Two weeks ago, America thought it was leading the AI race. Then out of nowhere, an unknown Chinese start-up turned the American stock market—and that assumption—on its head. DeepSeek, a Chinese company founded less than two years ago, released a free AI chatbot that rivals the most advanced available open AI products. And they did it despite America’s prohibition on shipping our most advanced microchips to China. America was caught flat-footed, asking how did this happen? And could we actually lose this tech war? Now, if your understanding of computers stops at the term hard drive, don’t worry. Today on Honestly, Bari has two incredible guests, experts on both AI and China, who are going to break it all down for you. Tyler Cowen is an economics professor, AI expert, and a must-read writer at his blog, Marginal Revolution. He is joined today by Geoffrey Cane, an expert on China and the author of The Perfect Police State: An Undercover Odyssey Into China’s Terrifying Surveillance Dystopia of the Future. Today, how this happened and what it means. And can America win the AI war with China? Header 6: The Free Press earns a commission from any purchases made through all book links in this article. If you liked what you heard from Honestly, the best way to support us is to go to TheFP.com and become a Free Press subscriber today. Learn more about your ad choices. Visit megaphone.fm/adchoices
From the Free Press, this is Honestly, and I'm Barry Weiss. Two weeks ago, America thought it was leading the AI race. Then, out of nowhere, a totally unknown Chinese startup turned the American stock market and that assumption on its head.
Yeah, John, we've got a bit of a tech sell-off this morning, and it's being caused by earth-shattering developments in the AI space. So let's take a look at this. Did you see the Dow? Down more than 140 points. On track for one of its worst days in the past two years. And here's why.
DeepSeek, a Chinese company founded less than two years ago, released a free AI chatbot that rivals the most advanced available open AI products.
Let's talk about DeepSeek because it is mind-blowing and it is shaking this entire industry to its core.
And they did it despite America's prohibition on shipping our most advanced microchips to China.
Buzz started growing about DeepSeek's latest AI model being more efficient while running on a lot less, a lot less advanced NVIDIA chips. Released last week, it's already being called what is a major breakthrough, as the app shows. It's work, it's reasoning.
I mean, this tech route and massive market sell-off, it's a wake-up call. It is for me.
America was caught flat-footed, asking, how did this even happen? And could we actually lose this tech war? Now, if your understanding of computers stops at the word hard drive, don't worry. I have two incredible guests, experts on both AI and China, who are going to break it all down for you.
Tyler Cowen is an economics professor, an AI expert, and a must-read writer at his blog Marginal Revolution. He is the author of over a dozen books, but frankly, I've lost count. He is joined today by Jeffrey Cain, an expert on China and the author of The Perfect Police State, an undercover odyssey into China's terrifying surveillance dystopia of the future.
We talk about how this happened, what it means. Is it a 21st century Sputnik moment? And if so, is China now officially ahead of us in the AI race? Stay with us. Tyler Cowen and Jeffrey Cain, welcome to Honestly.
Happy to be here. Good to be here, Barry.
Okay, the biggest news in the world of AI dropped, at least from my perspective, out of nowhere last week. And that news did not come out of California. It didn't come out of Silicon Valley. It came out of China. Tyler Cowen, tell us about DeepSeek, what it is, and why it matters.
DeepSeek is online. It is free. It is a very fast model. I would say it's the most fun model to speak with. It is more arbitrary and subjective and personality-rich than the other AI models. It's also open source, which means anyone can copy it. So there's no way of stopping it from spreading, even if you shut down that original website somehow.
So it's not as good as the best open AI products, but it is changing the world. And it's probably what a lot of the world is going to use.
You seem to have been using DeepSeek before it was even a headline in the news. Did you somehow get early access to this product?
No, I was using it before the R1 update came out. But I'm in a lot of chat groups. And if there's something new in the world of AI, odds are I'm going to hear about it. So I was just playing around with it. And it was so much fun. I kept on going.
When I wrote you about it, here's what you emailed me back. DeepSeek is not so PC like the American programs, though, of course, it is packed with Chinese restrictions, which we'll talk about in a second. But it's really fun to chat with, has lots of personality, a kind of Barry Weiss LLM. I guess I should be flattered by that.
But tell me about the personality and tell me a little bit about the restrictions.
Well, it will take on any personality you tell it to, which is part of the fun. But the default is simply that it's charming. It's more poetic. It writes better fiction, probably, than the American models. It's just a whole different vibe, you could say.
Now, if you try to ask it questions about Chinese politics, Tiananmen Square, the Uyghurs, it either A, won't answer, or sometimes it starts to answer A, And then the cursor starts moving backwards and it erases whatever it wrote down. And a very literal act of Orwellian censorship.
Wow. Jeff, do you want to add a little bit to that before we talk about the implications of it? Have you yourself been using it?
I have been using it and I'm also a big user of AI. I do enjoy using these chatbots and seeing what comes of them. DeepSeek, it is a step forward in AI technology, but the problem here is that this is about much more than AI itself. This is companies in China going to an old playbook that they've been using for a while. against the United States and its companies.
And this playbook is taking technologies that already exist where the U.S. and its allies are on the leading edge and refining them, trying to find ways to execute them better, to incrementally improve them, to make them cheaper, easier to use, and then releasing them on the mass market, in this case, in an open source way.
So this is an example of China using that old playbook and trying to catch up and overtake U.S. technology and When I say this, I'm talking about so much more than just Gen I. This is only one example of this technology leaping forward.
Okay, well, the major headline and the reason that this has caused such a stir – and this was the surface headline. We'll get into whether or not it was true – It appeared as if a small Chinese startup surpassed what we believed until that moment was sort of the bleeding edge of this technology and that they did it for $6 million over the course of about two months.
Now, for comparison, OpenAI reportedly spent $100 million training its last model. It has 20 times the number of employees. The numbers that we're talking about here just were a fraction of what we consider to be par for the course here in the United States.
Now, the question is whether or not that's actually true, whether or not they were actually able to do this on such a small budget and in a very short amount of time. So, Tyler, what's the truth there? Where did the headline get it wrong?
That's mostly not true. So it may be the case that the final training cost was very, very low. But you incur a lot of other costs in building the model. For instance, the company had 150 people working full time trying to develop the model. And their salaries and all of the fixed costs of the enterprise are not included in that sum.
