blast or shock wave dissipates according to an inverse-cube law, and heat and radiant light dissipate according to an inverse-square law. That means that if you hit something accurately, you can be much more effective.

In fact, in terms of ground warfare, we realized that precision warfare was more effective than using tactical nuclear weapons. Tactical nuclear weapons are big, but the inverse-square and inverse-cube laws are so powerful that precision weapons can actually give you more power than tactical nukes on the battlefield, in terms of destroying enemy vehicles, command posts, and things like that.

Precision weapons have become a cat-and-mouse game involving electronic jamming, hiding from things, GPS, sensors, computer vision, and so on. America has had the best precision weapons for a long time. That’s a military technological advantage that can partially cancel out greater production.

You see this on the battlefields of Ukraine. American precision weapons have been holding their own against Russian mass, even though the Russians fire 8 times as many artillery shells as the Ukrainians. The Ukrainians get more kills because they have American precision weapons.

There have been big fights over jamming—adaptations about how to hide things and how to do these things. We’ve seen that on the battlefields of Ukraine. It won’t solve the entire problem of underproduction, but it will partially solve it. It’s an offset. It will offset some of their power.

In order to maintain that military technological advantage in precision weaponry, it helps to have better chips, because chips become your onboard computer when you have a drone using computer vision to look for things.

Suppose you have a drone controlled by a remote operator. The drone is using radio waves to talk to the operator, who has a little joystick and is sending the drone around. It’s called an FPV drone. It’s a toy you can buy anywhere.

GPS is not a toy, and you can jam it. You can send out a radio wave. You can have a big radar device that sends out a large radio-wave signal to jam the drone’s connection to GPS, which means it loses its connection to the human operator. Then the drone just goes, “Where am I? I can’t find the thing,” or it simply falls out of the sky.

That’s the game people are playing in Ukraine. The way you beat that is to make the drone not need a signal. You have the drone use computer vision so it can look around and say, “There’s that tree. I see a Russian hiding under that tree.” Then it gets them, with no need for an operator.

Now we’re talking about the future of warfare as autonomous drone swarms. For that, you need lots and lots of chips. Once drones start fighting other drones, it becomes an arms race over who has the better chips on their drones.

Export controls are one way—not a sufficient way, obviously, but one of a number of ways—that we’re trying to maintain our edge in that future fight. The fight may not come, but if we would definitely lose such a fight, that takes away our power. It takes away our optionality. It allows China to do whatever it wants: conquer Japan, conquer India, maybe grab some of India’s territory, and reduce India to a minor power. China could probably conquer Japan and South Korea.

All of that means that we can’t stop China if it wants to do those things. Whether it then does them depends entirely on circumstances. We lose our vote.

The chip export controls have been effective, by the way. They didn’t stop China from developing LLMs, but they stopped China from manufacturing the best chips. They really have been effective in that.

How effective that is militarily, we don’t actually know, because we’re talking about weapons that haven’t been created yet. Autonomous drone swarms exist in research labs, but they don’t exist on the battlefield yet.

How effective those export controls will be at stopping autonomous drone swarms and making America’s autonomous drone swarms better than China’s is a question I don’t know the answer to. I suspect no one on the planet knows the answer, although someone out there knows better than me and isn’t saying.

The purpose of the export controls was not to prevent something like DeepSeek. Maybe someone hoped they would, but that was never the stated purpose. The purpose of the export controls was to stop China from being able to mass-manufacture leading-edge chips. That is not enough to maintain a military advantage over China, but it’s one piece of it.

But we’re also not selling them chips, right? I mean, there are multiple levels to the export controls.

That’s exactly right. But frankly, that part of the export controls is not going to work. The part where we don’t sell them the chips will fail.

So you think we can slow the development of China’s domestic chip industry effectively, but we can’t keep chips writ large out of their hands? That seems plausible to me as well.

Yes. In fact, had I been in charge of the export controls—which no one thinks to make me in charge of these things because I’m a random blogger who sits around with his pet rabbit—I would have focused not on chips. I would have kept selling them all possible chips, and I would have focused only on machinery, tools, software, and other things where we actually have a reasonable choke point.

I don’t know why they focused on the chips, and I don’t know why they added those controls. I don’t think it’s going to do us a huge amount of harm. I don’t think it’s going to be a big deal. I think what’s going to happen is that slightly-off-frontier Nvidia chips are going to be just as good as the frontier Nvidia chips, and we’re just going to keep selling those.

