How advanced is China’s AI, really? | Meanwhile in Asia

Exploring China's AI advancements, motivations, challenges, and US-China dynamics in the race for AI dominance and regulation.

Key Takeaways

China’s AI strategy is deeply tied to maintaining social and national stability.
The AI race is a high-stakes competition with enormous economic implications.
Both countries face challenges in advanced chip manufacturing, a critical AI layer.
China is rapidly advancing through a national plan and mobilizing major tech companies.
AI presents both transformative opportunities and significant societal risks.

What the video covers

China's AI guardrails are motivated by social and national stability rather than fear of existential AI risks.
The US and China are engaged in a competitive race for AI dominance, driven by a massive potential market valued at $50 trillion.
AI is described as a five-layer cake: power, chips, data infrastructure, models, and applications, with both countries lacking advanced chip manufacturing.
China launched a national AI development plan in 2017 aiming for global leadership by 2030, focusing on hardware, talent, and leveraging its large population.
Chinese tech giants like Baidu, Alibaba, ByteDance, and Tencent are heavily involved in AI development alongside innovative startups.
China excels in physical AI applications such as robotics and autonomous vehicles, while the US leads in sophisticated language models.
AI poses risks including job displacement, privacy concerns, disinformation, and potential misuse in cyber or biological attacks, though no catastrophic events have occurred yet.
There is optimism about AI’s benefits in China, but also concerns about economic and social impacts, especially on gig economy workers.
The US relies on Taiwan and South Korea for chip manufacturing, creating strategic dependencies in the AI race.
Collaboration between the US and China on AI regulation remains uncertain amid geopolitical tensions and differing priorities.

Chapters

What motivates China’s approach to AI regulation?

China’s AI guardrails are motivated primarily by the desire to maintain social and national stability, rather than fear of existential AI risks or systems going out of control.

Is there a real AI race between the US and China?

Yes, there is a competitive race for AI dominance involving development, infrastructure, and talent, driven by a potential market valued at around $50 trillion.

What are the main challenges both countries face in AI development?

Both the US and China face challenges in advanced chip manufacturing, relying on Taiwan and South Korea for production, which is critical for AI progress.

Full Transcript — Download SRT & Markdown

Speaker A

So what motivates guardrails in China? It's not fear of an existential, you know, end of days. It's not that. Or like an AI system going out of control. You just hinted at it. It's stability. It's social stability, and it's national stability. Get to

Speaker A

the gorgeous guest who knows things. Yeah. Hello, I'm Kunal Sehgal. Welcome to Meanwhile in Asia, a weekly show where we discuss some of the biggest stories happening in the region. Hi, and I'm Ivan Watson. And Kunal, I guess

Speaker A

this week we're talking about AI. And it's because it's a huge topic of discussion in the U.S. You've got some CEOs that are calling for guidelines and restrictions on AI development. They're calling out potential risk factors, race for AI with China and calling

Speaker A

for China to work with the US on regulating this new technology. Can the US and China actually work together on AI? And to answer that and to uncover all of this, we're joined by CNN's anchor and correspondent, Christy Lustau. Christy, thank you so

Speaker A

much for joining us. Now, you've been covering the rise of China's technological and AI rise for more than 25 years. Yeah, it has been a while. Yeah, I used to work in the Chinese internet scene in the late 1990s. I was working at

Speaker A

a company called Sohu.com. And while I was inside the company, I was also writing and reporting about tech trends in China. And I've been talking to some of China's top technologists and CEOs for over a quarter century. So I'm going to go first

Speaker A

to you with a question because not only do you know what we're talking about here, but you have opinions. Strong opinions. Strong opinions. So given some of the doomsday warnings that we've heard from people in the industry in the US, what is your

Speaker A

take on this? Is AI posing a potential threat, a human extinction level threat, as we've heard some people kind of invoke? Is Skynet about to come online? - Yeah, are the robots going to come kill us all, right? And the language that we've

Speaker A

heard has been extremely, extremely frightening, right? We're learning about examples of AI misbehaving. We're hearing about AI leaving the sandbox. We're hearing really smart individuals talking about how this could lead to a doomsday scenario. And we're gambling with our own

Speaker A

lives, right? And the risk is there of AI potentially being used to launch a cyber attack, a biological attack, an attack on critical infrastructure. But we need to underscore that hasn't happened yet. Right. And the same token, there are real

Speaker A

world harms caused by AI that's been going on for years. Right. Disinformation campaigns, scams. Look at what's happening with their social media posts, our privacy being infringed upon as AI systems are scraping our data, copyright infringement, let alone mass displacement