Now, it's still probably the case that their model is considerably cheaper. And some of that is that wages in China are lower. But they also had a number of innovations that just enabled the model to be done for a lower cost. So it is a breakthrough. But we will borrow from what they did, just as they have been borrowing from what we've done.
And keep in mind, the best products today are still the OpenAI products. and anthropics clawed, and those probably have what you would call a clearer path to the basket of making yet further improvements. So it has not surpassed us, no.
Well, there are a lot of theories floating around about how DeepSeek pulled this off. The company claims that they did this without what until now we had all viewed as the sort of crucial component, which are these advanced chips. Do you guys buy DeepSeek's claims here that they built this advanced AI without the chips?
Or did they find a way around the export control that had been set up by the Biden White House? Jeff, you want to take that one?
So there is evidence that they did get chips from NVIDIA that may have been smuggled through Singapore. This is according to one government report. And then on top of that, there's also evidence that surfaced that's been published in the Financial Times, which shows that they may have trained their chatbot, their Gen AI with ChatGPT with an OpenAI product.
Now, this would actually be against the license agreement agreement.
of chat gpt so they would have been breaking the rules in that case but on top of that the other problem here is that chat gpt and most american social media and ai websites are not available in china they're banned ostensibly because of censorship concerns the chinese government wants to be able to control what these ai can say about tiananmen square and the uighur genocide and so forth and so we now have a situation where china is making a breakthrough but the us can't even compete with china on its own turf
So that's the problem here. It's the borrowing of American ideas, of American technology, the imitation of it, but then also shutting out American companies from a vast consumer market that is China.
Just to underline the point, you don't think there's any chance here that they really, truly innovate it. You're saying that it's all basically a sort of rehash or a steal of what we've already created.
I would say it's a breakthrough in the sense that they can do it cheaper. They certainly need less computing power, as we call it in the field, compute. But I wouldn't call it a major breakthrough. And as Tyler said, they haven't overtaken American technology. They haven't leapt forward above open AI and what we're doing in America.
But they are catching up fast, and they're finding ways to do it cheaper, faster, and in more imitative ways that could be very damaging one day to American industry.
I think it's a bigger breakthrough than that. They had numerous algorithmic improvements. One estimate is that the cost of what they did is about 30x lower. Now, we don't know as to what chips they had, but we do know that before chip restrictions were placed on China, the company bought some very high-quality chips that are almost as good as the ones that are now banned from China.
So I think they certainly could have done what they did without illegally acquiring chips. But whether or not they did, I would just say we don't know.
I was texting with Neil Ferguson the day that this news came out, the historian and the free press columnist, and his basic take was a sentence, technological containment never works. Tyler, do you agree with that?
The word never makes me nervous, especially from a historian, but I mostly agree. This technology is going to spread. All the, oh, we need a six-month pause, do this, do that. It's not going to work. I'm not opposed to the chip restrictions on China. If I were President Biden, I would have done the same.
But I think we need to be realistic and understand they're only going to hold back China for a short period of time. And we need to be ready for the fact that they will likely be our peer in AI technology. And we'll move into a new world state where there's a kind of mutually assured destruction with these AIs. And it will be us and China.
And the key question is, what will the rest of the world use? I hope it's our products. Right now, DeepSeek is entirely accessible, totally free, open source. How the company will manage to fund that is a huge question. And we will see. But I think it's important that, say, as we go to Africa three years from now, I would strongly prefer they're using American AI.
So one of the things that happened, I think the day or within a few days of DeepSeek sweeping the world, at least it felt like that the day that it came out, was that NVIDIA stock tanked by like 17 percent, indicating that the market certainly thought that they did this without the chips. Is that how you guys understood that news? Yeah.
Well, the way I understood that news was that people might have assumed that, yes, they did it without the chips or they didn't need as much compute to create the AI. But that didn't really come as a surprise to me because there are a number of hedge funds, a number of these Wall Street funds that have been looking at AI recently and have been shorting. some of these stocks.
There are questions over, you know, does the earth even have enough natural resources and energy and compute to allow the next generation models to even exist in the future? I mean, 10 years down the line, are we going to run into the barrier, the barrier that does not allow us to proceed further with this AI technology, which is so energy intensive.
Personally, I think that we will find ways to make the computing needs less and less over time. I don't think that this is going to be a problem in the end. And I'm more of an advocate for staying ahead of the curve. I do think that we need a lot more government investment in the chip sector. We need a lot more drive in that area, the reshoring of American manufacturing.
So this is the problem where we stand now. We're now in a situation where compute is going down, where these chatbots are easier to make. And as Tyler said, the tech will spread. It's hard to contain. It can't really be contained. But we can do what we can to stay at the leading edge to ensure that the world hopefully one day is using American AI instead of Chinese AI.
There's some chance this is all good news for America. So if the game is simply who can throw up the most energy infrastructure as quickly as possible, I would predict we're likely to lose to China. And all this talk about big deals with UAE, with the Saudis, is an attempt to cope with the fact that we don't build energy supply quickly enough in our own country.
To the extent it's about algorithmic advances and you don't need as much brute force compute, that at least possibly is a plus for the United States.
Well, one of the assumptions that it seemed was shattered last week is this idea that largeness and scale meant being ahead, right? It seemed like the assumption or the consensus before this deep-seek innovation was that the large tech companies thought they were uniquely able to be ahead because of their size and their power.
And then you have this little small player like DeepSeek sort of turn that on its head. And it reminded me a little bit of, you know, a comparison between, let's say, SpaceX and NASA. For decades, we all thought only NASA could put a rocket into orbit. And then SpaceX, smaller, more agile, was able to do it. And now NASA seems more reliant on SpaceX.
So it makes sense for me, guys, of what this is. means in terms of the consensus about scale and whether or not that's actually the advantage?
So we do live in an age of technological democratization. We're seeing this with SpaceX. SpaceX, by the way, still a very resource-intensive company. It's not as if their investments are small by any means. We're seeing it in AI. We're seeing it in all kinds of technologies. Another big area where there have been massive advances, and especially in military technology, is drones. We saw a
cheap homemade drones coming out in Ukraine, some of them made in Turkey. And, you know, these are drones that anybody can snap a grenade on and you can, you know, fly it over a tank column and unleash weaponry destroy tank column.
So what this means in the big picture is that the traditional forms of industrial organization, of infrastructure, of warfare, military defense are potentially going to crumble in our age. We're seeing a new age where, you know, anybody can get their hands on these technologies. Anybody can innovate with them, find ways to make them cheaper and less intensive ways.
