YT YouTube2Text

YouTube2Text → Transcripts → Alex Karp on Why AI Will Hurt the Middle Class: You Can’t Have Billionaires Becoming 100x Wealthier

Alex Karp on Why AI Will Hurt the Middle Class: You Can’t Have Billionaires Becoming 100x Wealthier — Transcript

by The Builders with Walter Isaacson · 7,363 words · 1,237 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00How were we able to build the
0:01anti-terror product? That products
0:03changed the history of Europe. The
0:05problem was to build that product, you
0:08had to violate every single norm of how
0:11you built software. The biggest norm we
0:13violated is it took three years really
0:15roughly without revenue.
0:17Three years, no revenue. You would have
0:19to be basically insane to do that. And I
0:23do think it took a neurodivergent person
0:25at the helm to
0:27to fight for that.
0:34>> In the years after September 11th, Peter
0:37Thiel began thinking about a new
0:40problem. At PayPal, he and his
0:43colleagues had developed software that
0:45could identify fraud by finding patterns
0:49hidden across enormous amounts of
0:51information. Could a similar system help
0:54intelligence agencies connect data and
0:57discover threats before another attack?
1:00The idea became Palantir. And to lead
1:03the company, Thiel turned to an old
1:06friend from Stanford Law School named
1:08Alex Karp. Karp was an unusual choice to
1:12run a technology company,
1:14especially one working with intelligence
1:16agencies. He wasn't an engineer. He had
1:19studied philosophy in Germany.
1:22His politics leaned left. He was deeply
1:25suspicious of concentrated power and
1:28concerned about the threat of
1:30authoritarianism.
1:32Then he became the leader of one of the
1:34most powerful and controversial data
1:36companies in the world. So I wanted to
1:40ask him,
1:41who should be trusted with systems this
1:43powerful? Can artificial intelligence
1:46make governments and militaries more
1:49effective
1:50without making them less accountable?
1:52And when Alex Karp talked about
1:55defending the West,
1:57what exactly does he believe the is
2:00defending? I'm Walter Isaacson, and this
2:02is The Builders, my series of
2:04conversations with the most
2:06consequential minds shaping our future.
2:09Here's my conversation with Alex Karp.
2:12Alex Karp, welcome to The Builders.
2:14>> Very nice to be here. Huge honor.
2:16>> You know, one of my theories in
2:17biography is that it's all about dad. So
2:19tell me first about your father.
2:21>> My dad, he's a very smart person,
2:26very shaped by his German Jewish mother
2:31passing away very early, being raised by
2:34his father, and I think my parents were
2:38very much in love,
2:39but put it way it was not a coincidence
2:41that my dad married an a jarringly
2:44talented
2:46excruciatingly talented, somewhat
2:48difficult black woman. I think he loved
2:51the fact that it poked his stepmother,
2:54who was more status conscious in the
2:57eye. My mom is enormously just just
3:00capturing in her level of talent, and
3:04put it way my parents, had they been
3:06more what I think they could have been,
3:08they would have not birthed me.
3:10And I had this discomfort about the fact
3:13that, you know, this family that
3:14exploded
3:16in a messy divorce and with parents that
3:18are complicated
3:20created an artifact that is ideally
3:23morphed
3:24for building uh
3:27essentially pound here. You know, it's
3:29like if you take someone like me and you
3:32put it in this chaotic
3:34environment where you had the feeling
3:36you're on your own and you're
3:38subconsciously or partially consciously
3:40misaligned
3:42with your parents, we get a deeper trust
3:44in your own ability to survive, but
3:47maybe a latent mistrust of structures.
3:51>> How did being dyslexic give you wings to
3:53fly?
3:55>> You you always have to walk back from
3:56where you are to see where it got you.
3:58Like I I viewed it as a
4:00I mean, I was in the closet about being
4:02dyslexic until like six years ago.
4:04And then I just got tired. Honestly, I
4:06got frustrated. And then I kind of told
4:08a buddy of mine. But you know, it's
4:09interesting for me when I talk about to
4:12watch how people like this derided
4:14everything we did, derided everything I
4:16said, and now basically copy it
4:19it like verbatim.
4:21Now why does that happen? It happens
4:23because they're looking for a playbook.
4:26If you're dyslexic, the way I define it
4:28at least, it's because it's a broad
4:29term. I define it as having a um
4:32attenuate a a a a a non-one-to-one
4:35relationship
4:36with the taxonomy of a language, written
4:40taxonomy. So even now if you read and I
4:43read, there's more of a translation
4:44function even though I you know I did a
4:46PhD and
4:47I still do a lot of up doing a lot of
4:49the writing at Palantir. But why is that
4:51important? Because language, of course,
4:54you want to have you prefer not to be
4:56dyslexic, but the truth is if you are
4:59dyslexic, you're not colonized by the
5:01broader culture. So when I'm interacting
5:04with with with the broader culture or
5:06with a tech problem or with any problem,
5:09I'm interacting with the problem as it
5:11is. So how were we able to build the
5:13anti-terror product? That product
5:15changed the history of Europe. We've
5:17stopped
5:18hundreds of terror attacks, some of
5:20which would have have happened in other
5:22places, but many of which would would
5:23have happened in Europe. Think of the
5:25Europe you would have if that had
5:26happened. And then later our product is
5:29fundamental to the defense of Ukraine.