Speaker A

of jobs, right? But on the same token, new efficiencies have been introduced because of AI. Researchers who I've spoken to in fields like medicine or even space and marine exploration are saying this is a good thing. Look, we've heard doomsday language, really scary,

Speaker A

but it's important to underscore nothing has happened yet as of this taping. And we really want to focus on China and what it's doing in this sphere, right? Yeah, because I mean, there has been, I mean, a lot of conversation about US-China AI

Speaker A

race, right? We hear about this all the time. Is it actually a race? Yeah, I mean, it is a race and it's a race for AI dominance and AI development and AI infrastructure, right? And one of the reasons why is because there's big

Speaker A

money here. According to Joe Tsai, he's one of the co-founders of Alibaba, he put the figure of the total addressable market is some $50 trillion. Because AI, when dropped into a corporate setting, it's going to be replacing entire departments like HR, like accounting,

Speaker A

like coding and software engineering. And that's where we get to the threat to labor.

Speaker A

Absolutely. And that's why these companies are valued so highly too. And just a quick note, it's really hard for human brains to compute what a trillion means, right? Put it this way, one billion seconds is about 32 years ago. Guess how much one

Speaker A

trillion seconds is? That's 32,000 years ago. So we're talking about extreme exponential increases of wealth. So that's a motivating force here, right? Yeah, money. Money. But then we also need to break down what is AI, right? We keep talking about AI as this

Speaker A

big monolithic thing. One of the best explainers out there is from Jensen Huang from NVIDIA. And he says AI is a five-layer cake, right? The very bottom layer is the foundation and that's power. That's the electricity grid, the power infrastructure, right?

Speaker A

The level that's right above that is what turns raw electricity into computing power. That's the chips. Above the chips, you got another layer. That's the data infrastructure, the data centers. Then you have the models that could be your GPT, that could be your

Speaker A

LLM. And then on top, the icing on the cake is applications, right? Your chatbot, your autonomous robots, your self-driving taxis, all of that. China and the United States are lacking the same layer. And that layer is the advanced chip

Speaker A

making layer. Now, the United States has Nvidia, right? But that's a chip designer. It's not a chip manufacturer. And of course, as you well know, the United States is dependent on Taiwan and South Korea for that, right? And so you have little forces

Speaker A

like that that kind of motivate each other. China is trying to get more self-reliant on chips. The U.S. trying to make sure it keeps those alliances and it keeps its edge with those chips. So, you know, in this race, when did the starting

Speaker A

gun go off? I think the starting gun went off about 10 years ago in 2016. And that was when there was this AI called AlphaGo developed by Google DeepMind, beat the world's human champion in this Chinese chess game called Go. That was a

Speaker A

defining moment. We reported on it. We reported on it, talked to Demis Hassabis about it. He was amazed. He was exhilarated. He also thought that was a breakthrough that wasn't going to happen until 10 years later. China was watching very carefully. And within

Speaker A

a year, China reveals its AI development guidelines and says, "This is our plan and this is our goal to become a world AI leader by 2030." A national plan.

Speaker A

That's right. And it includes hardware, including self-reliance in hardware, includes talent and nurturing talent, and includes, "Do you know what? Let's mobilize our massive population so we can train our AI models on them too." This scaffolding is starting to build and it's

Speaker A

being introduced. And China is meeting those goals. Since then, its top tech companies have all been mobilized for the cause. Your Baidu, your Alibaba, your ByteDance, Tencent. And we've seen through the universities and labs, these incredible startups come up, including startups we all

Speaker A

know about these days, like DeepSeek, like Moonshot, like Zhipu and others. A quote from McKinsey here, AI ultimately by year 2030, that's the goal, will add $600 billion to China's GDP. So this is a 10-year-old national plan now that's been very

Speaker A

well underway. And I've watched as your colleague, as you've reported on different aspects of... the Chinese program, whether it's kind of robotics or drones or whatever. And some of it's been really visual and fun to watch, right? Yeah,

Speaker A

very visual, very fun to watch. And it underscores what China is good at in terms of AI. Kung fu robots. The US is good at the very sophisticated LLMs, right? And then China is good at the cheap and cheerful. On

Speaker A

open AI models and you talked about the Kung Fu fighting robots. Physical AI. Physical AI, exactly, the hardware. And yeah, so the last few years I've had a chance to go to this event called the Beyond Expo and it's one of Asia's leading

Speaker A

tech expos. And I've seen just how it's shifted from digital AI to physical AI.