But then the question is, how do you respond to all that? And it might simply be that chips are getting smaller and smaller and smaller. And what that means is that we're starting to hit some of the limits into what we can do in the physical realm with chips.
But we still have to find ways to stay ahead of that curve because the way to respond to these innovations to ensure that we're safe at home is to find ways to respond to these innovations. So for example, if there are drones attacking tank columns, making tanks obsolete. We need to find a way to respond to those drones. And that's not necessarily going to be a cheaper way of doing things.
We have to stay ahead of that curve.
Keep in mind, scale may still win out in China. So the major Chinese tech companies, say ByteDance, they all have their own AI models. No one's talking about them because of the Matthew effect.
Explain what the Matthew effect is, Tyler.
Well, once people start talking about something, everyone keeps on talking about it, right? It's like the Barry Weiss effect, we could call it. It's so crazy. But it's true. So the other Chinese models could win out. Keep in mind, the deep-seek people, they're not in with the Chinese state. It seems the CCP was as surprised as anyone. They're not part of the crony capitalist machine.
Politically, they could be the result of a clampdown. So I wouldn't necessarily bet against the larger Chinese companies because the communist system is set up to favor insiders.
Tyler, the speed of these innovations from companies like OpenAI is astonishing. I was just about to say that O1 was OpenAI's latest model, but you've corrected me and said that actually deep research came out yesterday. And by the time this drops on Thursday, who knows what will be out. But about O1, you wrote it was the smartest publicly issued knowledge entity that the human race has created.
Pretty unbelievable. Why haven't we been talking about that? Why is the thing that everyone's been talking about for the past week DeepSeek?
DeepSeek is more cool. And keep in mind, O1 Pro costs $200 a month. And it writes in a very objective style, which is good for scholarship, good for objectivity, but it's not as fun to play with for most people as DeepSeq. And so many of the advantages of O1 Pro, they're so sophisticated. Like, say, you could beat me in an economics test, but most people don't even see that difference.
So it doesn't appear smarter for most uses, even though it's clearly the better product.
Why is DeepSeq so much more fun?
I think a lot of it is a tone thing. The reinforcement learning, less was applied to it, so it slips more easily into different persona and characters. That's what most people want from their large language models, not better expertise than Tyler Cowen at economics.
Well, I want to talk a little bit about whether or not we're going to need PhDs at all in the future, including economists. But one thing that's really come out into sharp relief with the advent of DeepSeek is this question about open sourcing. Despite OpenAI's name and despite the way it began, which was the intention was for it to be open sourced, it is now closed source.
And Sam Altman has since said that he regrets deciding to make OpenAI closed source. I want you guys to think about a listener, for example, like Amy Weiss, my mom, who has no idea what open and closed source is, even though she's extremely smart. She just doesn't pay attention to these issues.
Lay out for us, Tyler, if you would, what is the difference between open and closed and what's actually the, like what's at stake existentially in the debate between open and closed?
Well, here's a simple example of how open source matters. The AI service, perplexity.ai, which many, many millions of Americans are using, just today, Monday, they put on their website access to a modified version of the R1 model. So they took or borrowed the model and made their own version of it.
The model they have up on their website, it will tell you the truth about the Uyghurs and Tiananmen Square. So they were able to modify it. And it's simply theirs. There's no infringement of property rights. It's truly open source.
Let's talk about the national security threat that DeepSeek and all of these other technologies or apps or AI models from China create. We know that Congress passed a law with overwhelming bipartisan support saying that TikTok presents a national security threat and it needs to be not banned.
I know that that language isn't correct, but needs to be sold just as Grindr was once sold to an American owner because of that threat. The Supreme Court went to the federal court, then the Supreme Court unanimously upheld that decision. If TikTok presents a national security threat, doesn't it follow that DeepSeek is also one, Jeff?
Oh, yeah, absolutely. So TikTok is a national security threat, and that's because it gives the Chinese Communist Party enormous algorithmic power in addition to data gathering abilities. The algorithmic power is really the important point that I think people don't emphasize enough.
It's the ability to reach billions of people on this app who might be young, who might be impressionable, who might fall under the sway of Chinese Communist Party propaganda and In the event of, say, an invasion of Taiwan, the Chinese government has threatened Taiwan. It has threatened war around the region. And we need to be ready and aware for that possibility.
It's not out of the realm of what is possible in the coming years. And I'm not being alarmist. That's literally what they say. So deep-seek is not – I wouldn't call it a national security threat yet, but an impending national security threat. And the reason I start on that assumption –
is because Chinese law requires everybody in China at the demand of the Ministry of State Security, which is China's CIA and FBI rolled into one, to hand over data, to cooperate in intelligence gathering, basically do what the Chinese government says or face jail time. This AI is a part of China's national strategy. It's not merely a matter of
making a chat bot and allowing people to play around with it. This is a part of China's national strategy, even though I think we both agree, you know, these were outsiders. This was a startup that wasn't really involved with the Chinese state. They will be soon without a doubt because the Chinese government will require it. And this isn't just about generative AI. There's lots of types of AI.
I saw it when I was in China, when I was investigating surveillance AI, so facial recognition technology, a lot of this being developed with the help of software developers who are trained by Microsoft and other American companies being sent out there to set up an AI system, which, by the way, the Chinese government literally called Skynet. I mean, it's like right out of Terminator.
And this is a predictive policing model that they use in Western China They would attempt to predict whether somebody would commit a terrorist act or a crime of any kind, send police to their house, take them away to a concentration camp for what they call re-education, which is a form of psychological torture.
This was being done with the help of a vast AI system gathering data on everybody through their text messages there. their apps, their police observations. So this is the kind of thing that we need to stand on guard for because the Chinese government has been trying to export a lot of this stuff.
It's been sanctioned by the U.S., but these are billion-dollar unicorns that are major global companies, and they're involved in vast human rights abuses and vast national security concerns in China. So that's why I'm concerned when the Chinese Communist Party gets involved, it is a national security threat because it simply can't be anything else.
The thing that I'm fascinated by is let's just take TikTok because deep seek, I don't think is on ordinary Americans radar in the same way yet. Maybe it will be in the next few weeks, given the speed of everything right now. But 170 million Americans know basically what you've said, right? They know China does bad stuff. They might not know all the details.
