Shawn Ryan Discusses AI w/ PALANTIR CEO Shyam Sankar

Thanks for having me, Sean. It's great to be back. I owe you a huge thank you.

I love it. But that was... Yeah, man. So, thank you. But yeah, it's good to have you back. I'm pumped about our conversation today. And I know you got the new book coming out and everything, but what have you been up to?

Sounds like it. Are you looking for real estate in Miami now?

Right on. Right on. What prompted that?

Yeah. So, you know, I think it's important to be in a state where your reps are actually going to rep you. And I think that's part of it that matters. And then you could think about, okay, well, what are the places we could go from Denver and Miami or really Florida had the best both combination of legal and positioning perspectives for us. It's the right place. We don't want to be like the 50th company to go to Austin or something. Austin's great. We have an office there. We love Austin. We love Texas. But Miami felt like the right home for us.

I haven't seen the latest. Catch me up.

I think it's very hard to separate fact from fiction with these things, you know, because you can kind of egg the agents on to doing very specific things to tell a dystopic story like what's going on in the prompting with like the Claude bots. My lived experience using these things operationally is that nothing crazy like this is happening. You know, that actually it's much more contained. It's much more sane. It really is more like an Iron Man suit for the American worker than it is a headless godless machine that's just roving around doing... you know, this is maybe we should just start here. It's like the ways in which I think the American people are really being lied to about AI is that you have on one hand incredible doomerism like hey this thing it's going to lead to mass unemployment 50% of entry-level jobs are going to be destroyed inside of a year or two and on the other hand you have essentially this fantasism it's going to lead to a utopia like untold abundance. I think neither of these things are right. They're wrong for the same reason which is they assume there's no human agency. You know, AI doesn't do anything. Humans use AI to do something.

And the reality is that the future of AI has not been determined. It is being determined every single day based on the decisions we're making. We can choose to use it to build AI slop or new forms of addiction and gambling. We can choose it to re-industrialize the country and bring prosperity to the American worker. Those decisions are being made every single day. We should use our agency as humans to decide what we value. And it's very clear what we ought to value. Then there's another part of this which is age-old. Who are we listening to in AI? We're committing the same fallacy we have in the past which is we're listening to the people who invented the technology not the people who are using the technology. You know, it is and by the way these inventors are geniuses. We need them in America. They are my heretics and hero archetypes that I talk about. But just because they have the genius to invent the technology doesn't also imply they have the genius to think about how to apply that technology, how to govern it. What are the consequences of it? The example I like to give people is the telescope. You know, Galileo did not invent the telescope. Galileo used the telescope to discover planetary motion. Who had a greater impact or a greater opinion of the impact of the telescope on physics and knowledge? Was it the person who invented it or the person who wielded it? So the people we ought to be listening to are exactly the people who are not inclined to give op-eds who are not on, you know, mainstream media. It's the American worker. It's the guy in the submarine industrial-based parts manufacturer. It's the ICU nurse. It's the factory worker making wires or machinery or equipment.

And ask them like what how has AI impacted your job? Has it replaced you or has it empowered you? And perhaps I think the most profound question that I always like to ask folks, how optimistic are you about your children's future in an America with AI? You'll be surprised how optimistic they are.

Yeah.

So let's come to each of these in turn. So let's start with the kids and then we'll come back to this idea of replacement. I have kids too, 13-year-old, 11-year-old. And so this is a personal question. It's not an abstract question to me. And what I think I want them to know is how to use this tool that it's a tool and it can unlock profound education for them. But the people who are going to succeed are going to have two things. They're going to have specific knowledge like who's winning right now at the front lines. It's the guy who has 15 years of experience. We'll talk about operators in a second, but like people who really know what they're doing, have unique knowledge and insight because what the AI doesn't know is that. But then with that insight and a bit of direction to the point of human agency, it is this incredible Iron Man suit for them. So that's one thing specific knowledge is going to continue to be valuable.

