In recent years, China has risen from being a Silicon Valley imitator to a formidable leader in technology, especially in the field of artificial intelligence. This ascent comes amid escalating Geopolitical tensions with the US and Europe, with AI becoming a catalyst for a new arms race in tech Innovation. The discussion explores China's transformation from a copycat to a pioneering force in AI Innovation, examining the challenges and strategies that have defined its growth.

Mustafa Suleyman, co-founder of DeepMind and CEO of Inflection AI, sheds light on the current dynamics of AI development and the global competition. While acknowledging China's progress and Innovation in AI, he emphasizes the strengths of the US in maintaining its leadership. With open-source technologies and large-scale corporate Infrastructure pushing advancements, Suleyman highlights the balance of power, regulatory strategies, and the potential future trajectories of AI.

Please remember to turn on the CC button to view the subtitles.

Key Vocabularies and Common Phrases:

1. Geopolitical [ˌdʒiːoʊpəˈlɪtɪkəl] - (adj.) - Relating to politics, especially international relations, as influenced by geographical factors.

Its ascent is peaking at a time when Geopolitical tensions with the US and Europe are intensifying...

2. Contender [kənˈtɛndər] - (n.) - A person or group competing with others to achieve something.

China has gone from a Silicon Valley copycat to a top Contender in technology on the world stage.

3. Imitation [ˌɪmɪˈteɪʃn] - (n.) - The action of using someone or something as a model.

China's playbook went from Imitation to Innovation.

4. Infrastructure [ˈɪnfrəˌstrʌktʃər] - (n.) - The basic physical and organizational structures needed for the operation of a society or enterprise.

The big companies have the resources and the people and the compute Infrastructure to be able to adapt and compete faster than ever.

5. Containment [kənˈteɪnmənt] - (n.) - The action of keeping something harmful under control or within limits.

...an appropriate Containment strategy, because much as this technology is manifested in software...

6. Proliferation [prəˌlɪfəˈreɪʃn] - (n.) - Rapid increase in the number or amount of something.

...adopt a kind of global anti Proliferation strategy...

7. Semiconductor [ˌsɛmɪkənˈdʌktər] - (n.) - A substance used in the production of electronic circuits, which conducts electricity but less easily than a full conductor.

China has definitely been investing for decades in trying to have their own independent Semiconductor supply chain...

8. Innovation [ˌɪnəˈveɪʃn] - (n.) - The action or process of innovating.

How did China go from being a copycat culture to a Contender when it comes to digital Innovation and specifically AI?

9. Ecosystem [ˈiːkoʊˌsɪstəm] - (n.) - A complex network or interconnected system.

The fundamentals in the US Innovation Ecosystem are unquestionably the best in the world.

10. Governance [ˈɡʌvərnəns] - (n.) - The action or manner of governing.

...improve our Governance and regulation, but also in actually building and making PI...

AI's New Frontier Exploring China vs. U.S. Tech Rivalry

Over the past decade, China has gone from a Silicon Valley copycat to a top Contender in technology on the world stage. Its ascent is peaking at a time when Geopolitical tensions with the US and Europe are intensifying and the rise of artificial intelligence has sparked a new arms race. The US is rushing to curb China's influence in AI, while balancing the need to regulate the powerful technology within its own borders. How America manages these competing priorities will have implications for where the balance of power in tech will shift to.

Next on this episode of AI IRL, we'll explore how China's playbook went from Imitation to Innovation, and discuss what's at stake in the competition between these two superpowers. Mustafa Suleyman, thanks so much for coming in. Now, you are the co-founder and CEO of Inflection AI and a venture partner at Greylock. Some people will know you as one of the co-founders of DeepMind, now owned by Google. Tell us what made you want to start fresh with Inflection.

Well, now is really the moment. I mean, it was very clear to me during 2020 2021 at Google when I was working on Lambda, the conversational AI that now powers Bard, that these large language models were truly going to be the future. And I wanted to start a company to build personal AI's, an intelligence that is completely aligned to your interests, on your team, and really functions as a personal assistant or a chief of staff. One that knows you intimately and over time can actually help you with practical tasks like scheduling, prioritizing, planning, and those kinds of things. And that's what you're working with. PI.