They know that by using TikTok, the Chinese Communist Party has access to important, maybe sensitive data, and 170 million Americans decided we don't really care. Tyler, what do you make of that? Like, people like Jeff spend a lot of their time writing important op-eds and books about why this should matter, and most people are just saying, but I want to watch my 10-second videos.
How do you square those things?
I was not happy with the congressional decision on TikTok. I agree with Jeff that it might be a national security threat, but we live in a nation that is governed by the rule of law, and I much would have preferred an agency process, a legal process, actual fact-finding by experts, not just a one-sided congressional vote.
And if TikTok has broken American laws, by all means, take appropriate action. And if those... Disruptions of the law are so extreme that it dictates banning the service. I would consider that. But we had a politically motivated congressional vote. I think it's a terrible precedent. We will live to regret that day.
And the real security threat in America right now is how many Americans are asleep at the wheel in the midst of the AI revolution. And DeepSeek actually is waking us up. So it's improving American national security from my point of view. It doesn't mean it will be doing that forever. But we need to actually listen to China more than what we're doing now.
Say more about that, Tyler. What do you mean a sleep at the wheel? And how do you figure that it's waking people up?
Most people do not understand how good current AI systems are and how rapidly they have the potential to improve. That will change over time our entire society. It also has major implications for national security. When GPT-4 came out, it was days and days until it was mentioned in the New York Times. Journalists are wary of AI. They don't understand it very well.
Well, they're wary of all new technology. Correct. And I think we're simply at a point where most people, they enjoy their chat GPT. It's fun to talk to it. You can help it name your dog or something. But they don't actually understand what the stakes are. And DeepSeek, by being so popular, you know, it was the number one download in the App Store last week. It's at least helping spread awareness.
And in an open society, you know, if we have higher awareness and we're still the country that by far attracts the immigrants who are going to do a lot of the creative work on future AI, I think we can win this battle, but we need to not be asleep.
Jeff, I'd love for you to respond to a thing Tyler said that I suspect you disagree with and would love to hear your take on it about The idea of listening to China more. And Tyler, maybe you want to put a little bit more meat on the bones there so Jeff can respond to it. But that sentence stuck out to me. What do you mean by that?
Well, I think of both Americans and Chinese as somewhat provincial, inward-looking, large countries that are relatively poor at dealing with the rest of the world. We're seeing this now in both places. And China is a country we could learn a lot from. They can still learn a great deal from us. They may or may not do that.
But we have the option of waking up and learning an incredible amount from them, not just on deep seek and infrastructure, but in many, many different areas of life, including the incredible number of first-rate students they produce. A lot of American AI is built by ethnic Chinese, right? Well, how'd that come to pass? What can we learn from that? So we need to prepare and be ready.
And that does not mean closing off to China.
I would actually reverse the argument. And I would say that actually it's China that needs to listen to the rest of the world, especially the democratic world, because this is a country where you can envelope yourself in propaganda from morning to night. And it's impossible unless you have a VPN, which is illegal in China. It's impossible to get access to social media outside of the country.
It's impossible to get access to chat GPT and to other GPTs made in America. So, you know, it's a country that has become enormously inward looking over the past decade that has marched in the steps of past authoritarian and even fascistic governments.
I often make the argument that the Chinese Communist Party has more in common with the fascist regimes of the past with Italy and Nazi Germany instead of the communist regimes, which were essentially socialist regimes. The Chinese Communist Party teaches a curriculum that
racial nationhood, this idea that the people of the nation come from a master race that's been misunderstood and that seeks to conquer lands around it.
It's very much a dangerous imperial system that we've allowed to get out of control by allowing them to join the World Trade Organization, by allowing the Chinese Communist Party to enter our international organizations, the World Health Organization, which censored and suppressed information originally about COVID because it might anger China.
We're now dealing with a behemoth, a giant of authoritarian potential worldwide that seeks to undermine democracy. And, you know, as long as our companies, as long as our people here are scared of criticizing China, which many companies are, you go to any big tech company, almost all of them will say nothing bad about China or about the CCP or the human rights abuses there.
the national security concerns there because they've been bought off by market access. They want to be able to manufacture their EVs and their cars and whatever else it is that they're making in China.
So we're now in a situation where we have to contend with a world power, an emerging world power that is antithetical to democracy in almost every single way that has become more authoritarian as they've opened up their markets and that seeks to counter U.S. power all over the world. It's a dangerous situation we're in.
Tyler, I suspect there's a lot there that you do not agree with.
a lot of adjectives in that statement. I would make a few points. First, there's plenty of propaganda here. I'm completely opposed to what the Chinese government has done to suppress free speech, but to think that we here are the ones who see the unvarnished truth, I think is very far from reality. For all the terrible things China has done, they are our only peer country.
They do many things better than we do, and we just have to learn from them. Many, many more Americans should visit China, should try to learn Chinese, should preoccupy themselves with China, but in an open way. China entering the World Trade Organization, I don't think that was nearly as significant a fact as people like to portray.
The chance that a cure from cancer comes from China is as high than the chance that it comes from the United States. Economically speaking, the two countries really cannot do without each other. We need to find a way to live with them.
And I think only having this negative attitude up front is a sure sign that we will create voter sentiment that doesn't actually allow for some possibility of peaceful coexistence.
They may be our only peer country. I think that that's probably objectively – I think we would all agree with that. But the thing is they perceive themselves or at least the CCP perceives themselves to be in a kind of war with us and they want us to lose and they want to win. Do you agree that that's true?
I think that's largely true. I think our attitudes toward them are not entirely different. The United States does want to, quote unquote, hold China down. I don't think that's an unreasonable view. But at the end of the day, it's a lot of talk on both sides, and we need to learn to live together for military reasons, trade reasons, tech reasons. The whole future of the world relies on that.
Well, the big comparison that many historians make is that we're in a new Cold War. Neil Ferguson has called it Cold War II. And that in the same way that the first Cold War was about sort of the nuclear arms race, this one is about the AI technological race. Tyler, do you buy that as a comparison?
That is likely true, yes.
So— Let's draw that analogy out. The way the Cold War was won was not by saying to our adversaries, let's find a way to live together. It was by beating them, wasn't it?
Well, what we actually did was built a much better system. We didn't attack them. We pursued detente. We tried to reach some kind of understanding. We sent them cultural exchanges. We sent them rock and roll and the Beatles. And over time, yeah, people in those systems ceased to believe in them. And that's how we won really remarkably peacefully.
