Palantir CEO Alex Karp On Government Contracts, Immigration, and the Future of Work

WIRED

WIRED's Steven Levy sits down for The Big Interview with Palantir CEO Alex Karp. From the benefits of embracing the government in an era when Silicon Valley kept a calculated distance to the criticism levied at Palantir from each end of the political spectrum, listen as Alex Karp breaks down the philosophies behind, and ambitions of, the controversial company he leads.  Director: Justin Wolfson Director of Photography: Matthew Caton Editor: Paul Tael; Cory Stevens Host: Steven Levy Guest: Alex Karp Line Producer: Jamie Rasmussen Associate Producer: Paul Gulyas Production Manager: Peter Brunette Production Coordinator: Rhyan Lark Camera Operator: Steve Pitre Assistant Camera: Joe Barnett Gaffer: Conner Wong Sound Mixer: Henrique Campos Ligeiro Production Assistant: Darwin Saquilayan Post Production Supervisor: Christian Olguin Supervising Editor: Eduardo Araujo Assistant Editor: Justin Symonds

Transcript

00:00We have to build systems like Palantir and we have to explain to people with working class skills,

00:04people with normal jobs, that their labor is going to become more important and more valuable over

00:08time with products like ours. If you run around saying that, you know, all the economy is going

00:12to completely shift to value creation only owned by 10,000 people, people on the left and the right

00:18are going to go nuts. I'm Stephen Levy, editor at large at Wired. I sat down with Alex Karp,

00:23CEO of Palantir, a controversial company that orchestrates data and solves problems

00:28for government and business. We discussed Palantir's work with ICE in Israel

00:32and why he feels the company has no competitors. Welcome to the big interview.

00:43Well, first of all, let me say welcome to the big interview, Dr. Alex Karp. You said that Palantir

00:50is at odds with Silicon Valley from the very beginning. You have kind of a harsh critique

00:55of Silicon Valley. You want Palantir to be different. What do you think the biggest problem

01:02Silicon Valley was? Well, we have a long and meandering relationship with Silicon Valley.

01:07In the beginning, we were at odds because we were pro-American, pro-West and pro-making the

01:12government functional. And that was very controversial in Silicon Valley because it equated to not making

01:17any money and being a loser. We won that battle. And I think Silicon Valley actually has become

01:21very, at least officially, and even behind closed doors, patriotic. And Silicon Valley has always

01:27been pro-meritocracy. And we're very aligned with that. Where we're currently at like a misalignment

01:33alignment apex again is we believe in using large language models in a way that creates actual

01:39empirical value, but also is very strong for workers, technical people, whether you're a plumber,

01:45carpenter, electrician, or you are a high school grad working in a plant where you'd normally be doing

01:49engineering things or couldn't do them. We see our products as being able to enhance

01:53your market value. And so we're kind of pro-labor and pro-military. So we're much closer to Silicon

01:58Valley on the patriotism thing. But for decades, we fought Silicon Valley about government is

02:06something we should support. We walked towards the government when everyone walked away. That's how

02:10we ended up powering Maven, which is the US government's military. Yeah, he took advantage when

02:15the Google workforce said, we don't want Google to work on Maven. And then we built it. Yeah.

02:22But explain what Project Maven is. Well, so to find an adversary on a battlefield from any center,

02:28including space, is a massive integration analytics project. So how do you find basically a needle in a

02:33haystack and where that needle is also complex because it may look like your adversary, but maybe

02:39that person's actually working for you. Maybe they're in your service. Maybe their phone has been

02:44compromised. Finding that needle in a haystack. So find a person that is your adversary you want

02:49to take out across a massive landscape where you have a human controlled in the loop, like targeting

02:54workbench. So you say, this target's been approved. Why? Under what conditions? The data protections that

03:00are necessary, the civil liberty protections, interestingly, that are necessary. And doing

03:03that at scale on the battlefield was not possible. And now it's been used in every operational context

03:08across the world. And, you know, there are many things that have led to America's ability to

03:14reestablish its deterrence, I would say, after the disaster in Afghanistan. But one of them was

03:19actually one of many corrosives of our soldiers. I think one of the other things is just the

