The Exchanges

Q Let me just ask, Emil, the important question that I think triggered a lot of the news this week is why then designate them as supply chain risk? Why not just abandon them, move on, use the other vendors? Like, why take this kind of punitive action?

A Yeah, so I don't view it as punitive, and I'll tell you why. It's if their model has this policy bias, let's call it, Based on their constitution, their culture, their people, and so on. I don't want Lockheed Martin using their model to design weapons for me. I don't want the people who are designing the things that go into the, the componentry to come to me, because if that poison, if you believe the risk of poisoning inside a threat, yes, it can enter into any part of the defense enterprise, but it's just the defense enterprise. So Boeing wants to use Anthropic to Build commercial jets. Have at it. Boeing wants to use it to build fighter jets. I can't have that because I don't trust what the outputs may be because they're so wedded to their own policy preferences.

AI assessment note: “I don't want Lockheed Martin using their model to design weapons for me.”

Q chance, according polymarket, and by the end of the year, 51% chance. So Emil, I guess there are two questions people really want to know. I'll leave off why we're doing this. I think President Trump has been pretty clear now. But how long is this going to take is the one question. And are we going to have to have boots on the ground? Maybe what is success here?

A Um, I think the, the president talked about this is a weeks, not months kind of operation, and it's aimed at essentially disarming the, the regime, uh, or the country from, uh, in such a way that they can't supply Hezbollah, Hamas, um, Muslim Brotherhood, all the kind of terror groups that get sponsored by weapons and money from Iran, not to mention the nuclear bit. And that's why you see from the reporting, they're going after the depots, the, the, you know, we went after nuclear sites before. They're a prodigious drone maker. These like huge one-way attack drones that can go, you know, hundreds and hundreds of miles. Um, lots of ballistic missiles, uh, that are aimed at every country in the Middle East, as you've seen, they've attacked them. So I think that's one. In terms of boots on the ground, there's no scenario where we have some protracted boots on the ground, Afghanistan, Iraq too, like scenario.

AI assessment note: “the president talked about this is a weeks, not months kind of operation”

Q was part of Dario's concern, is that when you press the button and hand over complete autonomy, and there's a kill action that you're now giving to a robot or to some autonomous system, do we then kind of have a moral issue at hand? And is that something that's kind of debated or discussed? And is that the right way to think about the framing of what goes on?

A I mean, we're not even close to there yet, right? We're like this, the systems are not, we wouldn't feel that a system, uh, that would have sort of like real risk for a civilian is ready to launch yet. So we're not even debating that. We're just trying to get basic autonomy in drones, basic autonomy in underwater unmanned vehicles, basic autonomy that You know, you've heard of this collaborative aircraft that fly along with the jet craft so that it has more firepower, but it's still tethered to what the jet does. So we're just at the very beginning of this stuff, but for Golden Dome is a good example of like, yeah, who can oppose that? Like it's the only way to get out of threat like that. Um, so who could oppose if you have a military base, you have a bunch of soldiers sleeping, that you have a laser that can take down drones autonomously, On that. So there's, it's, it's pretty scenario by scenario, but I don't, we're not having a lot of debate because the Skynet thing is so not a realistic thing at this moment, right? Except if one thing I did tell the anthropic guys, I was like, you know, or I'd tell us any company, your models are getting stolen by the Chinese. They're going to unguard rail them and use them against us. And then you want our models to be less capable against your model.

AI assessment note: “we're not having a lot of debate because the Skynet thing is so not a realistic thing”

Q key to a lot of this debate, and when will it be, you know, perfect to find as much better, and I guess this dovetails with the self-driving, You know, thoughts. It has to be a magnitude better than a human. So when will this be a magnitude more accurate than, you know, when we have, uh, make a mistake as a, as a military and we kill a civilian?

