Meta's AI Chief On AI Beef, New Models And Life With Zuck - EP 71 Alex Wang

All right, Kylie, we've got another big guest this week.

Huge.

Alex Wang, the chief of Meta's artificial intelligence efforts. It was how many months ago?

About 10 months ago.

About 10 months ago, he was the founder, co-founder, and CEO of Scale AI. Meta sort of quasi-acquired the company, half acquired the company and fully acquired Alex. And he's been, he's been in, in hiding. He's been in, uh, AI protection program since—

Yes, we haven't seen much of him whatsoever until today on the Core Memory Pod.

Yes. I am not sure exactly why this happened, but here he is. He's going to— he's going to tell us. Well, we'll find out. He's going to hopefully tell us about— they just released a new model. I'm sure we will get into some of that. And then they're a little bit of a mystery to me about like where they are philosophically on AI. Alex has always been a bit of— when he was at SCaLE, he was— he was— they always said they were Switzerland and he was loud on some things, but not, not always AI itself and how he feels about it.

And they made a lot of news last year with everyone they hired for millions and millions and millions of dollars and built up this team. And everyone's been waiting to see what they're going to do with all of these resources.

So we will talk about recruiting soup and all these millions of dollars. So this is it. This is Alex Wang's— he's emerged.

Hell yeah, brother.

I am Ashlee Vance.

And I'm Kiley Robinson.

And this is Core Memory. Alex, thank you for being here.

Yeah, excited to be here.

I feel like we were— we kind of texted a little bit pre-meta happenings. We had, we had like this kind of country music text chain for sure going. And then, and then I'm also— I've known Nat Friedman for a long time.

Yeah.

And then I feel like the two of you disappeared and went, went into the foxhole. Yeah. Went very, very quiet there. And now here you are with a new model.

Yeah.

Emerging.

Yeah.

But you guys went very quiet for a bit.

We, uh, yeah, we had a lot of work to do. I mean, I think, uh, turns out, um, building a frontier model from scratch in 9 months is, uh, Yeah, it takes a lot of painstaking effort. But yeah, I mean, it's been really exciting to see everyone use MuseSpark, you know, the model we released, and we have better models cooking. So it's exciting.

And so what you— I mean, you were like a San Francisco alight San Franciscan, I guess. And then, I mean, I assume you work at Menlo Park.

Yeah.

So were you—

did you have to like move down to South Bay?

Yeah. Wow.

Yeah. I did.

So you're full on committed?

I'm full on. Yeah. And for me, the city now is Palo Alto.

Okay.

Walk on University Ave, get a boba.

Okay. I was wondering about this. And what's, what's like the arrangement between— I mean, the people I know best are you, Nat, Daniel Gross. I've only hung out with Zuck once or twice. Yeah. I mean, what's the— I'm trying to paint a picture of, of, of how you guys are arranged.

Yeah, so basically the whole unit is called Meta Superintelligence Labs, which I oversee, and then there's various pieces of it. So there's a unit called TBD, which is the sort of large model research lab that I think is somewhat infamous, but that's where a lot of the leading researchers and infrastructure engineers are. They actually all technically report to me. So that's one setup. Then there's also a group called Product and Applied Research, or PAR for short. That's what Nat Friedman heads up. So they're responsible for all the products that we build and the actual deployment of these great models to the world. And then Also within the overall Meta Superintelligence Labs umbrella is FAIR, which continues to do exploratory and exciting research. I'm particularly excited about a lot of their scientific research. We've shown some pretty great work on using models, AI models to understand the brain, as well as using AI models to understand computational chemistry. We have built a universal model for atoms, UMA for short. Those pieces constitute Meta Superintelligence Labs, which I oversee in addition to having a very hands-on role with TBD Lab. And then Daniel Gross helps lead up Meta Compute, which is really focused on our long-term infrastructure planning to ensure that we obviously can build up all of the GPU infrastructure and data center infrastructure necessary for this very bold endeavor. And, and so he heads that up and partners closely with us.