The second part of it is do they know how to use AI, you know, so essentially like it is a bicycle. You have to learn how to ride the bicycle and that requires reps. I think that's why it's actually turning into a massive advantage for America because if you just compare it to say Europe, people are really thinking hard. How should we use it? They're just thinking, you know, and the American sensibility is to roll up your sleeves, get your hands dirty, play with the thing, experiment, try it out, do something, you know, a little bit of, and I mean this in the most positive possible way, the cowboy spirit. And that is something that the child's mind is very good at, you know. So I think one of the mistakes that our education system could make is try to restrict AI. In the early days, you're going to have people using it in stupid ways to write their essay for them. It's going to be a sloppy essay that you can tell is AI generated, but that might be what they're needing to get through that initial gate to go on to the more intelligent uses of it, which is, hey, I wrote a first draft of the essay. Critique it. What did I miss? What are other things I should think about? You know, help me elevate my own thinking. You know, it becomes a partner to do these things. I think that's really powerful.

Well, I think the most interesting use cases are going to start where you have the deepest domain knowledge. Like I think that's part of the reaction the American people have is they look at it and it seems like some of these things are kind of trivial. Like you're saving me a little bit of time here and then that's weighed against like crappy content that's being generated, misinformation, disinformation. So on balance it's like why should we believe in this future? Plus I got these data centers coming up and my electricity bill is going up. But if you look at it as hey this is the basis for re-industrializing the country. Like we're going to give the American worker superpowers. They're going to be 50 times more productive than any other worker in any other country. And that's not just a matter of pride. That economic leverage is how we bring back manufacturing to the US. We're not saying we're going to compete symmetrically. This is David's slingshot against Goliath here. Like, yeah, okay, the Chinese are the best at mass production today. What is our asymmetric approach to regaining the very thing that we once created? You know, we're going to reinvent production. So, an example from Panasonic Energy, they make every battery that goes into every Tesla at the Gigafactory in Reno in Sparks, Nevada. This is exquisite high-end Japanese technology that is operated, the employee base in that region of Reno, they're prior casino workers. So, the old apprenticeship journey to learn how to operate and maintain this equipment used to be three years. With AI, it's 3 months. That's leading to more employment, not less. And this is the way in which like it's like to my point of human agency, how do you choose to use this stuff? It really does matter. And I think on the consumer side, yeah, it's going to help me, you know, categorize my bills better. It's going to help me write responses. I think all those things are trivial. You know, the real stuff is going to start in the enterprise and work its way back.

Okay. I thought that's where you were going. I was like, "Holy shit." And it's really compelling to see how you have non-computer scientists. It really proves out the whole thesis here which is like experts, you know, the intel warrant officer the E4 the E8 who are inventing the future of how we're going to deter conflict and that poses lots of interesting challenges which I think apply just as much to the commercial sector as government which is like it breaks rank structure it breaks hierarchy like you're going to have to really embrace the internal disruption that's going to happen but I think this is a great thing for the American worker because for the last 100 years or so the managerial revolution has pulled power away from the frontline American worker towards the bureaucrat. AI is reversing that trend very quickly and that's very destabilizing to the middle managers where you're going to see a lot of resistance out of this. But what it gets at the core problem we have as a country which is the legitimacy of our institutions. You know why do doors fall off planes? Why aren't government basic government services provisioned in a way that we would all recognize as having basic competence? So you have two answers to that sort of question. You can be like, well, these people just don't care. Might be the case, but more often than not, my diagnosis of this, having done this across 50 different industries in the private sector and government, is that the people at the top, even when they do care, they have this steering wheel. They're trying to turn it very, very diligently, but what they don't know, it's actually a prop from the Jungle Cruise ride at Disneyland. Like that steering wheel is not connected to anything. And so that disconnection happens through this bureaucracy. And then you have the people on the factory floor, metaphorical or literal factory floor. They kind of look up and be like, "How can my leaders be so clueless? How do they not realize what's actually going on here?" That's really dangerous because it breeds nihilism, right? Like then you look at it and you're like, "Man, it's hopeless. We should give up. Let's burn it all to the ground. It doesn't really matter." And that's horrible because actually if you burn it to the ground, things will get worse. Like, yeah, everyone acknowledges it's not working right now. The answer is not burning it to the ground. It's fixing it. How are we going to fix it? What's our theory of change here? And the theory of change has to start both at the bottom and at the top. At the top, it starts with people who care. People who want to get things high agency leaders who care about the outcome. Then they need the tools. Everyone needs the tools to do this. And so then how do you empower the people at the bottom closest to the problems to actually go solve these tools? That I think is a quintessential American characteristic. Like we think about it as mission command. Give the intent. Let these people run. Let them cook. Don't overmanage them. Don't drain the creativity out of their souls. You think of every innovation on the battlefront. It was like the E4 rolling tanks across Europe in World War II discovered additional ways of getting through equipment, right? And the generals would let the soldiers cook. This is a powerful moment for the...

