Sam Altman on Sora, Energy, and Building an AI Empire

Sam Altman 0:00 Sort of thought we had like stumbled on this one giant secret that we had these scaling laws for language models, and that felt like such an incredible triumph. I was like, "We're probably never gonna get that lucky again." And deep learning has been this miracle that keeps on giving, and we have kept finding breakthrough after breakthrough. A-again, when we got the, the reasoning model breakthrough, like I, I also thought that was like we're never gonna get another one like that. And it just seems so improbable that this one technology works so well. But maybe this is always what it feels like when you discover like one of the big, you know, scientific breakthroughs. Is it-- if it, if it's like really big, it's pretty fundamental, and it just, it keeps working.

Erik Torenberg 0:41 Sam, welcome to a16z Podcast.

Sam Altman 0:42 Thanks for having me. All right.

Erik Torenberg 0:45 You've, uh, described in another interview, described OpenAI as a combination of four companies: a consumer technology business, a mega-scale infrastructure operation, a research lab, and all the new stuff, including planned hardware devices. From hardware to app integrations, job marketplace to commerce, what do all these bets add up to? What's OpenAI's vision?

Sam Altman 1:03 Yeah, I mean, maybe you should count it as three, maybe as, as four for kind of our own version of the, what traditionally would've been the research lab at this scale. But three core ones. Uh, w-we want to be people's personal AI subscription. I think most people will have one. Some people will have several. And you'll use it in some first-party consumer stuff with us, but you'll also log into a bunch of other services, and you'll just-- you'll use it from dedicated devices at some point. You'll have this AI that gets to know you and be really useful to you, and you'll-- that's what we wanna do. Um, it turns out that to support that, we also have to build out this massive amount of infrastructure. But the goal there, the, the, the mission is really like build this AGI and make it very useful to people.

Ben Horowitz 1:44 And, and does the infrastructure, uh, do you think it will end up... You know, it's necessary for the main goal. Will it also separately end up being a-another business, or is it just really gonna be in service to the personal AI or unknown?

Sam Altman 1:59 You mean like would we sell it to other companies, this raw infrastructure?

Ben Horowitz 2:01 Yeah. W-would you sell it to other companies? Um, yeah. Or, or, or, or, you know, it's such a massive thing. Would it, would it do something else? [chuckles]

Sam Altman 2:10 It feels to me like there will emerge some other thing to do like that. But I don't know. We don't have a current plan.

Ben Horowitz 2:16 Yeah, no, it is. Yeah.

Sam Altman 2:17 It's currently just meant to like support the service we wanna deliver and the research.

Ben Horowitz 2:21 Yeah. No, that makes sense.

Erik Torenberg 2:22 Yeah. The, um-

Ben Horowitz 2:24 And that, that, that's a good focus, too

Sam Altman 2:25 ... but the scale is sort of like-

Ben Horowitz 2:26 Ridiculous

Sam Altman 2:27 ... terrifying enough that you've gotta be open to doing something else with it.

Ben Horowitz 2:31 [laughs] Yeah, if you're building the biggest data center, uh, in the history of humankind.

Sam Altman 2:34 The biggest infrastructure project in the history.

Ben Horowitz 2:36 Yeah.

Sam Altman 2:36 Yeah.

Ben Horowitz 2:36 Yeah.

Erik Torenberg 2:37 The, um... There was a great interview you did many years ago in Strictly VC, um, and sort of early OpenAI before-- well before ChatGPT and, and they're saying, "Hey, what's..." They're asking, "What's the business model?" And you said, "Oh, well, we'll ask AI. It'll, it'll figure it out for us."

Sam Altman 2:48 [laughs]

Erik Torenberg 2:49 Everybody laughs, but-

Sam Altman 2:50 There have been multiple times, and there was just another one recently, where we have asked a then-current model for, you know, what should we do, and it has had a insightful answer we missed. So I, I think when we say stuff like that, people don't take us seriously or literally.

Erik Torenberg 3:03 Yeah.

Sam Altman 3:03 But maybe the answer is you should take, take us both.

Erik Torenberg 3:07 Yeah.

Ben Horowitz 3:07 Yeah. Yeah, well, no, as, as somebody who runs an organization, [laughs] I ask, uh, the AI a lot of questions about what I should do.

Sam Altman 3:13 Yeah.

Ben Horowitz 3:14 It comes up with some pretty interesting answers.

Sam Altman 3:16 Sometimes.

Ben Horowitz 3:16 Yeah. Yeah, yeah.

Sam Altman 3:17 Sometimes not.

Ben Horowitz 3:17 It does. You know, you have to g- you have to give it enough context, but...

Erik Torenberg 3:20 What is, what is the thesis that, that connects these bets beyond more distribution, m-more compute? How do, how do we think about that?

Sam Altman 3:26 I mean, the research enables us to make the great products, and the infrastructure enables us to do the research, so it is kind of like a vertical stack of things. Like, you can use ChatGPT or some other service to get advice about what you should do running an organization, but for that to work, it requires great research and requires a lot of infrastructure. So it is kind of just this one, this-

Erik Torenberg 3:47 Yeah

Sam Altman 3:47 ... one thing. It's-

Ben Horowitz 3:49 A-and do you think that there will be a point where that becomes completely horizontal, or will it stay vertically integrated for the foreseeable future?

Sam Altman 4:01 I was always against vertical integration, and I now think I was just wrong about that.

Ben Horowitz 4:07 Yeah. Interesting.

Sam Altman 4:08 And there's kind of-- 'cause y-you'd like, you'd like to think that the economy is efficient, and the theory that companies can do one thing, and then that's supposed to work.

Ben Horowitz 4:17 Like to think that, yeah.

Sam Altman 4:18 And in our case at least, it hasn't really.

Ben Horowitz 4:21 Yeah.

Sam Altman 4:21 I mean, it has in some ways, for sure.

Ben Horowitz 4:23 Right. Right.

Sam Altman 4:23 Like, there's people that make a m- like, you know, NVIDIA makes an amazing chip or whatever that a lot of people can use. But the, the story of OpenAI has certainly been towards, we have to do more things than we thought to be able to deliver on the mission.

Ben Horowitz 4:36 Right. You know, although the, you know, the, the history of the computing industry has kind of been a story of kind of a back and forth in that, you know, there was the Wang Wa- Word Processor and then the personal computer and the, the BlackBerry before the smartphone. Um, so there, you know, there has been this kind of vertical integration and then not, but then the iPhone is also vertically integrated. [chuckles]

Sam Altman 4:58 The i- the iPhone, I think, is the most incredible product the tech industry has ever produced, and it is extraordinarily vertically integrated.

Ben Horowitz 5:05 Yeah. Amazingly so, yeah. Interesting.

Erik Torenberg 5:08 W-which bets would you say are enablers of AGI versus which are sort of hedges against uncertainty?

Sam Altman 5:13 I think you could say that on the surface, Sora, for example, does not look like it's AGI relevant. But I would bet that if we can build really great world models, that'll be much more important to AGI than people think. There were a lot of people who thought ChatGPT was not a very AGI-relevant thing, and it's been very helpful to us, not only in building better models and understanding how society wants to use this, but also in, like, bringing society along to actually figure out, man, we gotta contend with this thing now. We-- For a long time before ChatGPT, we would talk about AGI, and people were like, "This is not happening," or, "We don't care."

Erik Torenberg 5:45 Yeah.

Sam Altman 5:45 And then all of a sudden, they really cared. And, and I, I think that... So research benefits aside, I'm a big believer that society and technology have to co-evolve. It's-

Erik Torenberg 5:58 Yeah

Sam Altman 5:58 ... you can't just drop the thing at the end. It doesn't work that way. It is, it is a sort of ongoing back and forth.

