Mark Zuckerberg — AI will write most Meta code in 18 months

1. 0:00 – 12:20

   How Llama 4 compares to other models

   1. MZMark Zuckerberg0:00

      (instrumental music plays) I would guess that the world is gonna get a lot more... a lot funnier and, like, weirder. If you think that something someone is doing is bad, and they think it's really valuable, most of the time, in my experience, they're right and you're wrong.

   2. DPDwarkesh Patel0:12

      I am worried that we're just removing all the friction between getting totally reward hacked by our technology.

   3. MZMark Zuckerberg0:19

      We are trying to build a coding agent that advances LLaMA research. I would guess that, like, sometime in the next 12 to 18 months, we'll reach the point where, like, most of the code that's going towards these efforts is written by AI. I tend to think that, for at least the foreseeable future, this is gonna lead towards more demand for people doing work, not less. If you've gotten the cost of providing that service down to one-tenth of what it would have otherwise been, maybe now that actually makes sense to go do.

   4. DPDwarkesh Patel0:47

      All right, Mark, thanks for coming on the podcast again.

   5. MZMark Zuckerberg0:49

      Yeah, happy to do it. Good to see you.

   6. DPDwarkesh Patel0:50

      You too. Last time you were here, you, um, you had launched LLaMA 3.

   7. MZMark Zuckerberg0:54

      Yeah.

   8. DPDwarkesh Patel0:55

      Now you've launched LLaMA 4.

   9. MZMark Zuckerberg0:56

      Well, the first version.

   10. DPDwarkesh Patel0:57

       That's right. What's new? What's exciting? What's changed?

   11. MZMark Zuckerberg0:59

       Oh, well, I mean, uh, the whole field's so dynamic, so, I mean, I, I feel like a ton has changed since the last time that we talked. Um, Meta AI has almost a billion people using it now, monthly, so that's, um, that's pretty wild. Um, and, you know, I think that this is gonna be a really big year on all of this, because, um, especially once you start getting the personalization loop going, uh, which we're just starting to build in now really, um, from both the context that all the algorithms have about what you're interested in feed and all your profile information, all the social graph information, but also just what you're interacting with the AI about. I think that's just gonna be kind of the next thing that's, um, that's gonna be super exciting, so really big on that. The modeling stuff continues to make really impressive advances too, as, as you know. Um, the LLaMA 4 stuff, uh, I th- I'm pretty happy with the first set of releases. You know, we announced the, um, we announced four models, and we released the first two, the Scout and Maverick ones, which are kind of like the mid-size models, mid-size to small. Um, it's not like... You know, actually, the most popular LLaMA 3 model was, um, was the, the eight billion parameter model.

   12. DPDwarkesh Patel2:10

       Mm.

   13. MZMark Zuckerberg2:10

       So we're, we, we've, we've got one of those coming in the LLaMA 4 series too. Um, our internal code name for it is Little Llama.

   14. DPDwarkesh Patel2:17

       (laughs)

   15. MZMark Zuckerberg2:18

       But, um, but that, that's, that's, that's coming probably, you know, over, over the next, over the coming months. But the, um, the Scout and Maverick ones, um, you know, I mean, they're good. They're, they're some of the highest intelligence per cost that you can get of any model that's out there, natively multimodal, very efficient, run on one host, um, designed to just be very efficient and low latency for a lot of the use cases that we're building for internally, and, you know, that's our whole thing. We, we basically build what we're, what we want, and then we open source it so other people can use it too. So I'm excited about that. Um, I'm also excited about the Behemoth model, which is, is coming up. Um, that's gonna be our first model that is, uh, sort of at the frontier. I mean, it's like more than two trillion parameters, so it is... Y- y- I mean, it's, you know, as, as the name says, it's g- it's quite, quite big, um, so we're kind of trying to figure out how we make that useful for people. It's so big that we've had to build a bunch of infrastructure, um, just to be able to push train it ourselves, and we're kind of trying to wrap our head around how does the, like, like, the average developer out there, how are they gonna be able to use something like this, and how do we make it so it can be useful for, um, distilling into models that are of reasonable size to run? 'Cause you're, you're obviously not gonna wanna run, um, you know, something like that in a, in a consumer model. But, um, but yeah. I mean, it's, there, there's a lot to go. I mean, as, as you saw with the, with the LLaMA 3 stuff last year, the initial LLaMA 3 launch was, um, was exciting, and then we just kind of built on that over the year. 3.1 was when we released the 405 billion model. 3.2 is when we got all the multimodal stuff in. Um, so I, we basically have a roadmap like that for this year too, so a lot going on.

   16. DPDwarkesh Patel4:01

       I'm interested to hear more about it. Uh, there's this impression that the gap between the best closed source and the best open source models has increased over the last year, where I know the full family of LLaMA 4 models isn't out yet, but, um, LLaMA 4 Maverick is 35 on, uh, Chatbot Arena, and then on a bunch of ma- major benchmarks, it seems like o4-mini or GP- uh, Gemini 2.5 Flash are, um, beating Maverick, which is in the same class. What, what do you make of that impression?

   17. MZMark Zuckerberg4:27

       Yeah, well, okay. There, there's a few things. I actually think that this has been a very good year for open source overall, right? If you go back to the, like, where we were last year, um, what we were doing with LLaMA was like the only real, um, super innovative open source model. Now you have a bunch of them in the field, and I think in general, the prediction that this would be the year where open source, uh, generally overtakes closed source as the most used model, models out there, um, I think is, is generally on track to be true. I think the thing that's been, um, sort of a, an interesting surprise, I think positive in, in some ways, negative in others, but, but I think overall good, is that it's not just LLaMA. There, there are a lot of good ones out there. Um, so I think that that's quite good. Um, then there's the reasoning phenomenon, which, which you basically are, are, are alluding to with talking about o3 and o4 and, and, and some of the other models, and I do think that there's this specialization that is, um, that's happening where if you want a model that is sort of the best at math problems or coding or different things like that, um, I, I do think that these reasoning models with, um, a lot of, uh, the ability to just consume more test time or inference time compute, um, in order to provide more intelligence is a really compelling paradigm. But for a lot of the applications that... And, and we're gonna do that too. We're, we're building a LLaMA 4 reasoning model and that, that'll come out at, at, at some point. Um, for a lot of the things that we care about, um...... latency and good intelligence per cost are actually much more important product attributes. Um, if you're, if you're primarily designing for, you know, consumer product, uh, people don't necessarily want it to wait like half a minute to go think through the answer. If you can provide an answer that's generally quite good too in like half a second, then that's great and that's a good trade-off. So I think that both of these are gonna end up being important directions. Um, I, I am optimistic about integrating the reasoning models with, um, kind of the, the core language models over time. I think that's sort of the direction that, um, Google has gone in with, um, with some of the more recent, uh, Gemini models, and I think that that's really promising. But, um, but I think that there's just gonna be a bunch of different stuff that goes on. Um, you also mentioned the whole, um, chatbot arena thing, which I think is interesting and it, it goes to this challenge around, how do you do the benchmarking, right? And, and basically, how do you know what models are good for which things? And one of the things that we've generally tried to do over the last year is anchor more of our models in, um, our Meta AI product north star use cases, because the, the, the issue with both kind of open source benchmarks and, you know, any given thing, like, like the, um, LLM arena stuff, is it's just, it's... They're often skewed for a, a, either a very specific, um, set of use cases which are often not actually what any normal person does in, in, in your product. Um, they are often weighted, um, c- kind of the portfolio of things that they're trying to measure is different, uh, from what people care about in any given pro- um, product. And, um, because of that, we've found that trying to optimize too much for that stuff has often, um, led us astray and actually not led towards the highest quality products, and the most usage, and best feedback within Meta AI as people use our stuff. So we're trying to anchor our north star in, um, in basically the product value that, that people, um, kind of report to us and what they say that they want and what their revealed preferences are in using, um, the experiences that, that, that we have. So sometimes I think, um, you know, sometimes these things don't quite line up and, and I think that a lot of them are, are quite easily, um, gameable, right? So I mean, I, I think on, on the, um, arena you'll see stuff like, uh, like Sonnet 3.7, it's like a great model, right? And it's, it's like not near the top. Um, and it was relatively easy for our team to tune a version of LLaMA 4 Maverick, um, that basically was way at the top, um, whereas the one that we released, that's the, the kind of the pure model, actually has no tuning for that at all so it, so it's further down. So I, I, I think you just need to be careful with, with some of the benchmarks-

   18. DPDwarkesh Patel8:54

       Mm-hmm.

   19. MZMark Zuckerberg8:54

       ... and, and we're gonna, we're gonna index primarily on the products.