5:31>> Oh wait, explain to me how that comes
5:32from being either not colonized.
5:34>> so first of all, if you go to school and
5:37how to learn software, build software
5:39company, you would have learned build a
5:41product that's not very valuable to the
5:43client where they're addicted to it,
5:45then hire lots of salespeople and get
5:47lots of people to buy it against their
5:49will. And that's how you build a
5:50software and then then the derivative of
5:52that is it's a simple but addicting
5:54product. So it doesn't really create
5:56value, but it it's hard to get rid of.
5:58It's like a venereal disease. Easy to
6:00get, maybe hard to get rid of. And what
6:02you had to avoid at all costs is
6:05anything that looked like labor. We
6:07spent three years building a product
6:10that was valuable. Okay, so then you get
6:11to what is value.
6:13We said value is both that you stop the
6:15terror tax and you protect civil
6:17liberties. Okay, that already you had
6:20two factions, one that wanted to stop
6:22terror tax and one to wanted to protect
6:24civil liberties. So the problem was to
6:27build that product you had to violate
6:30every single norm of how you built
6:32software.
6:33And it and then the biggest norm we
6:35violated is it took three years really
6:38roughly without revenue.
6:39Three years, no revenue.
6:41So
6:43you would have to be blank you'd have to
6:45be completely
6:47basically insane to do that. And I do
6:50think it took a neurodivergent person at
6:52the helm to fight for that. People
6:54always like I often get these comments
6:56in private from business executives how
6:59can you teach me how to be unpopular?
7:01>> You clearly grew up feeling a bit of an
7:04outsider. Another theory I have about
7:05biography and being an outsider can be a
7:07superpower in some ways in the sense
7:09that you aren't trying to fit in. You
7:11know inside
7:13>> Or maybe you've tried to fit in but it's
7:14failed.
7:15>> Did you try?
7:16>> Not no, but I think human beings
7:20as a species
7:21want to fit in.
7:23And I think most of us who are not
7:25trying subconsciously realize it would
7:28have failed to try.
7:30>> To what extent does that help you say as
7:32opposed to somebody like me who kind of
7:34feels oh I've always fit in to the child
7:36want to get along.
7:38>> Jewish guy in New Orleans. I I I think I
7:41love what yeah I I'm not sure that's
7:42true because you would not be able to
7:44write the things you've written if you
7:46were purely an insider. But any case
7:49that you'll I it's you're not me
7:50interviewing you, but I would if I were
7:52interviewing you I'd push back on that
7:53statement. No, I I think being an
7:55outside I don't think you get Palantir
7:58or my life story it's it's very unlikely
8:01you're going to get as an insider. And
8:02the and I think the inside track is
8:04overworn and not working.
8:07And so uh and then I think I think if
8:10you link it to what we've built at
8:11Palantir, the culture, the esthetics,
8:13the products, there is no
8:16there's not a single thing we've done
8:17that an insider would ever do. And most
8:20outsiders are pretending to be
8:22outsiders. Since there's so much value
8:23in being an outsider, you have lots of
8:26people who are like fake outsiders. And
8:28like
8:28>> But that's not you.
8:29>> Well, I mean I think people like me
8:32would I'd love to go on an insider
8:34vacation. Give me like the eight eight
8:36weeks a year where I can be an insider.
8:38I And again, I know it would be totally
8:40stupid.
8:40>> Mhm.
8:41>> But like I want to be reborn as like a
8:44my my reborn fantasy is Catholic
8:47Bavarian veterinarian.
8:50>> Well, getting to Bavaria and to a
8:52related concept which is a
8:55collaborationist instinct that's in the
8:57human species. You wrote about the Nazis
9:00when you were uh
9:02in graduate school, was it?
9:03>> I wrote I wrote about it was it was a it
9:06was a critique of Heidegger.
9:08For better or worse, it'd be hard to say
9:10Heidegger was not one of the two or
9:12three most important philosophers
9:15uh
9:16in the last couple hundred years along
9:18with other people, maybe Wittgenstein.
9:21Uh Heidegger
9:23developed a form of phenomenology, which
9:25is a fancy way of explaining how things
9:29in the world could impute value onto the
9:31world. And there are many different
9:32forms. His advisor was Husserl, who was
9:34Jewish. Heidegger became
9:36an ardent supporter of Hitler.
9:39Uh even
9:40though his mistress Hannah Arendt was
9:42Jewish and his advisor was Jewish. And I
9:44think Heideggerian philosophy is the
9:47most important political philosophy, I
9:50think for worse
9:52in the West now because most political
9:55movements that I would call regressive,
9:58far right, meaning they want to regress
10:00to something that obviously has never
10:02worked on the far right and on the far
10:04left
10:05pretend to be Marxist, certainly on the
10:07left, but are actually Heideggerian.
10:10And that's why I think, well, why I've
10:11been very early and for a year saying,
10:14like I'm still an unhappy Democrat. I've
10:16been warning the Democrats that a
10:17faction of their party is actually not
10:20progressive. They call themselves
10:21progressive and they want a form of
10:24communitarian
10:26uh economy that
10:28looks feudal.
10:31It's not even actually socialist and
10:33they pretend to be isolationist when
10:35they're actually just focused on
10:38the groups they're uncomfortable with.
10:39That's a right-wing party.