Speaker A

When you get to see the humanoid robots, the quadruped robots, the flying EVs, the autonomous EVs, also smart glasses. I can see the text right in front of me.

Speaker A

And then I go

Speaker A

that like it has dexterity, it can recognize what's in front of it, it can interact with someone. And what the company is trying to do is say, this is just a precursor of what's to come. This could become like an AI humanoid butler

Speaker A

for the home. Just a quick aside, when I got to play with this robot, I kid you not, Yao Ming, the Chinese basketball player. So all seven foot six of him, he rocks up. And so it was like this peak China moment in

Speaker A

Macau, playing dice with this AI robot. And then Yao Ming comes in. It's like, wow, that was amazing. Oh, hi. I mean, again, it's fascinating, Christy, because you've been covering this for so long. And I feel people are just discovering this right now.

Speaker A

And China has been at it for years. But there was one moment, right, when I feel like China's AI progress really came out to the world, to the US and everyone else. Do you want to explain to our audience what was that DeepSeek

Speaker A

moment? Yeah, and this was just a year ago. I mean, this was around like, what, January of 2025. And DeepSeek comes out with an AI that rivals ChatGPT, but at a fraction of the price. And the stock markets had a complete freakout. because

Speaker A

with this moment they were able to just sort of break the aura of invincibility of American tech and they were able to do this without advanced American AI chips chips And that was just a shock to the system. I recall the

Speaker A

DeepSeek founder, he was making some comments to the media at the time. And he said, look, traditionally, China Tech would be one to two years behind our U.S. rivals.

Speaker A

But I see it as a difference between originality and imitation. And China, we can't follow. We can't imitate anymore. And so we're seeing real innovation and that's helping China pull ahead. I remember playing with DeepSeek at the time because we were

Speaker A

all reporting on it here in Hong Kong. And one of the interesting things about it was that it had the built-in Chinese censorship. I'm asking DeepSeek, has Chinese leader Xi Jinping ever made a mistake? Sorry, that's beyond my current scope.

Speaker A

Let's talk about something else. So if you wanted to ask about the Tiananmen Square Massacre or ask something about Chinese leader Xi Jinping, it didn't answer really well. It kind of like deflected. You had AI kind of, hey, let's

Speaker A

look over here about sensitive political subjects inside China, which is... part of the hallmark of tech and AI in China, right? This is such an important point to make because AI has allowed China to enhance and automate its censorship as well as

Speaker A

surveillance. There was a survey that came out recently by a group called Lattice Flow AI. And they said, okay, we've done a study of Chinese AI and there is a clear political bias, especially in regard to political topics. Okay, no surprise there. But

Speaker A

they also added, as we find out Chinese AI become more advanced and more sophisticated, the bias is becoming stronger and stronger. And that's just a reminder for people all over the world who are choosing to use certain AI systems. You have to be

Speaker A

aware of the inherent bias. On the issue of surveillance, this is something CNN has reported on. There are hundreds of millions of surveillance cameras across China, many of these cameras are equipped with advanced AI capable of facial recognition and location tracking. This is

Speaker A

a tool that is supercharging, turbocharging surveillance and censorship. I know there was censorship around the DeepSeek model, but one of the things that was new for people to see about it was it was an open model. And this is what China does really

Speaker A

well, right? It develops these AI systems, but then it puts it out in the market so people can actually test it and you can remove things, you can play with it, right? So you can move that censorship in your models as more and

Speaker A

more you train it, other people, and build on it, which is what the American AI companies don't do. And haven't we heard of some companies that have chosen Chinese AI products over the American ones for cost-saving purposes? Yeah, I mean, that is one

Speaker A

of the advantages of Chinese AI. It's an open weight system, it's less expensive. And this is also one of the advantages of AI in general in China. It's cheap.

Speaker A

You look at when we're talking about the five layer cake, right? Land is cheap, data centers is cheap, electricity is subsidized by the government, there's plentiful amounts of solar energy, hydropower, etc. So building AI is much, much cheaper in China. And also on

Speaker A

top of that, Engineers are cheaper. It's a fraction of the cost of a Western engineer. And that gives China another AI advantage. You know, in the US right now, there is, the polls are showing, deep skepticism and mistrust, frankly, of everything

Speaker A

from AI to the data centers that you know a majority of American survey don't want built in their neighborhoods and in their towns for all sorts of different reasons and that's being reflected now in campaign races and stuff like that on the political

Speaker A

side. What is How are Chinese people viewing all of this technology? Do they share this suspicion? You know, it's interesting because there is so much optimism in China in regards to AI. There was a survey done every year. The Stanford release is an

Speaker A

AI index. And in 2025, it said 83% of respondents in China are optimistic about AI. And that's more than half the share in the United States. Recently, our team in Beijing led by Mike Valerio went out to get some public opinion out in

Speaker A

the street in Beijing. And they got some really interesting remarks. Let's listen in. I think we should actually accelerate our AI efforts and boost our overall competitiveness.

