Eric Schmidt谈AI、与中国的较量及美国的未来

[Music] I honestly believe that the AI revolution is underhyped. Now, why is this all important? >> Eric Schmidt is here. He's the former Google executive chairman and CEO. >> These agents are going to be really powerful, and they'll start to work together. We're soon going to be able to have computers running on their own, deciding what they want to do. Now we have the arrival of a new nonhuman intelligence which is likely to have better reasoning skills than humans can have. >> So if you were emperor of the world for 1 hour, the most important thing I do is make sure that the west wins. >> Ladies and gentlemen, please welcome Eric Schmidt. [Music]

Hi. Hi. Looking good. Good to see you.

Good to see you.

You're looking, Eric. You're looking. Very nice.

Oh my God. David, good to see you.

David Sax is here as well. It's like a reunion of all of our former companies. David, why did you quit after all?

My old boss.

What was it like working with Young Freedberg? Take us back.

Can I tell a story? We had to come down to Orange County, and they were like, “Hey, we're going to take the plane.” It was Eric's plane. We got on the plane, and then he went up and flew the plane. I'm in the back of the plane by myself.

Was it a King Air?

I was like, “The CEO of Google is flying me down to Orange County.” It was incredible. That was my first time actually hanging out with Eric.

It was my Gulfream.

That's right. He was way too smart.

Way too smart. Was he focused? Did he contribute? Did he move the needle?

But he was very smart. Okay, that's kind of our consensus of the pod as well.

Look, you guys know this guy well. He's really that smart. He taught me more stuff than most of the employees at Google, and then you left.

Well, tell us what you've been doing—no, no, wait. Before that, I have to ask you this question. There was a recently deleted video from Stanford.

Oh no.

You had a moment of clarity where you said, “Hey, at Google, people are like, ‘Too much work-life balance. They need to commit. They need to work harder.’ We had Sergey at the last event. He's going back to work, so Sergey got the message. Predicting Sergey's behavior is something I can fail at. I tried for 20 years. I am not in favor of essentially working at home.

“And the reason is, many of you guys work at home to some degree, but your careers are already established. Think about a 20-something who has to learn how the world works. They come out of Berkeley or Dartmouth, and they're very well-educated. When I think about how much I learned when I was at Sun, just listening to these older people who were 5 or 10 years older than I was argue with each other in person, how do you recreate that in this new thing?

“And I'm in favor of work-life balance, and that's why people work for the government. Sorry.”

That's right.

Sorry. If you're going to be in tech and you're going to win, you're going to have to make some trade-offs. Remember, we're up against the Chinese. The Chinese work-life balance consists of 996, which is 9:00 a.m. to 9:00 p.m., 6 days a week. By the way, the Chinese have clarified that this is illegal. However, they all do it.

That's who you're competing against.

I brought everybody back to the office. It's so much better.

So, let's just pick up on that theme. You don't need to defend the government.

No, no. Believe me, I don't see the need to. I'm an unpaid, part-time adviser to the government.

But we are in this high-tech competition with China. They obviously care about AI, too. They're trying to race ahead. I understand that you recently made a trip there. How do you handicap this competition?

You and I just talked about this as part of your incredibly important work in the White House. I had thought that China and the United States were competing at a peer level in AI, and that the good work that you and your predecessors have done to restrict chips was slowing them down. They're really doing something different from what I thought.

They're not pursuing crazy AGI strategies, partly because of the hardware limitations that you've put in place, but partly because the depth of their capital markets doesn't exist. They can't raise, based on a wing and a prayer, $100 million—or maybe the equivalent—to build the data centers. They just can't do it.

The result is that they're very focused on taking AI and applying it to everything. While we're pursuing AGI, which is incredibly interesting and which we should talk about—and all of us will be affected by this—we'd better also be competing with the Chinese in day-to-day stuff: consumer apps, which is something you understand very well, Chamath, robots, and so forth and so on.

I saw all the Shanghai robotics companies, and these guys are attempting to do in robots what they've successfully done with electric vehicles, right? Their work ethic is incredible. They're well-funded. It's not the crazy valuations that we have in America. They can't raise the capital, but they can win across that.

The other thing the Chinese are doing—and I want to emphasize this because it is a major geopolitical issue—is that my own background is open source. In the audience, you all know that open source means open code. Open weights means open training data.