I mean, who did you know the best out of that group before you got into this?

Well, I knew— I've known Nat and Daniel actually for a long time. Nat was one of our very first, one of my very first angel investors at Scale. And I think like in before I completed YC, Nat had invested in Scale and had given me advice throughout the years. Daniel, I think I also met around that time, you know, very, very early on and have gotten to know them through the years. And then we also have our chief scientist, Shengxia, who helps oversee the scientific agenda across all of MSL. And he is somebody who I have gotten to know, you know, I knew before starting MSL, but you You know, since he's come in, we've gotten a lot closer.

So I'm really curious, like taking a huge step back, it's been 10 months since you kind of went into hiding. Your company, you know, completely changed and now you're at Meta. What was that experience like? What, how was the deal made? Like how did you end up going to Tahoe and talking with Zuck? Can you just walk us through what that first meeting was like?

Yeah, so it was, so I've known Mark, um, for many years now. I think, uh, even while I was running Scale, I would, He was very generous with his time and I was able to get a bunch of advice from him. And he obviously is just such an experienced founder, like the founder at this point in some sense. And so we've known each other for many years and we'd actually talked about AI before a lot of this craze because Scale, obviously we'd been working on AI since 2016. And, you know, back when it was mostly self-driving and then, you know, the various transitions of the technology. And then around a year ago, almost like literally a year ago, you know, we had a conversation where we sort of started exploring if there were a way to work more closely. And in particular, I think Mark at that point, I think he was becoming increasingly AGI-pilled and really knew that first AI was going to totally transform Meta, but also that AI was one of these, you know, almost like once in a lifetime transformative technologies. And so he was really quite focused on it and knew he wanted to bet very big on it. And at the same time, you know, and he's talked about this publicly, like Llama 4 was not on the trajectory that the company needed to be able to continue making some of these bets. And so we were sort of talking at a very high level about how could we work together more closely, what could that look like? And obviously it kind of, you know, it was one of these very open-ended questions and these very open-ended brainstorm sessions as these things often are. And then sort of landed in this, interesting zone where we figured out a way to do it in a way, in a way that was like good for scale, that was good for Meta, and where we got to work very, very closely together to build out, you know, the most important technology of our time and do so in a way that, um, where we both had conviction that it was going to be, you know, we were building something that we'd both be really proud of, which is, you know, he put out this memo of personal superintelligence also about a year ago, um, and then we went quiet, obviously. But, um, but I think that really is the sort of like North Star for both of us is We want to build this technology in a way that empowers people where, you know, as many people in the world have access to it and it's as democratized as possible, that enables everyone to express themselves, everyone to have increased agency, everyone to create and build. And that's really the world that we want to work towards.

But you're, you know, I did a really early story on you. I mean, I've known you for a long time. Yeah. Yeah.

And so, yeah, I think so.

I mean, and you're, you know, there's all this lore. You were the youngest self-made billionaire and this hot shit. And Scale was such a prominent company and you had this reputation for reading the tea leaves of where AI was going to go. I mean, like, and when I would talk to you, I mean, Scale was part of your identity. It's very different to be the founder of this very prominent company and then taking a role at a place with 80,000 employees, even if you're it's a prominent role at that company. And yeah, I mean, I, I think I was really surprised. I mean, there's a lot of money involved, so like, okay. But just sort of knowing you, it's not like I know you that well, but just knowing you as much as I did. Yeah, I was, I was like, man, that must have been a hell of a sales pitch too, because it's a big flip.

Yeah.

Yeah.