10:42So in any case, that I wrote about
10:44Heidegger. Now, the truth is also I
10:46wanted to get out of jail. Like I
10:47realized halfway through my
10:48dissertation, this is not for me. I need
10:50to go and build something. To your
10:52thing, I'm a builder.
10:54The interesting thing about builders
10:56is it's elliptical. So the people you've
10:58written about, um Steve Jobs, Elon Musk,
11:02I think you will find in every case it's
11:04a combination of an outsider
11:08with an aesthetic
11:10proclivity that is as much humanities
11:13as
11:15tech. I think objectively, although many
11:17most people in tech would disagree,
11:19it's more
11:21aesthetic
11:23world importance and business acumen
11:27than tech, which is why the tech
11:28companies are built here and not
11:30unfortunately in Germany or France,
11:32because it's not that we're smarter,
11:35but we are producing
11:37>> Why did you found Pal Palantir? What was
11:39the point of it?
11:40>> Well,
11:41I mean my my version of why I founded it
11:44was I knew Peter in in law school.
11:46>> Peter Thiel, where did
11:46>> Peter Thiel. I was doing no work.
11:49Peter I think was doing much more work
11:50than I, but we we're in the same section
11:53and in the same dorm, which was like the
11:55dorm you were at if you didn't have a
11:57lot of money. And then we were both
11:59intellectually very curious and he came
12:01to me and was like, we have this idea
12:03to bring a methodology that was used at
12:05PayPal to the Intel space because
12:08basically it had worked to cut down
12:09cyber fraud and they obviously aren't
12:12good at this kind of tech.
12:13And I thought this is wildly interesting
12:16and Peter was is wildly compelling.
12:19And then I just said yes.
12:21I joined because I thought the idea
12:24Look, you can take a jingoistic version
12:27of America like everything. Of course
12:29we've made mistakes. And some of those
12:32mistakes are hard to understand in
12:33retrospect.
12:35Mistakes around integration respect I
12:38mean you know we could obviously
12:40slavery.
12:42I I think the class not
12:45not paying enough attention to class and
12:48not having the national project be
12:50stronger. There's a lot of lists, but in
12:52my view in the view of
12:55I think most Americans there's no
12:57project at scale. So you can't really
12:59compare America to like Sweden of the
13:02'70s, right? This is like a completely
13:05different the time completely
13:08homogeneous culture.
13:10>> Are you saying the diversity of America
13:11makes
13:12>> Well the size It's the size and grandeur
13:15of our project.
13:17>> Mhm.
13:17>> And the fact that we have this
13:19constitution
13:21that is still it's still fresh. Like you
13:25know again the first in my view the
13:28Second Amendment
13:30and the Fourth Amendment. So write a
13:31free write a free speech
13:34which basically also is a stand-in for
13:37freedom to decide your own religion,
13:39which is very radical. The Second
13:41Amendment Okay, I don't believe we
13:43should allow plastic 50 cow guns roaming
13:46through the streets, okay? But
13:48but I do believe that it's crucial to
13:50allow citizens the right to bear arms
13:53because
13:54it it's often reframed as self-defense,
13:56which I'm also in favor of.
13:58But it's really about putting the
14:00government on its back heel about
14:03mandating
14:05what official policies would be cultural
14:08>> Amendment pushes you back a bit.
14:10>> I know and I'm going to get to that, but
14:11I just want to say I just want to
14:12underscore.
14:14And again I I actually think I've I've
14:16never been a neocon. I never thought it
14:18made sense to make like the Muslim world
14:21it our some we have Muslim many Muslim
14:23country partners. It runs very well. And
14:26I'm not I Of course I want a different
14:28kind of government. I actually as much
14:30as I'm on the other side of the CCP, I'm
14:33want America to win more than I'm
14:35interested in how they run their world.
14:37That they It seems to they believe work
14:39is a I I'm never been of that school.
14:42But we have an ability I can say
14:44whatever I want here, and there is no
14:46government sanction to what I say. Okay,
14:48the Second Amendment does not exist to
14:50my knowledge anywhere else in the world.
14:52And so first and second, and they're
14:54underscoring
14:55still
14:57enormously radical. And I would say
15:00this is another reason I very skeptical
15:02of the
15:03fringe right and and not as fringe DSA
15:07people because the first thing they when
15:09they talk the first thing they want to
15:10do is get rid of our Constitution and
15:12their right to do that because our
15:13Constitution is the most productive
15:15thing. It's still keeping this country
15:17working. But all of that as necessary as
15:19it would would be first and second is
15:21not sufficient if you don't protect
15:23people's rights to privacy to use your
15:25Frank to rein that in.
15:28Then like what is your right of privacy?
15:31So your right of privacy is to conjure
15:33up things in your own space that might
15:36be completely heretical or offensive.
15:39>> So we're talking about the Fourth
15:40Amendment right to privacy. So many
15:42people would be talking about Palantir,
15:44the cameras gathering information on us.
15:47Doesn't that impinge on our right to
15:49privacy or do we overindex on privacy?
15:53>> No, well, the
15:55First of all, one of the things I think
15:57Europe got right was basically the
15:59European constitutional order focuses on
16:01privacy both in the commercial and in
16:04the government context.
16:06In America,
16:07our Constitution protects against go
16:09government overreach. And I would say
16:11culturally I think we're not sensitive
16:13enough to commercial overreach
16:15uh in the privacy.
16:16>> All
16:16>> students say they don't mind Google
16:18>> everything about them, but they don't
16:19want the government to know everything
16:21about them, which is odd.