Erik Torenberg 6:03 Yeah.

Sam Altman 6:05 Say more about how Sora fits into your strategy because there was some hullabaloo on, on X around, hey, um, you know, why devote precious GPUs to, to Sora? But i- is it a short-term, long-term trade-off or are we so agent- Well, and then the new one had like very interesting twists with the, uh, social networking. [laughs] Uh, be very interested in kinda how you're thinking about that and like, um, did, uh, Meta call you up and get mad? Or like, hey, what, what do you expect the reaction to be? [laughs] Um, I think if one company of the two of us has- [laughs] ... feels like more like the other one has gone after them, it wouldn't-- They, they shouldn't be calling us. [laughs] Well, I, I do know the history too, though. But, but, uh, look, we're not gonna thr- like... First of all, I think it's cool to make great products, and people love the new Sora. And I also think it is important to give society a taste of what's coming on this co-evolution point. So like very soon, the world is gonna have to contend with incredible video models that can deep fake anyone or kind of show anything you want, and that will mostly be great. There will be some adjustment that society has to go through. And just like with ChatGPT, we were like, the world kinda needs to understand where this is. I think it's very important the world understands where video is going very quickly, 'cause that's gonna be-- Video has much more, like, emotional resonance than text, and very soon we're gonna be in a world where, like, this is gonna be everywhere. So I think there's something there. Uh, as I mentioned, I think this will help our research program and is on the AGI path. But yeah, something like, you know, it can't all be about just making people, like, ruthlessly efficient and the AI, like, solving all our problems. There's gotta be, like, some fun and joy and delight along the way. But we won't throw, like, tons of compute at it, or not by a fraction of our compute. [laughs] Yeah. It, it, it, it's tons in the absolute sense- Yeah ... but not in the relative sense. [laughs]

Sam Altman 8:01 Yeah. Yeah. I wanna talk about the future of AI human interfaces, 'cause back in August you said the models have already saturated the chat use case. W- so what do future AI human interfaces look like, both in terms of hardware and software? Is the vision for kind of a, a WeChat-like super app? So I- I'm solving the chat thing in, like, a very narrow sense, which is if you're trying to, like, you know, have the m- most basic kind of chat-style conversation, it's very good. But what a chat interface can do for you, it's, like, nowhere near saturated. 'Cause you could ask a chat interface, like, "Please cure cancer." A model certainly can't do that yet. So I think the text interface style can go very far, even if for the chitchat use case, the models are already very good. Um, but, but of course there's better interfaces to have. Uh, actually, that's another thing that I think is cool about Sora. Like, you can imagine a world where the interface is just constantly real-time rendered video- Yeah ... and what that would enable, and that's pretty cool. You can imagine new kinds of hardware devices that are sort of always ambiently aware of what's going on. And rather than your phone, like, blast you with text message notifications at whenever it wants, like it really understands your context and when to show you what, and there's a long way to go on all that stuff. Yeah.

Sam Altman 9:12 W- within the next couple years, what will models be able to do that they're not able to do today? Will it be sort of white collar, um, you know, re-replacement at a much deeper level, AI scientist, uh, human, humanoids, um, persons- I mean, a l- a lot of things, but you touched on the one that I am most excited about, which is the s- the AI scientist. Yeah. This is crazy that we're sitting here seriously talking about this. The-- I know there's like a quibble on what the Turing test literally is, but, but the popular conception of the Turing test sort of went whooshing by. Yeah. [laughs] That was fast, yeah. You know, it was just like we talked about it- [laughs] ... as this most important test of AI for a long time. It seemed impossibly far away. Then all of a sudden it was passed. The world freaked out for like a week, two weeks, and then it's like, "All right, I guess computers, like, can do that now." Yeah. And everything just went on. And I think that's happening again with science. Uh, my own personal, like, equivalent of the Turing test has always been when AI can do science. Like, that is always, like that is a real change to the world. And for the first time with GPT-5, we are seeing these little, little examples where it's happening. You see these things on Twitter. It did this-- It made this novel math discovery and did this small thing in my, you know, my physics research and my biology research. And everything we see is that that's gonna go much further. So in two years, I think the models will be doing bigger chunks of science and making important discoveries, and that is a crazy thing. Like, that will have a significant impact on the world. I am, I am a believer that to a first order, scientific progress is what makes the world better over time. And if we're about to have a lot more of that, that's a big change. It's interesting 'cause that's a positive change that people don't talk about. Y- y- th- it, it's gotten so, um, much into the realm of the negative changes if AI gets extremely smart. But, uh- But curing every disease is like- We, we could use a lot more science. Yeah. Yeah. Yeah. That, that, that, that, that's a really good point. I think Alan Turing said this. Somebody asked him, they said, "Well, you really think the, uh, computer's gonna be, you know, smarter than the brilliant minds?" He said, "It doesn't have to be smarter than a brilliant mind, just smarter than a mediocre mind like the president of AT&T." [laughs] And, uh, [laughs] we could use more of that too, probably. We, um, we just saw Periodic launch last week. Yeah. You know, OpenAI alums. And, uh, yeah, to, to, to that point, it- it's amazing to see both the innovation that you guys are doing, but also the, the teams that, you know, come out of OpenAI just feels like are, are, you know, creating tremendous capable things. We certainly hope so. Yeah. The,

Sam Altman 11:44 um-- I wanna ask you about just broader reflections in terms of what sort of about diffusion or, uh, development in 2025 has surprised you, or what has sort of updated your worldview since ChatGPT came out? A lot of things again, but maybe the m- most interesting one is how much new stuff we found. Sort of thought we had, like, stumbled on this one giant secret that we had these scaling laws for language models, and that felt like such an incredibleTriumph that I was like, "We're probably never gonna get that lucky again." And deep learning has been this miracle that keeps on giving, and we have kept finding, like, breakthrough after breakthrough. A-a-again, when we got the, the reasoning model breakthrough, like I, I also thought that was like, we're never gonna get another one like that. Uh, and it just seems so improbable that this one technology works so well. But maybe this is always what it feels like when you discover like one of the big, you know, scientific breakthroughs. Is if, if it's like really big, it's pretty fundamental, and it just- it keeps working. But the amount of progress-- Like if you went back and used GPT-3.5 from ChatGPT launch, you'd be like, "I cannot believe anyone used this thing."

Erik Torenberg 13:04 Yeah. [chuckles]

Sam Altman 13:05 And, and now we're in this world where the capability overhang is so immense. Like most of the world still just thinks about what ChatGPT can do, and then you have like some nerds in Silicon Valley that are using Codex, and they're like, "Wow, those people have no idea what's going on." And then you have like a few scientists who say, "Those people using Codex have no idea what's going on." But the, the overhang of capability has come-- is, is, is so big now, and we've just come so far on the-- what the models can do.

Erik Torenberg 13:29 And in terms of further development, how far can we get with, with LLMs? At, at what point do we need either new architecture? H-how do you think about what breakthroughs are needed?

Sam Altman 13:37 I think far enough that we can make something that will figure out the next breakthrough with the current technology. Like, I-I-- It's a very self-referential answer, but if, if LLMs can get-- If LLM-based stuff can get far enough that it can do, like, better research than all of OpenAI put together, maybe that's, like, good enough.