A Yeah. No, it's, it's a good question. And, um, I don't know when that moment hits that FSD moment where it kind of gets better. It's certainly not there. And you wouldn't want to take huge risk with that. You know, there, there's a gradation of when you would use that and what kind of risk you were trying to take or not. If you were, Trying to take out a drone using AI, using a, like a laser or something, you'd be pretty like, okay, making mistakes because you just missed the drone, you know, like whatever with the, with the, with the laser and laser goes off, it's all over. Um, if you were doing something more sophisticated in the population area, have a densely populated area, you'd take less risk. So we're developing procedures, tactics for each scenario. And this is part of the debate I had with Anthropic, which is we need AI for things like Golden Dome. Chinese hypersonic Mixel comes up. You've got 90 seconds before it separates and all kinds of decoys, and you don't know where the actual payload is, and you want to get it, hit it from space. And a human can't, doesn't have the reaction time, doesn't have the, may not be able to discriminate with their own eyes what they're going after. That's a pretty, you know, low risk thing because it's in space and you're just trying to hit something that's trying to hit you. So I think in the next 10 years, you're gonna see a lot of …

AI assessment note: “I don't know when that moment hits that FSD moment where it kind of gets better.”

Q You can't do a war game scenario with it?

A You could do a scenario, but you can't like, let's suppose you're writing a plan saying like, if this happens, here's what we would do. And it might involve a kinetic strike, which causes harm to a human. So like, well, what do you think these folks do? You know, this is the Department of War. This is what we do. And so I said, okay, well, I've got to number one, have direct relationships with these companies, not just through Palantir, because I want to use it more broadly. And then number two, I need to have the terms of service be rational relative to our mission set. So we started these negotiations and took three months and I had to sort of Give them scenarios about like this Chinese hypersonic missile example. They're like, okay, we'll give you an exception for that. Well, how about this drone swarm? We'll give you an exception for that. And I was like, the exceptions doesn't work. I, I can't predict for the next 20 years what all the things we might do, use AI for. Um, and so, so all lawful use seems like a good thing. If Congress wants to act great, we have our own internal policies, like we'll follow them. We're not knuckle draggers here. We want, we don't want to hurt people unnecessarily. So, you know, it's our province to decide how we fight and win wars. Um, so long as they're lawful. And I think at some point it turned into a PR game for them because They were not…

AI assessment note: “You could do a scenario, but you can't like, let's suppose”

Q What's your view on how the industrial supply chain for hardware components and systems is coming along in the United States? Because my understanding is we're trying to reduce dependency on Chinese manufactured components. Where are we with respect to where we need to get to in the US manufacturing supply chain?

A We are early days, um, critical minerals. You see, you've seen the action around that. Um, you'll start to see, so I have the office of strategic capital, which has two hundred billion in lending authority. And what we're trying to do is, is it's like a treasuries plus a hundred bips loan to companies, show them that the department needs their solid rocket motors, their batteries, their fiberglass, like all the things that we That we're heavily dependent on for our defense industrial base that are completely outsourced to China and domesticate them here. Um, and we've got, uh, a bunch of great people running it. So I, but it's early days. It's gonna take for the rest of the term to get, um, I think we'll get critical minerals done before the rest of the term where, where we have the access to what we need to from us or allied countries. Um, uh, but from batteries is like the next problem I'm trying to solve. For example, batteries are totally outsourced both technologically and from lithium to China. Um, and there's like, you know, kind of call it 20 critical things. If I could get to all of them at some level, but then it'll take a few years for them to like build plants and do that stuff. But they're there. It's, it's very important. I hope whatever administration comes next continues it because I'm all free market, but, but we outsourced so much. That, um, you know, it cripp…

AI assessment note: “We are early days, um, critical minerals.”

Q Yeah. I mean, you also oversee DARPA, yeah? Yeah. DARPA is the father of the modern internet. And it's created a lot of really critical technologies. Can you talk about what's going on in there? Are there interesting things that you think our audience should know about that you're trying to push forward?