China is competing with open weights and open training data, and the US is largely, and for the majority, focused on closed weights and closed data. That means that the majority of the world—think of it as the Belt and Road Initiative—is going to use Chinese models and not American models.

I happen to think the West and democracies are correct. I'd much rather have the proliferation of large language models and that learning be done based on Western values.

Eric, we had a major open-source initiative with Meta. They have an incredible balance sheet and tremendous technical firepower, but they seem to have misexecuted and are now taking a step back and reformulating something that, to your point, looks a little bit more closed source.

It's not clear. Alex Wang's a good friend. He's come in, he's taken over, and he's obviously incredibly capable. I would not say that they're going fully closed.

I think they also got screwed up because the DeepSeek people, with R1, did such a good job, right? If you look at the reasoning model in DeepSeek, and in particular their ability to do reinforcement learning forward and back, forward and back, and forward and back, this is a major achievement. It appears that they're doing it with less precision—less numeric precision—than the American models.

As a bit of a technical thing, there's something called FP64, FP32, and FP16. The American models are typically using 16-bit precision for their training. The Chinese are pushing eight and now even four.

Is there something that the larger American companies need to be doing in open source so that we can actually combat this?

A number of the large companies have said that they want to be leaders in open source as well. Sam Alman indicated that the smallest version of the 03 model would be released, I believe, with open weights, and they have done so. He told me, anyway, that this model is much smaller than 10 the 26. It's much easier to train, and it will fit—or can fit—on your phone.

One path is to say that we'll have these supercomputers doing AGI, which will always be incredibly expensive and so forth. But we also have to watch to make sure that the proliferation of these models for handheld devices is under American control, whether it's OpenAI, Meta, Gemini, or what have you.

Recently, you took over Relativity Space, and I think for people who don't know, this is a business whose ambition is effectively to compete with SpaceX.

I think you were the first investor, or the earliest investor, in it.

I was.

I'm sorry.

It's okay.

You lost some money in the first round. What happened? Did you get crammed down?

No, no, no. I mean, all of us did.

All of us—I mean, look.

I've been very happily an investor in SpaceX, Swarm, and Starlink. Relativity was—

And by the way, Swarm is a big deal.

So thank you.

Yeah, Swarm has been a really great success for them and, I think, for the world.

But what I was going to ask you is, walk us through the evolution of the space market. Why did you decide—of all the companies you have the capital base to put your money into anywhere—why did you pick that business? Why now?

Rockets are really cool, and they're really hard. As you know, I'm a pilot and I know lots about jets, and I had assumed that rockets were as mature as jet engines. They're not. It is an art and a science. These things are very hard to do.

The amounts of power—I mean, in our case, the rocket is 4 million pounds of thrust. You have to hold the thing down to test it. You can't even hold it with metal things. You have to have other things to hold it down as well. There's so much force; otherwise, it will take off.

Another interesting thing about rockets is that a rough number is that 2% of the weight of the rocket is the payload, 18% is roughly the rocket, and 80% is the propellant. My reaction as a new person is, “You're telling me you can't do any better?” The physicists say, “After 60 years of physics, that's the best we can do to get out of the gravitation of Earth.”

And so, I think rockets are interesting and they're challenging. There's always an opportunity for competition. In Relativity Space's area, it's essentially a LEO company, a LEO competitor. So, low Earth orbit satellites, that sort of thing. The order book is full. We just have to launch the rocket, and this entry into space happened. I'm not sure how well known this is.

So, you can go as far as you want to go into this, but you've done a lot as well in next-generation warfare. Do you want us to talk about that, how you ended up there, and what role that plays, and just give us a landscape? Maybe David asked about the China question, but they're all kind of interrelated.

Well, in the first place, I'm a software person, not a hardware person. I explain to people that hardware people go to different schools than software people, and they think slightly differently. So, I'm always at a limitation in these new industries.

I had worked for the Secretary of Defense and had a top-secret clearance and all that. I was given a medal for trying to help the Pentagon reorganize itself. When the Ukraine war started, I was watching, and I thought, well, here's an opportunity to see how a country that has no navy and no air force does this with automation.