Yes. Yeah. So let's... there's this concept called Jevons's paradox. When we started inventing more efficient coal burning steam engines everyone thought that the consumption of coal would go down but the consumption of coal skyrocketed. Now that the engines were more efficient the cost to transport goods per mile was actually dropping. And so then the number of engines we wanted went way up. And the number of trains we wanted went way up. There's something like that that's going to happen here where if you look at something that is fundamentally demand constrained like actually if we made more no one would want it. Yeah, that's a problem. Like you getting more efficient is going to result in more jobs. But I don't think most things in the economy look like that. Most things in the economy, you look at healthcare, it's exploding. It's like 20% of our GDP. Healthcare costs might be our greatest national security risk. Like the solvency of our country depends on being able to deliver care to the American people at a better price and we're only going to need more care over time. So we know we're going to need more care over time for the same amount of money. How can we deliver more care? How can we get more efficient in doing that? That's Jevons's paradox. So an example of this with Tampa General, we're able to get sepsis deaths with the leading cause of deaths in the ICU down to zero from 50% of all deaths.

And there's no replacement of labor there. It's really automating the parts of the job that took the nurse away from the patient's bedside and then help them spend time with the right patients who had the greatest amount.

It is. That's exactly right. That's how we added a third shift to a submarine industrial-based parts manufacturer because the drag was the time they spent like dead weight loss in planning. Like I got to plan what to produce then I produce it. If the planning takes too long like tools down you can't if you don't have a plan yet you can't start making things. If I can shrink the planning process from a couple of weeks to a couple of hours, I have more time to make things. And then organically, the company's like, well, I have more work than I have workers. I need to go hire people. And that's the bounty, the American prosperity that we can see out of this. Now, I don't want to be too Pollyannish. I think there are things we need to make sure of. The most important thing I care about is reestablishing the connection between GDP growth and wage growth. Somewhere in the 70s something broke fundamentally where our GDP kept growing and wages stagnated. You know this has got to be this is the fundamental promise to the American people that the prosperity will be shared and the way in which that happens I call this the productivity dividend. The American worker at the front line who is using these tools to make their companies better. They need to participate in the economic upside of doing that. That's critical to not only the social stability of the country but the prosperity of the nation seizing this initiative.

I'm sorry. AGI, excuse me. I've always felt and I think the present moment kind of shows it that it's like this continuous journey you're on that maybe it's a frog boil, not in a negative sense where every version of the model is more capable than it was before. The models still have what they call jagged intelligence. They're savants at some things. They're not good at other things, but they're getting, you know, even the things they're getting they were not good at, they're getting better at. So, some people would say we're already there. Like for coding, they're so good. We're already there. I still think there's a fair amount of human agency that's involved in getting these things to work. As a consequence, it's valuing taste more than anything. You know like what to build, how to build it, how to think about the problem, what's the elegance of the solution and then it gives you a big lever to go after it. I think one of the challenges you could almost imagine entering a new dark age like the dark ages were caused because we kind of lost fundamental knowledge in the wake of the collapse of the Roman Empire. Even though we get to live, you know, not everyone has to know everything but at the end of the day someone has to know every... someone in society has to know about every part of it, right? I don't have to spend my time thinking about how to design chips I live somewhere else in the stack I get to write code I get to rely on the fact that someone else is a semiconductor expert and knows how to do lithography and make the chips that I depend on but at some point you can kind of see how it actually doesn't work for humans to not be involved with any of this stuff ever. That's the fantasist part of this stuff that if you don't know how it's made, you can't innovate on how it's made. You can't govern it. You can't understand or debug it. And so I think a much more reasonable path is what I call the inductive path. It's like we're on a journey. It's very dangerous to kind of skip steps and fantasize about the future ahead.