   20. DPDwarkesh Patel8:57

       Is... Do okay... Feel like there is some benchmark which captures what you see as a north star of value to the user which can be sort of objectively measured between the different models and you're like, "It... I, I, I need LLaMA 4 to come out on top on this"?

   21. MZMark Zuckerberg9:11

       Well, I mean, our benchmark is basically, um, user value in Meta AI, right? So it's-

   22. DPDwarkesh Patel9:17

       But you can't compare other models, uh...

   23. MZMark Zuckerberg9:18

       Um, well, we might be able to because we might be able to run other models-

   24. DPDwarkesh Patel9:21

       Mm-hmm.

   25. MZMark Zuckerberg9:21

       ... in that and, and be able to tell, and I think that that's one of the advantages of open source is basically you have a good community of folks who can like poke holes at, okay, where is your model not g- not good and where, where is it good? Um, but I think the reality at this point is that all these models are optimized for slightly different mixes of things. I mean, everyone is trying to, I think, go towards the same, um... You know, I think all the leading labs are trying to create general intelligence, right? And, um, super intelligence, whatever you call it, right? Like basically AI that can lead towards a world of abundance where like everyone has these superhuman tools to create whatever they want and that leads to just dramatically empowering people and creating all these economic benefits. I think that that's sort of... However you define that, I think that that's kind of what, what a lot of the labs are, are going for, and, um... But there's no doubt that different folks have sort of optimized towards different things, I think the Anthropic folks have really focused on kind of coding and, and agents around that. You know, the OpenAI folks I think have gone a little more towards reasoning, um, recently, and I think that there is a space which, if I had to guess, I think will end up probably being the most used one which, um, which is quick, is very natural to interact with, um, is very natively multimodal, um, that fits into kind of throughout your day the ways that you want to interact with it. Um, and I think you got a chance to play around with, um, with, uh, the new Meta AI-

   26. DPDwarkesh Patel10:51

       Yeah.

   27. MZMark Zuckerberg10:51

       ... app that we're, that we're releasing and, you know, one of the fun things that we put in there is the, um, the demo for the full duplex voice.

   28. DPDwarkesh Patel10:57

       Mm-hmm.

   29. MZMark Zuckerberg10:58

       And it's... I mean, it's early, right? I mean, it's not, um... You know, there's a reason why we haven't made that the default voice model in the app, but there's something about how naturally conversational it is that I think is just like really fun and compelling, and, and I, I think being able to mix kind of that in, um, with the right personalization is gonna lead towards a product experience where... You know, I, I would basically just guess that you go forward a few years, like we're just gonna be talking to AI throughout the day about different things that we're wondering and, um, you know, it's like you'll, um, you'll, you'll have your phone, you'll talk, you'll talk to it on your phone, you'll talk to it while you're browsing your feed apps, it'll give you context about different stuff, you'll be able to answer questions, it'll help you as you're interacting with people in messaging apps. Um, you know, eventually I think we'll, we'll walk through our, our daily lives and we'll either have glasses or, um, you know, other kinds of AI devices and, and just be able to kind of seamlessly interact with it all day long. So I, I think that that is... That's kind of the north star and whatever the benchmarks are that lead towards people feeling like the, the quality is... Like that's what they want to interact with, that I think is actually the thing that is ultimately gonna matter the most to us.

   30. DPDwarkesh Patel12:10

       Hmm. I got a chance to play around with b- both Orion and also the Meta AI app and th- the voice mode was super smooth. That w- it was quite impressive.I-

2. 12:20 – 27:22

   Intelligence explosion

   1. DPDwarkesh Patel12:20

      on the point of what the different labs are optimizing for, to steel man their view, I think a lot of them think that once you fully automate software engineering and AI research, then-

   2. MZMark Zuckerberg12:29

      Yeah.

   3. DPDwarkesh Patel12:29

      -you can kick off an intelligence explosion where you have millions of copies of these software engineers replicating the research that happened between LLaMA 1 and LLaMA 4, that scale of improvement, again, in a matter of weeks, uh, or months rather than years. Um, and so it really matters to just have... Close the loop on the software engineer and then you can ha- be the first to ASI.

   4. MZMark Zuckerberg12:49

      Yeah.

   5. DPDwarkesh Patel12:49

      What do you make of that?

   6. MZMark Zuckerberg12:50

      Well, well, I mean, I- I personally think that's pretty compelling.

   7. DPDwarkesh Patel12:53

      Mm.

   8. MZMark Zuckerberg12:53

      Um, and, and that's why we have a big coding effort too. I mean, we're working on, um, a number of coding agents inside Meta, um, y- you know, because we're not really a, um, an enterprise software company. We're primarily building it for ourselves.

   9. DPDwarkesh Patel13:07

      Mm-hmm.

   10. MZMark Zuckerberg13:07

       So we're... So again, you know, we, we go kind of, like, for, for, um, you know, the specific goal, we're not trying to build a general developer tool. We are trying to build a coding agent and an AI research agent that, um, that basically advances LLaMA research specifically. And, um, and it's, like, f- just fully kind of plugged into our tool chain and all this. So, I, I think that that's important and, um, and I think is going to end up being an important part of how this stuff gets done. I would guess that, like, sometime in the next 12 to 18 months, um, we'll reach the point where, like, most of the code that's going towards these efforts is written by AI. And I don't mean like auto-complete. I mean, right today you have, like, you have kind of, you know, good auto-complete.

   11. DPDwarkesh Patel13:54

       Mm-hmm.

   12. MZMark Zuckerberg13:54

       Like, you start writing something and it can complete the, the kind of section of code. I- I'm talking more like, you give it a goal, it can run tests, right? It can, it can kind of improve things. Um, it can find issues. Um, it writes higher quality code than, like, the, the average very good person on, on the team already. Um, and, and, like, I think that that's gonna be a really important part of this for sure.

   13. DPDwarkesh Patel14:20

       Mm.

   14. MZMark Zuckerberg14:21

       But I don't know if that's the whole game. I mean, I think that that's... That, I think, is gonna be a big industry, um, and I think that that's gonna be, um, an important part of how AI gets developed. But I think that there are still guy s- I think... I mean, look, j- I guess one, one way to think about this is this is a massive space, right? So, I don't think that there's just gonna be, like, one company with one optimization function that serves everyone as best as possible. I think that there are a bunch of different labs that are gonna be doing leading work towards different domains. Some are gonna be more kind of enterprise focused or coding focused. Some are gonna be, um, more productivity focused. Some are gonna be more f- social or entertainment focused. Um, within the assistant space, I think there are gonna be some that are much more kind of informational or productivity. Some are gonna be more companion focused. Um, there's gonna be a lot of the stuff that's just, like, fun and entertaining and, like, shows up in your feed. Um, and I think that that's... So, I think that there's just, like, a huge amount of space, and part of what's fun about this is, like, it's... Like, going towards this AGI future, there are a bunch of common threads for what needs to get invented. But there are a lot of things, at the end of the day, that need to get created. And I, I think that that's, um... I think you'll start to see a little more specialization between the groups, if I had to guess.

   15. DPDwarkesh Patel15:35

       It's r- it's really interesting to me that you basically agree with the premise that there will be an intelligence explosion and something like super intelligence on the other end. But then if that's the case, tell me if I'm misunderstanding you. If that's the case, why even bother with personal assistants and whatever? Why not just ge- get to superhuman intelligence first and then deal with everything else later?

   16. MZMark Zuckerberg15:52

       Well, I think that that's just one aspect of the flywheel. Right? So, uh, part of what, what I generally disagree with on the fast takeoff thing is it takes time to build out physical infrastructure.

   17. DPDwarkesh Patel16:04

       Mm-hmm.

   18. MZMark Zuckerberg16:04

       Right? So, if you want to build, like, a gigawatt cluster of compute, that just is gonna take some time, right? It, like, takes NVIDIA a bunch of time to, like, stabilize their new generation of, of, of the systems. And then you need to figure out the networking around it, and then you need to, like, build the building, and you need to get permitting, and you need to get the energy. And then, like, okay, you want, like, some... Whether it's gas turbines or, um, or green energy. You need to, like... There's a whole supply chain of that stuff. So, I, I think that there's, like, a lot of... And we talked about this a bunch on the-

   19. DPDwarkesh Patel16:38

       Mm-hmm.

   20. MZMark Zuckerberg16:38

       -the last time that I was, that I was on the, the podcast with you, and I think some of these are just, like, physical world, human time things that as you start getting more intelligence in one part of the stack, um, you'll basically just run into a different set of bottlenecks. I mean, that's sort of the way that engineering always works. It's like you solve one bottleneck, you get another bottleneck.