A I mean, there's so much, probably my favorite part of, of my, my office is like, because they're, That's where you, it's sort of the, it's still a very honored profession to be part of DARPA. Like, you know, being a, being in government service for a long time has sort of reduced in its stature since the Manhattan Project. Now, because now, now if you're a great ask, you know, uh, someone who wants to do rockets and stuff, you go to SpaceX. DARPA still has the best of the best. And so the most creative ideas happen there. One of the things that they're working on that's public is they're trying to use biology to synthesize critical minerals. So, so how do you, so how can you just pull them out of ground, use biology to do it so you don't need to do all this crazy, messy, dirty refining? That would like change the game big time on our ability to get the critical minerals we need faster and leapfrog the Chinese in terms of tech. Um, so they're doing a lot of that kind of stuff. They're deep in cyber. Cyber attacks are, are the next huge threat with AI, right? What we saw with creating all these agents to attack systems, The anthropic actually happened to them. Um, so they're, they're, they're working on this. That's, there's not a ton I can talk about kind of DARPA because it's so, it's so classified, but those are a couple of examples for you.

AI assessment note: “they're trying to use biology to synthesize critical minerals”

Q There you're showing drones that are sort of human operated, but how much of this should basically be AI so that it's just some computer vision. And again, back to what you said before, a model understands the rules and the red lines, but otherwise is beyond task and accomplish your mission. How much of it is one versus the other?

A I mean, it's, I believe that a sophisticated drone war is going to be drone swarms controlled by AI to some degree or another, right? To what degree the control matters. Like for example, Drones have decoys. They could spit out, you know, they could dazzle. They could put out things. So how do you discriminate what's a drone and how to hit it? You know, you can use AI for that because it's learned, you know, how to do automatic target recognition, for example. Um, and then also could it identify a person and that, and does that make it safer? So it's going after actually someone you want to get and not someone you don't want to get. Um, so there's a lot of uses for AI at the edge, if you will. Um, in the future here, the Ukrainians and Russians do something called, like, a kill box where they lose comms because it's jammed for this drone, and then it just starts going in a box and looking for, you know, the person they're trying to get, and they're trying, they're starting to use AI to do that.

AI assessment note: “sophisticated drone war is going to be drone swarms controlled by AI”

Q we're moving from the old primes to the new primes, that there's going to be a small set of big winners. And then obviously lots of seconds and subs and whatnot. Is that really how this market's going to evolve? So are we going to end up with Anderil, Palantir, maybe three or four others, and that's where most of the value is going to accrue from the market perspective?

A I mean, Andrew and Palantir want that, and I joke with them all the time about it, but I, I want, I definitely want at least a second layer that's innovative and trying to disrupt the first layer all the time. I've had a mom and pop, like, wholly owned company that, that makes these missiles called ERAMs that are, we really send to Ukraine, and they do it with, like, 30 people, and they can do a thousand a year, because they've designed a manufacturer, and it's awesome, so I want companies like that. To continue innovating. Maybe Andrel then buys them, but the, the, one of the reasons the primes are such a small number, it's not the only, but it's one, is they learn how to contract with the government. They learn how to go through the bureaucracy, and that became a competitive advantage. I'm trying to take that competitive advantage away.

AI assessment note: “I definitely want at least a second layer that's innovative and trying to disrupt”

Q a lot of conversation about space based lasers that could shoot ICBMs out of the sky to avoid, you know, global nuclear war. And we could always take out every nuclear warhead delivered on an ICBM. Is that technology feasible? Is there a place in the near future where we could see basically maximal global deterrence using these systems, either ground based or space based to take out hypersonic missiles?

A I think, I think the, the harder but more valuable problem to solve would be the space based way of doing it, because then, um, you could get at any kind of, almost any kind of threat that hits space, but you still need a ground layer because there's cruise missiles that could come at you. There's drones and so on. So, so we called it multi layers. Like, how do you, how do you get every kind of weapon at every layer? But, you know, directed energy lasers, as they get more powerful, you could take on a bigger weapon farther away, right? So those, uh, so those technologies as that, as they improve, it gets more and more capable. And I think all these defense systems, uh, are going to get more and more capable to get more and more of a variety of weapons that farther, farther standoff, which is what you want. You want it to, you don't want to shoot it when it's right over Tel Aviv. You want to shoot it, you know, When it's still over their, their land ideally.

