Ex-Palantir Vet: Your AI Problem Is Really a Data Problem

Ben Amor (0:00)

The thing that makes people go, oh, that's cool, was working with the Ferrari Formula One team.

Rob Kelly (0:06)

Can data win a race?

Ben Amor (0:08)

Absolutely.

Rob Kelly (0:25)

Welcome to Media and the Machine. I'm Rob Kelly. This show is about AI and how to profit from it. I believe this is the biggest technology shift of our lifetime. I'm a three-time founder and CEO of content-driven businesses, and each week I talk with the people on the inside of AI and content. The ones building this new world and learning what works firsthand. I'm an optimist about AI, but I'm also a realist, so every show ends with me asking our guests what AI means for our jobs, our families, and our future.

My guest is Ben Amor, co-founder and CTO of DealSage, a startup focused on making enterprise companies AI native. At DealSage, Ben's already used hundreds of billions of tokens across OpenAI, Google, and Anthropic, and he's bringing to his new company lessons from nearly 10 years at Palantir, the AI and data company serving governments and large enterprises that's now worth about $460 billion. Ben worked there as a forward deployed engineer years before that role became commonplace across AI, and he's generous at sharing what he learned there. We get into the ideas that shape Palantir under CEO Alex Karp and Chairman Peter Thiel, including Thiel's famous line that competition is for losers. The six or so books Palantir handed all employees, why they might hold 10 plus meetings before a pilot and walk away if there's no clear business outcome, Ben's work with Ferrari's Formula One team, why Silicon Valley dismissed Palantir's mix of software and consulting for years, and how amazingly Palantir went public with only about 100 clients. We also get into lessons Ben is applying now, where the real AI advantage lives, how companies can accidentally give private data to LLMs and how he's thinking about AI in school as a father of young kids, including whether to ban it. Please enjoy my conversation with Ben Amor.

What's the craziest data problem you worked on at Palantir?

Ben Amor (2:17)

I worked on a lot of different, very disparate types of problems. The thing that makes people go, oh, that's cool, was working with the Ferrari Formula One team. We were basically helping them take all of the sensor data that is streaming off the car as it goes around the track during the Grand Prix or during the prep, and in real time, analyze it to try and understand what was going on with the car throughout the race.

Rob Kelly (2:44)

I'm just curious, how many sensors would be on a Ferrari F1?

Ben Amor (2:48)

That's a good question, probably somewhere between like three and four hundred.

The most important one is the downforce.

So downforce is like how much force literally downwards they are managing to create, because the more downforce you have, the more friction you have, and so the faster you can take the corners without skidding out, and so kind of like in some ways the whole goal is to generate as much downforce as possible. The questions they wanted to be able to ask were things like, how does the average downforce that we are seeing on this particular corner of the track compare to the downforce we were getting when we ran this in the qualifying races, or when we ran this last year? And then tire wear is one that's very hard to measure. That's not a sensor that you can just look at. You kind of have to make estimates. Can you fit a better model to estimate how a tire is going to degrade across a race at a particular race course with a particular tire type? Because you want to be able to bring the car in at just the right time. You don't want to go too late, and then you start not being able to drive as fast, but you don't want to go too early, and then you lose momentum in the race, or the next set of tires is going to wear out too quickly.

Rob Kelly (4:02)

Who's seeing the data and when and where? Are they seeing it right there in the race?

Ben Amor (4:08)

Yeah, literally all of the engineers are looking at the data streaming in, and they're collecting the data like, how does this particular sensor on this lap compare to all of that historical data that we have?

Rob Kelly (4:20)

Where are those engineers during the race? Where do they sit?

Ben Amor (4:23)

Some of them are at the racetrack, and they're actually sat in the team area, but then the others are back in HQ, so in Maranello, which is where they're based.