Ben Horowitz: recommendation

Transcript context

…Four founders starting companies these days. When you look into the future of the AI industry, say in five, 10 years, how do you think things will play out slash where do you think the biggest opportunities remain? Where are you guys looking to invest most? In infrastructure. I think that there is obviously a real estate power cooling play. I think that's a little outside of hardcore technology investing that we do. But there's another layer which is take a given open source model, who can run it the cheapest with the lowest latency? And that's going to be extremely valuable, whoever has that. And Google has been historically very good at that and so forth, and Sam is really trying to build that now with Stargate. And so I think that's going to be a very important layer of value. I think that on the foundation model side, you have to be very selective as an investor. So in order to compete in foundational model world, our basic rule of thumb is you have to be able to, without much product progress, raise at least $ 2 billion because that's basically what it's going to cost you to train something that gets you competitive enough to make money. And there are just very few founders like that. So Ilya is one of those. Mary's one of those, Fei-Fei is one of those, but that's the kind of class of person you need. And there's whatever. There's certainly less than 10 of those in the world. And so that's kind of an important area, but a small area. And then I think the application layer is going to be very, very interesting. And I think that if you look at Sam, he's making most of his money off ChatGPT, almost all his money off ChatGPT now. And Chat GPT, like it or not, it's got a real moat. It's very hard to knock it off. It's perched. Everybody's taking a shot at it, people, great distribution like Google and Elon and Zuckerberg and everybody. And that thing just keeps going like it is. So I think the applications are both more complex and kind of stickier than people thought they were originally. The thing that people got very wrong is this whole thin wrapper around GPT, that's really wrong. In fact, here's how wrong it is. Back in the '80s, that same phrase was used, but it was thin wrapper around an RDBMS- Database.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.