03:24superiority of the targeting capabilities the U.S. government has. And, you know, there are other

03:29people and other companies, but we are the backbone of that. A lot of your government connections,

03:33they go back early. You know, you've worked for ICE. You've had a long-term contract with them,

03:38right? Well, that's actually later. I mean, our first contracts were at U.S. intelligence, so CIA,

03:44special operations, and FBI. And our biggest contracts ended up being in the DOD. And then later,

03:51we started working at Homeland Security under Democrats. Now it's very controversial because

03:55it's Trump, in my view. Well, it's controversial because he's doing things that we haven't seen done

04:00before. I want to ask you, do you keep an eye on that? Are you making sure that your products are

04:08being used according to your code, which supports democracy and, you know, says people can't be

04:14discriminated against? I'm wondering if there's a line for you where you would say to a president,

04:20no. We were the first, I mean, this is like, there's all these things that no one believes are

04:24true. I was the first, and I may be one of the few CEOs to say we would not build a Muslim database.

04:30Now, everyone thinks I'm like, and I am, you know, I've defended Israel. I think like people

04:36who wake up in the morning thinking about how Jews are running the world, and actually, you believe

04:40whatever you want to believe. I'm not so concerned about people who are skeptical about Jews or even

04:45they're anti-Semitic, but like people who wake up in the morning and the evening and night thinking

04:48about this. I think like, you know, Jewish derangement system is a little bit of a carcinogen. So I think

04:54I'm the last person in the world people kind of expect in the kind of rude and crude way people think of this

04:59to refuse revenue to build, you know, a Muslim database. And I would say more like at a high

05:06level, because I can't go into details, people internally know I pulled things from places where

05:11I thought there was anything like that going on in the U.S. Discrimination, yes, I'm against that. By the

05:16way, even if I was in favor of it, our products make that nearly as impossible as anyone's ever seen

05:22in a product. Because the way you do these things, the way you discriminate against people politically

05:27or based on their gender or race, is you do not have what are called ACLs. So you can't see where

05:32the data is flowing. What you can immediately see, not as a technical person, but as you,

05:37your cameraman, everyone watching this can go look and see how has the data merged? What kind of

05:42pipelines did you create? What were the assumptions that went into it? Where did the data flow? But where

05:47you and I disagree maybe is I spent half my life in Europe. I am an immigration skeptic. Like, and I

05:55personally think that U.S. people, meaning citizens have to decide by their vote, what our immigration

06:04policy is going to be. And I'll tell you, one of the reasons I'm very skeptical of my own party is

06:08that open border was not only a disaster, but it was tolerated. And that is massively corrosive

06:15and it's corrosive to democracy and it's corrosive to people. And leaving all the theory aside,

06:20like I was, I grew up in a progressive family. Classic progressives, my age and older, were

06:25massive immigration skeptics. Open borders is not a progressive policy. It is not a progressive

06:31policy. It never was. And then the fruits of this, you know, I spent most of my life, adult life in

06:37Germany. Look at what a version of open borders is a little different, has done to Germany. On any vector,

06:42rights of women, rights of gay people, future of society. So again, I just want to make it clear. I am against

06:48discrimination. I'm not saying you're saying this, but people tend to merge being against discrimination

06:53with being, to be against discrimination, I have to accept that there's nothing that can be done about

06:58illegal immigration in the West. And I just, I disagree with that. But more importantly than me

07:02disagreeing with it, the American people disagree with it. And I, quite frankly, you're not asking this. I think

07:07one of the reasons why the Democratic Party continues to fail and continues to fail and continues to fail

07:12is their talent managed. When you're like in businesses, like you have this problem where the people

07:17doing the work sometimes need to be told, no, you're wrong, and we're not doing that. And the Democratic Party and

07:23people like it, the Green Party in Germany, I don't view these policies as progressive.

07:26Well, I guess what I'm saying is, are you monitoring sort of what's happening with this government and

07:33democracy in this country and saying, maybe at some point, I have to look at...