   21. DPDwarkesh Patel16:58

       Yeah.

   22. MZMark Zuckerberg16:59

       Um, another bottleneck in the system, or another ingredient that's gonna make this am- like, work well, is basically people getting used to and learning and having a, a feedback loop with, um, with using the system. So, I don't think, like... Like, these systems don't tend to be the type of thing where, like, something just shows up fully formed and then people magically fully know how, how to use it. Um, and that's the end. I think that there is this co-evolution that happens where people are learning how to best use these AI assistants. On the same side, the AI assistants are learning what those people care about, and the developers of those AI assistants are able to make the, the kind of AI assistants better. Um, and then you're also building up this base of context. So now you wake up and you're, like, a year or two into it, and now the AI assistant can reference things that you talked about a couple years ago and, like, that's pretty cool. But you couldn't do that if it just, uh, you just launched the perfect thing on day one. There's no way that it could reference what you talked about two years ago if it didn't exist two years ago. So, um, so, I guess my view is, like, there's this huge intelligence growth.

   23. DPDwarkesh Patel18:14

       Mm-hmm.

   24. MZMark Zuckerberg18:14

       There's a, a very rapid curve on the uptake of people interacting with the AI assistants and, like, the learning feedback and, and kind of data flywheel around that. Um, and then there is...... also the buildout of the supply chains and infrastructure and regulatory frameworks to enable the scaling of a lot of the physical infrastructure. But I think at some level, all of those are gonna be necessary and not just the coding piece.

   25. DPDwarkesh Patel18:45

       Hmm.

   26. MZMark Zuckerberg18:46

       Um, I guess one, one specific example of this that I think is interesting... Actually, if- even if you go back a few years ago, we had a project on, I think it was on our ads team, to automate ranking experiments, right? That's, like, a pretty constrained environment. It's not like write open-ended code. It's basically look at the whole history of the company, every experiment that any engineer has ever done in the ads system and look at what worked, what didn't, what the results of those were, and basically formulate new hypotheses for different tests that we should run that could improve the, the performance-

   27. DPDwarkesh Patel19:22

       Mm-hmm.

   28. MZMark Zuckerberg19:22

       ... of the ads system. And what we basically found was, um, we were bottlenecked on compute to run tests, uh, based on the number of hypotheses. It turns out even with just the, the humans that we have right now, um, on the ads team, we already have more good ideas to test than you actually have either kind of compute or really cohorts of people to test them with, right? 'Cause I mean even if you have, like, three and a half billion people using your products, you still want each... You, you know, each test needs to be statistically significant, so it needs to have, you know, some number of whatever it is, hundreds of thousands or millions of people. And, um, and there's only, there's some, there's kind of only so much throughput that you can get on testing through that. So, we're already at the point even with just, like, the people we have that, um, that we already can't really test everything that we want. Um, so now, just being able to test more things is not necessarily gonna be additive to that. We need to get to the point where the average quality of the hypotheses that the AI is generating is better than what the, all the things above the line that we're actually able to test that, like, sort of the best humans on the team have been able to do before it'll even be marginally useful for it. So, I think that there's, like... It'll... We'll get there. We'll get there I, I think pretty quickly. But, um, but it's not like, "Okay, cool, the thing can write code," all of a sudden everything is just improving massively. There, there are, like, these real-world constraints that, um, that, that basically it needs to... First, it needs to be able to kind of do a reasonable job. Then it needs to, um, be able to... You need to have the compute and the, the kind of people to test. And, and then over time as the quality creeps up, I don't know, are we here in, like, five or ten years and it's, like, no set of people can generate a hypothesis as good as the AI system? I don't know, maybe, right? It's... Then, then I think i- i- in that world, obviously that's gonna be how all the value is created. But, but, but that's not the first step.

   29. DPDwarkesh Patel21:22

       Publicly available data is running out, so major AI labs like Meta, Google DeepMind and OpenAI all partner with Scale to push the boundaries of what's possible. Through Scale's Data Foundry, major labs get access to high-quality data to fuel post-training, including advanced reasoning capabilities. Scale's research team, Seal, is creating the foundations for integrating advanced AI into society through practical AI safety frameworks and public leaderboards around safety and alignment. Their latest leaderboards include Humanity's Last Exam, Enigma Eval, MultiChallenge, and VISTA, which test a range of capabilities from expert-level reasoning to multimodal puzzle-solving to performance on multi-turn conversations. Scale also just released Scale Evaluation, which helps diagnose model limitations. Leading frontier model developers rely on Scale Evaluation to improve the reasoning capabilities of their best models. If you're an AI researcher or engineer and you wanna learn more about how Scale's Data Foundry and Research Lab can help you go beyond the current frontier of capabilities, go to scale.com/dwarkesh. So, if you buy this view that, um, this is where intelligence is headed, the reason to be bullish on Meta is obviously that you have all this distribution, and... Which you can also use to learn more things that can be useful for training with the Meta... Yeah, you mentioned the Meta AI app is... Now has a, um, billion active users. Maybe-

   30. MZMark Zuckerberg22:48

       Not the app, not the app.
3. 27:22 – 35:56

   AI friends, therapists & girlfriends

   1. DPDwarkesh Patel27:22

      so on this point of AI generated content or AI interactions, already people have meaningful relationships with AI therapists, AI friends, you know, maybe more. Um, and this is just gonna get more intense as these AIs become more unique and more personable, more intelligent, more spontaneous and funny and so forth. How do we make sure people are gonna have relationships with AIs? How- how do we make sure that these are healthy relationships?

   2. MZMark Zuckerberg27:47

      Well, I think there are a lot of questions that you only really can answer as you start seeing the behaviors.

   3. DPDwarkesh Patel27:53

      Hmm.

   4. MZMark Zuckerberg27:53

      So, probably the most important upfront thing is just like ask that question and care about it at each step along the way. But- but I think also being too prescriptive upfront in saying, "We think these things are not good," um, often cuts off value, right? Because... I don't know, people use stuff that's valuable for them. I- I mean, one of my core guiding principles in designing products is like people are smart, right? They know what is valuable in their lives. Um, every once in a while, um, you know, something can- something bad happens in a product and you- you wanna make sure that you design your products well, um, to- to minimize that. But- but if- if- if you think that something someone is doing is bad and they think it's really valuable, most of the time in my experience, they're right and you're wrong, and you just haven't come up with the framework yet for understanding why the thing that you're doing is- is valuable and helpful in their life. Um, yeah, so- so that's kind of the main way that- that I- that I think about it. I do think that people are going to use, um, AI for a lot of these social tasks. Already one of the main things that we see people using that AI for is kind of talking through difficult conversations that they need to have with, um, with people in their life. It's like, "Okay, my- you know, my- my... I'm having this issue with my girlfriend or whatever. Like, help me have this conversation." Or like, "I need to have this hard conversation with my boss at work. Like, how do I have that conversation?" Um, that's pretty helpful. And then I think as the personalization loop kicks in and the AI just starts to get to know you better and better, um, I think that will just be really compelling. Um, you know, one thing just from working on social media for a long time is, um... There's this stat that I always think is crazy. The- the average American I think has... I think it's fewer than three friends.

   5. DPDwarkesh Patel29:45

      Hmm.

   6. MZMark Zuckerberg29:46

      Three people who they'd consider friends. And- and the average person has demand for meaningfully more.

   7. DPDwarkesh Patel29:52

      Yeah.

   8. MZMark Zuckerberg29:52

      I think it's like 15 friends or something, right? I- I guess there's probably some point where you're like, "All right, I'm just too busy. I can't deal with more people." But- but the average person wants more connectivity, connection than they have. Um, so there's a lot of questions that people ask of stuff like, "Okay, is this going to replace kind of in-person connections or real life connections?" And my default is that the answer to that is probably no. I think it- it- it... You know, I think that there are all these things that are better about kind of physical connections when you can have them, but the reality is that people just don't have the connection and they feel more alone, um, a lot of the time than they would like. So, I think that a lot of these things that today there might be a little bit of a stigma around, um, I would guess that over time...... we will find the vocabulary as a society to be able to articulate why that is valuable and why the people who are doing these things are, like, why they are rational for doing it, and like, and how it is adding value for their, for their lives. But, but also I think that the field is very early. So, um, I mean, it's, uh, like I, I think, you know, there are a handful of companies and stuff who are doing virtual therapists-

   9. DPDwarkesh Patel31:07

      Yeah.

   10. MZMark Zuckerberg31:08

       ... and, you know, there's, like, virtual girlfriend-type stuff. But it's, um, it's very early, right?