16:22>> Yeah, no, the
16:24I'm in favor of them not wanting
16:26government to know everything about
16:27them. I They're underestimating
16:29Basically, what they're often imputing
16:31is I don't care what commercial
16:33companies know about me cuz that's my
16:36choice, which by the way it is,
16:38obviously. But they're thinking of their
16:41opinions as being wholly their own.
16:44And at best, your opinions are a
16:45dialogue between your
16:48ideas and the outside consensus view.
16:52They're underestimating how
16:55much their thoughts are being influenced
16:58by
16:59the tech products they're interacting
17:01with.
17:02And because they underestimate that,
17:04they don't understand the in the the the
17:08the way in which they should be more
17:09concerned about privacy. In fact, where
17:11I'm sympathetic with some of our critics
17:14of Palantir, it's they've got the
17:15Palantir part wrong.
17:17But you'll see like Palantir
17:18surveillance state, blah blah blah blah.
17:20Okay, the the Palantir stuff is
17:21technically and in in practice
17:25the inverse of what they think or by the
17:28most extreme amount of any other tech of
17:30any tech company. The part that they're
17:32actually right about
17:33is that
17:35if you have companies controlling all
17:37data, they will control what you think
17:39or
17:40influ- heavily influence what you think.
17:42>> But isn't Palantir part of a
17:43surveillance culture and there's an
17:45upside to that, but also
17:47maybe loss of autonomy?
17:52>> it Of course, you have to get into
17:53definition of what we mean by
17:55surveillance.
17:56If you are an enemy of the US shooting
17:59at the US or planning to shoot at the US
18:02or trying to stop people who are coming
18:04into your country to dismantle
18:06>> Mhm.
18:06>> something you're doing that's nefarious
18:08and bad. Okay, but a lot of people don't
18:09agree with But okay. Yes,
18:12our products are used to make
18:14>> But what about when I'm walking down the
18:15streets in New Orleans and you have a
18:17system there that says, okay, here's
18:20where Walter was.
18:22>> Um, I mean, the domestic surveillance is
18:24where the issue comes in.
18:25>> I meant.
18:26>> But in a weird way, Palantir
18:28is used heavily in France
18:30and
18:31in other places in Europe where we're
18:34contractually not allowed to mention.
18:36There's a difference between
18:37surveillance under the context of a
18:39legal mandate
18:41and surveillance where there's no legal
18:43mandate. And this this distinction is
18:45actually both politically, morally, and
18:46technically important.
18:48So I think the political morality most
18:51people get that,
18:52but it's very hard to build a product
18:54that, for example, in the European
18:55context can only keep data where there
18:58is a warrant and then purges it where
19:00the warrant isn't
19:01uh
19:02expires or
19:04the hard part is you have warrant
19:06warrant warrant disappears. What happens
19:08to the data?
19:10Our products allow you to work in
19:11conformity with that. Some people say
19:13that's still surveillance. You're
19:14surveiling a
19:16someone the courts have decided is a
19:18criminal.
19:19I think there you get into where I would
19:21say, you know, you have people that are
19:23so pro you have the the right right
19:25version caricatured where they're like
19:27just kill the terrorists and criminals.
19:30Okay, that's not a society most of us
19:32want to live in. You have the left
19:33version that I think just brings the
19:35right, which is
19:36no, we're not going to surveil known
19:39hardened
19:40criminals. And where our product is used
19:42most effectively is in Europe on
19:45criminal gains, professional criminals,
19:48meaning they're organized and
19:51trained
19:52uh
19:53uh
19:54uh I imagine on border enforcement,
19:56which is because it's all done by
19:58professional criminals to get people in
20:01and on anti-terror, like on terrorism
20:03cases.
20:04And
20:04there you have in the case of domestic
20:07Intel, you have a
20:10essentially something like a probable
20:11cause and a construct that allows you to
20:14legally do it. So in practice is a form
20:17of law enforcement, I think everyone 90%
20:20of society supports. And I would say the
20:2210% that doesn't is basically going to
20:25get us
20:26some form of fascism. Like if the
20:27terrorist attacks in Europe that we had
20:29prevented had happened,
20:31you would have some form of left, far
20:34left, or far right fascism would amount
20:36to basically the same thing. You are You
20:38are tracked all the time.
20:39>> Well, let me Let me broaden it a bit
20:40more philosophical, not just to
20:42Palantir, but the technology we have
20:44today has enormous amounts of sort of
20:47data gathering capabilities, cameras and
20:50visuals and facial recognition. Do you
20:52think this whole explosion of data
20:56collection combined with AI in some ways
20:59impinges on personal autonomy?
21:01>> Well,
21:03the Look, obviously it can and it will
21:06if you're not If you don't push against
21:07it. So if now we go into the AI debate,
21:10AI is
21:12in my view best conceived as a raw
21:14material
21:15that can be processed for good or bad
21:18broadly broadly speaking.
21:21So one of the most negative ways you
21:23could use these harness these powers is
21:27for surveillance.
