Finding common ground in the U.S.-China AI rivalry

The U.S. and China want AI safety, but no one wants to slow down. From American public media, this is Marketplace Tech. I'm Megan McCarty-Kareno. AI is the big item on the agenda this week, as Chinese President Xi Jinping meets with President Trump in Washington, D.C. The talks come after a firestorm of news that has brought AI safety, front and center. The hack of hugging face by rogue open AI agents, the resignation of an anthropic researcher, and calls by AI leaders to pace the frontier. Always met with the rejoinder, what about China? Sam Sachs is here to tell us more about China's approach to these AI talks. Jesus, senior fellow at the Think Tank, New America. Both sides share immense vulnerabilities when it comes to artificial intelligence. So whether it's we're talking about job loss, youth addiction online, or whether we're talking about can humans and governments even control AI agents left unmonitored and wreaking havoc on global systems? That sense of shared vulnerability is co-existing from the government leadership level with the sense that if China gets a decisive strategic advantage in AI, it could be game over for the US. So that competitive dynamic is existing alongside really deep concern about what this technology means for the future of humanity. The main sort of strategic way that the United States has attempted to sort of enforce that competition with China is export controls on chips, right? How is that sort of affected the relationship? Starting in the Biden administration, the US government put controls not just on advanced semiconductors going to Chinese end users, but also on the manufacturing equipment. The Chinese side has not responded favorably to this. It has curtailed their access to the compute needed to train new models, but also to do what's called inference where you ask a model to reason and perform tasks on your behalf. So this is one of the main inputs that any country needs to develop and make use of AI, but what's also happened is that in the absence of US technology, the Chinese domestic industry has kicked into gear, and I think they're increasingly weaning off reliance of the US. So it's become counterintuitively less of a choke point than I think the US side may have intended. The environment and discourse around AI are very different in China in some very important ways. What do we get wrong in the US about the Chinese AI industry and how leaders there are thinking about this technology? A lot of people outside of China would be surprised to learn that Chinese companies, AI labs as well as tech internet platforms are very heavily regulated in China and they have been for several years. What that looks like is very different in China's political system than it looks in ours. But to give you an example, the Chinese public have been very concerned with teenagers, young people staying up all night talking with AI chatbots or companions. And so the government regulated it. They outlawed agents that developed human emotional characteristics that led people to an addictive state. I was in China over the summer and I heard that many of the Chinese AI companies were so concerned about complying with this regulation that they actually removed many of their AI agents so they were no longer available as a kind of over compliance. Now that's not to say that it's an area where the Chinese government is always acting in the best interests of the public. The Communist Party of China and its political ruling power is also very much front and center. So online censorship and making sure that AI models are not talking about things that would be subject to the great firewall are also a source of regulation. So I don't want to make it out to be too rosy a picture as in something the US would want to copy and paste entirely. But it is a picture that has to be taken to account. Why? Because there's a narrative in the US that we can't regulate AI because it would undermine the ability of US companies to innovate and to compete with their Chinese rivals. But the reality is Chinese companies are heavily regulated and they're innovating and they're in part looking to us and saying look you're not regulating why should we limit our own development. We'll be right back. You're listening to Marketplace Tech, I'm Megan McCarty Carino. We're back with Sam Sachs, Senior Fellow at the Think Tank New America. So realistically what kind of expectations should we have about what might actually be possible in these meetings this week? I do not expect a grand bargain on AI for China to come out and say we are also going to pace the frontier. And frankly I don't think it needs to happen. So I think what I'm expecting to happen based on the meeting that happened in New York City with Besant and Holy Fung. I expect to have meetings to talk about meetings. Let's move Trump and she aside on this topic. They're not going to figure it out. They're not going to agree on a grand bargain. But if they can signal we endorse our technical experts having a way where they can share information about some of these issues that will make the world a safer place. How robust have discussions between technical experts been between these countries so far? Considering all of the heated rhetoric and the rivalry, I think many people would be surprised to hear that quietly behind the scenes, there have been informal channels for technical experts to talk. I've been in a number of what are called track two dialogues where that's people who are not in the government coming together with academics, with industry practitioners from both sides to talk about shared concerns, to talk about policies. In terms of the technical experts, that's sometimes where the rubber meets the road. It's important to have conversations about policy and law. It's also important for people who understand the technology from an engineering standpoint who can talk about standards for them to have dialogue. That's actually happening. It's informal. I think it was notable a few weeks ago. There was an incident where a major vulnerability was found in WeChat, which is a super app in China that everyone uses to communicate and conduct daily life in China. A US security company actually found a vulnerability in WeChat that if left undetected and exploited could have wiped out the entire hundreds of millions of devices using WeChat, not just in China but around the world. This US security company actually went to Tencent and together they collaborated on patching, mitigating the vulnerability. That's an important example of non-government, non-state, private actor collaboration, sharing technical information behind the scenes to make systems more secure that don't stop at state borders. That's the kind of thing that I think needs to continue to happen and has been really hard because of the current geopolitical situation. That's Sam Sachs at New America. Jesus Alvarado produced this episode. I'm Megan McCarty Carino and that's Marketplace Tech. This is APM.