I mean, you want to be optimistic about it. You want to be able to see it, but you don't want to be reactionary to it. Things like UBI, I think, are reactionary to this totally unproven idea that there's going to be so much bounty that we're going to be reduced to being as useful as house cats.

Universal basic income. Basically, this idea that everyone should just have an income provided by the government that we fund through taxes because there will be no jobs.

Lots of people on the inventor side of AI talk about that. It's a little bit to my point of should we be listening to the people who are using it and wielding it?

They got the bomb because we had geniuses working on the Manhattan project. And a small number of them thought, some of them very famous, like Niels Bohr, one of our greatest physicists, thought, you know, we should tell the Soviets that we're building this thing. If we tell them, they won't be scared. And then another guy, Theodore Hall, one of the youngest members of the Manhattan Project, he was 18 years old working on this.

PhD from Harvard. He thought, well, in my infinite wisdom, because I'm so good at physics, I think that if two countries had the bomb, that would ensure global peace and stability. So, he actually walked to the trade mission in New York and told them, hey, I'm building this bomb and then subsequently he went back in there with technical specifications. So, this is a lack of epistemic humility, right? It's like just because you're a genius in one area doesn't mean you're a genius in another area. And I think it'd be a fair accounting to say every death due to communism since 1949, some of that culpability is on the hands of the Manhattan Project scientist who usurped the chain of command and thought, 'Hey, I'm just going to do this thing unilaterally.'

Well, I think the most optimistic case is really what is happening today with the American worker. So we've started running what we call American tech fellowships. We take people on the factory floor in the front lines and we put them through a six-week nights and weekends boot camp to learn how to build their own AI apps. These are not computer scientists. We're not even trying to make them computer scientists. There are people who have deep domain knowledge of what they do. They manufacture wiring. They are ICU nurses. One guy's a potato farmer in North Dakota. They know their craft. And I'm supercharging them with this. One of the most exciting new American tech fellowships we did was specific to veterans.

Yeah. It's been really rewarding for us to do. I mean, part of the thesis is like we want to fuel the disruption. And the disruption isn't just technical. It's also mindset. You know, one of these guys, prior enlisted Navy, a sailor. He now works at a manufacturing company. Grew up in a rural part of Georgia, dirt poor. No one ever told him he was supposed to be smart. No one ever told him he could do these things. And so just having someone lean in and say, 'We believe in you. We're going to give you the tools to unleash your human agency.' Watching this guy cook, he's improved. So he's reduced downtime on machines by 50%. He's improved yield on the factory floor by 20%. These are big numbers. It's shocking and you're really looking at the output of one person. One person who all that potential was latent. What the AI did is it removed all the drag to your point that all the ideas he had he could now realize and that's going to have a compounding effect. This is how we grow our GDP again. This is how we become a prosperous nation. Probably the part that I think the American people should feel most gaslit about over the last 30, 40 years with globalization is this idea that somehow you're not smart enough that there are people elsewhere who are going to work harder, work for less and they're better than you. And that's just not true. And I think part of this comes down to a belief in oneself and what is the message that we're giving them. And if you let these guys cook, it's eye-watering. I'm learning from them, not the other way around.