And, indeed, it has been a spectacular success as a matter of innovation, even though Ukraine was outnumbered 3 to 1, with huge differences in kinetic strength, weapons, mobilization, and so forth. Ukraine has held on really quite well.

What's happening now is you're seeing essentially the birth of a completely new military-national-security structure. One way to think about it is that, first of all, I've seen it live, and I will tell you that real war is much worse than the worst movies you have ever seen about war. That's all I'll say. It's really horrific, and it's to be avoided at all costs.

And then, for obvious reasons—and I love all these people who say, "Well, you know, the warmongering talk"—be careful what you wish for, because the other side gets a vote. When I started working and trying to understand what Ukraine was doing, Russia was pushed back, and they've come back with a very, very strong second and third round. So, the enemy gets a vote in this situation.

But to go on, the rough way in which war will evolve is that, first, things will have to be very, very mobile and very much not in fixed places. This takes out most of the military infrastructure that exists in the world.

Things like tanks, of which we're now building a whole bunch more—stronger tanks here in America—don't make any sense in a world where a 2-kilogram payload from a well-armed drone can destroy the tank. It's called the kill ratio. That drone costs $5,000 or $4,000 at retail. The American tank costs $30 million. You can send an awful lot of those drones to destroy those tanks.

The likely evolution goes something like this. First, people learn that drones are like rifles and like artillery. It's more efficient to use drones now than to use mortars, grenades, or artillery. That's clear if you just look at the economics, in terms of cost-effectiveness, as it's called.

The next thing that happens is that both sides develop drone capabilities, which is what you're seeing now, and each then becomes a war of drone against drone. So, you have drone against anti-drone. The shift then moves to: How do you detect the enemy drone, and how do you destroy it before it destroys you?

The doctrine ultimately is that the drones are forward and the people are behind. I've seen operations, for example, sitting in Kev, where the Ukrainians are commanding things over Starlink, I might add, at a distance in the drone war, and they're very, very effective. So, we've solved the latency problems, we've solved the timing problems, and so forth in that area.

The ultimate state is very interesting, and I don't think anyone has foreseen this. If you go back to our conversation about RL and planning, which is what you're seeing with AI, let's say that we're on one side and we have 1 million drones, and there's another side over here that has another 1 million drones.

Each side will use reinforcement-learning AI strategies to develop battle plans, but neither side can figure out what the other side's battle plan is. Therefore, the deterrence against attacking each other will be very high.

Today, the way military planners operate is that they count weapons. They say, "Well, you have this many and I have this many, and you can do this kind of maneuver," and so forth. But in an AI world where you're doing reinforcement learning, you can't count what the other side is planning. You can't see it. You don't know it.

I believe that will deter what I view as one of the most horrendous things ever done by humans, which is war. Because unless there's a perfect balance between the sides, there will be some mutual destruction of the drone supply, like there would be with any artillery stock in traditional warfare, and whoever's left ends up winning.

Well, it's very important to understand that there are no winners in war. By the time you have a drone battle of the scale I'm describing, the entire infrastructure of your side will be destroyed, and the entire infrastructure of the other side will be destroyed. These are lose-lose scenarios.

Isn't there an equilibrium, though, that this can also create, where, because of that mutually assured destruction, there's a deterrent?

Well, I'm arguing that it's not a deterrence.

Right.

Deterrence can be understood as: I want to hit you—which I don't—but I want to hit you so much that, if I do that, the penalty is greater than the value of me hitting you.

Right.

And that's how deterrence works.

But that seems like a great advantage and upside of this move toward drones and automation that we don't have today.

Well, there are many advantages to moving to drones and automation. One, they're much, much cheaper, right? They're much, much cheaper.

Yeah.

And two, you can stockpile algorithms. You can essentially learn and learn and learn. And remember, you can also build training data that's synthetic, so you can be even better than the others.

The final question I've been asked by our military is: What's the role of a traditional land army? I wish I could say that all of these human behaviors can occur without humans being at risk. I don't think so.

I think that the way robot war—essentially, drone war—will occur is that there will be these destructive waves, but eventually humans are going to have to cross a line.

After we've depleted them. So, you're investing in this drone technology, and then do you think Optimus and humanoid robots are the next volley in this new warfare?