I think it’s cosmetic. I think that part of the export controls is, for all intents and purposes, useless. Maybe it serves as a distraction, but I think the equipment export controls are effective. They’re already proving extremely effective. The software and personnel export controls are effective too.

The most effective part of the export controls isn’t even the machinery. It’s that we said, “If you’re an American citizen, you can’t work in the Chinese chip industry.” We’ll find you, put you in jail, or whatever. Talent is everything. Basically, we put a stop to a lot of talent going over to China’s side.

It’s not a complete stop, because all these things are porous to some degree. You can’t build a complete wall. I don’t know whether the very complex export controls released in the last days of the Biden administration—which designate all these third countries and so forth—will apply. I don’t know whether they’ll actually work.

They seemed rushed and a little too overreaching. The stuff we were doing before was probably working better than people realized.

It’s like a planned economy, almost.

Yes, and I think that’s going to need to be reevaluated. But the export controls over machinery were effective. China tried to pull this stunt by making a 7-nanometer chip, which they turned out not to be able to manufacture in any appreciable volume. Their costs were really high, their reliability was low, and the quality of those chips was terrible. They’re still dependent on TSMC.

That stuff works, and it will slow them down. It’s just one part of an overall strategy to maintain conventional deterrence over China for 10 years.

After 10 years, the threat of conventional war will probably diminish. It won’t go away, but I think the next 10 years are the peak risk. As Hal Brands has put it, we’re in the 1950s again in terms of the threat of war, like we were with the USSR—except now we’re more in the position of the USSR, because they’re the ones with the manufacturing capacity, while we’re the ones with the ideology we proselytize.

Our goal should be to freeze everything and deter China from launching a giant World War III for 10 years. Export controls are not sufficient to do that by any means, but they’re one piece of how we do it.

I think Xi Jinping is not going to last much longer after that. He’ll be 80 by the end of that period, and his fourth term will be up. The bellicosity and the dream of overthrowing the U.S.-led order that he has will diminish after 10 years, just as Soviet expansionism and aggression diminished a lot after Stalin’s death.

There was a little more aggression under Brezhnev, and there were some nuclear scares, but we were never in as much danger of fighting the Soviet Union as we were in the 1950s. World War III came very close. The Cuban Missile Crisis was the peak danger of a nuclear launch, but in terms of conventional war, it wasn’t even close to the peak.

We need 10 years of deterrence. That’s what we need. It will make the situation a lot better. Then we can work on repairing that relationship with China and going back to being friends, as we were.

Remember that we had a 15-year rupture in our relations with China under Mao—maybe 20 years—and we repaired it. We patched it up, and we’ll patch it up again someday. We’ll be friends with China again someday.

But unilaterally dropping deterrence and weakening all our deterrence methods is not the way to patch that up more quickly. Establishing deterrence and stability is the quickest path to repairing our relations with China.

Let me ask a couple of questions. We don’t have too much more time, so I want to spend the remaining time on this critical period. I agree that it’s a critical period for AI reasons, if not for plenty of other reasons.

It seems to me that we’re on a pretty risky path. We’ve got China and Taiwan. Taiwan, obviously, is the object of their desire, and it’s where the chips are made. Cutting China off from the chips made in Taiwan lowers the barrier to attacking Taiwan. If they’re not getting the best stuff that’s made there, it makes it that much easier to attack. You could even say that’s a reason to attack: “We’re getting the stuff from there, and they’re not.”

We’ve now got both sides announcing multihundred-billion-dollar build-out plans. Most concerningly to me, we’ve gotten multiple Western AI leaders—Alexandr Wang from Scale AI and Dario Amodei from Anthropic—saying that our plan should be to build a giant global alliance, apparently under Trump’s second term, to box out the autocracies, primarily China and Russia.

Eventually, we’ll make them an offer they can’t refuse. He doesn’t use exactly those words, but that’s basically how I read the essay: We’re going to come around and say, “Okay, we’ve cut you off, we’ve advanced our technology, and we have overwhelming AI superiority. Here are the terms on which you get to be part of the AI-enabled world order.”

That seems kind of crazy to me. I don’t know if you think it’s crazy, but it seems like it increases their incentive to do something crazy sooner rather than later.

You might be right, and you might be wrong. I’m not well placed to evaluate this because I don’t understand the technology well enough. Again, I don’t see DeepSeek as a big threat to America, but I could be wrong because I might not understand how these things work well enough.