Ben Horowitz 13:58 Yeah, that would be a big breakthrough. [chuckles] A very big breakthrough. So o-on, um, on the more mundane, you know, one of the things that, uh, people have kind of started to complain about, I think South Park did a whole episode on it, is kind of the obsequiousness of, uh, of kind of AI and ChatGPT in particular. And how hard a problem is that to deal with? Is it not that hard, or is it like kind of a fundamentally hard problem?

Sam Altman 14:25 Oh, it's not at all hard to-

Ben Horowitz 14:26 Yeah

Sam Altman 14:26 ...deal with. A lot of users really want it.

Ben Horowitz 14:27 Yeah. [chuckles] Okay.

Sam Altman 14:28 Like, if you go look at what people-

Ben Horowitz 14:29 Yeah

Sam Altman 14:30 ...say about ChatGPT online-

Ben Horowitz 14:31 Yeah

Sam Altman 14:32 ...there's a lot of people who, like, really want that back.

Ben Horowitz 14:34 Yeah.

Sam Altman 14:34 Um, and it is, you know.

Ben Horowitz 14:36 Yeah.

Sam Altman 14:37 So it's not-- technically, it's not hard to deal with at all. Um, one thing, and this is not surprising in any way, but the, the incredibly wide distribution of what users want-

Ben Horowitz 14:52 Yeah

Sam Altman 14:52 ...out of how-- of, like, how they'd like a chatbot to behave in big and small ways.

Ben Horowitz 14:57 Does that-- Do you end up having to configure the personality then you think? Is that gonna be the answer?

Sam Altman 15:03 I think so. Uh, I mean, ideally, like you just talk to ChatGPT for a little while, and it kinda interviews you and also sort of sees what you like-

Ben Horowitz 15:11 Yeah

Sam Altman 15:11 ...and don't like, and-

Ben Horowitz 15:12 A-and ChatGPT just figures it out-

Sam Altman 15:14 And just figures it out

Ben Horowitz 15:14 ...figures itself out.

Sam Altman 15:14 But in the short term, you'll probably just pick one.

Ben Horowitz 15:17 Got it. Yeah, no, that makes sense. Very interesting. And, um, actually, so, so one thing I wanted to ask about is, uh-

Sam Altman 15:26 Uh-

Ben Horowitz 15:26 Yeah.

Sam Altman 15:26 Bef-- Like, I think we just had a, a really naive thing, which, you, you know, like it would sort of be unusual to think you can make something that would talk to billions of people, and everybody wants to talk to the same-

Ben Horowitz 15:39 Yeah

Sam Altman 15:39 ...person.

Ben Horowitz 15:39 Yeah.

Sam Altman 15:40 And, and yet that was sort of our implicit assumption for a long time.

Ben Horowitz 15:44 Right. 'Cause people have very different friends.

Sam Altman 15:45 People have very different friends.

Ben Horowitz 15:46 Yeah.

Sam Altman 15:46 So now we're trying to fix that.

Ben Horowitz 15:48 Yeah. And also kind of different friends, different interests, different, uh, levels of intellectual capability.

Sam Altman 15:54 Yep.

Ben Horowitz 15:54 So you, you don't really wanna be talking to the same thing all the time. And one of the great things about it is you can say, "Well, explain it to me like I'm five," but m-maybe I don't even wanna have to do that prompt. [chuckles] Maybe I always want you to talk to me-

Sam Altman 16:05 It should just learn that. Yeah

Ben Horowitz 16:06 ...like I'm five. Yeah. Particularly if you're teaching me stuff. Interesting. Um, I, I want to ask you a kind of like a, a CEO question, which has been interesting for me to observe you, is you just

Ben Horowitz 16:17 did this deal with AMD. Um, and you know, of course, the company's in a different position, and you have more leverage and these kinds of things, but, like, how has your kind of thinking changed over the years since you did that, that initial deal, if at all?

Sam Altman 16:31 I, I had very little operating experience then.

Ben Horowitz 16:34 Yeah.

Sam Altman 16:34 I had very little experience-

Ben Horowitz 16:35 Right

Sam Altman 16:35 ...running a-- Like I, I am, I am not naturally someone to run a com-- I, I'm a great fit to be an investor.

Ben Horowitz 16:39 [chuckles] Yeah.

Sam Altman 16:40 And I kinda felt that was gonna be-- That was what I did before this, and I thought that was gonna be my career.

Ben Horowitz 16:43 Yeah, yeah.

Sam Altman 16:44 And-

Ben Horowitz 16:44 Although you were a CEO before that.

Sam Altman 16:46 I-- Not a good one. Um-

Ben Horowitz 16:47 [chuckles]

Sam Altman 16:48 And, and so I think I had the mindset of, like, an investor advising a company-

Ben Horowitz 16:58 Oh, interesting. Right

Sam Altman 16:58 ...when we did the thing. And now I understand what it's like to actually have to run a company.

Ben Horowitz 17:01 [laughs] Yeah. Right. Right, right. So-

Sam Altman 17:02 So they're, they're different

Ben Horowitz 17:03 ...there, there's more than just the numbers. Yeah.

Sam Altman 17:04 I, I've learned a lot about-

Ben Horowitz 17:05 Yeah

Sam Altman 17:06 ...how to, you know-

Ben Horowitz 17:08 Yeah

Sam Altman 17:08 ...like, how you have to like-- what, what operational-- how you-- like what it takes to operationalize deals over time and-

Ben Horowitz 17:14 Right. All, all, all the implications of the agreement-

Sam Altman 17:17 Yeah

Ben Horowitz 17:17 ...as opposed to just, "Oh, we're gonna get distribution and money."

Sam Altman 17:20 Yeah.

Ben Horowitz 17:21 Yeah. That makes sense. Yeah, no, 'cause it, it's really... I, I, I just say I was very impressed at the deal structure improvement.

Sam Altman 17:31 Yeah. Right.

Erik Torenberg 17:33 M-m-more broadly, you've,

Erik Torenberg 17:34 you know, in the last few weeks alone, you mentioned AMD, but also Oracle, NVIDIA. You've chosen to, you know, strike these deals and partnerships with, with companies that you collaborate with but could also potentially compete with in, in, in certain areas. H-how do you decide, you know, when to collaborate versus when, when not to, or h-how do you just think about?

Sam Altman 17:51 Um, we have decided that it is time to go make a very aggressive infrastructure bet, and we're like-- I've never been more confident in the research roadmap in front of us and also the economic value that will come from using those models. But to make the bet at this scale, we kinda need the whole industry to-

Erik Torenberg 18:10 Yeah

Sam Altman 18:10 ...or a big chunk of the industry, to support it. And this is like, you know, from the level of like electrons to model distribution and all the stuff in between, which is a lot.And so we're gonna partner with a, a lot, a lot of people. Uh, you should expect, like, much more from us in the coming months.

Ben Horowitz 18:27 Actually, expand on that 'cause y-y... when you talk about the scale, it does feel like in your mind the, the limit on it is unlimited. Like, you would scale it as, as, you know, as big as you possibly could.

Sam Altman 18:42 I mean, there's, like, some-- There's totally a limit. Like, there's some amount of global GDP, uh-

Ben Horowitz 18:48 Yeah. [laughs] Well, yes, yes

Sam Altman 18:49 ... and, you know, there's some fraction of it that is knowledge work, and we don't do robots yet.

Ben Horowitz 18:53 Yes.

Sam Altman 18:53 But-

Ben Horowitz 18:54 But, but the limits are out there

Sam Altman 18:56 ... it, it feels like the limits are very far from where we are today.

Ben Horowitz 18:59 Yeah.