It's going to be a long time before we see humanoid robots, which is what we see in the movies all day, right? It will be a very long time before we see that.

What you're going to see is very, very fast mobility solutions, right? Air-based solutions and also hypersonics.

Hypersonics, also things underwater. There's a lot of that going on. It's a different domain. If you look at the muro and some other boats that the Ukrainians used, they have essentially used USVs to destroy the Russian fleet in the Black Sea.

This was crucial for them because they needed to be able to export the grain from Odessa, and it's like 6% or 10% of their economy. It's a very big deal, and they did that with drones.

Eric, it seems like there's this overarching worldview that you have. You have this view on AI. There's all the stuff you're doing now in drones, in warfare, and in rocketry. It all converges, quite honestly, because in the next 5 or 10 years, these things will all come to pass.

How do you view the world? What is the role of America? What is your role as a capitalist, as a technologist, as a statesman?

I want America to win, right? I am here because of the American dream. The people who invested in me—in my case, Berkeley and so forth—people took a chance on me. I want the next generation to have that.

I also want you all to remember that I was just in Honolulu as part of the World War II surrender ceremony, and they talked about fighting tyranny. We forget that our ancestors, our great-grandparents, or whatever, fought the Great War to keep liberalism and democracy alive. I want us to do that.

How do we do that as Americans? We use our strengths. What are our strengths? We're chaotic, confusing, loud, but we're clever. We allocate capital smartly. We have very deep financial markets. We have this enormous industrial base, universities, and entrepreneurs, which are represented here.

We should celebrate this. We should stoke it. We should make it go faster and faster. I spend lots of time in Europe because of the Ukraine stuff. They are so envious of us. When you're in Asia, they are envious of us. Don't screw it up, guys. That's what I want to work on.

Can I just ask you about something outside of this external conflict? We had a conversation with Alex Karp today, and we actually had Tucker Carlson here yesterday. Some of the dialogue was around—I don't know if the right term is—the erosion of the West, and that there may be social issues brewing in the West that may be hurting us from the inside.

How much do you observe or spend time on these issues? The metric that often is cited now is declining birth rates in the West, and that our population—and we're going to talk with Elon in a few minutes about this—

Oh, sorry.

Oh, we just ruined my surprise. My bad. Oops.

Sorry. There’s your surprise guest. Sorry. Elon is a good friend, and he’s addressing this issue of population directly himself, trying to problem-solve for it. Good for you.

Is it a reflection of something going on? There’s a rise of Mandani getting elected in New York. Some of the historic values of the West seem to be under a state of transformation right now.

One metric of the success of a society is its ability to reproduce. I think this is a legitimate concern for the West. It’s much worse in Asia. The Chinese number is about 1.0 0 for 2 parents. In Korea, it’s now down to 78 for 2.

It’s really important to recognize that we as humans are collectively choosing to depopulate. The numbers are staggering, right? Imagine a situation where, instead of having growth, you have shrinkage. Furthermore, they’re getting older.

As a business, all of a sudden, your revenue is declining. There’s nothing you can do because you can’t innovate with fewer and fewer customers. If you put it in a business context, ignoring the moral issues, which are all very real, it’s just bad, right? We have to solve that problem.

I happen to be broadly in favor of immigration because I think immigration helps us solve that problem. But as a global mechanism, we have to address that.

In any case, from my perspective, you’re going to have these issues, but America is organized around the concept of American exceptionalism. As long as we understand that the way we make progress is by investing in the right people and the right businesses, having a strong capital market, and investing in the infrastructure that they need, we’ll be fine.

That is my actual opinion. Can we go back to AI for a second?

Eric, I think you can help us get to, let’s call it, a bipartisan understanding of these issues. I think you think really clearly about this. In the wake of Chad GBT launching at the end of 2022, I think the discourse was really dominated in 2023 and 2024 by this idea of AGI, and that AGI was imminent.

I think it created almost a panicky atmosphere in Washington among policymakers. You saw things like, “We’ve got to restrict open source, because then China will get it.” This was before DeepSeek launched, and then we saw that actually they’re ahead of us on open source.

It feels like there’s been a pullback a little bit from the AGI narrative, which I think is actually a good thing. I think it’s more conducive to calm, rational policymaking. What’s your perception of AGI right now? Where are we on that whole train?