   11. DPDwarkesh Patel31:13

       Yeah.

   12. MZMark Zuckerberg31:13

       It's, I mean, the, the embodiment in the things is, is pretty weak. A lot of them, like, you, you open it up and it's just like a- an image of, of like, of the therapist or the person you're talking to or whatever. I mean, sometimes there's some very rough animation, but it's not like an embodiment. I mean, you, you've seen the stuff that we're working on in reality labs where, like, you have the codec avatars and it, like, feels like it's a real person. I think that's kind of where it's going. You're gonna, you know, you'll, you'll be able to, um, basically have like an always on video chat where it's like, oh, and also the per- the, the, um, the, uh, the AI will be able to, you know, uh, uh, the gestures are important too. Like more than half of communication when you're actually having a conversation is not the words that you speak. It's all the non-verbal stuff.

   13. DPDwarkesh Patel31:53

       Yeah. I, I did get a chance to check out, um, Orion the other day and I thought it was super impressive. And I'm mostly optimistic about the technology just because generally I'm, as y- as you mentioned, like libertarian about if people are d- doing something, probably to think it's good for them. Although I actually don't know if it's the case that if somebody is using TikTok, they would say that they're happy with how much time they're spending on TikTok or something. So I'm mostly optimistic about it also in the sense that if we're gonna be living in this future world of AGI, we need to be, in order to keep up with it, humans need to be upgrading our capabilities as well with tools like this. Um, and just generally there can be more beauty in the world if you can see-

   14. MZMark Zuckerberg32:25

       Yeah.

   15. DPDwarkesh Patel32:25

       ... Studio Ghibli everywhere or something. I was worried that one of the flagship use cases that your, your team showed me was, "I'm sitting at the breakfast table and on the periphery of my vision is just a bunch of reels that are scrolling by. Um, maybe in the future my AI girlfriend is on the other side of the screen or something." Um, and so I, I am worried that we're just removing all the friction between getting totally reward hacked by our technology. Um, how, yeah, how do we make sure, like, I don't know, th- this is not what ends up happening in five years?

   16. MZMark Zuckerberg32:56

       Uh, I mean, again, I think, I think people have, have a good sense of what they want. I mean, that- that experience that you saw is mo- th- that was a demo just to show multitasking and holograms, right?

   17. DPDwarkesh Patel33:05

       Yeah.

   18. MZMark Zuckerberg33:05

       So, I, I mean, I, I, I agree that, like, I don't think that the, the future is like you have stuff that's trying to compete for your attention in the corner of your vision all the time. I don't think people would like that too much.

   19. DPDwarkesh Patel33:17

       Mm-hmm.

   20. MZMark Zuckerberg33:17

       Um, so it's actually, it's one of the things as we're designing these glasses that we're really mindful of, is like probably the number one thing that glasses need to do is get out of the way and be good glasses, right? And, um, as an aside, I think that's part of the reason why the Ray-Ban Meta product has done so well, is like, all right, it's like great for listening to music and taking phone calls and taking photos and videos, and the AI is there when you want it. But when you don't, it's like a great, you know, good looking pair of glasses that, that, that, that people like and it, and it kind of gets out of the way well. Um, I would guess that that's gonna be a very important design principle for, for the, um, the augmented reality future, right? I, th- the main thing that I, that I see here is, you know, I think it's kind of crazy that for how important the digital world is in all of our lives, the only way we can access it is through these like physical, you know, digital screens, right? It's like you, you have like a phone, you have your, your, your computer. You can put a big TV, it's like this huge physical thing. Um, it just seems like we're at the point with technology where the physical and the digital worlds should really be fully blended, and that's what the holographic overlays allow you to do. Um, but I agree. I think a, a big part of the design principles around that are gonna be, okay, you'll, you'll be interacting with, with people and you'll be able to bring digital artifacts into those interactions and be able to do cool things like very seamlessly, right? It's like if I want to show you something, here, like here's a screen. Okay, here it is. I can show you, you can interact with it. It can be 3D. Um, we can kind of play with it. Um, you wanna, you know, like play a card game or whatever. It's like, all right, here's like a deck of cards.

   21. DPDwarkesh Patel35:17

       Yeah.

   22. MZMark Zuckerberg35:17

       We, we can play with it. It's like two of us are here physically, like you have a, a third friend who's just hologramming in, right? And th- they can, they can kind of participate too. Um, but, but I think that in that world people are gonna be, you know, just like you don't want your physical space to be cluttered. It's sort of like a, you know, it, it just kind of has like a, it wears on you psychologically. I don't think people are gonna want the digital kind of physical space to, to feel that way either. So, I don't know. Th- that's more of an aesthetic and, and, and one of these norms that I think will have to get worked out, but, um, but I, I, I think, I think we'll figure that out.

   23. DPDwarkesh Patel35:55

       Mm-hmm.

4. 35:56 – 40:35

   DeepSeek & China

   1. DPDwarkesh Patel35:56

      Uh, going back to the AI conversation, you're mentioning how big of a bottleneck, um, the physical infrastructure can be. Related to other open source models like DeepSeek and so forth, DeepSeek right now has less compute, uh, than a lab like Meta, and you could argue that it's competitive with the LLaMA models. Um, if China is better at, you know, physical infrastructure, industrial scale ups, um, getting more power and more data centers online, how worried are you that this will, they might beat us here?

   2. MZMark Zuckerberg36:24

      I, I mean, I think it's a, it's like a real competition.

   3. DPDwarkesh Patel36:26

      Yeah.

   4. MZMark Zuckerberg36:26

      I mean, I think that you're seeing the, the industrial policies really play out, um, where, yeah, I mean, I think China's bringing online more power and because of that, I think that the US really needs to focus on streamlining the ability to build data centers and build and produce energy, um, or, or I think we will be at a significant disadvantage. Um-At the same time, I think some of the export controls on things like chips, I think you can see how they're clearly working in a way.

   5. DPDwarkesh Patel37:00

      Mm-hmm.

   6. MZMark Zuckerberg37:01

      Because, you know, there was all the conversation with DeepSeek about, how they did all these, like, very impressive low level optimizations. And the reality is, they did, and that is impressive, but then you ask, "Why did they have to do that when none of the, like, American labs-"

   7. DPDwarkesh Patel37:18

      Mm-hmm.

   8. MZMark Zuckerberg37:18

      "... did it?" And it's like, well, because they're using, like, partially nerfed chips that are the only thing that NVIDIA's allowed to sell in China because of the es- export controls. So, so DeepSeek basically had to go spend a bunch of their calories and time doing low level infrastructure optimizations that the American labs didn't have to do. Now, they produced a good result on text, right? It's like, um, DeepSeek is text only. Um, so the infrastructure's impressive, the text result is, is impressive, um, but every new major model that comes out now is multimodal, right? It's image, um, it's voice, and, and, and, and theirs isn't. And now, the question is, why is that the case? I don't think it's because they're not capable of doing it. I think that they basically had to spend their calories on doing these infrastructure optimizations to overcome the fact that there were these export controls. Um, but when you compare, like, LLaMA 4 with DeepSeek, I mean, our reasoning model isn't out yet, so I think that the, the kind of R1 comparison isn't, isn't, isn't clear yet. But, um, but we're basically, like, e- effectively same ballpark on all the tech stuff as what DeepSeek is doing, but with a smaller model, so it's, it's much more kind of efficient per, um... The, the kind of cost per intelligence is lower with what we're doing for LLaMA on text. And then all the multimodal stuff, we're effectively leading that and it just doesn't even exist in their stuff. So, um, so I think that the LLaMA 4 models, when you compare them to what they're doing, are, are good, and, and I think generally people are gonna prefer to use the LLaMA 4 models. Um, but I think that there is this interesting contour where, like, it's clearly a good team that's doing stuff over there, and I think you're right to ask about the, um, accessibility of power, the accessibility of compute and chips and things like that, um, because I think w- the, the kind of work that you're seeing the different labs do and play out, I, I think is somewhat downstream of that.

   9. DPDwarkesh Patel39:20

      (air whooshing) Freemium products attract a ton of fake account signups, bot traffic, and free tier abuse, and AI is so good now that it's basically useless to just have a CAPTCHA of six squiggly numbers on your signup page. Take Cursor. People were going to insane lengths to take advantage of Cursor's free credits, creating and deleting thousands of accounts, sharing logins, even coordinating through Reddit, and all this was costing Cursor a ton of money in terms of inference compute and LLM API calls. Then they plugged in WorkOS Radar. (logo chiming) Radar distinguishes humans from bots. It looks at over 80 different signals, from your IP address, to your browser, to even the fonts installed on your computer, to ensure that only real users can get through. Radar currently runs millions of checks per week, and when you plug Radar into your own product, you immediately benefit from the millions of training examples that Radar has already seen through other top companies. Previously, building this level of advanced protection in-house was only possible for huge companies, but now with WorkOS Radar, advanced security is just an API call away. Learn more at workos.com/radar. All right, back to Zuck.