I'm not saying we'll have a copy of that history with China, but we have the better system. Most of the world knows that. And we have the possibility of winning this next Cold War.
But the thing is, I guess, you know, I remember the days that I was at the Wall Street Journal editorial page, and they very much believed in this analogy. And therefore, they believed that by opening up trade with China and by liberalizing China's economy, political liberalism or some kind of thing akin to democracy would necessarily follow in the same way that it did with the former Soviet Union.
But that didn't turn out to be true.
It didn't turn out to be true in Russia either. It was a mistake that people ever thought that. We don't know what the future holds for China, but I think it is likely that the more interaction they have with the West and the United States, it's likely to go better rather than worse.
All right, here's another way that I think that that analogy maybe doesn't hold up. And Jeff, maybe you could think about this one with me out loud. Some people argue that this comparison is wrong because, one, there's no endgame to AI in the same way that there's a nuclear bomb at the end of the nuclear arms race.
And two, and this is something you might disagree with, AI can't be regulated in the same way. What do you make of those two arguments?
So nuclear weapons, significantly different technology from AI. We still don't fully understand what AI has the potential to do. Whereas with the nuclear weapon, that became quite obvious with Hiroshima and Nagasaki, the sheer destruction of The guaranteed destruction of World War III was clear and evident right there. But today with AI, you know, this is not a physical bomb.
It's not a physical weapon. It's something that can certainly reshape our society, reshape the way we do things, even the way that we do wage war, but it's not necessarily going to lead to that kind of world cataclysmic event. It could. I mean, I'm not ruling that out, but it's certainly possible.
I do think that there is a key point of significance when making these comparisons between the Cold War I and Cold War II. A lot of Cold War I was won by the U.S. and the West, our allies, because of our technological innovation. And one of the big innovations was in semiconductors with massive government support.
The Silicon Valley that we know today did not exist in the same way in the 1950s to the 80s. It was a site of military testing for semiconductors to try to create the most cutting edge technology that then we could use in our tech.
Guys like Steve Jobs, Bill Gates, the people who invented the Internet and so forth, a lot of them would not have been able to execute their vision, which was moving towards technological change to the miniaturization of technology and PCs without the innovations that had come before that through government funding, through government support. And, you know, just imagine a world today.
So we're now in Cold War II, as Niall is saying. Imagine if in Cold War I we had all this fantastic technology that was going to win the war, And we signed a free trade agreement with the Soviet Union and just gave it all to them and said, here's our tech. We're giving it to you. Use it well. This is free trade. We're going to help each other. We're going to coexist.
That would have been devastation because they would have been able to unleash all kinds of things that would compete with us, that would guarantee potentially their victory. We're now in a really unusual stage in human history. I has just been unprecedented in our age today, the last 30 years. We've just seen so much in the period that we live in.
But now we're at a stage where we're handing a hostile authoritarian government cutting edge technology that they intend to use against us, that they intend to use to dominate the global order. And that's a very dangerous situation to be in.
I think it's great that China has AI. AI is radical. It's the opposite of small C conservative. It will overturn status relationships and every society it enters. The CCP is terrified of AI, and they should be. It could disrupt their entire power, their hold on Chinese society. DeepSeek in particular is super easy to jailbreak, get it to tell you what you want it to tell you.
The notion that there are already entities out there in China smarter than the Chinese Communist Party is probably the single greatest threat to CCP rule. We have done it with this Trojan horse. We will see how it develops. But the fact that, oh, they're not reciprocal and it's been done through a Chinese company rather than an American. You know, fine. Things like that are going to happen.
And I think they're the ones who have a lot to worry about, not us.
Tyler, just to table set for a second. You agree that we're in a arms race or whatever you want to call it. Okay.
And we should spend more on defense. Strong agree.
Okay. And do you think that we're winning that race right now?
We're not losing it. You know, what we're hoping for is an ongoing perpetual stalemate of sorts. If we can win it and China becomes a mature democracy, a more populous Canada, that would be wonderful. But I'm not expecting it.
What's at stake in us winning it and how will we know? Because, you know, to go back to the analogy, we knew the space race was over when Neil Armstrong walked on the moon. We knew we'd won the race to build an atomic weapon when we set one off in Nevada. How will we know? Like, will there be a moment? How will we know if we won this? What does winning look like?
And I guess the deeper question is, like, what's at stake in us winning here?
I don't think you ever win. We don't know the space race is over. It's possible that either Russia or China develop space weapons that we have trouble countering, or that they're able to use space, say, to blow up our satellites, which our military is too dependent on. And the nuclear weapons race, well, Russia has more of them than we do.
And arguably, they have a greater willingness to actually use a tactical nuclear weapon. So at the moment, at least arguably, they're winning that one. So no, I don't think it ever ends. It's tough. I mean, you have to be vigilant forever. It's a very hard problem for free societies to do that and also to stay free.
Maybe let's take a step back. When Sam Altman, who runs OpenAI, was recently on the show, he was making a lot of this race and we were talking a lot about it. And I said, what's at stake? What happens if China wins?
And he basically just said they can do whatever they want, which I thought was a pretty chilling answer for those of us who are worried about what the CCP wants in the world, not just for China, but beyond its borders. What's at stake, Jeff, do you think in this race that we're in?
Well, the big question that everyone is asking is, will we be able to reach what is called a general AI or an AI that's able to essentially think and switch between tasks similar to how the human mind might take care of things? So this would be an AI that could, you know, at the same time, surveil you, gather data on you, and then go to, you know, generative mode and then go to
do this and that, make your dishes, run a tank. General AI is something that can do just about anything in theory, but it's not totally clear yet whether we're ever going to reach that stage of AI that is so all powerful. What I think victory looks like, if it is possible to reach that stage, victory will be whoever hits general AI. And if China is able to reach that stage before the United States
it would be very alarming because it would mean that there would be an all-powerful intelligent AI that can hack systems, that can wage cyber warfare, that can maneuver around an American telecommunications infrastructure, that can take control of missiles. Now, I don't mean to sound like it's the movie Terminator or something, or the robots are going to come alive.
I'm not sure that's actually going to happen. But if AI does get that powerful, that it can take on many tasks and exponentially learn from
from them and China gets there first or any hostile government gets there first, that would be a serious problem for democracies everywhere because we would have to contend with something that's infinitely smarter than us that even our systems cannot respond to.