07:38I would say the more important question is, have I ever worked against our commercial interests

07:42because it violated our norms? Yes. Have I done this in governments? Yes. Have I refused? We get no

07:48credit for this, but we almost went out of business because we were not working in Russia, China, or anyone

07:52else. So yes. Have we refused to give our product to foreign governments because we didn't agree with them?

07:58Yes. Do I agree with your implicit assertion that what's going on in immigration, as you formulate

08:04it, has never been done before? No. I think actually that's completely crazy. You know, when you're in

08:08Japan, if you miss your visa, do you know what happens to you? I didn't overstay my visa. Okay, great.

08:15I'll tell you what happens. You politely are put on a plane. Is Japan not a democracy? Do you know

08:20what happens in like Singapore? Well, I think... It is. But I just know, I think this is a very valuable line

08:27of questioning. You're asking, does our product allow for civil rights abuses and will I intervene?

08:33Yeah. Our product is the hardest in the world to violate. Will I intervene? Yes. Do I agree with

08:39some of the assertions you just made? No. The legitimate criticism could be, hey, these are

08:43two very rational people. You know, you're bringing up exactly the core and right question. And I'm telling

08:49you that I have done this. I will do it. And I continue to do it. And by the way, not that people our age

08:55like the term constructive engagement. I am as obnoxious and private as I am with you,

08:59with people I agree with. So like when they're asking me to do stupid shit, I can assure you,

09:04I'm telling you... Wait, is this stupid shit you're saying? No, no. I'm not saying you're asking me to do

09:07it. But I can assure you, one of the things you can be a certain of 100% is when somebody calls me

09:13and asks me to do some completely illogical, mostly it's illogical before it's immoral. That's the

09:17thing people understand. It's illogical. It will not get you the result you think. I am telling you,

09:22them, this is illogical and we're not going to do it. By the way, there's variants of this

09:26commercial. I have commercial clients, this is not even moral, who are like, hey, could you give

09:30me 50 engineers to do some crazy thing that will not help their business? And you know what I'm

09:34telling them? No. One of the things that happens when people work with me is they see I'm meeting

09:39clients and they'll ask me to do something that makes no sense. And I'll tell them, I can't go to my

09:44engineers and tell them to do something that makes no sense. And this is, I think a lot of these

09:48questions come down to that. Now, where we have to expand the prism of dialogue is, I think in

09:54complete honesty, there is a place where you and I, forget you and I, you step in, you're just

09:59honestly talking to you. It's just like talking to my family. It's like where me and my family disagree.

10:05And I'll tell you where those places are. They're on the border, they're in Israel, and they're in

10:09Ukraine. And what's super surprising is I get yelled at all day about those three issues. And by the

10:15way, the issue I don't get yelled at about the elite is about Ukraine. The issue I get, obviously,

10:20the Israel thing is like, you know, only thing anyone ever wants to talk about. You would not,

10:24it's like somehow people have forgotten it's a country with a GDP smaller than Switzerland. But,

10:28you know, and then my progressive friends, it's all ice all day. Some people might actually not even

10:33know what Palantir does. In the shortest possible way, do you explain to people who just know it as

10:40sort of something they see on the page and don't like and saying, hey, this is what it really is?

10:44Or love. We have a lot of people who love us and some people don't. But yeah, we tend to elicit

10:49strong reactions on both sides. So that's definitely fair. If you're internal or external

10:54intelligence, you're using us to find terrorists and find organized criminals while maintaining

11:00security and maintaining the data protection of your country. So like the highest data protection

11:05environments in the world, Western Europe, would love to get rid of us. They try every day to get rid

11:10of us. But building a data protection environment that actually allows you to do anti-organized

11:15criminality environment that also does data protection at any level of granularity, at any

11:20level of complexity is very, very hard. Ask the German government. They would love to find another

11:26alternative. And honestly, I'd have no problem with them finding it. So that's one thing we do.