5. 40:35 – 55:01

   Open source AI

   1. DPDwarkesh Patel40:35

      (air whooshing) So Sam Altman recently tweeted that OpenAI is gonna release an open source, uh, SOTA reasoning model. I think part of the tweet was that, "We will not do anything silly like, um, say that you, you, you can only use it if you have, uh, less than 700 million users." Um, DeepSeek has, uh, the MIT license, uh, whereas LLaMA ha- I think a couple of the contingencies in the LLaMA license require you to say, "Built with LLaMA," on applications using it, or any model that you train using LLaMA has to begin with the word "Llama." Yeah, but wh- wh- what do you think about the license? Should it be less onerous for developers?

   2. MZMark Zuckerberg41:07

      I, I mean, look, I mean, I... We've basically pioneered, like, the open source LLM thing, so I mean, I, I, I don't, I don't consider the, the license to be onerous. I, I kind of, you know, think that when we were starting to push on open source, it was this, (clears throat) it was this big debate in the industry of like, "Is this even a reasonable thing to do?" Like, "Can you do something-"

   3. DPDwarkesh Patel41:32

      Hmm.

   4. MZMark Zuckerberg41:32

      "... that is safe and trustworthy with open source?" Like, is, um... "Will open source be able, ever be able to be competitive enough that anyone will even care?" And, and basically when we were answering those questions, which, you know, a lot of the hard work that, you know, I think a lot of the teams at Meta... Although th- there are other folks in the industry, but really, the LLaMA models were the ones that I think broke open this whole open source, um, AI thing in a, in a, in a huge way. Um, you know, we were very focused on, "Okay, if we're gonna put all this energy into it, then at a minimum, you know, if you're gonna have these large cloud companies like Microsoft and Amazon and Google turn around and sell our model, that we should at least be able to have a conversation with them before they do that, around, um, a- around basically like, 'Okay, what kind of business arrangement should we have?'" But, but our goal with the m- with the license isn't, um... You know, we're generally not trying to stop people from using the model. We, we just think, like, "Okay, if you're, like, one of those companies, or, I don't know, if you're Apple, um, you know, just come talk to us about what you wanna do, and, and let's find, like, a productive way to, to do it together." So, I think that that's generally been fine. Now, um, if the whole open source part of the industry evolves in a direction where, you know, there's, like, a lot of other great options, and if, if, like, the, you know, the license ends up being a reason why people don't want to use LLaMA, then I don't know, we'll, we'll have to reevaluate the strategy, whether, you know, w- w- what it makes sense to do at that point. But-I, I just don't think we're there. That's not, in practice, a thing that we have seen, um, companies coming to us and saying, "We don't wanna use this because your, your license says that if you're, if you reach 700 million people, you have to come talk to us." So, um, it's, i- at least so far, it's a little bit more something that we've heard from, like, kind of open source purists. Like, "Is this as clean of an open source model as you, y- we, as, as you'd like it to be?" And, and look, I mean, I think that debate has existed since the beginning of open source with, like, um, you know, the, you know, just a- all the GPL license stuff-

   5. DPDwarkesh Patel43:41

      Mm.

   6. MZMark Zuckerberg43:41

      ... versus other things. And it's like, okay, does, like, does it need to be the case that anything that touches open source can, has to, has to be open source? Or can people just take it and use it in different ways? And I'm sure there will continue being debates around this. But I don't know. If you're, if you're spending many, many billions of dollars training these models, I think asking the other companies that, um, are also huge and similar in size and can easily kind of afford to have a, a relationship with us to talk to us before they use it, I think it seems like a pretty reasonable thing.

   7. DPDwarkesh Patel44:10

      If it turns out that you, you know, other models are also ... You know, there's, like, a bunch of good open source models, so that part of your mission is fulfilled, and maybe mo- other models are better at coding, is there a world where you just say, "Look, open source system, ecosystem is healthy. There's plenty of competition. We're happy to just use some other model, whether it's for internal, um, software engineering at Meta or deploying to our apps. We don't necessarily need to build with LLaMA."

   8. MZMark Zuckerberg44:37

      Um, well, again, I mean, we do a lot of things. So, it's possible that, you know, I, I, I guess let's take a step back. The reason why we're building our own big models is because we wanna be able to, like, build exactly what we want.

   9. DPDwarkesh Patel44:52

      Mm-hmm.

   10. MZMark Zuckerberg44:53

       Right? And, and none of the other models in the world are sort of exactly what we want. If they're open source, then you can take them and you can fine-tune them in different ways. But you still have to deal with the model architectures and, you know, they make different size trade-offs around, um, that affect the latency and inference cost of the models. And it's like, okay, at the scale that we operate at, um, that stuff really matters. Like, we made the LLaMA Scout and Maverick models certain sizes for a specific reason, because they fit on a host and we wanted certain latency. Especially for the voice models that we're, that we're working on, that we wanna just basically have purveyed and, and be across everything that we're doing. From the glasses, to all of our apps, to the Meta AI app, and, and, and all this stuff. So, um, so I think that there's a level of kind of control of your own destiny that you only get when you, when you build this stuff yourself. That said, there are a lot of things that ... Like, AI is gonna be used in every single thing that every company does. When we build a big model, we also need to choose which things, which use cases internally we're gonna optimize for. So, does that mean that, for certain things, we're not gonna, you know, think that, like, okay, maybe Claude is better for building this specific development tool that this team is using? All right, cool. Then, like, use that. Fine. Great. Um, I don't think ... We don't wanna fight with, you know, one hand tied behind our back. Um, we're, we're doing a lot of different stuff. Um, you also asked would we maybe ... Would it not be important because other people are doing open source. Um, I don't know. On this, I'm a little more worried.

   11. DPDwarkesh Patel46:23

       Mm-hmm.

   12. MZMark Zuckerberg46:23

       Because I think you have to ask ... For anyone who shows up now and is doing open source now that we have done it, there's a question which is would they still be doing open source if we weren't doing it. And, like, I think that there are a handful of folks who see the trend that more and more development is going towards, um, towards open source. And they're like, "Oh, crap. Like, we kinda need to be on this train or else we're gonna lose. Like, we have some closed model API." And like, increasingly, a lot of developers, that's not what they want. Um, so, so I think you're seeing a bunch of the other players start to do some work in open source. But it's just unclear if it's dabbling or fundamental for them-

   13. DPDwarkesh Patel47:05

       Mm-hmm.

   14. MZMark Zuckerberg47:05

       ... in the way that it has been for us.

   15. DPDwarkesh Patel47:06

       Mm-hmm.

   16. MZMark Zuckerberg47:07

       And, you know, a good example is, like, what's going on with, like, Android. Right? It's like, Android started off as the open source thing. There's not really, like, any open source alternative. Like, I think over time, Android has just been kinda getting more and more closed. Um, so I think if you're us, you'd kinda need to worry that if we stopped pushing the industry in this direction, that, like, all these other people maybe are only really doing it because they're trying to kind of compete with us and the direction that we're pushing things. And, you know, they've, they already have their revealed preference for what they would build if open source didn't exist and, um, and it wasn't open source. Right? So, um, so I, I, I just think we need to be careful about relying on that continued behavior for the future of the technology that we're gonna build at the company.

   17. DPDwarkesh Patel47:58

       I mean, another thing I've heard you mention is that it's important that the standard gets built around American models like LLaMA. Um, I guess I wanted to-

   18. MZMark Zuckerberg48:07

       Yeah.

   19. DPDwarkesh Patel48:07

       ... understand your logic there. Because it seems like with certain kinds of networks, it is the case that the Apple App Store just has a big contingency around what it's built around. Um, but it doesn't seem like, you know, you, you could ... If you build some sort of scaffold for DeepSeek, you couldn't have easily just switched it over to LLaMA-4. Especially since between generations. Like, LLaMA-3 wasn't MoE, LLaMA-4 is. So, things are changing between generations of models as well. So, what's the reason for thinking things will get built out in this contingent way on a specific standard?

   20. MZMark Zuckerberg48:34

       Um, um, I'm not sure. What do you mean by contingent?

   21. DPDwarkesh Patel48:36

       Oh, as in, like, it's important that people are building for LLaMA rather than for LLMs in general. Because that will determine what the standard is for the future.