I don’t know. Let me talk to those guys, see what they’re thinking, read their stuff, and think a little more about this. Then I’ll get back to you. I’ve had a lot of time to think about precision weaponry, the effect of export controls on things like drones and missiles, the war in Ukraine, and conventional deterrence with respect to Taiwan and the whole thing I just gave you.

What you just talked about with the whole AI-supremacy thing is funny, because my friend Grace wanted to have this big discussion with me about it. I said, “Grace, I don’t know. My blog is called Noahpinion for a reason.” I have plenty of opinions, but the set of things on which I don’t have opinions is vastly larger than the set of things on which I do have opinions.

I don’t have a strong opinion on this one way or another. I don’t know if what these guys are talking about makes any sense. I’d like to talk to Alexandr Wang. I can’t talk to Dario Amodei—he came on our podcast before—but we can talk to him about this. In fact, we should probably just get him back to talk about it, and maybe you could even debate him.

I’ll ask him about it the next time I see him, but I don’t know. It sounds hokey to me. The idea that we could maintain AI supremacy in the LLM field, at the country level or the company level, sounds unrealistic. I think DeepSeek probably shows that it’s unrealistic.

LLMs are a borderless, commoditized, universal thing. All the AI I was thinking about in terms of chips was autonomous onboard compute for autonomous drone swarms and precision weaponry. That’s a very different kind of AI. It’s not even the same kind of model: computer vision, convolutional neural networks, and things like that.

Maybe if you used an LLM to say, “Hey, robot, go kill the Russians. Find a Russian and kill them,” maybe you could do that. But that’s icing on the cake—slightly better control interfaces.

I know that locky groom is very interested in using LLMs for voice-controlled robots. Maybe that’s useful, but I don’t know how LLMs are going to fight wars. Are we going to say, “Hey, Claude, tell me how to kick the Chinese’s ass”?

When I was a college kid, I wrote a silly short story about an AI war between the United States and China, between American and Chinese AGIs. After achieving sentience, the U.S. and Chinese AGIs both decided it would be better not to fight, but to be stoner kids—to watch old movies, get stoned, and just chill.

The human minders, whose job was to make the AGIs fight each other, traded romantic advice with the AGIs in exchange for collaboration. In exchange for advice on their romantic relationships, the human minders overlooked the fact that the U.S. and Chinese AGIs had become friends and were not fighting each other.

That was after the spy-plane incident. I don’t know if you’re old enough to remember that. There was a spy plane—or some kind of radar plane; I didn’t understand things in those days—that got shot down. There was a brief ratcheting-up of U.S.-China tensions, and I was thinking about this story.

I don’t have any idea how LLMs will figure into the military balance. I don’t understand their military applications, and I don’t know if anyone does. Looking at how things are going, trying to maintain supremacy in LLMs at a company or country level probably isn’t going to work. But again, I’m not an expert. I’m just throwing ideas out there.

The model starts to get very AGI- or even ASI-like, with superintelligence. Then the path seems to be: We’ll have this superintelligence, it will figure out how to make the best drones, it will figure out how to hack their electrical grid better than they can hack ours, and we’ll have them cornered on all these fronts.

But the time window in which that’s supposed to happen is very short.

Maybe. I can’t sit here and say that’s obviously wrong. If I said that, I’d be overstepping my knowledge. I can’t tell you they’re wrong about that, but I also can’t tell you they’re right about it. I don’t even know if it sounds legitimate.

It doesn’t sound impossible, given the trajectory we seem to be on. But the time delta between when we have those things and when everybody else has those things does seem like it could be fairly short.

Eric, what do you think? You’ve been so silent this whole time. What do you think about this? Help me out here.

I don’t know what I’m talking about. I wish I could help. These things are beyond my pay grade to some degree, but I think it’s a good call for us to bring on more experts in the space, as you were saying before.

Yeah. At a minimum, this has been a good foundation. I think more time should be spent in the American discourse just getting clear on what our complaints are and what the threat from China is.

It’s not nothing, but that part of the conversation is so often skipped over into, “Are the chip controls working or not?” I feel like we’ve so often skipped the question of what we were trying to do, whether that has morphed during the time the policy has been in effect, and whether there’s a lot of post hoc justification going on.

There’s also a lot of confusion because DeepSeek has definitely burst onto the scene and confused a lot of people. I’m with you: It’s hard to look at something open source that looks more like a peace offering to me than a threat.

Noah, to that point on the peace offering, have you read Xi’s speeches, where he’s very direct about his goals for China? Are you familiar with their own rhetoric about how they conceive of the United States? Do you take them literally?