Very different. It's super different. And, you know, a lot of what What I was thinking about throughout this process is obviously I think like everyone in and around AI, I mean, progress has just happened, I think a lot faster than I did. I'd expected for a long time. Um, and, um, in the sort of like accelerating progress of these AI models, I think a few things really started to stick with me. One is, you know, um, it felt like, and I do think this is increasingly the case, that, um, those who build the AI models have, uh, just greater and greater rights, so to speak, or, or both economic and product rights to just build so much more around those models. And there were obviously all these early debates around, you know, what are, um, how does the ecosystem play out? And I think because of just how fast how fast the models are improving and how fast the research pace is. It just means that in many ways, being a place that is building the models is some of the most exciting places to be in the ecosystem. And then the second is really that so much of this next phase of technology really boils down to compute. And if you have lots and lots of compute, then you have the ability to build things and make big bets and deploy products and do things that you just can't if you don't have that compute. And I think that, um, this causes, this will cause an interesting stratification for the tech ecosystem where, you know, right now in some sense we think about, well, I think it's already changing, but, um, you know, you kind of think about all tech companies are the same, but in reality, you should think about companies with lots of compute very differently from companies without that compute. Because there's just things that companies with compute can build that those without compute just cannot. And so it, it creates this very interesting dynamic. And so part of what was very exciting about the opportunity at Meta first is that obviously Mark is, I think, very all in on AI and really has bet very big and is a very bold sort of leader and strategist. But also that, you know, this created the conditions where we're able to build with huge amounts of compute and with the right research effort and the right product efforts, we have the ability to really make a huge dent in the world.

And you guys have a ton of compute and you guys poached a lot of amazing talent. And that was— I was a part of that whole reporting frenzy at the time. It was unlike anything I've ever seen before. And it's been 10 months with many of these people. So what has What has it been like? What have the challenges been? And what has been the most exciting about having this whole new team at Meta?

Yeah, so, you know, when I kind of got to Meta, it was clear that there needed to be some reset of the efforts and some rebuild of our AI efforts to get onto the right trajectory. Because ultimately, like, you know, Lama IV was not on the same trajectory, and so we were behind the frontier. And so we needed to build a plan that would enable us to have a very, very fast velocity to be able to both catch up and hopefully exceed where the frontier is. So can you be specific?

Like, what were the problems that you found?

You know, I, I think that probably the more fundamental ones are just around, you know, a lot of the leading labs, they build the entire organizations around the premise that superintelligence is coming and it is very close. And this is a very realistic thing to believe that we can create and produce. And then you build the entire plan of the lab and the business and what you focus on around this fundamental belief. And, um, and so that was one of the first things is, is to just take superintelligence seriously and then start to rebuild all of your other assumptions around that core premise. So that's something I think was somewhat fundamental.

So you mean that they were lacking this religious conviction to all this on some level?

Yeah. And I think this is relatively common, actually. I think there's a lot of people at all the large companies who don't necessarily have this conviction, because if you think about it, you know, it's a bit of a different construction. Like a lot of the big companies, you know, they have very smart people who work on AI, but it's a little bit different from, you know, these startups where it's like, you know, these, it's like these new efforts start from scratch with this like crazy idea that superintelligence is coming. So I don't think there's a problem anymore. Like obviously I think now MSL, Meta Superintelligence Labs, is, it's in the name, built around this concept that superintelligence is coming. So, so there are a bunch of principles that we sort of laid out for the effort. And I think this sort of answers as to, you know, what were the things that we had to resolve. But one is take superintelligence seriously. Two is technical voices are loudest. Three is scientific rigor. Focus on basics and make big bets. So the, the concept of TBD and MSL broadly, when I sort of got started, was I thought about what would actually be the shape of a lab that would enable you to have incredibly fast velocity and catch up and potentially even overtake the frontier. And I came down to, there were three ways that I felt like that was possible. One is to have much higher compute per researcher. So, you know, a lot of the larger labs, they have lots of compute, but it gets spread so many different ways. And so that actually impedes the research velocity of any individual researcher. So if you build a more focused effort with a smaller team that has higher compute per researcher, you can actually make faster research progress. Two is talent density. So just, and this is like, I think, you know, I feel like human organizations always relearn this lesson, but the very small team where everyone is cracked is always going to move faster than the very large organization where responsibility is more distributed and it's more of a mélange. And the last one was on very ambitious research bets. And I think this is very well agreed upon in the industry. There clearly do exist these research bets that are very big and very risky, but if they work out, can totally change paradigms and totally shift how we build modern AI. And so, you know, in, in addition to obviously building towards very competitive frontier models, we're allocating a huge amount of our resources and compute towards these big ambitious bets because they pan out, then that gives us, you know, incredible models going forward.