21:29I would
21:31argue that most of that is going to
21:33happen in the commercial context, not in
21:36the government context. But it could
21:38happen in the government context, and
21:39I'm very in favor of actual
21:41Like I think we should restrict some of
21:43these technologies. Like For example, if
21:44you just take AI that we use on the
21:47battlefield, we should have legislation
21:49that prevents that technology from being
21:51used domestically. We need it on the
21:53battlefield or We have no choice because
21:55our adversaries are building stuff, and
21:56if we don't build it, we're going to be
21:58just demolished. And I would say if
22:00you're anti-war, in my humble opinion,
22:03looking at any part of history, you have
22:04to have a strong military. So the idea
22:07you're going to weaken the military and
22:08have more peace is
22:10a fantasy of people who are half idiots
22:12or actually who want war. But then the
22:14domestic part that you're flagging is a
22:16very real issue.
22:18Like what what are the technologies
22:21we're going to unleash? Like again, the
22:23famous You know, fascist countries have
22:26no crime.
22:27No crime. When the Nazis had power,
22:30the only crimes that were committed were
22:31by them.
22:33That that's how it works. There there is
22:35only one criminal band. It's the
22:37government.
22:38And there will be no other criminal
22:39band, and they will take people out and
22:41shoot them, and lo and behold, there's
22:42no crime. And they In order to do that,
22:45they shoot a lot of innocent people,
22:46too.
22:47You know, criminals are rational actors,
22:49and probably they're the people who get
22:50shot the least.
22:52The people get shot the most are people
22:53with political views like me, honestly.
22:56That's if you go tie this back to my
22:57outsider thing. I always thought
23:00from the time I can remember thinking,
23:03I will be the first person on the wall.
23:06And I
23:07believe I will be the first or second
23:08person on the wall if we have no crime.
23:12You know, and because the the first
23:13thing that they do is define you as the
23:16people with an actual view. I'm talking
23:18about things that like think kind of
23:19things I say that are outside the norm
23:21and are transgressive and a kind
23:23building a company or having too much
23:25money or you know, if you don't protect
23:27the rights of the people you don't like,
23:29the unapologetically, unabashedly
23:32opinionated people that also have money,
23:34you you you're going to end up in a
23:36society
23:36>> And you think we have an intellectual
23:38fragility that hurts us from doing that?
23:41>> I think I think that we're If we're not
23:44very careful in this country, we are
23:46going to end up with a country run
23:50heavily influenced by the extremes. Like
23:52take wealth tax. There are class
23:55inequities in the society that have to
23:56be addressed. And AI is going to
23:58exacerbate them. I mean, I believe AI
24:00will create more jobs or a bit higher
24:03quality jobs, especially for working
24:05class vocational people, and many people
24:07will get retrained. But even if their
24:09salaries go up twice inflation, they're
24:12unlikely to stay at pace with people who
24:15are in tech and in AI because those
24:17people could become 10, 20, 50 times
24:19wealthier.
24:20Okay? That Then the reaction to this is
24:23to impose something that just hurts
24:26just hurts the wealthy. It doesn't help
24:28the poor. And by the way, it's going to
24:29hurt the middle class for lots of
24:31reasons.
24:31>> Well, wait a minute. If you agree that
24:33AI is going to lead to greater
24:35disparities of wealth, then you go back
24:37in history, 1890s we had just as big of
24:39a tech boom, led to great disparities of
24:42wealth. Isn't there some reason to say
24:45let's put more in the commons that
24:47reduce
24:48disparity of wealth?
24:49>> That's what I'm saying. You have to do
24:51something.
24:52>> Yeah.
24:52>> But then like
24:54Well, my version would be you go to you
24:57basically say you have a 50%
24:59uh
25:00death tax.
25:01>> Mhm.
25:01>> I would go to people and say, okay, for
25:04every
25:05point you pay in, you can have some
25:07points. So you you could basically you
25:09pay 25% now and in But the government
25:11can only collect
25:13by doing half as much waste
25:15identification. Cuz the left's
25:17legitimate critique I agree with, most
25:19people agree with is you can't have
25:20people becoming a hundred times
25:21wealthier when they're already
25:23billionaires and everyone else is
25:24becoming 5% wealthier, okay? So that's
25:27legitimate.
25:27>> Mhm.
25:28>> The the legitimate critique of the
25:30people who
25:31are economically literate is, yes, but
25:35this isn't really going to help people.
25:36This is half of it is being wasted. So
25:39the way I would propose it is great. You
25:41have to tackle both. And it's the same
25:43thing as you have to protect civil
25:44liberties and fight terrorists. The
25:47other problem with the wealth tax, by
25:48the way, that I for some reason people
25:50haven't talked about is
25:52to enforce a wealth tax, you're going to
25:55need surveillance
25:56because I'm going to need to know that,
25:58you know, what is the bench in your
26:00house in New Orleans worth?
26:02>> Let's leave aside the practicalities,
26:03which seem very impractical at assessing
26:06each person's wealth and
26:07>> No, but no, sorry, sorry. Let me just on
26:08one this one point push back.
26:10They're not impractical with AI.
26:13It's exactly the point you were making.
26:15They're not impractical with AI. AI
26:17could impute every single thing you own,
26:20especially with all the data about It
26:21can impute the things
26:22>> Well, then it's wrong with then taxing
26:24people's wealth.
26:26>> Because it does not A, it doesn't help
26:28the poor.
26:30It breaks consensus in the society. It
26:32breaks the tech culture because people
26:34are liquid. So all those have been
26:35talked about. And the logical
26:38consequence of it is surveillance into
26:40your personal life. What do you do you
26:42>> give me a good Palantir way to say we
26:44have too much economic disparity. We
26:46want to have more common ground in our
26:48society. How do you do it besides uh
26:51maybe more government efficiency being
26:53the tradeoff?
