Alex Karp on Palantir, AI Dominance and Domain Skills

TL;DR

Palantir gives a battlefield commander a single pane of glass showing all assets, including data shared by allied foreign countries, plus where planes should fly, where troops should go, and which missiles to pair with which targets. Karp argues the primary risk of AI is social instability, not surveillance, and that workers with specific domain or vocational expertise will out-earn generalists.

Transcript

I know you're used to CEOs who don't answer your question. I'm going to directly answer your question. I get yelled at all the time for Ukraine, ICE, Israel, and generally telling the truth, which I would call okay. The chance of world survival goes up as America becomes stronger, more dominant. Alex Karp, thanks for having us into the Palunteer la... Read More

Key Insights

• Palantir's battlefield product gives a commander access to all of their assets for the first time, including assets of foreign countries that share a subset of their data, presented through what Karp calls a single pane of glass.
• The system shows where planes should fly, where troops should be, and where they should not go, running off data sets that were not available to the commander the previous year, and it can predict where assets should be placed.
• Missile-to-target matching is a capability Karp says the product will soon handle, deciding whether expensive missiles go on expensive targets, with cheaper effectors assigned to less important targets.
• Supply chain orchestration is what Karp calls one of the most overlooked parts of the product: what is being built, how, where, by which companies, and what those companies' costs are, which those companies currently find very hard to determine.
• Almost all of Palantir's clients started off as Palantir skeptics, according to Karp, and the business got off the ground through war fighters who told generals they did not care what stars they had because they intended to come home alive.
• The primary evidence for a surveillance state in the West is companies knowing every action a person takes at all times, not government watching consumers, since people carry electronic devices that monitor everything mostly so advertisers can sell them things.
• Pattern of life surveillance is described as a very important way of finding terrorists and criminals because everyone has a pattern, though Karp says such decisions require very precise tools and that Palantir monetizes the difficulty of those decisions.
• Large language models commodify general knowledge, so people with classically high IQ and generalized but non-specific knowledge are in trouble, while those with deep vocational or domain training will be paid downstream of the value they create.

Questions & Answers

Q: What does Palantir show a commander on the battlefield?

According to Alex Karp, a commander gets access to all of their assets for the first time, and that includes assets of foreign countries they work with that are sharing a subset of the data. Presented as a single pane of glass, it shows, for a complex operation, where the plane should fly, where the troops should be, where the muscle should go, and where they should not go. It draws on data sets that were not available the previous year and can predict where assets should be placed.

Q: How does Palantir help decide which weapons to use on which targets?

Karp says the product can and will soon be able to pick up what kind of missiles go on what kind of targets, for example whether really expensive missiles should be put on really expensive targets. The practical effect for a commander is being ahead of the adversary and able to predict their actions, putting the best effectors on the most important targets and cheaper ones on less important targets. He adds that the most important thing this does, honestly, is prevent conflict.

Q: Why does Alex Karp say the supply chain feature matters?

Karp calls supply chain orchestration one of the most important things that is overlooked about the product. It covers what you are building, how you are building it, where it is being built, which companies are doing it, and understanding the cost for those companies, which he notes is currently something that is very, very hard for those companies themselves to work out. He frames this alongside what he calls the sexy violent part of war as an underrated capability.

Q: What is Alex Karp's response to people who are suspicious of Palantir?

He pushes back on the premise and asks which people are suspicious. He says the people he cares about are those on the front line, in factories, driving trucks, and the war fighters who used the products. He recounts that the business got off the ground when war fighters went to generals and said they did not care what stars the generals had because they were coming home alive. Almost all of Palantir's clients started off as Palantir skeptics, and he invites people to spend 10 minutes looking at the product and decide.

Q: Is Alex Karp worried about a surveillance state?

He redirects the question, saying the primary worry should be whether adversaries are so scared of the US that they do not fight it, which he calls the single best way to stop wars. On surveillance itself, he argues the primary evidence for a surveillance state in the West is not government watching consumers but companies knowing every single action you take at all times, via the electronic devices people carry, largely so they can sell you products like cornflakes.

Q: What is pattern of life surveillance according to this interview?

Karp describes pattern of life surveillance as watching a person's patterns of behavior, and calls it a very important way of finding terrorists and criminals in general, because everyone has a pattern. When asked whether that means it is worth it, he says he is not saying that here. His position is that these things require very, very precise tools, and he acknowledges that Palantir is monetizing the fact that these decisions are difficult, hard, and important.

Q: What does Alex Karp say is the biggest risk from AI?

He names social instability as the primary risk of AI, rather than surveillance. His reasoning is about wealth distribution: if Palantir grew 10x, the question is how much real wages grow for the person on the factory floor, and if the answer is roughly 5%, about level with inflation, those workers cannot be blamed for being upset. He predicts this produces populist movements that make no sense, such as government running grocery stores, leading to dysfunction and then violence.

Q: Which careers does Alex Karp say will do well in the AI economy?

He argues that the commodity nature of large language models commodifies general knowledge, so someone with classically high IQ and generalized but non-specific knowledge is in trouble. People with deep specific or vocational training will earn much more, because you get paid downstream of the value you create. His examples include bending the metal of a ship's hull, writing a targeting script in Palantir, diagnosing a complex device as a high school graduate, and pouring cement precisely enough to build a Taiwan-style factory in America. An electrician, he says, will be better off than a Yale graduate without domain expertise.

Summary & Key Takeaways

• Karp describes Palantir as growing the GDP of the US and as the part of the AI economy where things are useful on the battlefield or commercially, and as the only AI company with what he calls a real spread function to working-class Americans. He opens by saying he directly answers questions, unlike CEOs the interviewer is used to.

• On the battlefield product, a commander sees all assets on one pane of glass, including allied data, with guidance on flight paths, troop positions, and no-go areas. Karp says it will orchestrate the supply chain and match effectors to targets by value. Beyond the violent part of war, he argues the most important thing it does is prevent conflict.

• Karp reframes the surveillance debate: the real question is whether adversaries are too scared to fight the US, and the West's actual surveillance comes from companies tracking every action to sell products. He prefers the word adversaries to enemies, and says Palantir does not sell to China, Iran, or Russia.

• The dominance argument: either America or China becomes the dominant AI player, and the rules differ depending on who wins. Karp says Europe has given up on technology, and that the Arab and non-Arab Middle East are closest in AI adoption, leaving America as the leader of its coalition by far.

• On AI risk, Karp names social instability as the primary danger. If Palantir grew 10x and factory-floor real wages grew only around 5%, roughly at inflation, he says workers could not be blamed for being upset, and predicts populist movements, dysfunction, and then violence.

• His career advice: a few schools such as Yale and Stanford are worth attending, otherwise go to the cheapest school. Specific skills like bending metal on a ship hull, writing targeting scripts, fixing complex devices, or pouring cement precisely enough to build a Taiwan-grade factory in America will pay. An electrician will be better off than a Yale grad without domain expertise.