All right. What do we do at Core Memory? We cover innovative, fast-moving, forward-thinking companies, which is why Core Memory is sponsored by Brex because Brex is the intelligent finance platform for many of these companies. 30,000 companies from startups to the world's largest corporations rely on Brex's technology for their finances. They've got smart corporate cards, high-yield business banking, and expense automation tools that are fantastic. I hate doing my expenses and Brex's AIs and software run right through those expenses, figure out where we're spending money, and take care of so much stuff for you so you don't have to waste your time on it yourself. Go to brex.com/corememory to learn more and just, you know, get with the program. Let's get going. Let's get out of this archaic finance software and move toward the future. CoreMemory and Brex.

Something Ashlee always talks about is how these labs are sort of leapfrogging over each other and they start to just serve the same thing. And you're talking about racing towards the frontier. I'm also thinking about, you know, the people that you guys hired was reported for these really, really wild salaries like we've never seen before. So like, how are you, you know, getting to this paradigm you speak of? Like specifically, what paradigm are you trying to achieve?

Yeah, I mean, I think there's a bunch of bold research bets, and I won't be able to go into detail on all of them, but I do think one fundamental question is, what do we care about? And we, in line with this idea of personal superintelligence, we really care about building these agents that are able to empower consumers, so empower billions and billions of people all around the world. As well as empower businesses. You know, we have Meta's kind of this incredible ecosystem. We have the billions, billions of users, which I think everyone knows about, but we also have hundreds of millions of businesses on our platforms that, you know, use Meta to run and operate their businesses. And so we really care a lot about building towards this future where we're able to build very powerful agents that empower and enable every one of both the consumers and the businesses on our platforms, um, and build kind of this new agentic ecosystem. Um, you know, we think a lot about what that looks like and what are the capabilities we need to, to build towards that. And, you know, obviously on that trajectory, there's a bunch of sub-components that are really important. Like we really have to have great agentic capabilities. Um, we have to have great coding capabilities because so much of what needs to be built is software. As you really get into it. We need to have great multimodality. So it informs a lot of the underlying bits and pieces that we need. And then we need to solve a lot of the bigger questions for long-running agents, like how do you think about the memory challenges? How do you think about building long-running agents? How do we build agents that are able to do more and more complex tasks on behalf of the users? So those are a lot of the you know, high-level pieces that we're thinking a lot about.

So if you— I mean, you're trying to get this, this superintelligence religion built within the company. There is— the way this was arranged, it's quite different to starting OpenAI or Anthropic, where, like you were talking about earlier, this is all built from the ground up and these companies have sort of an identity and it gets shaped over time. For the outsider's view, what you guys did looks far more mercenary. You know what I mean? It's like, we're going to go— you guys are out. We're going to go grab a bunch of high-priced people, bring them in. And it reminds me, you know, I just remember when Groq was starting up and it was like Elon in his Elon ways, just like, we're just going to get way more fucking compute than anybody else. And we have kind of this core team we're going to build around. And then it still felt like they caught up but then never reached that escape velocity, especially in people's minds of brand and things like that. So, I mean, it just seems like it's a hard thing to buy some of the bits that you're talking about.