26:54>> do a tradeoff where taxation goes up. My
26:57I gave you my version, but any other
26:58version besides a wealth tax, it goes up
27:01under the condition that for every unit
27:04of tax you raise a half of unit of waste
27:07is identified.
27:09And that transfers a lot more to the
27:11government because the problem you're
27:13trying to solve two problems,
27:16the inequity, which is a problem, and
27:19the inequity uh which is also a problem,
27:21which is how do you make the life of a
27:23poor person better? They're not the same
27:25problem. I'm in favor of people like me
27:28paying a greater share
27:30under the condition
27:32that it
27:33doesn't destroy American tech, which a
27:35wealth tax would do, does not expand the
27:37surveillance state, which it absolutely
27:39would do.
27:41And it lands in a way that helps working
27:45class and
27:47mid upper middle class and below. If you
27:49pick one, you just divide society. And
27:52the people who pay for that division are
27:54the people who have no voice. And de
27:57facto, you have no voice in the society
27:59if you're actually poor.
28:01Because the people repping you are
28:03interested in killing the wealthy.
28:05They're not interested in helping the
28:06poor structurally.
28:07>> Let's talk more about uh digital uh
28:10training on information, surveillance.
28:12Uh in some ways you've been talking very
28:14strongly about digital sovereignty now
28:17that the big LLM companies are not only
28:20sucking up all our data and training, uh
28:23they're doing it without our permission
28:24now building on the back of us.
28:26>> Well, US enterprises are livid and
28:29enraged, apoplectic. I don't know what
28:31the
28:32because what they've discovered is
28:34they're paying a lot of money for a
28:36product that is not created a lot of
28:38value engineered.
28:39>> Talking about LLMs at the moment.
28:40>> LLMs.
28:42Uh
28:43Yeah, and then lo and behold you find
28:45that while you're actually paying for
28:47them, we're training the model. Okay, so
28:49they are training the model. How are
28:50they training the model? They're
28:51training the model on your data. What
28:53data is most valuable? Data that's
28:55proprietary. De de facto this is what
28:57happens if you use any of these products
28:59without essentially
29:01a protective layer.
29:02>> ways like the old McKinsey and Companies
29:04management consultants
29:05>> Yeah, well yeah, but think times um
29:095 million and scaled. Yeah, it's like
29:12and and then and then it actually shows
29:14up in your business is theirs and in any
29:16case they're particularly angry uh
29:20at Anthropic. I
29:23it's funny I spend half I feel like I
29:24spend 10% of my day explaining to
29:27Enterprises Dario is not the caricature.
29:29He's not that you think he's not Lord
29:31Vader.
29:32Uh
29:34they have a philosophy that they believe
29:37and they ask well if they're so ethical,
29:39how can they take my data and turn it
29:41into a product that is competing against
29:43me while I'm paying for it.
29:45>> So how does Palantir fight against that?
29:47>> Well, you know our first of all our
29:49sovereignty thing is primarily we the
29:52secret of Palantir that is most obvious
29:55and most hard to replicate is we
29:58actually are aligned with the interests
30:00of our partners.
30:01So mostly we're like we're educating you
30:04know people are asking us well how do we
30:05do this? What do the contracts look
30:07like?
30:08How do we interact with closed models,
30:09open models? We're publishing document
30:12after document after document not just
30:14for our clients and partners. With our
30:15clients and partners
30:17uh
30:18we're you know they use what what's
30:21called and broadly called an ontology
30:23now a concept that we developed that was
30:26also derided that now everybody talks
30:28about but you can think of a contract an
30:29ontology
30:31as a a a contextual framework
30:34where that contextual framework is de
30:36facto what you're doing. You're taking
30:38random things I'm saying and putting
30:40them into a frame where somebody would
30:42find this interesting and intelligible.
30:44So basically the ontology is both a
30:47alpha creator, meaning it takes these
30:49raw products and makes them valuable,
30:51and a uh and then a protective layer
30:55um which means you can interact with it
30:57without being burned. But the larger
30:59point of the sovereignty thing is
31:02we believe in sovereign institutions and
31:05sovereign individuals. Then how do you
31:08apply that? Well, you know, that's
31:10complicated, but in the end,
31:12you know, whether it's visa vis consumer
31:14or visa vis government, the questions
31:15you've asked, it should still end up
31:17being there's an imputed version of a
31:20first amendment and a fourth amendment
31:22in your life. And then there's a
31:23national version like
31:25like in the in the the stuff we're doing
31:27at the Department of War which is
31:30very powerful or the stuff that's being
31:32used in Ukraine or in Israel.
31:34Like they have ways of doing things and
31:36data sets and ways of understanding the
31:38data and ways of implementing it on the
31:40battlefield that no one else should
31:41understand.
31:42>> Do you think by the way that uh
31:45more American companies should be
31:47willing
31:48to enlist in uh our military and
31:52national ambitions? And what do you
31:54think of when Anthropic or even Google
31:56or people say no no no, we don't want to
31:58be a part of
31:59>> The the the Silicon Valley is a
32:01wondrous, unique
32:04the world is jealous of place
32:06for good reason.
32:07>> Mhm.
32:08>> Uh it is
32:09a highly insulated island. The only
32:13institution that
32:15a super majority of Americans think is
32:17indisputably noble in this country is
32:19the US military.
