How AI is Reshaping Modern Warfare: Eric Schmidt Explains

China's robot cost advantage

00:00And I'm sorry to upset people in Texas. Um it sure looks to me like um China is going to dominate the low-end robotic space. They're driving the cost of humanoid and nonhumanoid robots down to thousands of dollars from 10 $20,000. Their hardware, which is derived from their electronic vehicle work, is extraordinary. We in America don't have access to the Chinese EV cars, but they're phenomenally good at it, and they're getting better and better, and they're doing it the Chinese way, which means it scales. So, ex Google CEO Dr. Eric Schmidt just gave another controversial take about the AI and automation race between the US and China. It's so profound that this will define the future of warfare with AI

The Ukraine war as AI test bed

00:47advancements, and the Ukraine war is a perfect example. Let's unwrap the big picture with Eric before we dissect deeper into how AI is already changing real wars. >> I think the Ukraine war is a very good test bed to try to understand how AI will change war. And I'm not trying to minimize the horrific nature of the war. I've been in it. I've been on the front lines. Uh real wars are much worse than you all think they are if you've not been in them. It's really horrific. Um I think that there's a couple of principles that I would offer. The first is that there's no place to hide. Um that modern sensors see everything. So, for example, um if you're on the front, uh which is a horrible place to be, you

01:36can only come in and out of your foxholes at dawn and dusk every 12 hours because those are the only times when you can move without being detected by the uh daytime cameras or the infrared nighttime cameras. And the infrared cameras are getting so good now that the visibility at night is very similar to what you need for the day. I mean, it's astounding. Um, so if there's no place to hide, then why are you building aircraft carriers? The US Army just announced a brand new huge tank, uh, which is a sitting duck for the next generation of drones. And by the time the tank is deployed, the drones will be incredibly powerful. You have the same principle with aircraft carriers and so forth. Um, it's my own view is that the

No place to hide: sensors expose tanks and carriers

02:24fastest learner wins. What's happening on Ukraine's side is that they're outnumbered three or four to one and Ukraine is moving very very quickly to innovate ahead of the Russians. The Russians are good, but the Ukrainians are better and that's why the fight is kind of balanced right now. It's it's not a stalemate. It's a tough fight, but Russia has not been able to achieve her objectives. Uh, which in my view is very good. Um, you know, if you go back to all of this hardware, when you go to the actual front, there's nothing on it. They don't have tanks or armored personnel carriers. All those things that are sitting in warehouses, you can't use them. They're too dangerous to be in. And I think what's going to

The fastest learner wins

03:04happen is that the autonomy aspects will make automatic killing machines in war much more likely. And I'm not talking about the robots that you see in movies. I'm talking about the equivalent of drones whether they're sea drones or air drones that are highly automated with highly automated kill chains. And I think the final observation, sorry to go on, is that this is integrated in an AI software framework using reinforcement learning and that integrated software framework with reinforcement learning means that you can have optimal battle plans. My observation of military having spent lots of time with them is they have thousands of people who consider themselves planners. Why do we need planners when we can have the computer

03:48do a better plan than humans can? The planners can obviously oversee what the computer's doing, but computers are going to be better at planning and vision and frankly killing. Just so you know, reinforcement learning AI is terrifyingly powerful in war. Basically, it works as simply as this diagram. But the scary part is it doesn't just follow human rules. It learns by simulating millions of battles, discovers strategies no human planner would ever think of, and then executes them at machine speed. And here's a real example from Eric, which sparked from the famous Google Alph Go move 37. So, I was present when AlphaGo won the first Korean uh championship against the human and I was literally in Korea and

Reinforcement learning and the AlphaGo move 37 lesson

04:35there was a famous move called move 37 and go has been a game that's been played for 2,500 years which made no sense and I was with the room with the humans and the human says proof proof that the game is going to lo the computer's going to lose to humans because no human would make such a stupid mistake. When they re when they reverse engineered the algorithm, the computer had discovered a new move. And this was the first example that I'm aware of in a gaming context of a completely new strategy being invented that was not visible to humans who were real experts in it. So I'm of the opinion that if you have a hundred humans who are busy controlling all the drones and these are smart people and