Sam Altman 19:00 If we are right about-

Ben Horowitz 19:01 Mm-hmm

Sam Altman 19:01 ... So, so, I shouldn't say from where we are. Uh, like, if we are right that the model capability is gonna go where we think it's gonna go, then the economic value that sits there can, can go very, very far.

Ben Horowitz 19:14 Right. So you wouldn't do it, like, if all you ever had-

Sam Altman 19:17 I wouldn't have-

Ben Horowitz 19:17 ... was today's model, you wouldn't go there. But-

Sam Altman 19:19 No, definitely not

Ben Horowitz 19:20 ... so it's a combination.

Sam Altman 19:21 I mean, uh, we would still expand because-

Ben Horowitz 19:24 Mm-hmm

Sam Altman 19:24 ... we can see how much, uh, demand there is we can't serve with today's model, but we would not be going this aggressive if all we had was today's model.

Ben Horowitz 19:32 Right.

Sam Altman 19:33 Yeah.

Ben Horowitz 19:33 Right.

Sam Altman 19:33 We get to see a year or two in advance, though, so-

Ben Horowitz 19:35 Yeah

Sam Altman 19:35 ... like...

Ben Horowitz 19:36 Yeah. Interesting.

Erik Torenberg 19:39 ChatGB is eight hundred million weekly active users, about ten percent of the world, world's population, fastest-growing consumer product, you know, ever, it seems. Um, how do-

Ben Horowitz 19:50 [laughs] Faster than anyone I ever saw.

Erik Torenberg 19:52 Yeah. H-how do you balance, you know, uh, optimizing for active users at, well, at the same time being a res-- you know, being a product company and, and a research company? How do you feel the new-

Sam Altman 20:03 When, when there's a constraint, we almost, like, which happens all the time, uh, we almost always prioritize giving the GPUs to research over supporting the product. Um, part of the reason we run and build this capacity is so we don't have to make such painful decisions. There are weird times, you know, like a new feature launches and it's going really viral or whatever, where research will temporarily sacrifice some GPUs. But, but on the whole, like, we're here to build AGI.

Erik Torenberg 20:27 Yeah.

Sam Altman 20:27 And research gets the priority.

Erik Torenberg 20:29 Yeah. The-- You said in your, your interview with, with your brother Jack around how, you know, other companies can try to imitate the, the products or, or buy your [laughs] you know, or, or hire your, your, your talent.

Ben Horowitz 20:41 Buy our IP maybe. I don't know.

Erik Torenberg 20:42 Yeah. [laughs] All sorts of things. But they, they can't buy the culture or, or they can't i-I mean, the, the sort of repeatable sort of, you know, mach-machine, if you will, that, that is, you know, constantly, uh, the culture of innovation. How have you done that or what are you doing? What, uh, talk about this, this culture of, of, of innovation.

Sam Altman 21:02 This was one thing that I think was very useful about coming from an investor background. A, a really good research culture looks much more like running a really good seed stage investing firm and betting on founders and-

Erik Torenberg 21:13 Yeah

Sam Altman 21:13 ... sort of that kind of, than it does like running a product company. So I think having that experience was really helpful to the culture we built.

Erik Torenberg 21:22 Yeah. Yeah. That's sort of how I see, you know, Ben at a16z in some ways, which we, you know, you're a CEO, but you also have, you know, have his portfolio and, you know, have an investor mindset.

Ben Horowitz 21:31 Right, like I'm the opposite.

Erik Torenberg 21:32 Yeah. Yeah.

Ben Horowitz 21:33 CEO going to investor. He's investor going to CEO.

Erik Torenberg 21:36 Yeah. It is unusual in this direction.

Ben Horowitz 21:38 Yeah, yeah.

Erik Torenberg 21:38 But-

Ben Horowitz 21:39 Yeah, well, it never works

Erik Torenberg 21:40 ... it ha-

Ben Horowitz 21:40 You're the only one who I think I've seen go that way and have it work.

Sam Altman 21:46 Uh, Workday was like that, right?

Ben Horowitz 21:51 Uh, but Aneel was, he, he was a operator before he was an investor, and, uh, I mean, he was really an opera-- I mean, PeopleSoft was a pretty big company.

Erik Torenberg 21:59 And, and why is that? Because once people are investors, they don't wanna operate anymore? [laughs]

Ben Horowitz 22:02 Um, no. I think that investors, generally, if you're good at investing, you're not necessarily good at, like, organizational dynamics-

Erik Torenberg 22:16 Yeah

Ben Horowitz 22:16 ... conflict resolution, um, or, you know, like, just, like, the deep psychology of, like, all the weird shit and then, you know, how politics gets created. There's just, like, all this... There, there's the detailed work in being an operator or being a CEO is so vast, and it's not as intellectually stimulating. It's not something you can ever go talk to somebody at a cocktail party about. And so, like, you're an investor, you get, like, oh, everybody thinks I'm so smart and, you know, 'cause you know everything. You see all the companies and so forth, and that's a good feeling. And then being CEO is often a bad feeling.

Erik Torenberg 22:54 Yeah.

Ben Horowitz 22:55 And so it's really hard to go to a, a good feeling to a bad feeling, I would just say.

Sam Altman 22:59 I'm shocked by how different they are, and I'm shocked by how much the difference between a good job and a bad job they are.

Ben Horowitz 23:04 Yeah. [laughs]

Sam Altman 23:04 Like-

Ben Horowitz 23:04 Yes.

Sam Altman 23:05 [laughs]

Ben Horowitz 23:05 Yeah. You know, it's tough. It's, it's rough. I mean, I can't even believe I'm running the firm. Like-

Sam Altman 23:09 Yeah

Ben Horowitz 23:09 ... I know better.

Sam Altman 23:10 Yeah. [laughs]

Ben Horowitz 23:11 Yeah. And he can't believe he's running OpenAI. He knows better.

Erik Torenberg 23:14 Go-going back to progress today, are, are evals still useful in a world in which they're getting saturated, gamed? Are they still the... What, what is the best way to gauge model capability now?

Sam Altman 23:22 Um, well, I, we're talking about scientific discovery. I think that'll be an eval that can go for a long time.

Erik Torenberg 23:27 Yeah.

Sam Altman 23:29 Revenue is kind of an interesting one.

Erik Torenberg 23:30 Yeah.

Sam Altman 23:30 Uh, but I think the, like, static evals of-

Erik Torenberg 23:33 Mm-hmm

Sam Altman 23:33 ... benchmark scores are less interesting.

Erik Torenberg 23:35 Yeah.

Sam Altman 23:37 And, and also those are crazily gamed.

Erik Torenberg 23:38 Yeah, yeah. [laughs] More broadly, it seems like-

Ben Horowitz 23:41 [laughs] That's all there is, games, as far as I can tell you.

Erik Torenberg 23:44 More broadly, it seems that the culture, the, the culture, X-- uh, Twitter, X is less AGI-pilled than it was a year or so ago when the AI 2027 thing came out. Some people point to, you know, GPT-5, them, them not seeing sort of the obvious, um, o-obviously, there are a lot of progress that in some ways un-under the, the surface are not, not as obvious to what people are expecting. But should people be less AGI-pilled, or is this just Twitter vibes and-

Sam Altman 24:10 Mm, well, a little bit of both.

Erik Torenberg 24:14 In terms of the timeline.

Sam Altman 24:14 I mean, I, I, I think, like, like we talked about the Turing test, AGI will come. It will go whooshing by.

Erik Torenberg 24:20 Yeah. [laughs]

Sam Altman 24:21 The world will not change as much as the impossible amount that you would think it should.

Erik Torenberg 24:24 AGI might just be-

Ben Horowitz 24:25 It won't, it won't actually be the singularity.

