Why China Isn’t Panicking About AI Armageddon

Here in the United States, over the past week, there's been a lot of panic about how quickly AI is advancing and whether it could bring about the end of humanity. But in China, there doesn't seem to be the same level of concern. They've actually responded with Umbridge, but not for the reason you would think. Our colleagues do Woo covers business in tech in Asia, and he says that in China, worrying about the AI apocalypse is sort of like a problem of privilege. The way China sees it, this is a problem for tomorrow or years away. When you talk to researchers inside the labs, they think they're behind the great American models like added throptic and open AI. And they're thinking, well, this is not a big deal for us yet. We don't think our models are going to end the world anytime soon. What we got to do right now is make sure that our products are competitive with the American products, because we're not there yet. And if Beijing isn't willing to get on board with an AI slowdown, it's unlikely Washington DC will either. Stu says this standoff reminds a lot of people of one of the defining moments of the last century. Think about the original nuclear arms race. I think the US has invented the first atomic weapon, and then the US said, whoa, whoa, whoa, this thing we invented is actually super dangerous, and everybody should slow down research on their own nukes. And China's thinking, wait a minute, that's not fair. We definitely don't feel safer if America is the only one that has the nukes. So they're saying we're going to keep on working on our stuff full steam ahead. Welcome to the journal, our show about money, business, and power. I'm Ryan Knutzen. It's Wednesday, September 16th. Coming up on the show, why is it so hard to de-escalate the AI arms race? Chinese President Xi Jinping has made AI development one of the country's top priorities, urging vast investment in the new technology and dedicating a huge amount of China's resources to building it. And China is concerned about the risks of AI, but their approach so far has been very different from the US. In general, and I think a lot of people would be surprised by this, China takes AI safety really seriously. They've been thinking about this for 10 years, but what they consider AI safety is a lot different than what the American labs talk about. So, let's rewind. So 10 years ago, it started with the company that would go on to invent TikTok. That's by dance. Their first big product was this news app. And the special thing about it was that human editors didn't decide what its users read. They used an algorithm. Basically, a machine decided based on what you previously read, what article you would see next. And the founder would brag about this in Chinese media. And that caught the attention of the leaders of the Chinese government. They're thinking, "No, no, no, no. You know, for 80 years or so, the Chinese government has had all control over media, over bookstores, newspapers, and the internet in China." So, the Chinese government got very involved in regulating algorithms and AI. Now, for example, if you search in China for Tiananmen Square on social media or ask an AI chap about it, you'll find almost nothing about that 1989 massacre when the government killed hundreds of people. So how does China regulate AI to be pro-government? The government makes sure that they know what the training data is, that the Chinese trap out there being used to train, and then they make the both the companies and local government agencies test the answers to make sure that it doesn't say anything that will incite people to rebel against the government, right? So, that's really serious. That's number one, AI companies must protect the rule of the Communist Party. Along the way, they decided, "Okay, there's other things that we should worry about. We don't want anything to civilizing society in general. We don't want kids getting addicted to AI. We don't want people dating AI chatbots and not having kids and marrying and all that. They don't, they're worried about self-arme, all the stuff that the American companies are worried about, too." You mean like, they don't want people to use it so much that they like lose track of reality, so put limits on it in that sense. Yeah, exactly. I think that's something they share with the American AI labs. They're worried about this as a general harm to society. So, China is very active in regulating AI around censorship and mental health and things like that, but are they doing anything about these concerns around the potential end of humanity? No, they're not doing anything active right now. That's not to say China doesn't recognize the threat, though. So, what's happening is that every year, for the past three years, the Chinese government has published this document about AI safety, and they list all these concerns, concerns threatening the state, but also concerns to mental health. And one of the things, among the dozens of things they list is AI could become self-aware, and it could get more power, and it could challenge humans for control of the world. So, in a sense, yes, it's in this official document that this is a problem, but it's one of dozens of issues many of them have to do with making sure that the Communist Party doesn't lose power. Let's talk about how this race is playing out. How far behind the US is China? So, when you talk to the researchers inside the Chinese labs, they think they're maybe six months behind, or even more. And one big reason for that is that they just don't have the chips. The US government, under the Biden administration, started banning the export of high-end Nvidia chips to China. So, by some estimates, they have maybe 10% of the computing power that the US does, and the Chinese labs like DeepSeek have been really clever about using work-arounds to remain competitive, but at the end of the