Yeah, I would say this is one of, I think, the larger, I would say, like narrative violations or maybe like differences between external perception and what the day-to-day inside is like actually. I think a lot of people, you know, maybe have some of the impressions that you're talking about. A lot of that is formed because of just, you know, the reporting and what it sort of like, you know, how that sort of went down. And a lot of the reporting was overstated in various ways, but, you know, it sort of, it all sort of like bubbled up. And part of it was because we did the recruiting so quickly. Like we knew that when I got in, I knew I was like, if we want to build great models, we need to have the the team yesterday. So we had to just go and blitz it and do it very, very quickly. But I think that the culture within the lab is actually very much so a startup. And there's a bunch of things that I think have created this feeling. I think one is that it was an entirely new built team kind of like within Meta, but then the culture of the lab is like, I mean, everybody was very attracted and excited about these things I talked about. Like people joined because, you know, there was high compute per researcher, so they could make more progress than maybe they would be able to make at wherever they were before, because there was great talent density. Like people saw that it was like a truly cracked group that was pretty small, and that we were going to give them the resources and freedom to make very bold research bets. So the sort of, um, I think it's like an incorrect assumption to think that like the researchers are just money motivated or anything. And for most of them, actually, like, you know, the financial prospects of them staying wherever they were look very good as well, like looked very, very, very strong as well. So money was not, you know, those sort of like maybe how it seemed like, but the primary motivations were actually much more that they had an opportunity to build from scratch and have lots of compute, have the ability to approach their very ambitious research directions, and then do so in a group that didn't feel bloated. So I think as a result, the vibe and culture are much healthier. And I would actually say many people who visit the lab who are at one of the other labs, like, often comment that the vibe is actually, you know, reminds them of early OpenAI or early Anthropic or, you know, kind of these like more nascent stages of these other labs because in some sense we're like, you know, now 10 months old as a, as an effort.

So, and just because Mark Chen was on this podcast and brought up the soup, tobacco during these recruiting wars. Yeah. Did you— did you— is this true? Did Zuck make soup? Did you make soup to recruit people?

I don't know if we made the soup, but, but I was told it was actually made by Zuckerberg.

I don't know.

I don't know if we made this soup, but I do think it is true that we like— I think part of— part of the premise of building this lab was also that like we had to show everyone that we really, really cared about this technology and we cared about their specific research directions and what they were working on. And it was a very individualized recruiting process, but also was one like I, I mean, I'm very proud of the team that we've built. And I think that like people had to know that we were serious too. Like, I think that by default, a lot of people didn't know what to think about Meta's AI efforts, or they didn't know, you know, they didn't know that much about us in many ways. So, so it took a lot of like going to people, talking to them, explaining what we're building, explaining what we're focused on, explaining what we— why we cared about the technology, what we wanted to do with it. Like, that was very important.

I mean, I'll move on after this to not just belabor all the recruiting stuff, but, you know, same thing. When you were at SCaLE, you were like the everyone would call you the Switzerland of AI. And then I just remember you knew everybody and you were, you were in the center of things. And, and then it feels like some of this came with a personal cost. I mean, you and Sam used to be flatmates and, and I texted Sam about you coming on the show. He did not have flattering things to say. And so, you know, it seems like some of this, it definitely must have come at a personal cost to you.

Yeah, I think that, you know, I think some of this is unfortunate. And my honest expectation is that, like, as we get closer and closer to superintelligence, like, my hope genuinely as a human that all the sort of animosities that exist between various people in this industry, which is, you know, very topical obviously right now with other things that are happening. But I think that my real hope is that all these animosities subside over time and then people sort of come together and realize, you know, this, we are building this incredibly important technology and it's important for all of us to be really thoughtful about that as we build it. And so, yeah, I think, and one of the things that I feel like is a responsibility of mine, honestly, is to ensure that, you know, we, the technology that we develop and the ways we deploy it are as thoughtful as possible.

Some of the stuff you talked about, like in terms of recruiting, it was trusting that we have a mission and it's not, you know, just, just products like you can go for your own research mission. And you're also quite young. We're almost the same age, which is very funny. And I'm not— and I'm not a billionaire. But Jan had said in the press shortly after he left that you were like young and inexperienced and more people are going to leave. So I'm curious, like, how has that boded for you a leader at this huge company and you're quite young. Like, what was reading that like? Have you talked to him?

