Joint military exercises between the United States and South Korea drew to a close on Friday, nearly a week ahead of schedule at the orders of Donald Trump. In a post on Truth Social last weekend, he wrote that he was scaling back the drills because they sent a signal that is, quote, "totally inappropriate and hostile," unquote, to North Korea, a country he wrote was, quote, "unthreatening and respectful," unquote. The exercises, which have in the past involved tens of thousands of troops, are designed to test South Korea's response to a potential North Korean attack. The decision appeared to take South Korea by surprise.
Joining me now to discuss all of this is Victor Cha. He was a member of the last Republican president, George W. Bush's National Security Council from 2004 to 2007, where he was involved in negotiations with North Korea. He's also the co-author of an important book, Korea: A New History of South and North. Victor, welcome. First, give us a sense of how this is being received in South Korea. They must be startled, to say the least.
Yes, Fareed, they're definitely startled by it. But for the South Korean president, who does have a good relationship with Trump, if what Trump is doing leads to a meeting between him and the North Korean leader, for the South Korean president, that's a good thing. He wants to see that. So, it's a bit of a mixed bag.
And is there — I mean, the symbolism, of course, is terrible. Is there an actual military price that gets paid here? Because my understanding is that these kind of military exercises do serve a useful purpose.
Yes, they do, Fareed. This exercise in August is the single largest annual military maneuver exercise that the US and the South Korean military do together. And if you don't do it, then the military doesn't have the same level of readiness, which affects deterrence on the peninsula.
Give us a sense of what has happened in North Korea over the last few years. I mean, is there any softening in North Korea, either in terms of external policy, fewer missile tests? Internally, I mean, they — I think they have still 100, 150,000, somewhere between those two numbers, political prisoners. So, I mean, is there any reason to be — is there a thaw that you see that, you know, maybe this is directed toward?
It's an interesting question. I would say there are two things that have changed from the general situation with North Korea. The first and probably the most significant is this relationship that Kim Jong-un has created with Vladimir Putin. They are basically allies in the war in Ukraine. This is something that started in about 2022 and has really flourished. So, that's a major, major development. The second, with regard to any sort of softening, is that, as President Trump mentioned, Kim Jong-un has been relatively quiet in terms of nuclear tests and ICBM tests during President Trump's time in office in his second term. During the Biden administration, there was no diplomacy and there were many, many North Korean provocations. And even in the first year of Trump's first term in office, there were many, many provocations. So, they have seemed to take their foot off the accelerator, if you will, when it comes to these weapons demonstrations, at least in the first year and a half or so of the second term.
And Trump seems to be looking for a deal. He mentioned that North Korea has 57 nuclear missiles, a kind of very specific number that I hadn't heard before, and that he's okay with essentially normalizing them as a nuclear weapon state with 57 nuclear missiles.
Yeah, so it's very interesting that President Trump takes two different approaches to Iran and North Korea. In Iran, he's very absolutist — they cannot have a nuclear weapon. But in the case of North Korea, he's demonstrated some flexibility. He's referred to North Korea as a nuclear power. He also calls for denuclearization in the end, but I think he's taking a much more pragmatic approach which says, you know, I need to deal with the threat and I can't sort of wait for denuclearization to happen because that's not going to happen anytime soon.
Do you believe, having negotiated with the North Koreans, that you could imagine a significant deal where they were to agree to any kind of path even to denuclearization?
It will not be easy, for sure, to try to negotiate denuclearization with North Korea. Having said that, there are a lot of immediate things that we could try to address, like trying to get them to stop producing more fissile material for nuclear weapons, trying to attenuate the relationship with Russia, and trying to ensure there is no nuclear accident or nuclear first use in the region as North Korea grows as a nuclear power. So, there's a lot of homework that could be done in initial negotiations that may not say denuclearization, but they are very important and tangible for US national security.
While Silicon Valley remains dominant, last month the Chinese startup Moonshot released an AI model that rivals those from Anthropic and OpenAI. AI has also been widely incorporated into everyday life in China, from robot cashiers to highly advanced surveillance systems. So, how has China managed such feats? I'm joined by Evan Osnos, staff writer of The New Yorker, long-time China watcher, who lived there for 8 years. Welcome, Evan. You went back and shed some light on this question that many people are wondering about, which is: China spends maybe a tenth as much as the US on data centers, it doesn't have the most advanced computer chips, and yet its best models seem very close to the best American models, maybe 6 months away. How has it managed to do this?
Well, this has caught Americans by surprise a couple of times over the last couple of years. Remember, there was of course the DeepSeek moment. This was the first time a Chinese model was demonstrating that kind of capability, and then most recently from a firm called Moonshot. And this is especially striking because they have less capital at their command. There are also export controls, of course, on some of the most powerful chips in the world. So, look, they have been accused by American AI makers of what's called distillation — of essentially unfairly copying American products — and at the same time there is a recognition now across Silicon Valley and in Washington that China has been able to make much faster gains than we had expected. And I think that's partly because of a kind of focus at the leadership level. They have been dedicated to this really for 10 years. And this 10-year period turns out to be decisive because they have been able to make much faster progress than the United States expected.
And you know, what are the three things you need for AI? You need compute, that is chips, you need energy, and you need human talent. And it seems to me that China has lots of human talent at the very highest levels in computer science. It clearly has much more energy. I think it has 10 times as much installed electricity than the United States or some such number. What it didn't have was the most advanced computer chips. What did they say about how they got around? Is it that they're managing to get it in because they imported via, you know, Southeast Asian countries, or is it that the indigenous chips are good enough to get them a long way?
It's a combination of both of those, but it is also a larger strategic shift. In some ways, this is a story that started during the first Trump term. That was a period when, you'll recall, the first trade war was a real wake-up call in Beijing because they said to themselves, "Look, we are entering a period of greater strategic competition with the United States. We have to become more self-reliant. We don't want to have to rely on them as much as we do now." At the moment, there was a kind of memorable contrast where the United States under Donald Trump the first time was setting aside what had been the first federal plan for AI, because it had been conceived under the Obama administration, but at the same time China was adopting its first federal plan for AI. And 10 years later, we see some of the fruits of that. So, look, I think even if you set aside the hype — and China is very good at hype — we have to be clear-eyed about the progress they've made or we're going to be caught flat-footed.
Is it also possible that this whole idea of a race is kind of misconceived because we're going to have AI, they're going to have AI, the whole world is going to use AI for all kinds of things? I mean, you know, people often make the electricity analogy. It seems to me nobody really won the electricity race, even though Thomas Edison did start it and he built a company, General Electric, that did well, but I mean, the Soviet Union electrified. What do you think of the idea of this as a race, and do the Chinese think of it as a race?
Yeah, in some ways the conversation in Silicon Valley is very different than the conversation you hear in places like Beijing and Shanghai. In Silicon Valley there's a lot of talk about achieving artificial general intelligence or a decisive strategic advantage — the idea that there might be someday where we cross the tape and therefore the game is ours. In Beijing they see it quite differently. They say what we're doing is trying to get it out into the economy already as diffused as possible, and I saw it on the ground. Look, you can go to a Chinese car factory today and you'll see that they have made this shift towards robots much more decisively than we have in the United States. There's a lot of pros and cons to that. I'm not saying this is an unambiguous good, but it is giving them an economic advantage that we're wise to take seriously. That means their unit per production is lower and that means then that they're more competitive around the world, and that in some ways is the ballgame. It's not just in this US-China race as you say, but it's in developing countries. It's in Africa and Latin America and many other places to come.