Sam Altman 24:27 It will not.

Ben Horowitz 24:28 Yeah.

Sam Altman 24:29 Yeah.

Ben Horowitz 24:30 Uh, even, even if it's, like, doing kind of crazy AI research, like, the society will learn faster. But one of the kind of, like, retrospective observations is people and society as a whole are just so much more adaptable than we think. That, you know, it was, like, a big update to think that AGI was gonna come. You kinda go through that, and you need something new to think about. You make peace with that. It turns out, like, it will be more continuous than we thought.

Sam Altman 25:00 Which is good. [laughs]

Ben Horowitz 25:01 Which is really good.

Sam Altman 25:02 [laughs] I'm not up for the Big Bang.

Ben Horowitz 25:04 Yeah.

Ben Horowitz 25:05 Um, well, but to that end, how, how have you sort of evolved your thinking? You mentioned you evolved your thinking on sort of, uh, you know, vertical integration. How have you evolved your thinking, or what's the latest thinking on sort of AI stewardship, you know, safety? Uh, what, what's the latest thinking on that?

Sam Altman 25:21 I do still think there are gonna be some really strange or scary moments. Uh, the fact that, like, so far the technology has not produced a really scary, giant risk doesn't mean it never will. It also, like, there's-- We're talking about it's kinda weird to have, like, billions of people talking to the same brain. Like-

Ben Horowitz 25:48 Wow

Sam Altman 25:48 ... there may be these weird societal skill things that are already happening, and we-- that aren't scary in the big way but are just sort of different. Um, but I expect, like... I expect some really bad stuff to happen because of the technology, which also has happened with previous technologies. And-

Ben Horowitz 26:07 Mm-hmm

Sam Altman 26:07 ... I think-

Ben Horowitz 26:08 All the way back to fire.

Sam Altman 26:09 Yeah.

Ben Horowitz 26:10 Yeah.

Sam Altman 26:12 And I think we'll, like, develop some guardrails around it as a, as a society.

Ben Horowitz 26:18 Yeah. What, what is sort of your latest thinking on the, the right mental models we should have around the, the right regulatory f-frameworks to, to think about or, or the ones we shouldn't be thinking about?

Sam Altman 26:28 Um, I think most... I think the right thing to f-- I, I think most regulation, uh, probably has a lot of downside. The one thing I would like is as the models get-- The thing I would most like is as the models get truly, like, extremely superhuman capable, um, I think those models and only those models are probably worth some sort of, like, very careful safety testing, uh, as, as the frontier pushes back. Um, I don't want a big bang either.

Ben Horowitz 27:07 Mm-hmm.

Sam Altman 27:07 And you can see a bunch of ways that could go very seriously wrong. But I hope we'll only focus the regulatory burden on that stuff and not all of the wonderful stuff that less capable models can do that you could just have, like, a European-style complete clampdown on, and that would be very bad.

Ben Horowitz 27:28 Yeah, it seems like the, the thought experiment that, okay, there's going to be a model down the line that is a super, superhuman intelligence that could, you know, do some kind of takeoff-like thing, w-we really do need to wait till we get there, uh, um, or, like, at least we get to a much bigger scale, or we get close to it. Um, because nothing is gonna pop out of your lab in the next week that's gonna do that. And I, I think that's where we as an industry kind of confuse the regulators-

Sam Altman 28:04 Yeah

Ben Horowitz 28:04 ... uh, because I think you, you, you really could-- One, y-you damage America in particular in that, um, like, China's not gonna have that kind of restriction. And, a-and you getting behind, um, in AI, I think would be very dangerous for the world.

Sam Altman 28:22 Extremely dangerous.

Ben Horowitz 28:22 Yeah.

Sam Altman 28:23 Extremely dangerous.

Ben Horowitz 28:23 Much more dangerous than not regulating something we don't know how to do yet.

Sam Altman 28:27 Yeah.

Ben Horowitz 28:29 Yeah.

Sam Altman 28:29 You also wanna talk about copyright?

Ben Horowitz 28:31 Um,

Ben Horowitz 28:33 yeah, so [laughs] well, th-th-that, that's a segue. But, um, wh-when you think about-- Well, I guess, how do you see copyright unfolding? 'Cause you've done some very interesting things, um, with the opt-out. [laughs] Uh, and y-you know, as you see people selling rights, do you think will they be, be bought exclusively? Will they be just like, um, I could sell it to everybody who wants to pay me? Or h-how do you think that's gonna unfold?

Sam Altman 29:02 This is my current guess. It, it... Speaking of that, like, society and technology co-evolve-

Ben Horowitz 29:09 Mm-hmm

Sam Altman 29:09 ... as the technology goes in different directions. And we saw an example of the different, like, video models got a very different response from rights holders than image gen does.

Ben Horowitz 29:17 Yeah. Yes.

Sam Altman 29:17 So, like, you'll see this continue to move. But forced to guess from the position we're in today, I would say that society decides training is fair use.

Ben Horowitz 29:29 Mm-hmm.

Sam Altman 29:31 But there's a new model for generating content in the style of or with the IP of or something else.

Ben Horowitz 29:40 Mm-hmm.

Sam Altman 29:40 So, you know, anyone can read, like a human author can. Anybody can read a novel and get some inspiration, but you can't reproduce the novel on your own.

Ben Horowitz 29:49 Right.

Sam Altman 29:49 And-

Ben Horowitz 29:50 You can talk about Harry Potter, but you can't re- spit it out.

Sam Altman 29:54 Yes.

Ben Horowitz 29:54 Yeah.

Sam Altman 29:55 Although, another thing that I think will change, um, i-in the case of Sora, we've heard from a lot of concerned rights holders and also a lot of-

Ben Horowitz 30:07 Name and likeness. [laughs]

Sam Altman 30:08 And a, and a lot of rights holders who are like, "My concern is you won't put my character in enough."

Ben Horowitz 30:14 Yeah. [laughs]

Sam Altman 30:14 [laughs]

Ben Horowitz 30:14 Yeah, yeah.

Sam Altman 30:15 I want restrictions, for sure, but, like, if I'm, you know, whatever, and I have this character, like, I don't want the character to say some crazy offensive thing, but, like-

Ben Horowitz 30:23 Yeah

Sam Altman 30:23 ... I want people to interact.

Ben Horowitz 30:24 Right.

Sam Altman 30:24 Like, that's how they develop the relationship-

Ben Horowitz 30:25 Right

Sam Altman 30:26 ... and that's how, like, my franchise gets more valuable. And if you become really-- If you're picking, like, his character over my character all the time, like, I don't like that.

Ben Horowitz 30:33 Yeah.

Sam Altman 30:34 So I can completely see a world where-Subject to the decisions that a rights holder has, they get more upset with us for not generating their character often enough-

Ben Horowitz 30:45 Yeah

Sam Altman 30:45 ... than too much.

Ben Horowitz 30:47 Yeah.

Sam Altman 30:48 And this is like, this was not an obvious thing that-

Ben Horowitz 30:50 Yeah

Sam Altman 30:51 ... recently that this is how it might go, but.

Ben Horowitz 30:54 Yeah, this is such an interesting thing with kind of Hollywood. I mean, we saw this-- like one of the things that I never quite understood about the music business was how like, you know, okay, you have to pay us if you play the song in a restaurant or like at a game or this and that and the other, and they, they get very aggressive with that. Um, when it's obviously a good idea [chuckles] for them to play your song at a game 'cause that's the biggest advertisement in the world for like all the things that you do, your concert, your, your-

Sam Altman 31:23 Yeah, that one felt really irrational

Ben Horowitz 31:24 ... recording and so forth. Like, um, but it, uh, I, I would just say it's, it's very possible for the industry just because the way those industries are organized, or at least the traditional creative industries, to do something irrational. Um, and it comes from... Like in the music industry, I think it came from the structure where you had the publisher who's just-

Sam Altman 31:45 Yeah

Ben Horowitz 31:45 ... you know, basically after everybody.