Speaker A

We should encourage everyone to embrace AI. It feels like an unstoppable force. After all, look at me. I'm over 50 and I use AI too.

Speaker A

So there is a sense of optimism there because people believe that AI is reshaping society. It's also freeing them from dangerous work and dirty work.

Speaker A

But there are worries, you know, parents are worried. What are their sons and daughters going to study, right? Already youth unemployment is so high in China, right? I wanted to ask about that. If you have this kind of record high youth unemployment and

Speaker A

you're building armies of robots that can do Kung Fu and dance and stuff like that, I mean... Is China creating a future robotic workforce that will take jobs away from young people who don't have enough job opportunities to begin with?

Speaker A

That's the danger here. Already hundreds of millions of people are working in gig economy jobs. These are jobs that could be easily done by an autonomous drone or autonomous taxi or delivery vehicle one day. They're going to lose their jobs.

Speaker A

What's interesting is this is also being reflected in some of the science fiction that's being written in China. And I really believe if you want to understand technology and tech development in China, you got to read its sci-fi authors. You got to read

Speaker A

Three-Body Problem. You got to read Chen Xiufan and a work that he did with Kai-Fu Lee, the AI expert in China called AI 2041. And it talks about potential harms, like building a 21st century caste system, where you're going to have an

Speaker A

AI technocratic elite, And then you'll have maybe the former gig economy workers who are out of work and looking for meaning in their lives. Right. And the book also talks about, you know, other issues like, you know, the danger of autonomous weapons, you

Speaker A

know, as well as, you know, mass job displacement. So there is a discussion happening in China. And that's in a book that's published in China. Yes, it is. It is. It is sold in stores. It is. So But then like we also know

Speaker A

that there is no tolerance for dissent, open dissent in China. We know all of the data center protests in the United States, right? And the U.S. state of Texas put a moratorium on all data center being built in the state. I mean, you

Speaker A

can't imagine Zhejiang province doing anything like that. Well, I think considering China a one party state, I mean, dissent is not tolerated. And just like you have whistleblowers in the U.S. coming out, like, you know, recently talking about what these leading AI companies

Speaker A

are doing. I mean, you cannot imagine that probably happening in China, even if some people who are working there feel that way. Yeah. I mean, There are, as I mentioned, concerns about job losses. There are concerns, you know, as kind of implicitly mentioned

Speaker A

in sci-fi in China about what could happen to society. But also the Chinese government is introducing guardrails to anticipate these problems, right? I wanted to ask about that because the CEO of Anthropic, Daryama Day, in his kind of memo that set off a

Speaker A

lot of the discussion in the U.S., he was calling for the U.S. to work with China on establishing guardrails on AI. And I imagine that was not welcomed in Beijing. No, it wasn't welcome because yeah. For a lot of good reasons possibly. Yeah,

Speaker A

they thought number one, fear mongering, stop it. A little bit of like, don't tell us what to do. And also it's a self-serving comment. Of course, the US AI giants want to slow down because they want to lock in their pole position while

Speaker A

they're on front, right? But also, Beijing has its own guardrails in place. It was in 2023, just after ChatGPT comes out, China becomes one of the first countries in the world to introduce guidelines on generative AI, saying that it must be

Speaker A

in line with socialist values. I know that China has these guardrails and they've had this for years, but Isn't it a bit different on why China has those guardrails as compared to what the US companies are asking? Because for China, it's

Speaker A

more about controlling information and circulating something for political stability. Let me rephrase it this way. Is there any chance that Chinese leadership would allow a technology to move forward in their country without a kill switch? It would definitely have to have a kill

Speaker A

switch. Yeah. I mean, it's what, so what motivates guardrails in China? It's not fear of an existential, you know, end of days. It's not that. Like an air system going out of control. You just, you just hinted at it. It's stability. It's social

Speaker A

stability and it's national stability. And that's why we've seen rules in place for generative AI, rules in place against deep fakes. Also the latest rules that came out about companion AI to make sure minors are protected from friendly AI chatbots and that adults