05:23they're well trained and so forth, my computer using reinforcement algorithms can actually come up with moves that you cannot anticipate and I can beat you. That is my opinion. So my guess is that the key things you need for for winning a war on a go forward basis is you're going to need millions of drones of many different kinds. by the way, not you know, you think of one type drone, different sizes and power that the algorithms will run in simulation and they'll learn what combinations to do based on what the enemy is doing and the the battle will occur very quickly. So the enemy comes in with a particular force projection and the computer says ah the optimal winning algorithm is to do A, B and C. The humans will watch

06:10this and go, "Holy crap, look at what it just did." Um, and this is equivalent of of uh sacrificing your queen in chess. But as a mathematical uh ability, the goal in war is to increase the probability of winning from 50/50, say, all the way up to 100%. We we know how to do that algorithmically. Eric is quite confident about the Ukraine war, but China as well as its AI and autonomous power, it's a different story. >> Well, it's interesting. I was just there. Um, I visited a whole bunch of robotic factories and um, and I'm sorry to upset people in Texas. Um, it sure looks to me like, um, China is going to dominate the low-end robotic space. They're driving the cost of humanoid and nonh humanoid robots down to thousands

China's hardware scale vs US software lead

07:03of dollars from 10 $20,000. Their hardware, which is derived from their electronic vehicle work, is extraordinary. We in America don't have access to the Chinese EV cars, but they're phenomenally good at it, and they're getting better and better, and they're doing it the Chinese way, which means at scale. So, my conclusion was that the ideal scenario is Chinese hardware and American software. which is geopolitically not possible, but that's the way I would phrase it. On the AI side, and I've written publicly about this, it sure looks to me like China's pursuing a different strategy than the US. The US is, and I'm very excited about this, working toward general intelligence. If you look at the planning gains and the

07:48incredible look at whether it's the latest 2.5 Gemini um OpenAI's successor to 03, it just goes on and on. Look at the coding capabilities of Claude. Um it it's unbelievable what they're doing that and and by the way that's at huge cost, right? Hundreds of billions of dollars of capital. You know the story about we're running out of electricity. We need God knows I testified in Congress 92 gawatt more power. An average nuclear power plant is 1.5 gawatt. You see the problem. China has all the power that it needs. It's built the infrastructure by the way using solar largely. Um and they have all the power they need. They don't have the chips because of chips resorts. And they're also don't have the depth of the

08:36financial markets to fund $50 billion worth of hardware. At least today. It looks to me like because of all of those factors, they're going to focus on consumer products and enterprise applications for the rest of the world. I don't think that America will allow those things in. There's a pretty good chance that Chinese software is going to power the majority of the world's, you know, um human interfaces with computers and business interfaces. Uh that's of concern. I think it's very important that US technology dominate all of the spaces. I want America to win and I think we need to win in all of those. I think we're not spending enough time on the more mundane stuff. We're spending as we're spending a good amount of time

09:20on the exciting stuff which is great. We need to spend more time on the mundane stuff as well. >> Economically, Eric sees a split. The US leads in cutting edge, very expensive models. China leads in scaled, cheap hardware and may capture massive global market share in day-to-day AI applications. But what about Chinese military AI? Um there's no hint at all. Um the people I was with were not military. The Chinese military does not welcome me to chat with them. Um I always when I visit a Chinese company, I always believe because of civil military fusion. I'm in one building and I always believe in my mind that there's another building next door which I'm not allowed into where they do all the military

Chinese military AI hidden behind civilian firms

10:01work. So remember in their model the the companies the the national champions do both military and and civil stuff together presumably in different buildings with different classification. So you have to assume that every product that I see which includes by the way the robotic stuff there's a military division as well. I did not see them. >> So it's shrouded in secrecy which is understandable cuz yeah it's a state military secret. But Eric also assumed that civilian AI and robotics advances in China are tightly coupled to military programs, even if that part is kept hidden. Unfortunately, the US is not keeping up in this future of warfare. Here's why. Well, the the US is at the moment well behind in the areas that I'm speaking