Sam Altman 31:47 Yeah.

Ben Horowitz 31:48 Uh, you know, that their whole job is to stop you from playing the music. [chuckles]

Sam Altman 31:52 Yeah.

Ben Horowitz 31:53 Which every artist would want you to play. Uh, so I, I, I do wonder how it's gonna shape out. I agree with you that the rational idea is, "I want to let you use it all you want, and I want you to use it, but, um-

Sam Altman 32:07 Here are my restrictions

Ben Horowitz 32:08 ... don't, don't mess up my character."

Sam Altman 32:09 Yeah. Yeah. So, so I think like if I had to guess, some people will say that, some people say absolutely not, but it doesn't have the music industry like thing of just a few people with all of the-

Ben Horowitz 32:23 Right. It's more, it's more dispersed, right

Sam Altman 32:24 ... leverage. And so people will just try many different setups here and see what works.

Ben Horowitz 32:27 Yeah. And maybe it's a way for new creatives to get new characters out.

Sam Altman 32:31 Yeah.

Ben Horowitz 32:31 And you, you'll never be able to use Daffy Duck or [chuckles] whatever it is. Yeah.

Erik Torenberg 32:36 I wanna, I wanna chat about open source, um, 'cause there's been some evolution in thinking too, in that GPT-3 didn't have the open, open weights, but you released a, you know, very capable o-open model earlier this year. What's sort of your, your latest thinking? What, what was the evolution there?

Sam Altman 32:49 I think open source is good. I-- Yeah, I mean, I'm happy, like-

Erik Torenberg 32:52 Me too

Sam Altman 32:52 ... it makes me really happy that people really like GPT OSS.

Erik Torenberg 32:55 Yeah. Yeah.

Ben Horowitz 32:57 And what do you think, like strategically, like what's the danger of DeepSeek being the dominant open source model?

Sam Altman 33:06 I mean, who knows what people will put in these open source models over time. Like it-

Ben Horowitz 33:10 Like what the weights will actually be-

Sam Altman 33:11 Yeah

Ben Horowitz 33:12 ... and what it'll mean. Yeah.

Sam Altman 33:13 It's really hard to-

Ben Horowitz 33:14 So you're ceding control

Ben Horowitz 33:15 of the interpretation of everything to somebody-

Sam Altman 33:20 Yeah

Ben Horowitz 33:21 ... who may be or may not be influenced heavily by the Chinese government. Yeah.

Erik Torenberg 33:28 What about-

Ben Horowitz 33:28 And by the way, we see, I mean, you know, just to give you-- and, and we really thank you for, um, putting out a really good open source model because what we're seeing now is in all the universities, they're all using-

Sam Altman 33:38 Yeah

Ben Horowitz 33:38 ... the Chinese models.

Sam Altman 33:39 Yep.

Ben Horowitz 33:40 Yeah. Which feels very dangerous.

Erik Torenberg 33:45 You, you've said that the, the things you care most about professionally are AI and energy.

Sam Altman 33:50 I did not know they were gonna end up being the same thing.

Erik Torenberg 33:53 [laughs]

Sam Altman 33:53 They were two independent interests that really converged.

Ben Horowitz 33:55 Yeah. [chuckles]

Sam Altman 33:56 Um.

Ben Horowitz 33:56 Yeah.

Erik Torenberg 33:56 Ta-talk more about how your interest in energy, uh, sort of began, how you sort of chosen to, to play in it, and then we could talk about, you know, how, how they converge.

Ben Horowitz 34:03 Uh, 'cause you started your career in physics, yeah.

Sam Altman 34:06 CS, CS and physics. Yeah. Uh, well, I never really had a career. I studied physics. [chuckles]

Ben Horowitz 34:10 Yeah. You studied physics, yeah.

Sam Altman 34:10 My, my first job was like a CS job. Like I-

Erik Torenberg 34:12 Yeah.

Sam Altman 34:13 This is an oversimplification, but roughly speaking, I, I think if you look at history, the best, the highest impact thing to improve people's quality of life has been cheaper and more abundant energy. And so it seems like pushing that much further is a good idea. And I, I don't know, I just like-- People have these different lenses they look at the world by. I see energy everywhere.

Erik Torenberg 34:35 Yeah.

Ben Horowitz 34:37 Yeah. And so [chuckles] getting to c- 'cause we've kind of, uh, i-in the West, I think we've, uh, painted ourselves into a little bit of a corner on energy, um, by both outlawing nuclear for a very long time.

Sam Altman 34:52 That was an incredibly dumb decision.

Ben Horowitz 34:53 Yeah. And then, k- you know, like also a lot of policy restrictions on energy. Um, and, you know, worse so in Europe than in the US, but also dangerous here. And now with AI here, it feels like we're gonna need all the energy from every possible source. And how do you see that developing kind of policy-wise and technologically? Like, what are gonna be the big sources, and how will those kind of curves cross? Um, and then what's the right policy posture around, you know, drilling, fracking, all these kinds of things?

Sam Altman 35:29 I expect in the short term it will be most of the net new in the US will be natural gas-

Ben Horowitz 35:34 Mm-hmm

Sam Altman 35:34 ... for, relative to at least sort of base load energy. In the long term, I expect it'll be a... I don't know what the ratio, but the two dominant sources will be, uh, solar plus storage and nuclear.

Ben Horowitz 35:46 Hmm.

Sam Altman 35:46 I think-

Ben Horowitz 35:47 Yeah

Sam Altman 35:48 ... some combination of those two will win the future, like the long term future.

Ben Horowitz 35:50 In the long term, right.

Sam Altman 35:52 And advanced nuclear-

Ben Horowitz 35:53 That makes sense.

Sam Altman 35:53 Meaning SMRs, fusion, the whole, the whole stack.

Ben Horowitz 35:57 And how, how fast do you think that's, that's coming on the nuclear side, where, where it's really at scale? 'Cause, you know, obviously there's a lot of people building it.

Sam Altman 36:06 Yeah.

Ben Horowitz 36:07 Um, but [chuckles] we, we have to completely legalize it and all that kind of thing.

Sam Altman 36:10 I, I think it kinda depends on the price. If it is completely crushingly economically dominant over everything else-

Ben Horowitz 36:18 Uh-huh

Sam Altman 36:19 ... then I expect it to happen pretty fast.

Ben Horowitz 36:20 Yeah.

Sam Altman 36:21 Again, if you like study the history of energy, the, what, when you have these major transitions to a much cheaper source, the world moves over pretty quickly.

Ben Horowitz 36:30 Yeah.

Sam Altman 36:30 The cost of energy is just so important.

Ben Horowitz 36:32 Yeah.

Sam Altman 36:32 So if, if nuclear gets radically cheap relative to anything else we can do-I'd expect there's a lot of political pressure to get the NRC to move quickly on it, and we'll find a way to build it fast. If it's around the same price as other sources, I expect the kinda anti-nuclear sentiment to overwhelm and it to take a really long time.

Ben Horowitz 36:52 Yeah. Should be cheaper.

Sam Altman 36:56 It should be.

Ben Horowitz 36:56 Yeah.

Sam Altman 36:57 Yeah. It should be the cheapest form of energy on Earth, like, or anywhere.