   22. MZMark Zuckerberg48:44

       Sure. Well, well, look. I mean, I think these models encode values-

   23. DPDwarkesh Patel48:47

       Mm-hmm.

   24. MZMark Zuckerberg48:48

       ... and ways of thinking about the world.

   25. DPDwarkesh Patel48:49

       Mm-hmm.

   26. MZMark Zuckerberg48:50

       And, you know, we had this interesting experience early on where we took an early version of LLaMA and we translated it. I think it was ... It might have been into French or s- some other language. And the feedback that we got, um-... i think it was, I think it was French. Hmm. From, from French people was, "This sounds like an American who learned to speak French. Like, it doesn't sound like a French person." And it's like, "Well, what do you mean? Does it not speak French well?" It's like, "No. It speaks French fine, it's just, like, the way that it thinks about the world is, like, seems slightly American." I- so I feel there's, like, these subtle things that kind of get built into it. Um, over time, as the models get more sophisticated, they should be able to embody different value sets across the world, so maybe that's, like, a very kind of, um, you know, not particularly sophisticated example. But I think it sort of illustrates- Hmm. ... the point. Um, and, you know, some of the stuff that we've seen in testing some of the models especially coming out of China is like, they sort of have certain values encoded in them. And, um, and it's not just like a light fine-tune to get that to feel the way that you want. Now, the stuff is different, right? So, I think language models, um, or something that has like a, kind of like a world model embedded into it, have more values. Reasoning, I think, is, I mean, I guess there are kind of values or ways to think about reasoning but one of the things that's nice about the reasoning models is they're trained on verifiable problems. So, do you need to be worried about, like, cultural bias if your model is doing math? Probably not. (laughs) Right? I think that that's... You know, I think it's like the, the chance that, like, some reasoning model that was built elsewhere is, like, going to kind of incept you by, like, solving a math problem in a way that's, that's, um, devious seems low. Um, there's a whole set of different issues I think around coding, um, which is the other verifiable domain, which is, you know, I think you, you kind of need to be worried about, like, waking up one day and, like, does a model that have some tie to another government, like, can it embed all kinds of different vulnerabilities and code that then, like, the intelligence organizations associated with that government- Hmm. ... can then go exploit, so now you sort of like, all right, like in some future version where you have, you know, some model from some other country that we're using to, like, secure or build out a lot of our systems, and then all of a sudden you wake up and it's like everything is just vulnerable to, um, in a way that, like, that country knows about but, but like you don't? Or, or it turns on a vulnerability at some point. Th- those are real issues. Um, so what we've basically found is, um... No, I mean, I'm, I'm very interested in studying this because I think one of the main things that's interesting about open source is the ability to distill models. Um, you know, most people, th- the primary value isn't just like taking a model off the shelf and saying like, "Okay. Like, Meta built this version of LLaMA. I'm gonna take it and I'm gonna run it exactly in my application." It's like, no. Well, your application isn't doing anything different if you're just running our thing. You're at least gonna fine-tune it or try to- Yeah. ... distill it into a different model. And when we get to stuff like the Behemoth model, like, the whole value in that is being able to basically take this very high amount of intelligence and distill it down into a smaller model that you're actually gonna wanna run. But this is like the beauty of distillation, and it's like one of the things that I think has really emerged as a very powerful technique in the last year since the last time we, we've sat down is, um... And I think it's worked better than most people would predict, is you can basically take a model that is much bigger and take probably like 90 or 95% of its intelligence and run it in something that's 10% the size. Now, do you get 100% of the intelligence? No. But, like, 95% of the intelligence at 10% of the cost is, like, pretty good- Yeah. ... for, for a lot of things. Um, the other thing that's interesting is now with this, like, more varied open source community where you... It's not just LLaMA, you have other models, you have the ability to distill from multiple sources. So, now you can basically say, "Okay. LLaMA's really good at this, like maybe the architecture's really good because it's fundamentally multimodal and fundamentally more, um, inference-friendly and more efficient." But, like, let's say this other model's better at coding. Okay. Well, just, you can distill from both of them and then build something that's better than either of them for your own use case. Um, so that's cool. But you do need to solve the security problem of knowing that you can distill it in a way that is safe and secure. And so this is something that, that we've been researching and have put a lot of time into, and what we've basically come to is like, look. Anything that's kind of like language is, is quite fraught because there's, like, a lot of values embedded in that. So unless you don't care about having the values from whatever the model is that you got, you probably don't wanna, like, distill the straight, like, n- l- language world model. Um, on reasoning, I think you can get a lot of the way there by limiting it to verifiable domains, running, um, kind of code cleanliness and security filters, like, like whether it's like the LLaMA Guard open source or the Code Shield open source things that we've done that basically, um, allow you to incorporate different, different, um, input into your models and make sure that the... that both the input and the output, um, are secure. And then just a lot of red teaming to make sure that you're, you're, um, like, y- you just have people who are experts who are looking at this and saying, "All right. Is this model doing anything-" Mm-hmm. "... that isn't what I want after distilling from something?" And I think with the combination of those techniques, you can probably distill on the reasoning side for verifiable domains quite securely. Um, that's something I'm, I'm pretty confident about, and it's something that, that we've done a lot of research around. But I think this is a very big question, is like how do you do good distillation? Because there's just so much value to be unlocked. But, at the same time, I do just think that there is some fundamental bias in the different models. Speaking of value to be unlocked-

6. 55:01 – 59:18

   Monetizing AGI

   1. DPDwarkesh Patel55:01

      What do you think the right way to monetize AI will be? Because obviously digital ads are quite lucrative, but as a fraction of total GDP, it's small in comparison to, like, all remote work. Um, uh, like e- even if you can increase its productivity and not replace, uh, work, that's still worth tens of trillions of dollars. So, is it possible that ads might not be a r- yeah, how do you think about this?

   2. MZMark Zuckerberg55:23

      Uh, I mean, like we were talking about before, there's gonna be all these different applications, and different applications tend towards different things. Um, ads is great when you wanna offer people a free service, right? Because it's free, you need to cover it somehow.

   3. DPDwarkesh Patel55:35

      Yeah.

   4. MZMark Zuckerberg55:35

      Ads is like, okay, it's ads solves this problem-

   5. DPDwarkesh Patel55:38

      Right.

   6. MZMark Zuckerberg55:38

      ... of like a person does not need to pay for something, and they can get something that is like amazing for free. Um, and also, by the way, with modern ad systems, a lot of the time people think that the ads add value to the thing-

   7. DPDwarkesh Patel55:50

      Mm-hmm.

   8. MZMark Zuckerberg55:50

      ... if you do it well, right? It's, um... You know, you need to g- be good at ranking, and, and you need to be good at having enough liquidity of advertising inventory, um, so that way... You know, if you only have five advertisers in the system, no matter how good you are at ranking, you may not be able to show something to someone that they're interested in. But if you have a million advertisers in the system, then you're probably gonna be able to find something pretty compelling if you're good at, at picking out, you know, the different needles in the haystack that that person's gonna be interested in. So, I think that definitely has its place, but there are also clearly going to be other business models as well, including ones that, um, just have higher costs-

   9. DPDwarkesh Patel56:32

      Mm-hmm.

   10. MZMark Zuckerberg56:32

       ... so it doesn't even make sense to offer them for free. Um, which, by the way, there have always been business models like this. There's a reason why social media is free and ad-supported, but then if you wanna watch Netflix, um, or like ESPN or something, you need to pay for that. It's okay, 'cause the content that's going into that, like they need to produce it, and that's very expensive for them to produce, and they probably could not have enough ads in the service in order to make up for the, the cost of producing the content, so basically you just need to pay to, to access it. Um, then the trade-off is fewer people do it, right? It's like they're talking about hundreds of millions of people using those instead of billions, so it's, it's... There's kind of a, a value switch there. Um, I think similar here. You know, not everyone is gonna want, like, a software engineer or a thousand software e- engineering agents or whatever it is. But if you do, that's something that you're, you are probably gonna be willing to pay thousands or tens of thousands or hundreds of thousands of dollars for. Um, so I, I think that this just speaks to the diversity of different things that need to get created, is like there are gonna be business models at each point along the spectrum. And at Meta, um... Yeah. For the consumer piece, we definitely wanna have a free thing, and I'm sure that will, will end up being ad-supported. But I also think we're gonna wanna have a business model that supports people using arbitrary amounts of compute to do, like, really even more amazing things than what it would make sense to be able to offer with a free service. And for that, I'm sure we'll end up having a premium service. But, but I mean, but I think our, our basic, you know, values on this are we wanna serve as many people in the world.