US procurement fails and the 17-year weapon cycle

10:46about. Um, the Secretary of Defense has issued a very important and positive order to get organized around drones. The US spends I think it's a trillion dollars a year roughly in military national security spending. It's it's on that order. Um so there's plenty of money in the system. One way to understand the US military is and thank God we're at peace is it's sort of a large it's sort of a jobs subsidy program for existing industries. If we're really at war, which hopefully never will occur, if we're really at war, I think you would see our military change very, very quickly, we have plenty of US companies that understand what I'm talking about, many of which are in the room, and they have trouble

11:36getting big contracts because that's not how the primes work. The US military does not have very good software capabilities compared to their hardware capabilities. The procurement systems are not organized around these. So again, if it were possible, I would accelerate this transition to offset X, autonomy, and drones. I see no reason why America should give an inch. Um I want us to uh the phrase is we bow to no one, right? America's dominance in the world is very important to American exceptionalism, to American economy, to freedom. I think America is committed to fighting against tyranny. That's why we fought in World War II. I'm a strong supporter of all of this. I worry that we're still spending 99.9%

12:26of the money on stuff that's not very important. Um, another example for the last 15 years that I've worked on this, when you have a person with a rifle in in the enemy, they typically use a missile from a jet to to kill that person. That's that doesn't make any sense. Um, in the initial war in Ukraine, before the Ukrainians had what they have now, they would use milliondoll U S300, S400 missiles to shoot down shaheds, which cost about $27,000 at a time. Again, the math doesn't work. It's really important in this new world to have defensive systems, which are, by the way, dual use, that you can massroduce at scale. What I'd like to see in America is a commitment to large manufacturing capabilities in the United

13:19States with stockpiled parts. I I was talking to some of my general friends and they pointed out that in a China conflict, which again we hope will never occur, we're going to also want to have those kinds of factories in the Philippines, Japan, Korea, you know, our allies as well because of the tyranny of distance, which makes sense to me. It's a it's a complete rethink of this Well, you know, our military, these these are real heroes. I I I love working with them and they have a doctrine of fight tonight. They have enormous courage, much greater courage than I do. We do not have the tools and the systems that they need. We are failing our leaders uh people in this room. Part of the reason we're having this meeting is

14:05because collectively I think we all kind of agree with this in philosophy. And so America's strength is innovation. So why are we not u applying the same principles of innovation that are in the consumer sector to the military? And it's because of the procurement systems. Right? The average time to weapon system is something like 17 years now. Right? By then the thing is is obsolete. Um I can't remember which of the jets it is but it has some it has like a pennium 4 processor in it. It's still programming. When I was uh flying, when I was in the Pentagon, I discovered a burrow 6600 deep in the bow of someone where people were afraid to turn off. Like the programmers who wrote that are literally

14:48dead, right? So, come on, guys. We can do better than this. And so, this part of the conversation, part of what you're trying to do, Gene, is to to get everyone to talk about this. Um, it seems to me that the way America changes is either from a crisis, which we do not want because it's war, or because we collectively agree that we want to win in a new space. And for all of you, there's enormous revenue here. Now, you have this explosion in um fantastic startups and I want them to I want them to get through the valley of the death. I want the government to have a way of working with them. I want them to know who to call. The thought experiment that I used to give is let's imagine that I collect all the venture capitalists in

Fixing procurement, startups and the Cold War analogy

15:34the Bay Area in a room for a military procurement venture conversation. Who from the government is going to come and present to them and what is he or she going to say? Right? If you can't answer that question, right, then you're not going to win. Now, our venture capital community loves to talk to CEOs and they imagine who the customer are, but this is a single source customer, right? For good reason, by the way. We don't want separate militaries. The the military has got to learn how to work with us. And of course, they're very law-abiding. The laws have to change to allow for this. And I'm excited for that to occur. I'd like it to occur during my lifetime. To me, it's almost like China and the US

16:18are in an AI race similar to the Cold War arms race. Except this time, instead of the US duping the Soviets into spending themselves into oblivion, developing a tech that won't actually benefit their population while severely neglecting their infrastructure, cost of living and corruption, China is the one pulling these moves on the US. Seems to me like China just read the playbook on how the last challenger to world hegemony was overcome and is turning it on the current hegeimon. But you know, the future is always hard to tell. So stay tuned and thanks for watching.