11:30Then you have the battlefield, the Project Maven, and also organizing the battlefield, what the

11:34special forces use it for. How do you know where your troops are? How do you bring them home with

11:38all the data that you'd have from any source environment? How do you manage assets while

11:42doing that? Meaning assets from the other side, assets meaning spies, basically. How do you get

11:47the human? How do you prepare your egress? How do you get in and out of the battlefield

11:52safely, as safely as possible? So avoiding mines, avoiding enemies. Those things change rapidly in

11:58the battlefield. That's a solver problem that's powered in great part by us across America and with

12:04our allies. How do you supply targeting capabilities to allies, Ukraine, Israel, so that they get the

12:11product. But obviously, we don't have access to the full data they use because we're not part of

12:14those countries. That's a Palantir thing. On the commercial side, it's, you know, how do you,

12:20to make, I mean, the current version would be, if you want AI to work, you're going to have to have

12:25precursor things. You can think of it as like the precursors to make anything. What are those things

12:28going to be? You're going to need high fidelity data. To have high fidelity data, you're going to

12:32need to be able to integrate your data with a model you understand. You're going to have to

12:35build something called pipelining, meaning you can serialize and de-serialize a technical way

12:40of taking apart and putting together the assumptions on the data. You're going to have to be able to

12:43maintain your security model and maintain the logic and actions of your business. And you're going to

12:47have to be able to abstract that solontology. So shorthand, if you're doing anything that involves

12:53operational intelligence, whether it's analytics or AI, you're going to have to find something like

12:59ontology or foundry and, you know, and FDEs to install it. That's a Palantir.

13:04It's interesting because you, going back a little, you implied that, you know, the rest of

13:08Silicon Valley hasn't figured out how to do that. You're unique in exploring it.

13:17Yeah. What I'm really saying is we know how to do it. If you find someone else who can do it

13:22and you don't want to work with us, buy it from them.

13:24I mean, is there anyone you consider a competitor?

13:26Our competition is actually political. The woke left and the woke right wake up every day figuring

13:31out how they can hurt Palantir. And if they get into power, they'll hurt Palantir. How could you

13:36build a society with unfettered migration in Europe and not have Palantir? I don't know how you're going

13:40to do that, but they think they can do that. If they win, if the communists in France win,

13:44they will pull us out. If the, if like the Madani win of the democratic party takes over,

13:49I viewed that as my party, but if that's the party, I'm not in it. Or if the right woke wing,

13:55which is like, everything is a conspiracy. Any use of technology is actually going to only be used

14:00to eviscerate and attack us as like our data protection pro part of Palantir is built into

14:06the software. It is literally the hardest software to abuse in the world, but they don't seem to want

14:10to do it. But if you like, don't want meritocracy, either left or right, you don't like the consequences

14:15of it. You hate Palantir. And that is actually our competition. And by the way, not that they

14:20believe it. The thing that is protecting those people, the people that hate us the most is actually

14:25our product, because it is the hardest product to misuse, to, to take away the rights of human

14:30beings that's ever been built. And don't believe me, just go look at it, spend 10 minutes looking at

14:35it, find a technical person and look at our architecture or look at it yourself. And that's

14:40actually our competition. And, you know, in a weird way, we embrace the competition because they say

14:44these crazy things about us. And then the smart people go look at it.

14:48My head's spinning a little. How about now we get in there?

14:50I would say some of this is downstream from if you explain to the world that labor is going to be

14:55valueless, people are going to elect the most ridiculous people ever. We have to build systems

15:00like Palantir and we have to explain to people, but also let them see the results, the fruit of our

15:05labor, that their labor, meaning people with working class skills, people with normal jobs, that their labor is

15:12going to become more important and more valuable over time with products like ours, which happens

15:17to be true. If you run around saying that, you know, all the economy is going to completely shift

15:21to value creation only owned by 10,000 people, people on the left and the right are going to go

15:27nuts. And they're going to vote for people who are like, basically, yeah, nothing I've said has ever

15:31worked in history. It has never worked in history. You cannot find a time in history where any of that

15:37stuff has ever worked ever. They still will gravitate to it because they're afraid of what

15:42could work, which is like an AI driven AGI environment where no one has a job. And the

15:46only people who make money are like people sitting out here. And so that's part of the problem. Now,

15:52also, honestly, I mean, you know, I think universities and elite institutions have played a really

15:57corrosive role here. People are teaching pagan religion views. Again, pagan religion as it's a new

16:05of religion with sacrifices. Who's the sacrifice? Me. I'm the sacrifice. And that, again, it's also

16:10like it's a form of thought that has never worked. You don't seem constrained to me. You're doing great.