Trust is earned. One, I mean just an example I saw this morning, I saw a clip of myself on X or something. Yeah, it was X and somebody said, 'Hey, Grok, what's this from?' And it said the Joe Rogan Experience. And I'm like, shit, I mean that's pretty basic stuff. How do you not know? And so it makes me wonder, shit, what else is this stuff getting wrong? Because we rely on this for a lot of things here. And so when I see a simple mistake like that, it just makes me wonder what other mistakes are we getting out of this.

I would think about it as how do we trust the humans? You know, if you ask the human, 'Hey, what is this from?' In that particular case, we'd expect them to be a lot better, but there's an element of what is this person uniquely credible at? And you develop priors on it, like where are they able to help me, where are they not? Maybe I'm not asking the question in the right way, maybe I'm not providing enough context to doing it. So this is why I think the specific... so if you think about trust in the general, like any given human you ask them a bunch of questions they're going to get some of this stuff wrong, they're going to be pretty convinced they're even right about some of the stuff they get wrong. I think AI, it's the same thing. So it's about us having enough... this is the point of rolling up your sleeves and playing with it is like, hey, where do you believe this thing, where have you seen it being good or not. And then you develop... the technical term for this is eval. But you develop a set of tests that you're constantly running it through to understand like when they release a new model, is this model at least as good as the old model? Is it better? How much better is it? Where can I... is there new trust I can extend to it? Again, to the point that trust has to be earned. So you're not going to get any of this stuff for free where it's just like YOLO, ask a question, blindly trust it. It's going to be like, hey, this is a new teammate. This is a fresh second lieutenant. I don't trust them with anything. We're going to build a relationship together solving problems together and we're going to see where you're really a rock star and where you can help me be more effective and working backwards from the problems we have. That's a much narrower scope. It's like, hey, I make wires, I'm a potato farmer, these are my problems. How do I develop trust in this domain with you where I'm the expert actually? And so I'm going to be a really good judge of whether the danger is like using it where you actually have no knowledge. So, you're a blank slate and you're going to by default maybe trust it way more than you should.

Thank you.

Yeah, great question, Derek. So, the first thing you have to understand is we don't collect any data, right? We're a software company. We provide our software to the government or to a private sector like to a manufacturer. So the only data that is going to be in the software is the data that the organization has access to. If you're a manufacturer, that's your supply chain. That's your production data. That's your customer orders. If you're the government, it's what you have lawful authorities to have access to. Then there's the question of safeguards, which is where we think Palantir is the worst platform to try to abuse civil liberties in because we have immutable audit logs. We have purpose-based access control, role-based access control, classification-based access control. If you misuse the data in the platform, first of all, all the controls that are in there prevent misuse. But if you try to circumvent these things and misuse it, there's an immutable audit trail of what actually happened. You are going to be caught. Now, I can tell you anecdotal stories like there are institutions that don't want to work with us because sometimes those sort of protections are too strong. Sometimes that's uncomfortable. Sometimes you don't want to know what the data... or you actually already know what the data is going to tell you and you don't want protections that strong coming in. But that's the core thesis. So you have to go back to the founding precept of the company which is like, you know, politics is structurally zero-sum. People just like to argue about who's right or not.

You know, it's like you kind of get nowhere by just arguing to the nth degree. The question is how do we move out the efficient frontier? You know, like if we go back to the Palantir story in particular in the post-9/11 world, everyone was like, what's more important, privacy or security? I don't know. As an American citizen, that sounds really stupid. I kind of want both. Why can't we have more of both? Who's working on the technologies that mean that for a given level of privacy, I can have more security or a given level of security, I can have more privacy? So how do you bring more nuance to the question? So if you can protect data in a more fine-grained way. If you can attest to the purpose with immutable audit trails, maybe you can have reasons to have access to the data on a temporary basis. Maybe you can have condition-based access control where given what's happening in the world and a precept of human intelligence that's telling us about something, we're in a new regime for a limited period of time. Having a system that allows all that to happen, if you really go back to a pre-9/11 world, it was essentially binary. You either share all the data or none of it, which then biased towards either gross violations of oversharing or not being able to connect the dots because people didn't share at all. That's insane. That's the position you don't want to be in. Now, the reason we tend to get attacked is we're living in the messy reality of the arena. If you're looking at this politically and you just want to argue about who's right, if you can't see that actually both sides are right about something, that actually there's a kernel of both, how do we bring a synthesis of these perspectives, how do we do that with technology so we're not reliant on the fallibility of humans or people rotating in and out and loss of knowledge and transfer. It's about restoring human agency. So for a given set of policies that a democratic society wants to enact, do you have the capacity to enact it? This goes back to my point of the broken steering wheel, you know, okay, if you have a broken steering wheel, how good can the institution actually be? How nuanced can your policy be? Like if we make that work, it can be responsive to the American people, being responsive to the electorate.