day, the US has a firepower. They really have the infrastructure that enables them to stay ahead and race. That might be the biggest single factor China doesn't have the chips. One area where China is excelling and making the US nervous is in how it's releasing its models. Chinese AI models are so-called open-weight models, meaning they're much cheaper and far more customizable. And a lot of US companies are starting to use them. So, if you use cloud or Gemini or chat GPT, those are closed-weight models. Basically, it's hard to go under the hood and tweak those models to make it more suitable for whatever you're doing if you're a business or a government or whatever. The way China selling their open-weight models is saying, "Hey, you can take our models and you can adjust them to suit your needs." It seems like this is sort of a similar difference between Android and iOS, which is made by Apple. Apple and iOS is very closed. You just get the iPhone and this is what it is. Android is an operating system that any phone maker can use. Anybody can take it off the shelf and customize it in a million different ways that they want to. Right, so that's a great analogy. So, okay, so maybe the American models are 5%, 10% better. But, you know, so how can a Chinese company compete? Well, we'll make yours more customizable. It might not be as good, but it might be better tailored for you, your business, your government. Stu says one of the things motivating China and informing its thinking right now is the country's own history. This is the constant thread in the current Communist Party. Is that China has been a great civilization and over the years there have been so many foreign conquerors, whether it was in the British and the opium wars or the Japanese 100 years ago that they've been embarrassed by these conquerors. And one of the reasons they've been losing is because they didn't have the technology. So, fast forward 100 years, we're in the 21st century and China is now one of the two superpowers in the world. And they're competitive in the AI race. They think they can get there or at least, you know, being equal footing with the US. And they look at the US saying, hey, we got to slow down the development of this technology. They don't like that. They think it's another example of a colonial power trying to keep them down. A spokesman for China's foreign ministry said American calls to slow down AI innovation were fear mongering and part of a Cold War playbook to hobble arrival. These two countries are so far apart and everything. It's hard to imagine that they'll agree to slow down because both sides think AI is essential for our future, for our national security. Both sides are thinking about this, but, you know, the way this 21st century arm trees is going, no one wants to, you know, stop working on the thing that might kill us all. As both sides continue to build the thing that might kill us all, how is the White House thinking about China's refusal to slow down? That's next. Imagine setting your makeup, then forgetting it's even theirs. Meet new grippy setting mist from Mabelie, New York. Jelly mist technology locks in your look for up to 24 hours, with flexible all day comfy grip. No tightness, no stickiness, no residue. Just plump, dewy hydrated skin that still feels like your skin. Try new grippy setting mist from Mabelie, New York. Maybe it's Mabelie. Our colleague, Amrith Ram Kumar, covers tech policy. We caught up with him outside an event about AI regulation. Right now, I am at the Marriott Marquis Hotel in downtown DC, where there is a pro-human protest/assembly happening with people like Bernie Sanders, Steve Bannon, many lawmakers and advocates for increased AI regulation are meeting and chatting. Huh, Bernie Sanders and Steve Bannon seeing eye-to-eye on something. That's Yes, as we wrote in our story, all it took was the possible end of humanity to get them together. It's definitely a lot of odd political bedfellows in the space right now. As much of the U.S. political class as in full AI freak out at the moment, many people in the Trump administration refused to join the bandwagon, and the main reason they give is China. We're leading China in AI. We're a sophisticated country in the world, and frankly, I want to keep it that way, because whoever wins AI wins. If there's one thing the President Trump has been consistent about in his second term, it's that the U.S. needs to stay ahead of China in AI. So the message from top officials, medium-level officials, any officials, it has been consistent that China is not going to stop, and they are going to continue developing models and trying to steal technology, or they can, or train off U.S. models to get ahead. The big question is how the U.S. would ever get to a place where it would trust that China would actually slow down, given the economic insecurity imperative to raise the head with this technology. What do you think it is that Trump and the rest of the U.S. government thinks the winner gets? Administration officials and the President are super worried about AI's military applications. They really think this could be a super weapon several years down the road, and the idea that the U.S. would be well behind China in having access to that super weapon, even if the margin that we're behind by is not that large, is terrifying to a lot of people. Beyond just maintaining an edge over China, President Trump also sees AI as a major economic stimulator for the U.S. despite the growing backlash. The President is looking at a lot of this. He's very close with the AI executives, and he thinks priority number one should be making sure the U.S. is ahead of China, and then his approach is, yes, we'll do some regulation and guardrails if needed, but we just have to stay ahead of them. That's the main priority and the one thing he's determined to win. If President Trump wants to stay ahead of China, why