   11. DPDwarkesh Patel58:10

       Lambda is the cloud for AI developers. They have over 50,000 NVIDIA GPUs ready to go for startups, enterprises, and hyper-salers. Compute seems like a commodity though, so why use Lambda over anybody else? Well, unlike other cloud providers, Lambda's only focus is AI. This means their GPU instances and on-demand clusters have all the tools that AI developers need preinstalled. No need to manually install CUDA, drivers, or manage Kubernetes. And if you only need GPU compute, you can save a ton of money by not paying for the overhead of general purpose cloud architectures. Lambda even has contracts that let enterprises use any type of GPU in their portfolio and easily upgrade to the next generation. For all of you wanting to build with Llama 4, Lambda has a serverless API without rate limits. It's built with rapid scaling in mind. Users have 1,000x their inference consumption without ever having to apply for a quota or even speak to a human. Head to lambda.ai/thorakesh for a free trial of their inference API, featuring the best open source models like DeepSeek and Llama 4 at the lowest prices in the industry. All right,

7. 59:18 – 1:02:50

   The role of a CEO

   1. DPDwarkesh Patel59:18

      back to Zuck. How do you keep track of... Um, you've got all these different projects, s- some of which we've talked about today. I'm sure there's many I don't even know about. Um, a- as the CEO overseeing everything, there's a big spectrum between like go- you know, going to the Llama team and, "Here's the hyper-parameters you should use," to just giving like a mandate, like, "Go make the AI better." H- and there's many different projects. How do you think about the way in which you can best deliver your value add, um, and oversee all these things?

   2. MZMark Zuckerberg59:44

      Well, I mean, a lot of what I spend my time on is trying to get awesome people onto the teams.

   3. DPDwarkesh Patel59:48

      Hmm.

   4. MZMark Zuckerberg59:48

      Right? I mean, it's, um... So there's that, and then there's stuff that cuts across teams. It's like, all right, you build Meta AI, and you wanna get it into WhatsApp or Instagram. It's like, okay, then now I need to get those teams to talk together.

   5. DPDwarkesh Patel1:00:03

      Mm-hmm.

   6. MZMark Zuckerberg1:00:03

      And then there's a bunch of questions like, okay, I was, um... You know, it's like, okay, do you want the thread for Meta AI and WhatsApp to feel like other WhatsApp threads, or do you want it to feel like other kind of like AI chat experiences? There's like different idioms for, for those. And so I think that there's like all these interesting questions that sort of need to get answered around like how does this stuff basically fit into all of what we're doing.

   7. DPDwarkesh Patel1:00:29

      Hmm.

   8. MZMark Zuckerberg1:00:29

      Then there's a whole other part of what we're doing which is basically pushing on the infrastructure. If you're... if you wanna stand up a gigawatt cluster, then first of all, that has a lot of implications for, um, for the way that we're doing infrastructure build outs. It has sort of political implications for how you engage with the different states where you're building that stuff. Um, it has financial implications for the company in terms of, all right, there's like a lot of economic uncertainty in the world. Do we like go double down on infrastructure right now? Um, and, and if so, what other trade-offs do we wanna make around the company? And th- those are things that, like...... it's tough for other people to really make those kind of decisions.

   9. DPDwarkesh Patel1:01:09

      Mm-hmm.

   10. MZMark Zuckerberg1:01:09

       Um, and then, and then I think that there's this question around, like, taste and quality, which is, like, when is something good enough that we wanna ship it?

   11. DPDwarkesh Patel1:01:20

       Mm.

   12. MZMark Zuckerberg1:01:20

       And, and I, I do feel like, in general, I'm the steward of that for the company, although, you know, we have a lot of other people I think have good taste as well who are also filters for, for different things. But, um, but yeah, I think that those are, those are basically the areas. But I think, um, AI is interesting, because more than some of the other stuff that we do, it is more research and model-led than really product-led. Like, you can't just, like, design the product that you want and then try to build the model to fit into it. You really need to, like, design the model...

   13. DPDwarkesh Patel1:01:55

       Mm.

   14. MZMark Zuckerberg1:01:55

       ... first and, like, the capabilities that you want, and then you get some emergent properties, then it's, "Oh, you can build some different stuff, because this kinda turned out in this way." And I think at the end of the day, like, like people wanna use the best model, right? So, that's partially why, you know, when we're talking about building the most, like, personal AI, um, the best voice, the best personalization, um, and, like, also a very smart experience with very low latency. Those are the things that we basically need to design the whole system to build, which is why we're working on full duplex voice, which is why we're working on, like, the personalization to bo- both have, like, good memory extraction from your interaction with AI, but also be able to plug into all the other Meta systems, and why we design the, um, specific models that we design to have the kind of size and latency parameters that they do.

   15. DPDwarkesh Patel1:02:45

       Hmm. Speaking of politics, um,

8. 1:02:50 – 1:07:56

   Is big tech aligning with Trump?

   1. DPDwarkesh Patel1:02:50

      there, there's been this perception that some tech leaders have been aligning with Trump. You and others have donated to his inaugural event, and were on stage with him, and I think you se- settled, like, a lawsuit, which gave, uh, resulted in him getting $25 million. I wonder, um, what's going on here? Is it ju- does it, does it feel like the cost of doing business with the administration or... Yeah, what, what's, what's one sort of thing about this?

   2. MZMark Zuckerberg1:03:11

      I mean, my, my view on this is, like, he's the President of the United States. Our default as an American company should be to try to have a productive...

   3. DPDwarkesh Patel1:03:19

      Mm-hmm.

   4. MZMark Zuckerberg1:03:19

      ... relationship with whoever is running the government. Um, I would do this, you know, uh, like, we- we've tried to offer to support, um, previous administrations as well. I've, I've been pretty public with some of my frustrations with the previous administration...

   5. DPDwarkesh Patel1:03:33

      Mm-hmm.

   6. MZMark Zuckerberg1:03:33

      ... and how they basically did not engage with us or the business community more broadly, which I think, frankly, I think is going to be necessary to make progress on some of these things. Like, we're not gonna be able to build the level of energy, um, that we need if there, if you don't have a dialogue and they're not prioritizing trying to do those things. So, um, but fundamentally, you know, look, I mean, I think a lot of people wanna write the story about, like, like, you know, what direction are people going.

   7. DPDwarkesh Patel1:04:01

      Mm.

   8. MZMark Zuckerberg1:04:01

      I, like, I just think it's, like, we're trying to build great stuff. We wanna work with, have a productive relationship with people, and that's sort of, that's, that's how I see it, and it is also how I would guess most others see it. Um, but obviously, I can't speak for them.

   9. DPDwarkesh Patel1:04:19

      Um, you've spoken out about how you've re-thought some of the ways in which you, um, engage and defer to the government in terms of moderation stuff in the past. How, how are you thinking about AI governance? Because if AI is as powerful as we think it might be, the government will want to get involved. What, what is, like, the most productive approach to take there, and wha- what should the government be thinking about here?

   10. MZMark Zuckerberg1:04:40

       Yeah, I guess in the past, I, I probably just, um, I mean, most of the comments that I made, I think were in the context of content moderation.

   11. DPDwarkesh Patel1:04:47

       Mm-hmm.

   12. MZMark Zuckerberg1:04:48

       Right? Where, um, you know, it's been an interesting journey over the last 10 years on this, where there have... It's obviously been an interesting time in history. There have been novel questions raised about online content moderation. Um, some of those have led to, I think, productive, um, new systems getting built, like our AI systems to be able to detect nation states trying to interfere in each other's elections. I think we will continue building that stuff out, and that, that, I think, has been net positive.

   13. DPDwarkesh Patel1:05:16

       Right.

   14. MZMark Zuckerberg1:05:16

       I think other stuff, we went down some bad paths. Like, I just think the fact-checking thing...

   15. DPDwarkesh Patel1:05:21

       Yeah.

   16. MZMark Zuckerberg1:05:22

       ... was not as effective as Community Notes, 'cause it's not an internet-scale solution. There weren't enough fact checkers, and, like, people didn't trust the specific fact checkers. They, like, you, you, you wanted a more robust system. So, I, I think what we got with Community Notes is the right one on that. But, um, but my point on this was, was more that, um, that I think, historically, I probably deferred a little bit too much to, um, either the media in, in kind of their critiques, or the government on things that they did not really have authority over, but just as, like, a central figure. Um, like, I, I, I think we tried to build systems that where maybe we could not have to make all of the content moderation decisions ourselves or something. And I, I guess I just think part of the, the growth process over the last 10 years is just, "Okay, like, we're a meaningful company. We need to own the decisions that we need to make. We should listen to feedback from people, but shouldn't defer too much to people who are not, who do not actually have authority over this, because at the end of the day, we, like, we're in the seat, and we need to, like, own the decisions that we make."