32:21And
32:21>> And the valley was born out of
32:24>> valley Well, it's the hypocrisy would
32:26not even exist without the US military.
32:28It's Thank you for reminding everyone of
32:29this.
32:30The valley was built as a hybrid
32:33uh basically university, government
32:36capital.
32:36>> Of a Neuberger model in which yeah,
32:39academia and
32:40>> It was mostly Stanford, the government,
32:43various departments, and people who came
32:46up with the venture model.
32:48>> And so what happened?
32:49>> Well,
32:50a lot happened and we made a lot of
32:52mistakes where, you know, I think I do
32:54think the government lost its coolness.
32:57I think Elon,
32:59I believe Palantir and others, Palantir
33:02graduates have helped bring that back in
33:04the context of
33:06defense, defense tech, defense tech
33:08adjacent. But I think a wave entered
33:12this country that's unfortunate where
33:15one of the ways you can
33:17signal to your community you're smart is
33:19by being craven, meaning you make other
33:22people feel smart while you transfer
33:23their IP to you. And that model's been
33:26very very successful and it the unit
33:28economics of it are very good, like
33:30unusually good.
33:32Uh crazily good. Do you get if you're a
33:35venture capitalist
33:37or you're involved in this, you can make
33:39a hundred, a thousand times your money.
33:41And by the way, that leads also to
33:42things that are legitimately good. If if
33:44if there was not so much wealth creation
33:46in the valley, the the bastard misfits,
33:50Palantir being the most
33:52best example of this, would never be
33:54able to raise money. We raise money on
33:56the back of the fact that everyone's
33:57getting wealthy, but we are the last
33:58people to get the money.
34:00>> But I'm go go back to this notion of
34:02technology engineering being in service
34:05of a greater
34:06>> but by not by but by breaking
34:09by breaking the explicit or definitely
34:12tacit covenant
34:14between wealth creation and creation of
34:16value,
34:18you break other tacit covenants in the
34:21society, meaning you get to get wealthy
34:24and we'll leave you alone
34:26with your vast wealth because you're
34:28also creating value in the society. And
34:30if you break that, you end up with a
34:31fracturing of the society.
34:32>> show me how we broke it. In other words,
34:34was the wealth creation done on
34:36frivolous dumb things that didn't add a
34:38lot of value?
34:39>> I I these things do create value, but if
34:41you're the average, say you're you're
34:43sitting in
34:44Louisiana,
34:46and you're a carpenter,
34:49uh
34:50do you feel like your life is better,
34:53the materially better than it was 30
34:55years ago? I don't think so.
34:57And so I think
34:58right or wrong,
35:00I think the perception is more right
35:01than wrong. You your perception is a
35:03small group of people have gotten a
35:05hundred, 200 times wealthier,
35:07and your life hasn't changed that much.
35:09>> And it's based on consumerism and ads.
35:11>> Consumerism, ads, but but in the end of
35:15the day,
35:16these things become micro
35:19subcultures and they're very divisionary
35:21where everyone in that micro culture is
35:24in the culture because they are the
35:25same. But the difference the difference
35:28is they keep you get you kicked out of
35:30the culture get smaller and smaller and
35:32smaller. Witness
35:33many Democrats refused to accept me as a
35:35Democrat because I thought they were
35:37part of the party that was going nuts.
35:39But we we we are struggling to find
35:41commonalities.
35:43Uh okay, so that's the backdrop. Then AI
35:45comes. You're taking already a fire
35:48that's burning and you're dumping
35:50gasoline on every part of it because the
35:52people who running are running these
35:54labs
35:55are saying your life is going to suck.
35:58Okay, so by the way, the people doing
36:00most damage here you in a normal
36:02political movement
36:03you'd have the people
36:04doing the movement saying hey this is
36:06going to be good. No, you have the
36:07people in the movement like your house
36:09is going to suck.
36:11You're going to have no job.
36:13And when you when you have no job you
36:15can rely on me to give you money. Oh but
36:18by the way, I'm going to be worth you
36:19know.
36:20>> But it seems we've lost actually two
36:21things. One, your life will be much
36:23better that we're not just creating some
36:25consumerism ad driven stuff. And
36:27secondly, a national ambition, a
36:29national greatness. Both those things
36:31seems to have been lost.
36:32>> linked because if you have a national
36:34ambition, you have to talk about things
36:36that will actually make your life better
36:38today, tomorrow.
36:40Not about the way how bad your life is
36:42going to be building a dream that you
36:44don't belong to. And by the way, they're
36:46linked because part of the reason the
36:48valley goes so wrong on these things is
36:50they're not talking to anyone in America
36:51and they don't care. They there's like
36:53it's an island. On that island, you're
36:55cool.
36:56>> Talk about the anti-government
36:57sentiment, especially coming out of the
36:59'60s and '70s in Silicon Valley, but
37:03there's a strong anti-government
37:04sentiment on the right now. Where's that
37:07come from and is that dangerous?
37:09>> Well, again, you know, I we at Palantir
37:11are always interested in what is the
37:12legitimate version of maybe a retrograde
37:15argument. So let me start with
37:19if you have a rise of a powerful
37:21technology that could be used to abuse
37:23you, used to undermine the imputed
37:25rights that you actually have under our
37:27Constitution
37:29directly or indirectly.