It seems to me that one of the big debates is around, and I don't think there's a clear answer in MAGA world about where they fall, is about decoupling or not. Rand Paul was on this show a few months ago, and he had a pretty strong argument against decoupling based on the idea that the more we trade with China, the lower our appetite is for armed conflict.
How can you go to war with someone when you're changing so much money with them? Where do you guys fall on that argument?
That's bullshit. Total bullshit. Say more. Yeah, that's the lead up. That's what people said before World War I. It was a time of wonderful flourishing of technology, of free trade, of really the first vestiges of what would today look like globalization. I mean, that was late Belle Epoque.
when the world was becoming more prosperous, more industrialized, and traditional borders were not looking the same as they had in the past. But that was the lead up to World War I. And part of the reason is because of national pride, nationalism. It was the world going to war because, you know, for no particular reason at all, it wasn't even clear why the world was at war.
But that pride that comes before the fall is something to certainly be wary of.
I think you're far too pessimistic. The stronger China gets, the more small-c conservative it will become. I can quite readily imagine a future where the US and China have like two-thirds overlapping common interests, and we to some extent work together
to put down lower-scale AI mischief-makers in other parts of the world, because the US and China both will have a strong stake in stability, stable oil supplies, things working out a particular way. Yes, China does want to take Taiwan. They are not, in general, an expansionist power that wants to take over the whole world. So I'm much more optimistic about the prospects for future cooperation.
But stability for them looks very different than stability for us.
Of course. They're more dependent on foreign energy, for one thing.
But also they're more willing to clamp down on things that we regard as basic freedoms in order to attain what they perceive as stability.
On their own people. They would say we're much more willing to be interventionist in other countries to the detriment of other citizenries to get what we want. That's somewhat of an unfair criticism, but it's also not entirely untrue.
Well, let me put a simple question to you both. Right now, today, two of the top 10 apps in the Apple Store are Chinese, with Deep Seek being the most popular. If you could wave a magic wand and we didn't live in a democracy, or let's just say you were president tomorrow, would you ban those apps from being in the Apple Store?
I don't necessarily think that we should ban it. I wouldn't use the word ban, but I do believe in reciprocity when it comes to trade. And I do think that we should look at finding ways to separate Chinese AI companies that are accessible in the US. Those should ultimately be controlled by either friendly countries or US companies. And the reason is because of reciprocity.
Almost everything that we produce online is blocked in China. When Americans like us go to China, and say free trade is great, that only works if both sides are playing by the same rule. This is an example of one side.
And when I say this, I'm talking about before the recent tariffs that went up, but one side largely playing by the rules of free trade and one side essentially creating a closed market and exporting dumping materials on the rest of the world. So it's a completely unbalanced system. It doesn't work in anybody's favor outside of China.
And we need to do what we can to separate some of these industries to ensure that what's happening in China is reciprocal to what's available in the U.S.
I say don't ban them. But if they break the law, throw the book at them.
But DeepSeek has admitted, Tyler, that it's going to save user data in China where the government there can access all of it at any time.
China already hacks into many, many of our systems, whether it's OPM or email systems through cloud computing. So it's a myth to think that all this data to begin with is secure.
So is privacy just over online and should we just assume China has access to everything?
You should assume China has access to everything. I hate that.
Should it bother us? Should it keep us up at night?
Yes, it should bother us, but it's not going to change most likely.
I mean, one of the things that I thought was interesting is after TikTok was banned, a lot of people said it doesn't matter if China has our data because American companies take our data, too. Is that moral equivalency, Jeff, or is there a meaningful difference between the two?
It's a false equivalence, and it's something that a lot of people do say, but it's just it's flat out wrong. American companies do take data. And there are terms in the agreement that are guardrails against how your data is used. On top of that, we have a system of checks and balances of courts and independent judiciary in America. So if something were to happen, you do have some kind of recourse.
To say that the United States, the government, the democracy of the United States is comparable to a one-party authoritarian regime is just flat out wrong. And the thing about China is that that data that you put in, whether it's in DeepSeek or it's in TikTok or whatever it might be, that data is accessible in China. The Chinese Communist Party, if they want it, they can get their hands on it.
You don't know how they're using it, but we do know that this is a party that puts millions of people in concentration camps. Do you want those people having access to your data, to your kids' data? Whether or not you know what they're doing with it, you can see that this is a pretty evil regime, and we do want to separate... the data that exists here in America versus what's accessible in China.
The real risk to America mostly is American stupidity. If you look at defense spending, in real terms, it's remarkably low right now and we just let it fall. And we should raise that significantly and also make our defense procurement much better, make our science policy much better. Not doing that, those are the real risks on my list of worries that China has data about me, other Americans.
I really don't like that fact, but it's just not that huge a problem compared to these other issues.
Put the data to the side. You know, lots of people have compared the kind of content that American kids are served on TikTok to digital fentanyl. And we know that we get videos about drinking Tide Pods or whatever, and Chinese kids get like Khan Academy, how to do calculus at the age of five. Does that bother you, Tyler? And should that bother the ordinary American?
The best parts of our education system are better than the best parts of their education system. And you can see that by comparing the performance of ethnic Chinese in America to ethnic Chinese in China. That said, we have a significant problem for the bottom half of our system. I don't have any quick fixes on how to make it all better.
But the major AI advances have, in fact, come in this country. And they've come largely through immigrants, individuals who really did not ever want to go to China.
Just to pan out for a second, there's a lot of Americans who don't use Claude, Perplexity, any of these tools yet, Tyler and Jeff, in the way that you guys do. Tell them, those of them who are listening, how these tools have changed your life already.
It's smarter than me, and it knows more than I do. So I have a near infinite demand to use it. I just keep on asking it questions all day long. I hardly use Google anymore, except like to buy tickets. And just yesterday, I started using Deep Research.
Deep research can write, say, a 10-page paper for you in five or six minutes, where the quality of the paper is what a PhD-level assistant would give you if that person had a week to work on it. And you get this in five or six minutes. I just started this yesterday. That is amazing. Most people don't know this. It's incredible. It's going to change our whole world.
It is incredible. So just to interject one thing, I've been working on a PhD for a long time, and seeing these advances, it feels as if You know, that years and years of work on a PhD, it's like, oh, man, you know, a chatbot can do this. What was the point? Yeah, what was the point of all that?