I’ve read very little, so I wouldn’t even pass myself off as a hobbyist in this domain, certainly far from an expert. But since you asked, I did happen to read one, because I’ve been asking everybody, “What am I missing that should make me afraid of China to the point where I want to do all these things?”

I think most of these things probably won’t work, except to keep beneficial AI out of the hands of the Chinese public. That seems like the biggest effect we’re likely to have.

That won’t happen.

I hope not. There’s no way that will happen. That’s impossible. The only person who could do that is Xi Jinping, and he won’t do that. Maybe a little bit, but he mostly won’t. In any case, they’ll have all the AI they can buy.

Yeah, well, then I guess I still don’t know what we’re hoping to accomplish, other than that we just seem to be escalating. But I read one speech that was chosen to try to persuade me, and it felt like a mirror-image sort of speech. It was mostly domestically focused, and it included an explicit claim from Xi, which you could say is a lie or whatever, that China is not seeking to export its model to the rest of the world.

It said that China wants to resist and overcome containment, which was the translated word, I think, for what they perceive us to be doing to them. But it didn’t seem particularly warlike, hateful, or scary. When it was sent to me, I expected it to sound a lot more frightening than it ultimately did in translation.

The guys who were scary-sounding were the Wolf Warrior diplomats. Starting in 2019, but really in 2020, Xi Jinping put these guys—named after some silly action movie—in charge of the Ministry of Foreign Affairs. They went around saying a bunch of very bellicose things: “You are failing. We are right. We will bury you,” and so on.

They sounded like Kaiser Wilhelm in 1913. You can read what these people said. Just Google “Wolf Warrior diplomats.” It was incredibly aggressive.

Xi Jinping himself is not that aggressive, but I would urge you to read the book The Long Game by Rush Doshi. It’s not an alarmist book. It’s one of the driest books you’ll ever read, but it is relentless and well argued. Rush Doshi understands Chinese strategic thinking better than you or I will by the time of our deaths. He already understands it better.

If you want to see examples of actual Chinese bellicosity, look up the Wolf Warriors and read some of the things they said a few years ago. But in addition to the words, I would urge you to look at the actual actions China has been taking.

The hacks of critical infrastructure have been happening daily—not just privacy invasions, like listening to your phone calls over your cell phone or reading your text messages, but penetrating critical infrastructure. Look at some of the things China has been doing by cutting internet cables around the world. That’s a real action. Cutting internet cables is a real thing that you can do.

The actions are bellicose in addition to the words. China is actually doing aggressive things. Look at its territorial incursions in the Philippines and some of the things it has done toward India along the border—slowly taking territory. There’s a lot of aggression happening out there that isn’t just words.

Will that blossom into the conquest of the world? Maybe not. Our job is to stop it and minimize the probability that it happens.

World War I taught us that you shouldn’t go to war, that you should hold your fire and not be an idiot. World War II taught us the opposite: Holding your fire is appeasement, and when you have an enemy that relentlessly wants to take over more and more and will never stop, you shouldn’t appease it. That just makes it harder to fight later on.

Which is the right lesson with regard to modern China? I don’t know. It may be some mix of the two, or some third thing altogether.

The Cold War taught us that deterrence works. There were fights in Vietnam and Korea, but at the end of the day we didn’t nuke the world. We didn’t have a big conventional war against the Soviet Union. We didn’t pull out the tanks, helicopters, and everything else for the big one. It didn’t happen.

All through that time, the world got more peaceful and stable—not uniformly or monotonically, but the Cold War was a success despite all the dumb mistakes we made and the horrible things that happened.

The Cold War was a bigger success than World War II. In World War II, we beat the bad guys and won, but 55 million people died. That sucked.

With the Cold War, we learned from World War II. We didn’t appease; we deterred and contained. Most of all, we deterred the Soviet Union from military adventurism. After Stalin died, after that pivotal decade, the danger went way down.

That’s a success we can learn from. I think the Cold War—not World War I or World War II—is the best model for interaction with China. Despite the lack of an ideological component—we’re not ideologically at war with China—I think the Cold War is still the best model.

Deterrence is still the best option, because ultimately that’s how our relations improve in the future: We stabilize the world. Only in a stable world can China and the United States become friends again, as we traditionally were.

That’s my thesis. I do have to go, but that’s my final thought.

That’s a scary one, because I expect pretty choppy waters ahead for multiple reasons.