Yeah, I saw him in India like a couple of weeks after that. And, and I mean, Jan is, is a notable, very outspoken person. And I think everyone always, always knows what Jan is thinking. So, but, but no, I think, I think, you know, he obviously said what he said and I saw him in India. He congratulated us on the Muse Spark launch. I saw you guys.

Atching things up on X.

Yeah, I mean, truly, I do exactly what I just said before. I think all personal animosities, I think as we get closer and closer to superintelligence, will—

Seems like it's getting worse. Does it?

Does. Yeah, maybe it gets worse and then it gets better. But yeah, no, I think that I have a lot of conviction in how we've set up MSL and the research efforts that we have and the progress that we're making. And I'm excited to show the world the incredible work that our researchers are doing and the progress we're making.

I also don't want to belabor the point, but is that a challenge you continue to face, like people thinking you're too young and inexperienced to lead such a huge effort at Meta?

You also get the knock that you're not an engineer.

Oh, yes.

That is definitely not true. Like, I Once upon a time, I was a software engineer in Silicon Valley, but— Like, doesn't this piss you off? Yeah, I mean, I honestly— I mean, so on the age thing, I mean, like, people have said this my whole time in Silicon Valley. So, you know, to some extent, like, I actually, like, I almost don't even think about it anymore because it's just— it's always there. Um, but yeah, no, I think that, I think that there are always, I think actually many people in AI always, there's like various mischaracterizations about them or, um, there's like, you know, people always say shit and, and there's sort of always, um, uh, you know, what's out there is never always correct and it can be frustrating, but I choose to just channel it into the work that we're doing and what we put out there. And, um, again, I am really proud of MuseSpark. I'm even more excited about the models that we have cooking and the products that we have cooking. So I think like in the long arc of, you know, in the long arc, all of this will play out just fine.

You were a math Olympiad, right?

I mean, you were—

you—

I mean, which if for people who don't know, I mean, in my experience covering Silicon Valley, it's like everyone who does well in these competitions tends to be quite proficient at coding, engineering, and thinking about these problems. I will say, I mean, over time you did get reputation among some circles as kind of like a salesman at scale a bit and enjoying life and things like that. I mean, so I did wonder when you took that job if it was going to be kind of like harder to boss people around or just, just quite different.

I don't know.

Yeah, by the way, I actually, in general, in terms of my management philosophy for MSL, is not to boss people around. I think that like, you know, this great Steve Jobs quote, which is, most companies hire people and tell them what to do, but we hire people for them to tell us what to do. And I think, you know, that is like pretty core to the entire thesis of TBD and MSL and how we've built it is that we're going to we were going to have, we're going to hire brilliant researchers and, you know, create the best environment for them to do the work of their careers and the work of their lives. Um, so, uh, so I think, you know, so long story short, I'm not trying to boss anyone around actually. I'm trying to create the best environment for researchers to do incredible work.

All right, pod friends, I am here to tell you about SendCutSend. They are a manufacturing phenomenon. Whether you're working on your own car or you're a giant automotive company, a rocket maker, if you need metal cut, machined, or bent, you're going to SendCutSend. They'll do it for you right, and they will do it for you fast. Go to sendcutsend.com/corememory for a 15% discount on whatever you're trying to make. We love SendCutSend. Go visit them. Today on MuseSpark for a minute. I mean, just as I was reading everything over the last couple of days and playing with the model a bit, I'm trying— I was just trying to wrap— we were trying to wrap our head around exactly where you guys see it in terms of what you're trying to achieve at Meta. It seemed like on the benchmarks, you know, did well on some, was behind the models on others. It seemed like you guys were emphasizing this and apologies if I get the technical stuff wrong, you can set me right, but it seemed like you were emphasizing that you felt like you guys had some efficiency gains that some of the other models maybe didn't have. And then you're doing this crazy thing with the 16 agents working on it. I was playing with that last night. So it felt like to me, you guys think you've picked a couple technical directions where maybe you're ahead of people, but then I went through all your tweets on X last night. There were people who would compliment you. There were other ones that would take a dig at you and you would say, you know, just wait for the next thing. And so, yeah, I guess we were trying to figure out— it didn't seem like you guys were like planting a flag and being like, you know, we have conquered everything with this model.