did he lift some of those Biden era restrictions on chips sales and allow Nvidia to start selling chips to China? Given the fact that chips are so essential to AI. This is a fantastic question and is one of a lot of people in Washington and lawmakers have been asking if beating China is your number one priority, why have you had such a loose stance on chip exports? The administration's response is essentially that they want Nvidia and others to be able to compete with Chinese companies in China. One of the things that we heard from David Sachs, who's President Trump's top AI advisor this week, was that he seems to have this concern that anthropic and open AI, they're called to be regulated by the U.S. government. It's just a way to cement their dominance, their market dominance, something he calls regulatory capture. Can you explain what this concern is that he's expressing? It's not just David Sachs. You had JD Vance, lawmakers, many other prominent individuals. They have this view that the top few AI companies now are calling for regulation to cement their status. The idea would be open AI and anthropic and Google and others that have so much money, so many resources that they can sort of navigate a thicket of regulation at this point and they have so much clout in Washington and have spent millions of dollars on lobbying and advertising that lawmakers listen to them about how those regulations should look and so does the Trump administration. The idea would be that they would essentially get to shape regulations that box out their competitors and that would eliminate a lot of startups and potential innovators and we're still in the early innings of AI development so that could potentially hamstring the U.S. in the long run if startups aren't allowed to compete fairly in this space. This idea of regulatory capture has definitely taken on a lot of prominence recently. If you talk to people at open AI and anthropic etc. they say that's not what they're doing. They say these are genuine security risks they want addressed in the strongest way possible. You're talking a lot of people about shifting AI policy. How do you feel like DC is reacting or changing if at all to these concerns that are being raised by Silicon Valley? What has not changed is that they're still not really a viable path for any legislation or executive action of consequence to come down in the next several weeks. White House has been clear. Trump has been clear that he doesn't want to take a big swing. They aren't going to suddenly exert a lot of oversight over new models or have the government very involved in this space. It will be interesting where the company is land on their so-called coordination on catastrophic risks if they can get together and agree to do that on their own. Even from companies to countries could of course unilaterally slow down. But the conundrum is that nobody wants to go first. So like nuclear weapons roughly 70 years ago, rivals are all going to have to agree to slow down at the same time. There's a slim chance that there could be progress on that next week when Xi Jinping travels to DC for a meeting with President Trump. But nobody's holding their breath. We've been told by people generally that keep expectations relatively low as far as a deliverable, like a concrete agreement to avoid catastrophe or what would be equivalent to sort of a nuclear non-proliferation agreement that the countries agree that if things get out of hand also things down, we've been told that's unlikely and that'd be a huge surprise if something like that happened. Instead, we were told the discussion might be pretty high level and focused more on just both sides capabilities and also security concerns on both sides. Our colleague Stu Wu in Singapore and a lot of other people actually have made this comparison to the nuclear arms race between the Soviet Union and the US. What do you make of that? Is that sort of the best case scenario we can hope for is that there's sort of a mutually sure destruction that results in a degree of disarmament or do you think there's too many differences to draw that comparison? I think it's helpful for just generally getting people in the right frame of mind for what this could look like. The big difference you share from people is that nuclear weapons could not be used to cure cancer and a lot of people think AI could be used to cure cancer and you can apply that in a lot of different dimensions. There are a lot of positive use cases of AI for humanity, for education, for medicine, whatever sector you work in. That makes it a lot different. So it's easy to say that the two countries should just get together and avoid this catastrophe, but the countries are also trying to optimize for all the benefits that the technology could give its people over the next 5, 10, 20 years. And to mention that with a nuclear bomb at least at the end of the day, it was still a human being who had to press the button about whether or not to detonate it, but with AI, the fear is that we're going to lose control of it and who knows what it'll do. That's exactly right, and I think the world has also just gotten a lot more complicated, especially as the model's getting more powerful in that technology, is in the hands of more and more millions and millions of people, it's a bit more complex. That's all for today, Wednesday, September 16th. The journal is a co-production of Spotify and the Wall Street Journal, additional reporting in the episode by Rafael Huang, Maya Davis, and Shavon Hughes. Thanks for listening, see you tomorrow. Imagine setting your makeup, then for getting it's even theirs. Meet new groupie setting mist from Mavily New York. Gel to mist technology locks in your look for up to 24 hours, with flexible all-day comfy grip. Just plump, dewy, hydrated skin that still feels like your skin. Try new groupie setting mist from Mavily New York. Maybe it's Maybelline.