Speaker A

are not allowed to have addictive relationships with these friendly AI chatbots. You know, just rule after rule after rule. all pointing towards, you want to guide this sector along, we will continue to make sure that all layers of that cake are going to

Speaker A

be there, but not at the expense of social and national stability. Well, then I think that leads us to this question. Can, in this period of incredible technological transformation, can the world's two largest economies, China and the U.S., cooperate when it comes to

Speaker A

AI? Christy? Just to answer that easier question in a word. We know there's like two schools of thought, right? So one camp says, no way. There's just no trust, right? There's no geopolitical trust. There's no technological trust. It doesn't help when you have

Speaker A

a US tech ban in place. Also, U.S. tech companies are angry at Chinese AI companies for using distillation. That's the process of how they train their models by distilling the models of the U.S. larger LLMs. They say, that's copycat. You can't do that

Speaker A

anymore. So there's that lack of trust element. Right. And then on top of that, in the US, there is no agreement between Silicon Valley and the Trump administration. Right.

Speaker A

And then you have everyone in Beijing looking at Silicon Valley and saying, we don't want to slow down because, you know, we don't believe in that. It's not fair.

Speaker A

That's self-serving. So what's in the yes camp? Can can the US and trying to work together? I believe that they can because there is a dialogue underway between U.S. and China, right? Is there? Yeah, there is. You have the trade truce

Speaker A

is still in place. There's business exchanges, geopolitical exchanges, cultural educational exchanges. There's also the interdependence between U.S. and China. You know, you have generations of Chinese engineers who maybe they did their undergrad degree in China, but they got their

Speaker A

doctorate in the US, they worked in US labs and then went back to China. You also have what's at stake, right? I mean, these are two big powers, shared planet. we need to come up with some sort of

Speaker A

common guidelines to be able to manage what could turn into an existential threat. And one final point, if I may, I mean, I also think it's not an all or nothing exercise, right? I mean, these are two powerful players, a lot of talented

Speaker A

people, amazingly powerful technology that could be used for good or for bad. But through collaboration and competition could also push each other to new heights and to create something that could really be for the benefit of humankind. You know, you think of the

Speaker A

space race, you think of getting to the moon. Like AI can be used to, you know, come up with new therapies to come up potentially for a cure for cancer. You know, just imagine if we can just keep our eyes on that sort

Speaker A

of prize. I love that perspective. I have a different view of this. I do want, of course, everyone wants the U.S. and China to work together, right? You don't share Christie's optimism? No, I don't. And I wish they did, but there's such

Speaker A

a huge trust deficit that you mentioned between the U.S. and China. Yeah. I mean, the Biden era rules on chip bans, right, which China, at least from the commentary I read on Chinese state media, is basically they view US as someone

Speaker A

who don't want China to be technologically equal. And that's distilled to the AI race, especially. And then you have these leading US AI companies who say openly that China winning the AI war is what dystopian world looks like. You have

Speaker A

some of those people saying that. I mean, I really don't think China is going to come to the table and be like, all right, yes, let's regulate together. Let's sit down at a table. Let's sit down and discuss how do we regulate our

Speaker A

models or our companies or the AI systems. And I think one key question is who's going to be the trusted middleman to broker this agreement? Because US and China, for the sake of humankind, we do need to come up with guardrails for AI.

Speaker A

So is it going to be Pope Leo? Is it going to be Jensen Huang?

Speaker A

Probably not. Probably not. Is it going to be an attentive journalist from not China or the United States, perhaps from India, who could ask the right questions and nudge people along? Could be you or Christy Lustad. Man, I've learned a lot here at

Speaker A

Christie today. And what I really appreciate is that for now, we humans are sitting around the table having this discussion. We haven't been replaced just yet. Yeah, not yet.

Speaker A

I really appreciate this discussion and speaking to a real expert. And that's all we have for this week's episode of Meanwhile in Asia. Iwan wants to plug in something.

Speaker A

Yeah, I'm going to show my age and what a lad I am. If you're enjoying this, feel free to write us on an old school email account. Not AI. Not AI. You can use AI to write your email. MeanwhileInAsia at CNN.com. We'd love to hear from you. That was

Speaker A

nice.

Topics:China AIAI raceUS-China relationsAI regulationAI technologychip manufacturingAI applicationssocial stabilityAI riskstech innovation

Get More with the SozAI App

Transcribe recordings, audio files, and YouTube videos — with AI summaries and speaker detection. 30 minutes free.

App Store Google Play

Or transcribe another YouTube video here →

Free tools: TXT to SRT · SRT Validator · Merge SRT · Subtitle to Text · All tools