Well, what is the point? I mean, one of the things that, you know, the doomers, those that are really, really worried about how this technology will upend our society is not just that it will put a lot of people out of work, but that it will somehow create a kind of existential crisis about our purpose in the world when we're around something, Tyler, as you said, that's already happening.
smarter than one of the smartest people I've ever met in my life. How do you contend with those questions?
We'll do it bit by bit, piece by piece. I think we need to radically reorganize higher education and what a PhD is and means. I don't think it will be done in top-down, centralized fashion. It should not be. We need a higher education system with a lot more diversity, true experimentation. You can fail. You can go out of business. and set that process in motion.
And I'm confident the United States can do this, but right now it's a kludged cartel, and it operates terribly, and it's super conservative, and we're hardly making progress on this at all. I think it's one of the great national scandals.
After the break, Tyler Cowen and Jeffrey Cain on whether AI is going to become a new god, a new religion, plus tariffs. Stay with us. Last month, the Bank of Canada released this report explaining what the economic ramifications would be if Trump passed this 25 percent tariff on all goods coming out of Canada.
And what this big bank report came out with is that the Canadian GDP would decline by 6 percent. What the economist Stephanie Kelton did is she asked Deep Seek the same question. What would happen to Canada's GDP if Trump imposed the tariffs he had been threatening?
And she watched as Deep Seek reasoned through the problem and came up with basically the same conclusion in a matter of about 12 seconds. I mean, what's the point of these huge realms of the knowledge economy? Five years from now, Tyler, will all of that stuff, what will it look like? Will it go away? Who's going to win and who's going to lose?
I don't think we know yet. A lot of people who think they're so smart and know so much and write so well, their skills are not so unique anymore. I do think they will have future roles in guiding the AIs, verifying the outputs, judging what is relevant, deciding which questions to ask. But I think, again, it will be a radically different setup once we allow competition to operate.
With big losers, those big losers are people with very influential voices. They will hate it. They will go down screaming. Part of me may enjoy that a bit. I just think we need to start experimenting and get the competitive process going. I don't think you can predict the exact outcomes. What will you enjoy the most?
Just waking up every morning and knowing that whatever question I have, I now have a smarter entity to ask than I ever had before in my entire life. That thrills me every day.
No, I was asking more about Schadenfreude.
Oh, Gary Marcus. I mean, seeing Gary Marcus eat crow, that's like the single thing that will give me the greatest pleasure. Because he's been saying for so many years, none of this stuff works. He's so obviously wrong. He doubles down in such a convoluted way each time. It's a standing joke with many of the people I know. And just to see the final comeuppance for Gary Marcus, that will be amazing.
Forgive me for being petty, but that's the true, sincere answer.
You're almost never petty, so I really enjoy it. Okay, let's talk about these tariffs, okay? So we're talking on Monday around 4.30 p.m. This is a story that's moving so rapidly that who knows, Canada might be the 51st state by the time this podcast comes out in a few days. But Trump has instituted these 25 percent tariffs against Mexico and Canada.
And guys, correct me if I'm wrong, 10 percent on China. Am I correct?
Yes. That's the announcement. What it really means, who knows.
Okay. China, let's start there because that one is probably the most defensible and understandable. Jeff, where do you stand on that particular proposed policy?
So I think that's a good policy on China because that's the best tool that we have now to offset the effects of China's forced technology transfers of U.S. companies. It's dumping of products on the U.S. market and the various... forms of trade trickery that the Chinese government has used over the last 30 years. The tariff is good because it offsets all that while keeping the dollar strong.
That's one of the reasons why the Trump administration does like tariffs. And I should also add, we have done a lot to sanction certain industries in China. We've also gone after Chinese banks with sanctions that are operating in Mexico that are involved in getting the base chemicals for fentanyl across the border. So there's been a lot of that done already, the targeted work.
But now, because we're starting to reach the limits of what's possible with those sanctions, it's time to put a tax on Chinese products that are being unfairly dumped, that are forcing these technology transfers, stealing American tech. That's the way to offset all this.
Tyler, Chinese tariff, yay or nay?
I wouldn't do it. Look, it's not going to matter much. The Chinese will let their exchange rate depreciate. That will counteract the effect of the tariff. What we actually need to do is look at a few key sectors. Maybe it's components of drones. Make sure we're much less dependent on China. And if we use targeted tariffs to help us do that, that could make sense if we do it wisely.
But this general 10%, whatever they're going to end up making the number, it's all for show. It's Trumpian bluster. It's not really going to matter or help much of anything.
All right, let's steel man it. Why is Trump doing this? What is the best possible argument? Why is Trump creating, putting China to the side, a trade war with our neighbors in Mexico and Canada?
Trump likes to dominate the discourse, be the center of attention. And he also genuinely believes that trade deficits are bad. Economists pretty much universally have rejected that view. So he's bad at economics and he loves being the center of attention.
Jeff, I think it's to get concessions. And we've seen how quickly the Mexican government just sent 10,000 troops to the border. Now, I agree with Tyler that a lot of this is potentially bluster. Maybe it won't last. Maybe in a month or two, everyone's going to forget about this.
But I do have to say it is interesting how quickly the Mexican government was able was willing to accede to his demands once that tariff went on. And the reason is because America, one of the biggest consumer markets of the world, Mexico depends a lot more on America than America depends on Mexico. Same for Canada.
But that's Mexican bluster. They already have 15,000 troops on the border. I mean, both sides do this. Don't think the Mexican leaders are any more honest or straightforward than Trump. With Canada, the thing we actually ought to demand is more spending on Arctic defense against Russia and possibly China.
For us to get that to happen, we're in a better position if publicly we act nicer to Canada than what we're doing right now. I think from Trump, it's a huge mistake.
There's sort of a larger historical question here, which is that in the sort of post-World War II era, and especially since the fall of the Soviet Union, the idea is like America's not supposed to go around bullying smaller, weaker countries.
And the U.S., you know, at least ostensibly as the guardian of the Pax Americana, the guardian of the world order, is especially not supposed to act like that. How are we hurting ourselves and our position in the world from going around sort of beating up allies like Canada and Denmark. What do you think that does for us, Tyler?
I think we've always bullied smaller, weaker countries. The big difference is now we're doing it more openly. I don't endorse that. I actually think the world, to some extent, will get used to it, and it may matter somewhat less than Trump's critics think. But I don't see the upside. With Greenland, I see the upside. It would be great to have Greenland. It wouldn't be great to have Canada.