Exactly, yeah. PI stands for personal intelligence, and we see it as your personal assistant, essentially. How do you feel about the concentration of power in technology right now among the big companies coming out with their own AI models? Microsoft, Google Meta.

Well, I mean, I think you're right to point out that the big companies have the resources and the people and the compute Infrastructure to be able to adapt and compete faster than ever. Right. And that is a very impressive reality of how technology works these days. It requires a huge amount of compute. And at my own company, Inflection, we've managed to raise a lot of money and build one of the largest supercomputers in the world.

But on the flip side, the other trend that's taking place is that the open source movement is now almost at cutting edge and it's probably about twelve months behind the very largest models built at the big tech companies. And so that's kind of a contradictory trend. The big companies are getting bigger and better, faster. But also everybody is now going to be able to get access to these same capabilities in open source.

Let's fast forward a little bit into the present day where the world order has changed since you first entered the tech industry. How did China go from being a copycat culture to a Contender when it comes to digital Innovation and specifically AI?

Yeah, I mean, China is more than a Contender. I think in many respects, it's one of the leading players. Now. You know, they've always been incredibly creative, and I think that we may have, you know, sort of overstated the extent to which they were simply copying, you know, what we were doing. And I think it's important to remain humble about the fact that, you know, these ideas are going to be widely available to millions and millions of people. And, you know, in many respects, all of our cultures are, you know, equally creative.

And they've been able to take, you know, tools and ideas that are available both in open source, in the academic literature and move them forward, innovate at the cutting edge. They've produced a lot of very good papers, contributions to the academic conferences that have been peer reviewed. So I just think overall, it's better to take an approach of humility, acknowledging that they're also going to be a really significant player in this field over the next few decades.

How would you rank the United States in terms of technological prowess in that world order? Well, I mean, there's no question that we in the US and in the west more generally are currently in the lead. We have the very best, most innovative technology companies and technology hubs. Without doubt, Silicon Valley is leading the pack. And I think we'll continue to. I think we worry too much that we're going to sort of be dislodged from our podium position.

For how long, do you think? Well, I think the question is, what does it look like to be dislodged? And is that really the right framing in the first place? Like, what is the difference between first and third? And, you know, people talk about there being a race for AI, and it's very unclear what that looks like to come second or third in that race. I think it's sort of the wrong metaphor. We're talking about super powers, Clash of the Titans sort of stuff.

It reminds me of back when the previous US administration banned companies from doing a lot of business with Huawei, and Huawei ended up basically having to get a lot better very quickly at making its own chip designs and all that kind of stuff. And it feels like it backfired a little bit, yeah.

You've also advocated for the US restricting sales of some Nvidia chips, which are crucial in training AI systems to buyers who don't agree or who refuse to use them ethically and safely. So what do you think is the best way the US can actually flex its power? Is it through the technological Innovation, or by stifling effectively some of its adversaries through regulation?

The first thing I should say about that is I don't think now is the time to do that. I mean, I think in theory, when the time comes, if the time comes for us to adopt a kind of global anti Proliferation strategy, we're worried about technology being so powerful that we have to stop their spread. If that moment arises, and I think we're quite a long way from experiencing that, then technically that would be an appropriate Containment strategy, because much as this technology is manifested in software, it lives on the Internet, it still sits on chips, whether you're training a model or serving a model.

So in order to ask a model a query, you have to have access to one of those Nvidia GPU's. And for what it's worth, there are really only one major company in the world that produces those high end GPU's, Nvidia. And they manufacture the vast majority of their chips at TSMC in Taiwan that use one single company, ASML, a Dutch company that manufactures the core chip that produces the chips. And they sit in the Netherlands. So it's actually a very narrow supply chain. And so I think the point I was sort of trying to make is that if you wanted to adopt a Containment strategy, there is a supply chain where you can impose chokeholds on it.

In ten years from now, which country do you think will be leading the world in AI? Is it going to be the US? Will it be China? Will it be Europe? It's going to be the US, I'm pretty sure of that. The fundamentals in the US Innovation Ecosystem are unquestionably the best in the world.