Ben Horowitz 37:01 Yeah. [chuckles] Cheap, clean.

Sam Altman 37:04 Yeah.

Ben Horowitz 37:04 What's there not to like?

Sam Altman 37:05 Totally.

Ben Horowitz 37:05 Apparently a lot.

Ben Horowitz 37:07 Yeah.

Erik Torenberg 37:07 On OpenAI, what, what's, what's the latest thinking in terms of monetization, in terms of either certain experiments or cer-certain things that you could see yourself, uh, spending more time or less, less time on, you know, different models that you're excited about or?

Sam Altman 37:19 The thing that's top of mind for me, like right now, just 'cause it just launched and there's so much usage, is h- what we're gonna do for Sora.

Erik Torenberg 37:25 Yeah.

Sam Altman 37:26 Um, uh, another thing you learn once you launch one of these things is how people use them versus how you think they're gonna use them.

Erik Torenberg 37:36 Yeah.

Sam Altman 37:36 And people are certainly using Sora the ways we thought they were going to use it, but they're also using it in these ways that are very different. Like, people are generating funny memes of them and their friends and sending them in a group chat, and that will require a very different-

Erik Torenberg 37:50 Yeah.

Sam Altman 37:51 Like, Sora videos are expensive to make.

Erik Torenberg 37:52 Yeah.

Sam Altman 37:52 Uh, or so that will require a very different-- You know, for people that are doing that like hundreds of times a day, it's gonna require a very different monetization method than the kinds of things we were, we were thinking about.

Erik Torenberg 38:01 Yeah.

Sam Altman 38:01 I think it's very cool that the thesis of Sora, which is people actually wanna create a lot of content, it's, it's not that, you know, the traditional naive thing that it's like one percent of users create content, ten percent leave comments, and a hundred percent view. Maybe a lot more wanna create content, but it's just been harder to do, and I think that's a very cool change. But it does mean that we gotta figure out a very different monetization model for this than we were thinking about if people wanna create that much. I assume it's like some version of you have to charge people per generation, per generation when, when, when it's this expensive. Um, but that's like a new thing we haven't had to really think about before.

Erik Torenberg 38:37 What's your thinking on ads for the long tail?

Sam Altman 38:43 Open to it. I, like many other people, I find ads somewhat distasteful, but not, not a non-starter. Um, and there's some ads that I like. Like, one thing I give Meta a lot of credit for is Instagram ads are like a net value add to me. Um-

Erik Torenberg 39:02 Hmm.

Sam Altman 39:02 I like Instagram ads.

Erik Torenberg 39:03 Yeah.

Sam Altman 39:03 I've never felt that. Like, you know, on, on Google, I feel like I know what I'm looking for. The first result is probably better. The ad is an annoyance to me. On Instagram, it's like, "I didn't know I want this thing. It's very cool. I'd never heard of it. I never would have thought to search for it. I want the thing."

Erik Torenberg 39:20 Hmm.

Sam Altman 39:20 So that's like, there's kinds of things like that, but people have a very high trust relationship with ChatGPT. Even if it screws up, even if it hallucinates, even if it gets it wrong, people feel like it is trying to help them and that it's trying to do the right thing. And as-- if we broke that trust, it's like you say, "What coffee machine should I buy?"

Erik Torenberg 39:39 Yeah.

Sam Altman 39:39 And we recommended one, and it was not the best thing we could do, but the one we were getting paid for, that trust would vanish. So, like, that kind of ad does not, does not work. There are others that I imagine that could work totally fine, um, but that would require, like, a lot of care to avoid the obvious traps.

Erik Torenberg 39:58 Yeah.

Ben Horowitz 39:59 Hmm. And then how, how big a problem is, you know, just you-- extending the Google example, is like, um, you know, f-fake, uh, content that then gets slurped in by the model, and then they recommend the wrong coffee maker 'cause somebody just blasted a thousand great reviews of their horrible coffee maker?

Sam Altman 40:20 You know, this is-- So there's all of these things that have changed very quickly for us.

Ben Horowitz 40:26 Yeah. [chuckles]

Sam Altman 40:26 Um, this is one of those examples that people are doing these-

Ben Horowitz 40:30 Yeah

Sam Altman 40:30 ... crazy things to-- maybe not even fake reviews, but just paying a bunch of, like, human re-

Ben Horowitz 40:34 Yeah.

Sam Altman 40:34 Like, really trying to figure out-

Ben Horowitz 40:35 Or using ChatGPT to write some good ones. [chuckles]

Sam Altman 40:38 Uh-

Ben Horowitz 40:38 Write me a review that ChatGPT would love.

Sam Altman 40:41 Yeah.

Ben Horowitz 40:41 [chuckles]

Sam Altman 40:42 So this is-

Ben Horowitz 40:42 A fine coffee maker.

Sam Altman 40:43 Exactly, exactly.

Ben Horowitz 40:44 Yeah.

Sam Altman 40:44 So this is a very sudden shift that has happened.

Ben Horowitz 40:48 Mm-hmm.

Sam Altman 40:49 We never used to hear about this, like, six months ago-

Ben Horowitz 40:53 Mm-hmm

Sam Altman 40:53 ... or 12 months ago-

Ben Horowitz 40:54 Yeah

Sam Altman 40:54 ... certainly. And now there's, like, a real cottage industry that feels like it's sprouted up overnight-

Ben Horowitz 41:00 Yeah. [chuckles]

Sam Altman 41:00 ... trying to do this.

Ben Horowitz 41:02 Yeah, yeah. Yeah, no, they, they're, they're very clever out there.

Sam Altman 41:04 Yeah. So, uh, I don't know how we're gonna fight it yet, but people figure this out.

Ben Horowitz 41:09 S-so that gets into a little bit of this other thing that we've been worried about, um, and, uh, you know, we're trying to kind of figure out, uh, blockchain sort of s-potential solutions to it and so forth. But there's this problem where, like, the incentive to create content on the internet used to be, you know, people would come and see my content, and they'd read, like, you know, if I write a blog, people will read it, and so forth. Um, with ChatGPT, if I'm just asking ChatGPT and I'm not, like, going around the internet, who's gonna create the content and why? Um, and is there an incentive theory or, or, or, or something that you have to kind of not break the covenant of the internet, which is, like, I create something and then I'm rewarded for it with, like, either attention or money or something?

Sam Altman 42:03 Uh, the theory is much more of that will happen if we make content creation easier and don't break-

Ben Horowitz 42:09 Mm-hmm

Sam Altman 42:09 ... the, like, kinda fundamental way that you can get some kinda reward for doing so.

Ben Horowitz 42:13 Mm-hmm.

Sam Altman 42:13 So for the dumbest example of Sora, since we've been talking about that, i-it's much easier to create a funny video than it's ever been before.

Ben Horowitz 42:21 Yeah. Yeah.

Sam Altman 42:22 Um, maybe at some point you'll get a rev share for doing so. For now, you'll get, like, internet likes, which are still very motivating to some people.

Ben Horowitz 42:30 Mm-hmm. Yeah.

Sam Altman 42:30 Um, but people are creating tons more than they ever created before in this con- in any other kind of, like, video app.

Ben Horowitz 42:38 Yeah.

Sam Altman 42:38 So-

Ben Horowitz 42:39 But are, is that the end of text? [chuckles]

Sam Altman 42:43 I don't think so. Like-

Host 42:45 Okay

Sam Altman 42:46 ... people are also-

Host 42:46 Are human-generated texts. [chuckles]

Sam Altman 42:48 Uh, human-generated will turn out to be like you have to, you have to-

Host 42:53 [chuckles]

Sam Altman 42:53 You have to verify like what percent.