16:16I mean, I'm not complaining. I mean, it's like, we're not victims of Palantir. We're not playing

16:22the victimhood thing where we don't, I don't think I'm a victim. I don't actually believe we should,

16:27these things should be framed as victims. And yes, we're doing very well. And you know,

16:30we might even do better. You have this bonkers earnings call in February, huge, huge growth.

16:36Well, it's like people who are not financial, which is most people. And honestly, I wouldn't

16:41be paying attention to this if I wasn't running Palantir. Those numbers are a little bit like

16:45flying saucer took off numbers. So, you know, people are financial, many of whom doubted us are

16:51like, wait a minute, the plate dropped. Because what it basically, what non-financial people might not

16:56appreciate is we're not just growing, we're growing with high margins with, and that's at our

17:01scale, basically, in my view, never been done before, not in a similarly situated business.

17:06You said it was anomalous, like, you know, some sort of historic moment that the company is anomalous.

17:12You know, in a weird way, we undervalue these, or I undervalue the whole numeric side of our business,

17:17because it's downstream of value creation. So if you have, if you have, if you are doing anything like

17:24what we're doing, so call it 68% growth in the U.S. And again, we're doing it, you know, our way.

17:29Salesforce is anemic. We're not running around really trying to sell our wares. We have a conference

17:34here, people fight to get in, which is pretty cool. We have the best clients in the world, especially

17:39in America. Clearly, we've tapped into something. I'll tell you what we've tapped into. There's

17:43essentially a simple thing. LLMs and software-empowered LLMs create the expectation of

17:51prey to optimality between partner and vendor. You are going to get paid in the future a percentage

17:57of the value you create. And the value you create is going to be, you know, it's at some

18:02point objectively measured, and it's going to flow into your financials and vice versa.

18:06So you have this unique culture, and you even have called it a cult.

18:11It's been called a cult.

18:12I wonder how much you cultivate that outsider mentality. It reminds me, we're both from Philadelphia,

18:19of Jason Kelsey after the first Super Bowl saying, nobody likes us, we don't care.

18:24The thing people don't understand is there's a massive feature side to being an outsider

18:28or even popular. People forget, people always think of the bug size of being unpopular.

18:33Like, you know, you're unpopular. It's not pleasant. You're not going to like it. I don't

18:37like it, actually. But there are massive feature sides. Like, you get the best people in the world.

18:42Like, someone says something ridiculously stupid about Palantir. Five people look at it,

18:46and the fifth person who's exactly the kind of person you want inside your business or

18:51will be the person buying your product someday or investing in you says, hey, it can't be as

18:56simple as that. It's a conspiracy. It can't be as simple as they want to eviscerate my rights,

19:01because those rights would be the same rights that, like, an agent would need protected by the

19:05agency or something.

19:06Then those people investigate and investigate and investigate. And they're like, wow,

19:10this is a real interesting company. And they're thinking of the 10th derivative of this problem,

19:15not the simplistic bullshit that somebody online saying has no idea how our products work.

19:19And then you get the best people. Like, a really unexpected consequence of this,

19:24there's no country in the world where our brand is as bad as France. We have the best French

19:29employees in the world, especially in U.S. commercial. And why do they join? Because every

19:34time the Communist Party says, we are doing these crazy things, and they clearly know nothing about

19:40what they're talking about, the smartest person in France says, wait a minute, if it was that simple,

19:43any idiot would have done it. By the time they're done doing the diagnostic, they're like, hey,

19:47there's only one company in the world I want to work at. And that's how we have the best French

19:50employees in the world. The best French employees, arguably, in the world are at Palantir in America.

19:55And that's very counterintuitive. Working at a company that's apparently a CIA front, which is like,

20:00yeah, obviously, complete bullshit.

20:02Well, yeah, no, you may not be a front. Thank you so much, Alex.

20:04Thank you very much.

20:05Goodbye.

20:06Bye.

Category

Transcript

Be the first to comment

Recommended