AI assessment note: “directed energy lasers, as they get more powerful, you could take on a bigger weapon”

Q And then going back to the face of war going forward, my understanding is that there have been more drones deployed by the United States this past week than we've done in the history of military activity. Is that right? And like, how does that really change things going forward here?

A It changes it big to well. So So the predator drone was the first big drone program like 1015 years ago. It was this big honking drone. Um, and then if you remember, Obama would take out some of these Al Qaeda leaders with drones on their balcony and things like that. Uh, I think we, uh, President Trump took out Soleimani with a drone near his car. That was the beginning. And then the Iraq, uh, sorry, the Russia, uh, Ukraine war happened where it's like drone on drone. 70% of the, the casualties are for drone, but because of drones. So, um, drone on drone warfare, robot on robot warfare, those things are the future for sure, right? And that's why companies like Anderil are companies like Anderil is because they're making unmanned systems.

AI assessment note: “drone on drone warfare, robot on robot warfare, those things are the future for sure”

Q you can't tolerate that these folks will do that. It's too critical of a technology. By the way, this is de-platforming all over again. Remember what happened when you didn't like what was said, now all of a sudden you were de-platformed? This is that times a thousand, because this is not about posting on social media. This is about using fundamental technology to either advantage or disadvantage your business.

A Yeah, I mean, I think I described it the other way the other day as these, the leaders of these companies say they're going to cause 50% white collar unemployment. This is as powerful as a nuclear bomb. It's like 50,000 geniuses in a data center, so you could have a small country coerce the world into it, so whatever. So you're like, holy cow. All right, so this is a general substrate of intelligence, of technology that's applicable to a lot of things. Very generalized. It's not like Workday saw like HR software. We could just use competitor. This is going to be part of our everyday life in so many different ways. And the controlling the, what, whether it has a moral conscience. I mean, Anthropic has its own constitution. It has its own soul. It's not the U S constitution. So you're subject to that plus whatever whims and how that changes. And that's a scary thought for, for Americans generally, Um, and I think that did come through a little bit today, and in the coming years, it's going to be a bigger and bigger deal.

AI assessment note: “Anthropic has its own constitution. It has its own soul. It's not the U S constitution.”

Q I guess a dovetail to that is why couldn't this have been handled quietly? Is this Anthropic who made this a public spat or was it the administration that made it a public spat or two to 10?

A I mean, they have a very good sophisticated press operation and like really good and painting us as doing mass surveillance where, where, where your, their issue was like some commercial database thing that someone else could buy. They didn't want us to buy to use it, which I'm not even sure we buy them except to do recruiting for soldiers. And, you know, we run schools, hospitals. We do a lot of things that do do. We don't just fight wars. And, um, And the way they were able to characterize these two things, which are genuinely scary to people, but were not the real issues. Um, it was really the, you worry, I worried about them shutting off our system at a moment of need or them messing with our system at a moment of need. Those were, we were scared.

AI assessment note: “they have a very good sophisticated press operation and like really good and painting us”

Q Emile, how much does this have to do with China?

A I think, you know, my instinct is, and I'm not speaking for the administration on this, is that's a second-order benefit to some of these things. Like the, um, you know, and you said eight, eight conflicts. There have not been eight conflicts. There is, like, we inherited Gaza. We inherited Russia-Ukraine. Um, Venezuela was its own operation, and then you could sort of attach to it the drug boats that were coming out of that as, like, one big operation. Um, and then the Houthis was just, Biden was ignoring the Houthis. They were just shooting at our ships. So that was like very limited in terms of like, stop shooting our ships. We need freedom of the seas. And that wasn't sort of a, you know, so that's something any president should be doing generally. I think, um, the Iran being the one, you know, material conflict outside of Venezuela. So it's not, it's not that many. And, and how long did Venezuela last? It was one raid.

AI assessment note: “that's a second-order benefit to some of these things”