So, first of all, the speech that the president delivered about a month ago about AI strategy, which I think you probably wouldn’t say, but you kind of wrote it for him, was exactly right. So, thank you.

David collaborated with an amazing leader who we all respect and admire so much, Eric.

Yes. Nevertheless, saying I wrote it was way too strong. Actually, but anyway—

If you didn’t write it, then it must have been your twin. But in any case, you got the emphasis right, which was that investment in research and investment in the kind of stuff that we do is really, really important.

I don’t agree with you on this AGI thing because there’s this group which I call the San Francisco narrative, because they all live in San Francisco. Their narrative goes something like this: Today, we’re doing agents. The agentic revolution will change businesses, which I agree with. What happens is that the systems will become recursively self-improving, with recursive self-improvement, as it’s called.

If you have a scale-free problem—and a scale-free problem, for example, is programming or math, where you can just keep doing it—you get these enormous, fast gains if you buy enough hardware, do enough software, and so forth. That is still underway.

The collective view of that group is that in the next 3-ish years, they believe we will get forms of superintelligence. The way they define it is basically a savant: a chemistry savant, a physics savant, or a mathematics savant. I don’t agree with 3 years, but I do agree that it’ll be maybe 6 or 7 years.

But if it’s a savant in a particular area, is that general intelligence?

It’s not general intelligence yet. general internal intelligence is when it can set its own objective function.

Right?

And there’s no evidence of that. There’s no evidence right now of the ability to set your own objective function.

The thinking—and I’m writing a paper on this, so I’ve been studying it—is that the technical problem is non-stationarity of mathematical proofs. What you’re doing is trying to solve against an objective function, but the objective function keeps changing, which is how humans operate. Your goal changes every day, whereas computers have trouble with that. As a math problem, we don’t have an algorithm yet for LLMs that can do that. People are working on it.

The test will be: Can you, basically, using the information available in 1902, derive the same thing that Einstein did with special relativity, followed by general relativity? We cannot do that today.

Most people believe that the way this will be solved is through analogy. The theory of great geniuses is that they understand one area extremely well, and they’re so brilliant that the person can then take their ideas and apply them to a completely different domain. If we can solve that problem, then I think it’s over. Then we get to AGI, and it’s a whole different world.

I think one of the reasons why it’s hard to replace a human—and JK and I debate this—is that humans are end-to-end. We can do the whole job. You have a sort of complete understanding, and you can pivot very easily.

AI, at least as we know it today, is not end-to-end. It has to be prompted. You get an answer, and that answer has to be validated. Then you have to ask a new question because it never gives you exactly what you want. You have to apply more context and go through an iterative loop. Finally, you get to an answer that has business value.

The way biology puts it is that AI is not end-to-end; it’s middle-to-middle. Humans are end-to-end. As a result, instead of AI replacing all of us, AI will be very synergistic with humans because we can define the objective function, do the prompting, and work with it to iterate. It does a lot of the work in the middle.

That seems to me like a very optimistic, less dystopian take on it. What you just said is exactly what’s going to happen for the next few years: each of us will have assistants that, on our command and with our prompting, will be incredibly helpful for whatever problem we have.

People are using these things for relationship advice, for talking to their kids—I mean, it’s all crazy stuff. But the fact of the matter is, that’s it. To me, the real question is: When does it cross over to having its own volition, its own ability to seek information, and solve new problems? That’s a different animal.

But have we seen any evidence of recursive self-improvement yet?

Not yet. I’ve funded a number of startups that claim to be close to it, but of course these are startups, and you never know, which tells me it’s 5–10 years, cutting numbers.

What do you think Google’s doing on this front?

I’m not at Google anymore. Every version of Gemini is at the top of the leaderboard. Gemini 2.5 just overtook everybody, and I’m sure there’s another one coming. Demis is working really hard on this question about scientific discovery. That is a path to getting to AGI.

Eric, we appreciate the work you’re doing. We appreciate you being here with us. We appreciate what you’ve done—the impact you’ve had on Silicon Valley and society.

I am so happy to be part of this. You created this incredible community, and there are all these smart people who spend all their time listening to you.

Very concerning.

Wow, Eric. Thanks, Eric. Appreciate you. Cheers. All right.