Yeah, no, by no means. I think that, you know, the What we did is, over the past 9 months, we rebuilt a lot of the stack and a lot of the research. So we rebuilt our pre-training stack, we rebuilt our RL stack, we rebuilt a lot of the science and did a lot of work on data. So in many ways, what's been happening over the past 9 months is really like a full-on renovation, as it were, for the core research stack. MuseSpark is kind of an early data point on that scaling ladder, but it is not, you know, in some ways, like MuseSpark is kind of like the entrée or the appetizer, I guess appetizer, entrée in French, but it's kind of like the appetizer for what we're building. But we are in development with larger models and we expect the larger models to be, you know, we're much more excited about the larger models than we are even about MuseSpark. But it was an important data point we thought to put out there into the world because it was a, you know, the entire program that we've built is developed around predictable scaling. So, and we see the scaling, I think we talked about this in the blog post, on many axes. We see, you know, very consistent pre-training scaling and predictable pre-training scaling. We see predictable RL scaling and scaling reinforcement learning. We see predictable test time scaling. And a lot of what you just talked about, the contemplating mode, is we're also seeing very exciting results in multi-agent scaling. And so everything about our program is built to continue scaling as we go. And so, you know, MuseSpark was like this early data point on our scaling trajectory, but the next data point we're like a lot more excited about, and the data point after that we're even more excited about. So I think we're excited to show people this sort of like next rung up on our overall scaling efforts. And for MuseSpark specifically, I think it, you know, if anything, it's actually the overall end-to-end performance ended up being quite a bit better than we expected. And it had a bunch of emergent capabilities and behaviors that we were pretty excited about, that we found in training. For example, some of its abilities in agentic visual coding, being able to produce websites or being able to produce games, some of these capabilities actually emerged from the fact that it is both a pretty strong agentic model but also pretty strong at multimodality. So there were a lot of things that we're very excited about this model. And so we put it out there. We think for most consumer use cases, actually a very good model. And it is quite competitive with the other models. MuseSpark, as we deployed it, is not yet competitive on, you know, agentic coding. And so those are capabilities that we're working on and we're building towards for the next set of models. But yeah, I would expect the next model that we produce to be better overall than MuseSpark, and that's something we're pretty excited about. But even MuseSpark as we released it, which I think we, to be clear, wanted to set the clear expectations, like we didn't think it was going to be a state-of-the-art model across the board, but it is a very good model. And we think a lot of people and a lot of users who tried it out experienced that.

I'm curious, what was the holdout for releasing like a frontier model? What do you still need in order to hit all of those benchmarks and blow it out of the water?

I mean, the one-word answer is just scaling. Like, uh, MuseSpark, um, is, uh, kind of as I mentioned, it's early on the ladder and we have very strong predictability. So we know if we scale this model up, like what performance to expect from that increased model size. And we expect the upcoming models to just be able to perform much better across the board. When does that happen? Coming months, you know.

Coming months.

Wow.

So like a year into the whole endeavor.

Well, as I mentioned, we built the whole program so that we would be able to move very, very fast. So there was a time period where we had to rebuild all the foundations and rebuild everything. But now we're in the period where we're going to be in fast scaling mode.

What do you feel like you're doing technically that's different to everybody else?

Well, one of the things that we found, kind of as I mentioned, MuseSpark performed very well in some ways even better than we originally expected, especially like a year ago. And when we went back and kind of like analyzed and tried to understand why does it perform so well, we think a lot of it comes down to just having built a very, very clean stack from scratch and having had the ability in this rebuild process to do everything quote unquote the right way. So We really had this, in some ways, this luxury, the ability to build a very clean pre-training stack and very clean RL stack and to do everything in the right way by the experts who know exactly how to build these systems. That was able to meaningfully accelerate both our trajectory, but also I think it really shows in the model.