But other than that, we should mostly play nice to people and do hardball behind the scenes.
Okay, let's do a lightning round. Tyler, has anti-wokeness peaked?
Anti-wokeness is going to lose the quality of its reputation because people will see more of it. When the woke dominate and the anti-woke are the opposition, everyone loves the anti-woke. Now I think everyone will end up upset at both of them. So in that sense, it's peaked.
What's the next culturally high status move? What should anti-woke now morph into?
I think the entire terms of the debate will change. There'll be cultural chaos. The major issue will shift to AI and how we're going to deal with it and who will gain and lose in status. And we'll forget about all that nonsense, mostly for the better.
Jeffrey, who is the best writer about China right now? Where do you get your news about China?
Bill Bishop is the best writer about China, and I get a lot of news from his newsletter, Sinicism, which is excellent.
Seconded, by the way. And Jordan Schneider and Peter Hessler is a book writer. But it's hard to know what's going on in China. I'm hoping to go back there this year. That's the best thing to do.
Tyler, what's your favorite province to visit or city or town or street to visit in China?
Western China, Yunnan province. It has perfect weather for the entire year, arguably China's best food, incredible ham, amazing mushrooms. About half of the population is non-Chinese ethnic groups. And it's just mind-blowing, one of the very best places in the world to see.
Tyler, who is the best writer about AI right now? What do you recommend that we all read?
Twitter. You just have to follow Twitter. Don't click on Trump-related links. Click on AI-related links, and the algorithm driven by AI will take care of the rest. It's really the only way to do it other than being in private chat groups.
I got to say that's a hard recommendation because Twitter is just so overrun with – I hate even using these words, but just poor information, low-quality information. How do I know who to follow?
Well, on AI, you can figure out what's going on just by reading lots and lots and lots. And also, if you can, get to know some of the people. Have two separate Twitter accounts, one for politics, one for AI. There's plenty of incorrect information about AI, but it's not like American politics where it's all just or mostly just lies and misinformation.
A significant chunk of it is smart people trying to figure out what's going on. I'm not saying any of us get to the truth, but it's a great place to try.
Will we ever get to the utopian point where AI does all the work allowing all of us to exist on a permanent vacation, Jeff?
I don't think we're going to get there. I think that it will advance, and it already is advancing far. It is incredible what AI can already do, but the amount of compute that's going to be needed for that I think is just going to be off the charts.
Tyler? I think we'll get to a world where people who take good care of themselves live to 97 years old and die of old age, and that will be amazing. But people want to work, and there will always be work and paid work.
How will AI affect wealth inequality, something that a lot of people are really concerned about?
I think in the short run, it will raise people at the bottom who can't even write a decent essay. Longer run, you'll have lone individuals or groups of two and three people start what become rather quickly billion-dollar companies, and most of the staff will be AIs.
So people at the very top, I think, will be much richer, but the actual gains will be to people who are fairly poor, cannot currently get or afford a good medical diagnosis or good legal advice, and AI is going to give them all of that for either free or nearly free.
Tyler, one of the things I've noticed a lot recently is we've known this for a long time on the left and now we're seeing the exact thing on the right, is that all of the incentives and all of the excitement is on the extreme and the political center is collapsing.
As someone who I think situates yourself somewhere in the center, what do you make of that phenomenon and is there any way to make the center cool again?
I don't know what these terms mean anymore. To me, Trump in some, but not always, is very much in the center. He's pledged he doesn't want to cut the big government programs. He claims he wants to keep America out of war. He probably means that.
Yet there's something about the tone and the vibe that is so weird and sometimes so unkind and so unpleasant that it's hard to feel comfortable with it or want to endorse it. So I think we just need to keep our cool, try to be smarter, ask the AI when you need to, and not be taken in too much by movements or labels or ideologies, and actually figure things out.
Jeff, what's the last book you read that you'd recommend to our listeners?
I read a lot, so sometimes I even forget. So I'm actually, I just finished reading this book right here, which is Dostoevsky. It's Demons. And this is a wonderful novel about what happens when a society tries to replace the idea of God with ideologies, in this particular case, communism.
Tyler, does America need a religious revival?
America is seeing a religious revival, at least amongst its intelligentsia. We need this badly. I hope it spreads further.
Are we replacing God with AI, Jeff?
I think that we have the potential to, and I think that we already are doing that in many ways, and we need to keep the human aspect in the AI.
How do we do that, Tyler?
I think we'll have religions that emphasize AIs as oracles. They will nominally be the same as the major current religions, but they will feel quite different. I don't know how to stop that. I also don't know if it's good or bad. It may reach more people. It may be more persuasive. It may be more versatile. We'll see how it works out.
Tyler Cowen, Jeffrey Cain, thank you so much for coming on, honestly.
Thank you, Jeff. Thank you, Barry. Good to see you, Barry. Thank you, Tyler.
Thanks for listening. If you like this conversation, if it made you think differently about AI, China, tariffs, or any of the above, share it with your friends and family and use it to have an honest conversation of your own. Last but not least, if you want to support this show, there is one way to do it. It's by going to the Free Press' website at thefp.com and becoming a subscriber today.
We'll see you next time. Hi, Honestly listeners. Barry here with a really exciting announcement. The world today, as you know, is confusing, chaotic, unpredictable, bizarre, times scary, at other times just absurd. But chances are, whatever crisis is currently blowing up your phone, it's probably happened before, or at least a version of it.
And there's so much we can learn from looking to the past. That's why we have just launched a brand new podcast with free press reporter and history buff Eli Lake. It's called Breaking History. If you turn on cable, you're going to get superficial commentary on what's going on in the world today.
But if you listen to this show, you're going to learn deep things about history and about how whatever's going on right now is tethered to the past. Because before there was the Columbia Tentafada, there was the Weather Underground. Before there was Russiagate, there was the Red Scare. Before there was Donald Trump, there was Huey Long and Andrew Jackson.
As George Santayana warned, those who don't remember the past are condemned to repeat it. So tune in every other Wednesday as we remember, rediscover, or learn it for the very first time. Do me a favor, pause me right now, stop listening to me, and go subscribe to Breaking History on Apple, Spotify, or wherever you get your podcasts. I'll be waiting. Go ahead.
I am seriously so excited about this show. Eli's brilliant. I learn every single time I listen to him and I read him. And I know if you love this show, you're going to love his. Thank you. Breaking History every other Wednesday from the Free Press.