What flipped that balance of power? Well, I think it's hard to say. I mean, I really do think that the US is in the strongest possible position. At the end of the day, it's talent that drives this Innovation, and talent still concentrates around a small number of physical places. Despite everything that we've experienced in the pandemic and people wanting to work remotely, it's still really important to be physically proximate to other people who are working with you in the office. I think that that trend isn't going to change.

I do think China is going to catch up. I think they're going to develop their own AI chips, their own fabrication facilities, but that is going to take five years or so. The export controls that the US has implemented are real, and they really will prevent China from getting access to the next frontier of models. So the GPT five onwards. Right now, I don't really see how China could train those kinds of models.

Is there any downside, any unintended consequences that could come from this Containment strategy the US is employing? Yes, I think it is a very aggressive strategy. The counter attacks have already started. Right. China restricted gallanium, which is a raw material used in chip manufacturing, and I expect them to have their own reactions in time. We're very dependent on their supply chain. And so the idea that we could completely decouple, I think is misguided.

I'm going to just put you on the spot really briefly and wonder, can you tell because we're all friends here. Sure, go for it. Do you have a major regret about your work in AI? And do you also have a major single goal?

Yeah, my major single goal is to show that it is possible to create safe and ethical artificial intelligences which are accountable to humans. There is a possible trajectory where some groups may recklessly experiment with algorithms that can self-improve. And if in ten to 15 or 20 years time, those kinds of AI's are set loose on the open web in a reckless way, then I can imagine that possibly causing serious harm. And so for as long as I can remember, almost 15 years that I've been involved in AI, I've been trying to raise awareness and concern about the possible ways in which it can go wrong.

And part of doing that is to show a bunch of practical ways that it can go right, both with writing my book and laying out a series of Containment strategies and things that we have to do in politics to improve our Governance and regulation, but also in actually building and making PI, which I think is hopefully going to continue to remain a very safe and controllable personal AI that can work for you. Mustafa, thanks so much for coming in. Been a pleasure. Thanks very much for having me. It's been great.

Helen Toner, thank you so much for joining us. You're a director at Georgetown center for Security and Emerging Technology, and you're also an OpenAI board member. Dan Wong, thank you so much for joining us. You're a visiting scholar at the Tsai China center at Yale Law School, and you've been studying technology and China for many years now. And we've just heard from Mustafa Suliman and he thinks that China's emerged as one of the biggest contenders in AI.

So, Helen, I'm just curious to start with, do you think we're overestimating China's capabilities? It's certainly true that they're a big player and it's also true that we shouldn't underestimate them and that we have a bit of a history of underestimating China's ability to innovate. I mean, I think 10-15 years ago, the story, at least in Washington, DC, where I'm based, was, yeah, China can innovate. If they're producing things that look like new technology, it's because they've stolen them. I think that's clearly no longer true.

At the same time, I think it's also very possible to overestimate them. And sometimes I think I see that as well. So are they a serious competitor? Yes. Are they eating our lunch? No. So that copycat culture, it's fair to say that most people are thinking we're past that stage.

Dan, what do you think of that? Well, I think it's certainly the case that China focuses on different things. So President Xi Jinping is, I think, a little bit less enamored with the Internet platforms, with a lot of the cryptocurrencies, with a lot of the AI businesses that Silicon Valley specializes. He's also a little bit less enamored of Wall Street and financialization and hedge funds. And if I had to think about what his broad views are, it would be much closer to being something like capitalism with German characteristics, which is much more about building cars, building hydraulic pumps, building heavy machinery. The heavier, the better. About all sorts of these are major industrial projects.

And I think China is very much a leader in things that include electric vehicle batteries. Now, solar photovoltaics, it is making a lot of electronics products, and it is innovating on a huge sort of a lot of the automotive supply chain. But I think they have a little bit less of a focus on AI, and therefore there is not so much political direction anymore to really push forward to these consumer applications. I think it's fair, though, some of those areas, the US is also innovating very, very strongly.

So. Are there any areas where China is innovating significantly better than the US? Well, I would say something like electric vehicle batteries, where China has something like 70% of all global expansion, all global capacity of electric vehicle batteries is now mostly in China. When it comes to the clean technology supply chain, again, these very industrial projects these pieces of heavy machinery that is a little bit more the focus.

