Relentless
00:47:58 · Ty Morse, Sam Altman
[00:47:58] S0: When was the last time that you realized that you weren't, like, being ambitious enough? [00:48:04] S1: I mean, I definitely badly undershot on the compute investments. Could you [00:48:08] S0: have known going in with the right mental model? Yes. But I got, like, [00:48:14] S1: psyched out by the financial markets or something. I don't know. That was clearly a mistake. [00:48:19] S0: How do you correct that in the future? [00:48:22] S1: Well, I mean, hopefully, I'll learn from it, and I'll I'll make some new mistakes, but I won't make that one again. [00:48:26] S0: On the compute side, this is kind of, like, the biggest infrastructure project of maybe all time or is about to be. Yep. Are you gonna try and, like, vertically integrate under OpenAI's hood everything from, like, power generation to, like, token generation to, you know, serving some person in ChatGPT or Codex or something like that? Or are you gonna try and have a bunch of partners?
00:48:26 · Ty Morse, Sam Altman
[00:48:26] S0: On the compute side, this is kind of, like, the biggest infrastructure project of maybe all time or is about to be. Yep. Are you gonna try and, like, vertically integrate under OpenAI's hood everything from, like, power generation to, like, token generation to, you know, serving some person in ChatGPT or Codex or something like that? Or are you gonna try and have a bunch of partners? [00:48:45] S1: We're not literally gonna try to do it all inside of OpenAI, obviously, but we will try to do a better job of a well functioning supply chain than we have done so far. Like, you know, we we have brought chip design and model design together, and that's been good. Don't think we need to bring electron production together in the same way because that's more of a commodity. However, I am... When you really think about the whole supply chain that has to come together to produce intelligence, I would say there is relatively too much focus on algorithms that create better algorithms and not enough focus on data centers that can create more data centers, which in a world of robots and a truly automated supply chain, you can totally imagine doing. You can, like, spend a data center's thinking power to drive a fleet of robots to make more copies of the data center, and that's probably a very wonderful thing to do, relative to other uses of back compute if we're right about what that additional compute will eventually unlock.
00:48:45 · Sam Altman, Ty Morse
[00:48:45] S1: We're not literally gonna try to do it all inside of OpenAI, obviously, but we will try to do a better job of a well functioning supply chain than we have done so far. Like, you know, we we have brought chip design and model design together, and that's been good. Don't think we need to bring electron production together in the same way because that's more of a commodity. However, I am... When you really think about the whole supply chain that has to come together to produce intelligence, I would say there is relatively too much focus on algorithms that create better algorithms and not enough focus on data centers that can create more data centers, which in a world of robots and a truly automated supply chain, you can totally imagine doing. You can, like, spend a data center's thinking power to drive a fleet of robots to make more copies of the data center, and that's probably a very wonderful thing to do, relative to other uses of back compute if we're right about what that additional compute will eventually unlock. [00:49:58] S0: How much of your time is spent thinking about how to, like, successfully design that fully robot end to end future where intelligence is just producing
00:49:58 · Ty Morse, Sam Altman
[00:49:58] S0: How much of your time is spent thinking about how to, like, successfully design that fully robot end to end future where intelligence is just producing [00:50:06] S1: thinking about... Almost all goes into, like, execution. Like, the idea... That's a very obvious idea. It's very easy. You can, like, kind of relatively quickly say, here's the pieces that need to come together. But then to, like, get that whole machinery and all these companies working together, that that... You know, none of the glory of the big idea and trying to, like, think the big thoughts, but, like, a lot of grinding. [00:50:26] S0: When you think of, like, execution, what does [00:50:28] S1: that typically mean for you? Not... There's no typical there. Like, the... I mean, to the degree that there is, it's that you just, like, do whatever step is required in the problem to figure out how to make it happen. Mhmm. But, you know, figuring out how to, like, finance new fab build out is very different than figuring out how to get a great chip design team together and then get them to work with the research team and then actually get the supply chain running well. Like, each of those is, a very different approach.
00:50:28 · Sam Altman, Ty Morse
[00:50:28] S1: that typically mean for you? Not... There's no typical there. Like, the... I mean, to the degree that there is, it's that you just, like, do whatever step is required in the problem to figure out how to make it happen. Mhmm. But, you know, figuring out how to, like, finance new fab build out is very different than figuring out how to get a great chip design team together and then get them to work with the research team and then actually get the supply chain running well. Like, each of those is, a very different approach. [00:51:02] S0: One thing that I know that Brian Chesky does is he talks about, like, shamelessly trying to find whoever the expert is at any given thing, and then basically just go ask them whatever question he has if he's try... Trying to come come up to speed on something. What is your process for doing something like that if it's something new that you need to become good at in a very rapid period of time? [00:51:24] S1: Definitely finding the experts and talking to them, reading as much as I can. I don't have a lot of, like, great... I don't have a lot of, like, novel strategies here. I... I've always been really grateful and pleasantly surprised by how much experts are willing to help people if they ask. It's, like, a very nice thing about humanity. [00:51:46] S0: One thing I've heard you say is you should always ask for what you want. Yeah. Because not all the time you get it, but sometimes you do. [00:51:52] S1: Sometimes you do.
00:51:52 · Sam Altman, Ty Morse
[00:51:52] S1: Sometimes you do. [00:51:53] S0: And when you do, amazing things can happen. When was the last time that you asked for something that would seem absolutely insane to someone else and you got it? [00:52:01] S1: Actually, it was this recent time I got on a plane that I still can't talk about, but it did work. [00:52:06] S0: They don't work. Go back one more that you can talk about. [00:52:10] S1: I mean, in some sense, Codex was an example of this. Like, is not the most recent one, but it was one that I think it's instructive. We were way behind Claude Code, and it seemed like a kind of crazy kamikaze mission to try to beat them with a coding app. And, you know, the, like, the consensus is that this kind of thing never works and you just move on to the next one, but, you know, it's like a fool's errand to try to, like, win when someone else already has momentum in a particular product category. But we decided we thought it was really important. We asked a team to do it. They performed a legitimate, like, unbelievable, very rare in the history of business thing. And now it is the product that most of and model that most of the best coders that I know use. And that felt like an impossible thing. But if we hadn't asked the team, like, hey. We have a really important but extremely important mission for you, it just wouldn't have happened. [00:53:08] S0: Why'd you make that decision? Why didn't you just give up?