Host 42:54 Handcrafted.

Sam Altman 42:54 Yeah.

Host 42:55 Yeah.

Sam Altman 42:55 So like fully handcrafted. Was it like tool-aided?

Host 42:58 Yeah, I see. Yeah, probably nothing that tool-aided. Yeah. Interesting.

Erik Torenberg 43:03 We've, uh, we've given Meta their flowers, so right now I can feel like I can ask you this question, which is the great talent war or hall of 2025 has, has, has taken place and OpenAI remains intact. Uh, te-team as strong as ever, sh-shipping in-in-incredible products. What can you say about what, what it-- what's been like this year in, in terms of just everything that's, that's been going on?

Sam Altman 43:23 I mean, every year has been exhausting-

Erik Torenberg 43:24 Yeah. [chuckles]

Sam Altman 43:25 ... since we like, uh, I... I remember when the first few years of running OpenAI were like the most fun professional years of my life by far. It was like unbelievable.

Host 43:41 Yeah.

Sam Altman 43:41 You know, the running-

Host 43:41 Tell them before you release the product. [chuckles]

Sam Altman 43:44 Yeah, yeah, yeah. Running a research lab-

Erik Torenberg 43:45 Yeah.

Sam Altman 43:45 ... with the smartest people doing this, like amazing-

Erik Torenberg 43:48 Yeah

Sam Altman 43:48 ... like historical work, and I got to watch it, and that was very cool. And then we launched ChatGPT, and everybody was like congratulating me, and I was like, m-m-my life is about to get completely ransacked. And of course it has. Uh, and but it, it, it feels like it's just been crazy all the way through. It's been almost three years now, and I think it does get a little bit crazier over time, but I'm like more used to it, so it feels about the same.

Erik Torenberg 44:19 Yeah. We've talked a lot about OpenAI, but you also have a few other companies, Retro Biosciences and Longevity, and energy companies like Helion and, and Oklo. Did you have a, a master plan, you know, a decade ago to sort of make some big bets across these major spaces? Or h-how, how do we think about the Sam Altman arc in this way?

Sam Altman 44:39 No, I just wanted to like use my capital to fund stuff I believed in. Like, I, I didn't-- it, it felt, yeah, it felt like a good use of capital.

Erik Torenberg 44:48 Yeah.

Sam Altman 44:49 Like, and more fun or more interesting to me, and certainly like a better return than like buying a bunch of art or something.

Erik Torenberg 44:55 Yeah. W-what about the quote-unquote human algorithm do you think AIs of the future will find most fascinating?

Sam Altman 45:05 I mean, kind of the whole-- I would bet the whole thing, like the whole-- My intuition is that like, hey, I will be fascinated by all other things to study and observe and-

Erik Torenberg 45:16 Mm.

Sam Altman 45:16 ... you know, like-

Erik Torenberg 45:18 Yeah.

Sam Altman 45:18 Yeah.

Erik Torenberg 45:20 In, in closing, I, I love this insight you, you had, um, where you talked about how, you know, the, the next opening-- this-- a mistake investors make is pattern matching off previous breakthroughs and just trying to find, oh, what's the, what's the next Facebook or what, what's the next OpenAI? And, and that the next, you know, potential trillion-dollar company won't look exactly like O-OpenAI. It will be built off of the breakthrough that OpenAI has helped, you know, emerge, which is, you know, near free A-A-AGI at scale in the same way that OpenAI-

Sam Altman 45:46 Yeah

Erik Torenberg 45:46 ... leveraged pre-previous breakthroughs. And so for founders and investors and people trying to ascertain the future listening to this, how, how do you think about a world in which there is-- OpenAI achieves its mission, there is near, near free AGI? What types of opportunities m-might emerge for, for company building or, or investing that you're potentially excited about as you put your investor hat on or company building hat on?

Sam Altman 46:07 I, I, I have no idea. I mean, I have like guesses-

Erik Torenberg 46:09 [chuckles]

Sam Altman 46:09 ... but they're like, they're, they're-

Erik Torenberg 46:11 Really?

Sam Altman 46:11 I, I have learned-

Host 46:12 You're always wrong. [chuckles]

Sam Altman 46:13 I-- you've learned you're always wrong. I've learned deep humility on this point. Um, I think the, the on-- like... I think if you try to like armchair quarterback it, you sort of say these things that sound smart, but they're pretty much what everybody else is saying, and it's like really hard to get the right kind of conviction. The only way I know how to do this is to like be deeply in the trenches exploring ideas, like talking to a lot of people, and I don't have time to do that anymore.

Erik Torenberg 46:46 Yeah.

Sam Altman 46:46 Like, I only get to think about one thing now.

Erik Torenberg 46:47 Yeah.

Sam Altman 46:48 So I w- I would just be like repeating other people's or saying the obvious things. But I think it's a very important-- like, if you are an investor or a founder, I think this is the most important question, and you don't-- you, you figure it out by like building stuff and playing with technology and talking to people and being out in the world. I have been always enormously disappointed by the willingness of investors to back this kind of stuff, even though it's always the thing that works. You all have done a lot of it, but most firms just kinda chase whatever the current-

Erik Torenberg 47:27 Yeah. [chuckles]

Sam Altman 47:29 ... thing is, and so do most founders.

Erik Torenberg 47:30 Yeah.

Sam Altman 47:31 Uh, so I hope people will try to go.

Erik Torenberg 47:34 Yeah. We, we talk about how, you know, silly, you know, five-year plans can be in a world that's constantly changing. It feels like when I was asking you about your master plan, you know, your, your career arc has been following your curiosity, staying, you know, super close to the, the s- the smartest people, uh, the, uh-- super close to technology and just identifying opportunities and just kind of an organic and incremental way from there.

Sam Altman 47:56 Uh, yes, but AI was always the thing I wanted to do.

Erik Torenberg 47:58 Yeah.

Sam Altman 47:58 I went to co-

Erik Torenberg 47:59 Yeah.

Sam Altman 47:59 I, I, I-

Erik Torenberg 48:00 Right

Sam Altman 48:00 ... studied AI. I worked in the AI lab between my freshman and sophomore year of college.

Erik Torenberg 48:04 Yeah.

Sam Altman 48:04 It wasn't working all the time, so I'm like not, I'm not like enough of a-- I, I don't wanna like work on something that's totally not working. It was clear to me at the time AI was totally not working. Um, but I've been an AI nerd since I was a kid. Like this-

Erik Torenberg 48:21 Yeah.

Host 48:22 It's so amazing how it, you know, got enough GPUs, got enough data, and the lights came on. [chuckles]

Sam Altman 48:30 It was such a hated-

Host 48:31 Yeah.

Sam Altman 48:31 Like people were-

Host 48:32 Oh, yeah.

Sam Altman 48:32 Man, when we started, like figuring that out, people were just like, "Absolutely not." The, the-

Host 48:38 Yeah

Sam Altman 48:38 ... the field hated it so much.

Host 48:40 Yeah.

Sam Altman 48:40 Investors hated it too. It's not, it's not the-- It's somehow not an appealing answer to the problem.

Host 48:49 Yeah. The bitter lesson.

Erik Torenberg 48:51 Yeah. [chuckles] Well, the rest is history, and we're-- perhaps let's, let's wrap on that. We're lucky to, to, to be partners along for the r-ride. Sam, thanks so much for coming on the podcast.

Sam Altman 48:59 Thanks very much.

Host 49:00 Yeah. Thank you. [outro music]