I think the US is really strong in things that include semiconductors, artificial intelligence, biotechnology. These are things, the smaller things, where China is not so good, but there are big areas in which China is way outbuilding the US. I mean, the semiconductors is getting significantly better in China, partly maybe as a result of the Huawei issue.

Helen, what's your view on the Semiconductor side of this? So China's definitely been investing for decades in trying to have their own independent Semiconductor supply chain, and they've been making some progress. The thing to know about semiconductors is there's a huge range of sophistication from the really basic semiconductors, computer chips that go in basic devices, your smart fridge, your car, up to the kind of more advanced chips that might go in a brand new iPhone, up to very advanced chips that go in the data centers, gigantic warehouses full of computing, which is what you're training advanced AI systems on.

So China has made a lot of progress on the more basic end of the Semiconductor sort of scale and is still really struggling to hit the most advanced end. And that's where, I mean, it's not even, you can't even say the US is alone able to create those advanced semiconductors, but what you have is an incredibly advanced global supply chain.

Many describe artificial intelligence and its development across these two superpowers as an arms race. Is that framework misguided, do you think? I think it's a bit misguided. I think you really want to look at what kind of AI are we talking about and in what way might it be a race. I think the place where it's clear that it makes sense to think in a competitive framework is in kind of military domains.

So, you know, it's just true that, you know, the US military really wants to make sure that it is keeping Pea saw with, and ideally a little ahead of, you know, the Chinese or a lot ahead of the Chinese military in terms of what technology they have.

Dan, Washington has taken some steps to curtail some of China's ability to get their hands on the cutting edge chips that really power these AI platforms. What's your take on that? Is that an effective strategy, or will China just find a way to build their own? I think that China is going to try to figure out a lot of the way to build their own. A lot of the chipmakers now have been under very significant stress from us sanctions dating back to about a year ago now released by the Biden administration. That is really stopping their ability to access not just the high-end chips, but also the Semiconductor production equipment as well.

And I think, in general, it's really difficult to distinguish between what kinds of technologies should the US really not want China to be able to develop and where is it? Fine. So, you know, you mentioned, you know, clean technologies before where China, if you're looking at, you know, manufacturing solar panels, as you mentioned, EV batteries, a lot of other areas, China is doing amazingly at that. And what you're starting to hear in some us policy making circles is that maybe that's a bad thing, that China is making a lot of clean technology stuff.

And I find that a little sort of tricky to reckon with is at what point do we want to really shut them out or prevent them from being able to manufacture anything? If it's something good, useful that we can use, of course, if it's technologies that we think have really significant military value, I understand. But then it starts, the lines between those categories start to get very, very blurred.

It seems like these tensions are just ratcheting it up even higher as time goes on with each administration. They're intensifying. So to your point, if we do need these other technologies, solar technology, for this transition into a greener future, I mean, what kind of happens next? How much do we place on kind of the AI arms race that could then potentially work against us in climate change? Is that a concern or is that too far in the future still?

Well, a few months ago, Beijing did announce some export controls on certain solar technologies involved in making production wafers. So it's already happening. It is already happening that there are some threats out there, that relatively speaking, Beijing has been pretty restrained against retaliating against American companies of the likes of Intel, of the likes of Tesla so far. But these things could escalate more.

How does the relationship between the ruling party in China play out against the kind of the development of companies? I think the Chinese system is definitely more top down when it comes to AI development. You see less individual tech companies like OpenAI or Google taking the lead and more these kinds of multi-institution partnerships, or maybe you have a university and a tech company and a government department all collaborating on big AI projects. That's much more common in China.

At the same time, China is a huge country, a limit to how much the government can achieve from the top. So a lot of what you see is sort of Beijing providing direction, providing ideas, providing sort of signaling of this is what we and the central government would be excited about.

Dan, how would you compare how the Chinese government is handling AI with its people versus how they handled social media?

Well, I think President Xi Jinping doesn't necessarily view AI right now. A lot of these Chachi PT like applications as some of the most economically productive things that are coming into our society. So I think that President Xi looks much more at AI like he would at Facebook or Twitter, now known as x or whatever it is.