It would be insane not to worry about it. I mean, I think it'd be insane not to worry about it. It's not deep enough thought to think that we're somehow... we're actually the antidote to that. We're not the cause of it.

And that's my really point. When you're in the arena, you're going to be criticized for doing it. Like, hey, you should... it's almost like a purity test. Like, you shouldn't be touching this at all. It's like, I'm just saying, what's the counterfactual that letting a bunch of historical legacy contractors who aren't as sophisticated with technology trying to solve these problems? Who is leaning in and saying we want to make these institutions function better? That's the antidote to nihilism, you know, make things work. That's also the American builder way. We're not going to get out of these problems through policy alone, through politics alone. We got to build our way out of these things. We got to build a better future.

Amazing.

Thank you. Little something.

Right on. I too got you a gift here. We have a Mobilize ammo box here. So, this is an actual... it was made at the Lake City Army Ammunition Facility in Missouri. It's been in continuous operation since 1941, making ammo. This specific box was made in '73. And it held 20mm electrically primed shells. Don't worry, there's no shells in it now. But I have it here.

I have it in a little Mobilize swag. So, we have the book, we have a hat, some stickers, some patches. There we go. Yeah. And a nice little Mobilize jacket.

Well, I mean, if you think about it, so I was on in April, Operation Spider, 12-day war, Midnight Hammer, the skirmish between India and Pakistan, Maduro, you know, there's a lot going on in the world.

Sure feels like it. All of these things are happening against the backdrop of China still. So even though in some sense China's taking a back seat, it is the driving force here. You know, who's buying the Iranian oil that keeps the regime going? What is the industrial base that's supporting Russia's war machine? These pieces are interconnected here. And so the radical pace at which these things are happening, I think underlines the precept of the book and a lot of what I've been talking about, which is to really prevent World War III, we need to have a strong enough deterrence posture to make sure our adversaries don't want to mess with us. And I'd say things like Midnight Hammer and Maduro are really the first things we've done that have restored deterrence. This sense of, oh man, I have been underestimating the US. And we have to continue that trend where the kind of missing part of this I think if you thought about it as a spear that the pointy end of the spear is really good. Look no further than Midnight Hammer or Maduro to see that the shaft of the spear needs work. That's the industrial base. That's our ability to link the factory floor to the foxhole.

And a big part of the book, I spend time talking about the historic figures who threw away the process, who rebelled against the system and actually delivered the capabilities we need. And I think that's a really important narrative because honestly, everything that's ever worked was against the system. It was despite the system, not because of the system.

And having the courage to look at the American industrial base, whether it's Isaiah with Valor, you know, people trying to build nuclear reactors now, it's the heterodox thinking. It's not coming from the big companies. It's coming from the founder figures. It's coming from the crazy youthful energy of invention that has always characterized the American soul.

Yeah.

We're capable of it, though. I mean, I think that's just the going through it. Like, they're in that... first of all, huge credit to the department because if you went back even 10 years, none of these people existed. And it's not because we didn't have them in America. It's because there was no way the department was going to do it. Like, the business case didn't meet. No one was going to buy that. Now you have the department leaning in, recognizing that a maverick like Dino is not a problem. He's the solution. How do we make more bets?