37:32And you've seen mismanagement of issues,
37:34call it COVID, which I think most people
37:36would say in some form or another was
37:38mismanaged. And
37:40you have no
37:42the national purpose
37:44measurement
37:46score is maybe
37:48the lowest I've ever seen.
37:50So you have high mistrust in government.
37:52You have low national purpose. You feel
37:54like people who claim to believe in
37:56things are just liars.
37:58And you have
38:00uh the people
38:02who are doing the best seem to be not so
38:04tethered to you.
38:06You would be in you would be it would be
38:08a little bit weird if you didn't have
38:09skepticism of the government.
38:11Now where I think it's very
38:13nonproductive
38:15uh again on the right and on the left,
38:17but I again just for clarity, I think
38:19the right you're referring to and the
38:21left I'm referring to are not actually
38:24right. So I think that right is actually
38:26left and the left is actually right.
38:28And that's one of the reasons like
38:30fascist movements
38:32tend to do this. It's very somehow very
38:34charismatic.
38:35>> Did you learn that in Germany?
38:36>> I spent a lot of time
38:39with Nazis, ex-Nazis
38:41when I was in Germany
38:43because I wanted to understand like real
38:44ex-Nazis. You had to go to the
38:45countryside to meet them because in the
38:47city there were no ex-nazis. Everybody
38:48was like, we hated Hitler. Just in the
38:50country, people were honest about it.
38:52And
38:54most of the people who I actually think
38:56that the anti-semitism
38:58was a motivation, but I I'm honestly for
39:01the record, I think a lot of the
39:03motivation was the charisma of Hitler to
39:05them and the national abuse of a
39:08national mandate which is justified
39:11and abuse of patriotism which is
39:13necessary to humans. And you see this
39:15Why is the regressionary left and the
39:17regressionary right very different
39:19movements?
39:20Why are they so charismatic to their
39:22peers? It's because if you go to the
39:25right
39:26which has been indoctrinated with a
39:29capitalism is always good, which is a
39:31simplistic thing that the the correct
39:34narrative is it's better than the
39:37alternatives. That's very different than
39:38it solves all problems. And the classic
39:42right of center
39:44institutions
39:46sell capitalism as it solves all
39:48problems, when in fact
39:51that's obviously wrong. It's creates it
39:53solves
39:54problems that other systems don't solve
39:56in a way they don't solve them and
39:57therefore is better, but it has negative
40:00externalities. So
40:01>> Such as?
40:02>> You know, egregious wealth dislocations
40:06that then obviously
40:09and impinge on political power and a
40:12dislocation
40:13of societies and narrative because
40:17you're not living in the same world if
40:18you have a billion dollars or if you
40:20have a hundred dollars in your account.
40:22Capitalism inherently
40:24assigns
40:25a dollar value to human life.
40:28And that dollar value is a derivative of
40:31your net worth.
40:32If you get hit by a car and you go to
40:35court, the driver hits you will pay a
40:38different sum presumably depending on
40:40whether it's you or a working class
40:42person unemployed or me.
40:43>> Mhm.
40:44>> So it inherently does that.
40:46>> Let let me end then with a question on
40:48what you just talked about, which is
40:50what does it mean to be human?
40:52And you look at, you know, I'll go to
40:53Creature for Space where they now have
40:55drones, and you've worked with that
40:57quite a bit, where humans are almost not
41:00in the loop anymore. That to me seems a
41:02bit of a metaphor of what we're facing
41:05with this new technology, is to what
41:08extent do humans not only have to be in
41:10the loop on things like systems, but
41:13to what extent when there's
41:16intelligence and maybe even
41:18consciousness in machines, what is the
41:20role of humans?
41:21>> Well, what is special
41:23about humans
41:25when our core capacity that we've built
41:29our moral systems on and our claim to
41:31superiority, which isn't framed this way
41:34for reasons,
41:35has been our cognitive
41:38uh
41:39>> Intelligence, yeah.
41:40>> Yeah, but what happens when we birth a
41:42species that we're that's no longer
41:44true?
41:45And um and this is where I again think
41:49the founding fathers of our country were
41:52something beyond genius-level wisdom
41:54people because you have a a different
41:57setup in Europe where the Constitution
42:00is birthed by parliamentarians.
42:03Our rights, according to our
42:05Constitution in this country,
42:07uh are given to us because uh of a
42:10granting and
42:11presupposition of a higher figure. That
42:13means there is something in the human
42:16spirit as
42:17protected under law and internalized by
42:20Americans. A reason why I'm very
42:22skeptical of changing of bringing open
42:24migration because it's very special.
42:27Where we believe deep in our hearts
42:30because of because of what we were given
42:32in this country, that there is something
42:34in us above and beyond
42:36uh just our ability to speak and our
42:39cognitive abilities, which at some point
42:41will be surpassed, which is called
42:43human, which is artistic, individual,
42:46particular, special, and more valuable
42:49and needs to be protected.
42:51And that
42:53that setup for lack of a better word is
42:55unique to this country.
42:57And we should make sure we fight to keep
42:59it.
43:00>> Alex Karp, thank you so much.
43:02>> Thank you.

About this transcript

This page contains the full transcript of Alex Karp on Why AI Will Hurt the Middle Class: You Can’t Have Billionaires Becoming 100x Wealthier by The Builders with Walter Isaacson, generated from the public captions YouTube serves with the video. The transcript has 7,363 words across 1,237 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.