And these are platforms like TikTok that, if you're asking me is something like TikTok or Facebook really increasing my personal productivity? Well, I would say, actually, it is really horribly detracting from my productivity. And so I think that a lot of these AI platforms right now create speech issues. When an ordinary Chinese person sitting in Beijing or Shanghai asks AI, well, what happened in 1989? Well, what do you think of socialism? What do you think of President Xi's performance?

These AI tools are having a lot of challenges actually answering socialist correct values that are in line with the government's ideas. And so these are major speech platforms first and foremost, and perhaps not really the economically miraculous thing that raises our productivity growth at the very moment. Now, it might happen later on. Right now, I think the party views it as somewhere between economically not that useful and potentially very socially destabilizing.

So from a consumer standpoint, AI is almost, it can potentially create more problems for the Chinese Communist Party. But then if you look at it from, like, a surveillance standpoint or military type of AI applications, that's kind of in a different camp. Is that right?

Yeah, that's right. And I think there are definitely, I mean, one area where China is, I would say, leading in AI is on surveillance. They really have leaned hard into facial recognition, you know, voice recognition, even gait recognition. So how you walk in a crowd, can they pick you out by that? I think in the west, there's been some development of these technologies, but with a lot of criticism from civil society, a lot of hesitation, sort of, I would say, like very yellow light. And in China, it's been blaring green light.

On the topic of WeChat, I mean, when we spoke ahead of this, I think we touched on the fact that there are some generalized misunderstandings about China's approach to social media and data and AI. AI relying on huge amounts of data. And the perception is that because of things like WeChat and similar things in China, that there's this huge pool of data that the population in China is very happy giving over and is used to giving over, and therefore is just sitting around waiting to be exploited by AI. And I remember that you made a pretty significant point. It's like, that's probably not really the case, and people think that's the case, but it isn't.

So I wonder if you could touch on that. Yeah, I think people have this perception that because the Chinese government is very controlling and has a lot of access to what's going on in Chinese citizens private lives, that that means that there's these beautifully curated, you know, extremely complete, large databases of information that can be used to train AI systems or to do other things.

But for anyone who's ever worked with data, you know that cleaning data, structuring data, making sure that the right information is flowing in the right places is incredibly challenging. So it's both true that the Chinese government can get very detailed information on the private lives of its citizens if it chooses to, and also true that it's incredibly messy, the way that that works, and incredibly difficult for any given actor to get access to something.

So, for instance, a company trying to train some AI model using data is actually subject to a pretty significant data Governance setup in China. Data Governance regulations that make it really challenging to use customer's data for things it wasn't already designed for, that you didn't already get consent, for which we learned from Cambridge Analytica scandal.

Just as one case, and there are many others, that is not the case in the US, it's not the case in Europe. There's a lot going on. The EU's AI act is pretty widely praised by some people, not everybody, but it's quite broad and far reaching, but that's not the case in China in terms of what you're saying.

So China's setup is now more similar to the GDPR that exists in Europe. So around kind of protecting consumer data and giving people the right to opt in to different uses. Of course, in China, there's always the exception, unless the security services think that you might be a dissident, and then they're going to get whatever they want on you. But in terms of kind of companies, they're certainly relatively restricted in what they can do.

Switching gears a little bit, Dan, I'm curious about how China's sluggish economic growth recently weighs into the government's calculus and how they're thinking about AI's potential to spark social unrest. I think that the Communist Party is very much involved in trying to think about, you know, how to control possible social unrest. I was in Shanghai in November 2022 when there were protests in the streets denouncing zero Covid, denouncing the communist party by name that these were things that were erupting.

And so I very much signed on to Helens description of AI that China is leading in the AI of censorship, it is leading in the AI of facial recognition. That right now it looks very much like the US is a place where ordinary consumers can tinker with and play around on a lot of these AI tools, whereas in China, China is mostly reserving these very powerful tools for its own use potentially to control society. That right now China is in a moment of economic weakness. This might deteriorate further, this might get a little bit better, but there is always this presence hovering over there that the state security might ramp up its technology really to make sure that there are no socially distinctive stabilizing incidents.

Well, Helen, Dan, thank you so much for being here. It's been great to learn from the both of you. Thanks so much. Thank you very much.

Artificial Intelligence, Technology, Innovation, China Tech, Geopolitics, US-China Relations