   17. DPDwarkesh Patel1:06:34

       Yeah.

   18. MZMark Zuckerberg1:06:34

       And, um, so I think we probably, you know, it's be- it's been a maturation process, and, um, in some ways, painful, but, but I, I, you know, I think we're, we're probably a better company for it.

   19. DPDwarkesh Patel1:06:44

       Mm. W- w- will tariffs increase the cost of building data centers in the US and shift build-outs to Europe and Asia?

   20. MZMark Zuckerberg1:06:50

       It is really hard to know how that plays out. Um, I think we're probably in the, the early innings on that.

   21. DPDwarkesh Patel1:06:56

       Mm-hmm.

   22. MZMark Zuckerberg1:06:56

       And, um, it's very hard to know.

   23. DPDwarkesh Patel1:07:00

       Got it. Um, what is your single highest leverage hour in a week? What are you doing in that hour?

   24. MZMark Zuckerberg1:07:06

       I don't know. I mean, every week is a little bit different. Um, and it's probably gotta be the case that the most leveraged thing that you do in a week is not...... the same thing each week, or else-

   25. DPDwarkesh Patel1:07:16

       Mm-hmm.

   26. MZMark Zuckerberg1:07:16

       ... by definition you should probably spend more than one hour doing that thing every week.

   27. DPDwarkesh Patel1:07:19

       (laughs)

   28. MZMark Zuckerberg1:07:19

       Um, but... Yeah, I don't know. It's, it's part of the fun of, of, of, of both, I guess, this job, but also the industry being so dynamic is, like, things really move around, right? And, like, and the world is very different now than it was at the beginning of the year, than it was six months in- into the middle of last year. Um, and I think a lot has sort of... has really advanced meaningfully-

   29. DPDwarkesh Patel1:07:44

       Mm-hmm.

   30. MZMark Zuckerberg1:07:44

       ... and, like, a lot of cards have been turned over since the last time that we sat down. I think that was about a year ago, right?
9. 1:07:56 – 1:15:11

   100x productivity

   1. DPDwarkesh Patel1:07:56

      You know, you talked about these models being mid-level software engineers by the end of the year. Or, um... What would be possible if, say, software productivity increased, like, 100X in, uh, y- w- two years? What kinds of, uh, things could be built that we can't build right now?

   2. MZMark Zuckerberg1:08:09

      What kinds of things? Well, y- It's... I mean, that's an interesting question. Um, I mean, I think one theme of this conversation is that the, like, amount of creativity that's gonna be unlocked is gonna be massive. Um... (smacks lips) And if you look at, like, the overall arc of kind of human society and the economy over 100 or 150 years, it's basically people going from being primarily agrarian and most of human energy going towards just feeding ourselves, to that has become a kind of smaller and smaller percent, and the things that take care of, like, our basic physical needs are a smaller and smaller percent of human energy, which has led to two impacts. One is more people are doing kind of creative and cultural pursuits.

   3. DPDwarkesh Patel1:09:01

      Mm-hmm.

   4. MZMark Zuckerberg1:09:01

      And two is that more people... Uh, th- people on- in general spend less time working and more time on entertainment and culture. Um, I think that that is almost certainly gonna continue as, as this goes o- this isn't like the one to two year thing of what happens when you have a, a, like, a super powerful software engineer, but I think over time, you know, if, if you... Like, everyone is gonna have these superhuman tools, um, to be able to create a ton of different stuff. I think you're gonna get this incredible diversity. Um, part of it is gonna be solving, like, things that we hold up as, like, these, like, hard problems, like solving diseases or, like, solving different things around science or, um, or just, like, different technology that makes our lives better. But I would guess that a lot of it is going to end up being kind of cultural and social pursuit and entertainment, and, like, I would guess that the world is gonna get a lot more... like, a l- like, a lot funnier and, like, weirder and, and quirkier in a way that, like, the memes on the internet have sort of gotten over the last 10 years. And, um... And I think that that adds a certain kind of richness and depth as well that in kind of funny ways I think it actually helps you connect better with people, because now, like, I don't know, it's like all day long I just find interesting stuff on the internet and, like, send it in group chats to the people I care about who I think are gonna find it funny and it's, like... Like, the, the media that people can produce today to express very, very nuanced specific cultural ideas, um, I don't know. It's, it's, it's cool, and I think that'll continue to get built out, and I think it does advance society in a bunch of ways, even if it's not, like, the hard science way of curing a disease.

   5. DPDwarkesh Patel1:10:50

      Yeah.

   6. MZMark Zuckerberg1:10:50

      But I guess this is sort of, if you think about it, like the, the, like, meta social media view of the world is like, yeah, I think people are gonna spend a lot more time doing that stuff in the future, um, and, and it's gonna be a lot better and it's gonna help you connect because it's, like, gonna help express different ideas, um, because the world's gonna get more complicated, but, like, our technology, our cultural technology to kind of express these very complicated things, um, in like a very kind of funny little clip or something, um, are gonna just get so much better. So, I think that that's all great. Um... (sniffs) I don't know. Next year, for... I, I tend to... I mean, just to... I guess one other thought that I think is interesting to cover is, um... I tend to think that at for at least the foreseeable future, this is gonna lead towards more demand for people doing work, not less. Now, people have a choice of how much time they wanna spend working, um, but... Like, I'll give you one interesting example of something that, that we were talking about recently. We, um... So we have, like, three... almost three and a half billion people use our services every day, and one question that we've struggled with forever is, like, how do we provide customer support? Right? Today, like, you can, you can write an email. Um, but we've never seriously been able to contemplate having, um, you know, having, like, voice support where someone can just call in, and I guess that's maybe one of the artifacts of having a free service, right, is like the, the revenue per person's not so high that you can have an economic model that people can, can kind of call in. But also, with three and a half billion people using your service every day, I mean, you'd, uh, there'd be like a massive, massive number of people, like s- some, like, like the biggest call center in the world type of thing.

   7. DPDwarkesh Patel1:12:32

      Yeah.

   8. MZMark Zuckerberg1:12:32

      Um, but it would be like, wh- what... $10, $20 billion, something ridiculous, a year to, to kind of staff that. So we, we've never really kind of, like, thought too seriously about it 'cause it was always just like, no, there's no way that this kinda makes sense. But now, as the AI gets better, you know, you're gonna get to this place where the AI can handle a bunch of people's issues. Not all of them, right? 'Cause, um... A- a- and maybe 10 years from now or something it can handle all of them. But when we're thinking about, like, a three to five year time horizon, um, it'll be able to handle a bunch, kind of like self-driving cars can handle a bunch of terrain, but in general they're not, like, doing the whole route by themselves yet in, in, in most cases, right? It's like-People thought truck driving jobs were gonna go away, there's actually more truck driving jobs now than there were, like, when we started talking about self-driving cars, um, you know, whenever it was, almost 20 years ago. Um, and I think for, going back to this customer support thing, it's like, all right, w- it wouldn't make sense for us to staff out, um...

   9. DPDwarkesh Patel1:13:39

      Mm-hmm.

   10. MZMark Zuckerberg1:13:39

       ... calling for everyone. But let's say the AI can handle 90% of that. Then, like, and then if you, if i- if it can't handle it, then it kicks it off to a person. Okay, now, like, if you've gotten the cost of providing that service down to one-tenth of what it would have otherwise been, then, all right, m- maybe then now that actually makes sense to go do, and that would be kinda cool. So, the net result is, like, I actually think we're probably gonna go hire more customer support people, right? It's like, uh, like, the, the common knowledge or, like the, the kind of common belief that people have is that, like, oh, this is clearly just gonna automate jobs and, like, all these jobs are gonna go away. I, I actually just, uh, that has not really been how the history of technology has worked. It's, it's been, you know, you, you can, you, like, create things that take away 90% of the work, and that leads you to want more people, not less.

   11. DPDwarkesh Patel1:14:27

       Final question: who is the one person in the world today who you most seek out for advice?

   12. MZMark Zuckerberg1:14:32

       Oh, man. Well, I feel like it's m- part of my style is I like having a breadth of advisors. So, it's, it's not just, it's not just one person. But it's, um ... we've got a great team. I mean, it's, uh, that I'm, I'm ... and I think the- there's people at the company, people on our board, um ... and there's a lot of people in the industry who are doing new stuff. I, there's, there's not, there's not, like, a single person. Um, but it's, uh ... I don't know, it's fun. And, and also, as, when the world is dynamic, um, just having a reason to work with people you like-

Episode duration: 1:15:49