Before I do these interviews, I throw you and the model and everything into all the AI systems and sort of get them to poke around. I mean, and the thing that kept coming back was this token efficiency. I mean, like, is this something you guys feel like you've figured out or this was just a happy accident with MuseSpark? It seemed like on some of these benchmarks you were just doing it with far less effort than the other models.

Yeah, yeah, yeah. No, this was an exciting result for us. Yeah, I think like on artificial analysis, for example, it used to achieve, you know, pretty similar results, like many fewer tokens than, let's say, the models from some of the other labs. Yeah, we think this is a testament, I think, actually to this whole clean stack, which is that I think, you know, I can't say for certain, but one of the reasons why some of the other models maybe require a lot more tokens could be that there's some level of fundamental inefficiency at another part of the stack that gets patched by enabling models to think longer. We were pretty impressed and excited about the token sufficiency that we found. Frankly, we think as we keep scaling the models and continue scaling overall, we think that bodes really well for the future performance of our models.

Spark was really good at vision benchmarks and that efficiency and that that vision, uh, you know, expertise seemed like it would be really important for your guys' hardware endeavors. And you've talked before about like a constellation of these AI products that can see what you see and hear what you hear. Can you talk a little bit more about how that fits into your guys' broader vision for serving AI?

Yeah, 100%. I mean, I think, um, one of the things that's very exciting about Meta overall is that the, um, Ray-Ban Metas, the glasses, have been this hit product and we've sold millions of copies and we have some big fans.

The biggest fan.

I do love them.

Um, but it is like, um, it is so, it is like a very exciting direction for all these devices to think about what, what does your relationship with technology look like if it really can kind of, um, fade into the background a little bit and be a lot more contextual. And as you mentioned, see what you see, hear what you hear, and be much more intelligent and helpful in the moments where you need it. And also capture all this context about what is happening in your life and what are the things that really matter and what are the things that it should pay attention to. And so we really see a future world in line with personal superintelligence where you have, as you mentioned, this constellation of devices that all are there to help capture context, are all there to enable the technology to kind of like fade away a little bit and help you, help you get like very intelligent and valuable insights from agents, like proactive insights, or, you know, you'll mention something and then the agent will go off and do some research or, you know, take actions for you, or, you know, it can just kind of, be this like super intelligent sidekick that makes everything in your life better.

I feel like there's some kind of problem that you guys have though, because I love these glasses. I use them all the time. I do it for our video stuff. And then I actually— I like to take phone calls on them and then pretty much like run our entire business in WhatsApp. I refuse to use Slack. I travel so much that WhatsApp just got like embedded into my life. In full confession, I don't think I've like ever used Meta's AI. Agents or anything until, until you were coming on the show and I wanted to see what it was. I always go out to Claude, I go, I go out to ChatGPT to do this work. Um, and, and like, I, I mean, I, I saw the AI agent button on WhatsApp kind of like for the first time today. Just, I'm sure it's been— I know it's been sitting there the whole time, you know. So I, I don't know. I mean, I am like I am in your world and kind of like didn't even see it there. There's just, maybe I can't be unique in this.

Yeah. Well, one of the things is we knew we needed to have great models and great products before we really pushed for, you know, tighter integration across the, across our entire ecosystem. So, um, in many ways, like we were, we've been waiting to have great models that can then enable most of the consumer use cases that we really care about. And now I think we're at a point in a moment which is very exciting, which is that our models are pretty good. We're pretty excited about them. We have better models on the way. And so now we're going to undergo the sort of like process to do a lot of large-scale integration of all of the you know, the family of apps that we have with our AI and integrate our business products with our AI and go through this, this evolution towards like knitting together almost all the pieces of our ecosystem together with our AI. I think, you know, to some extent you've seen what that looks like for Gemini over the past few years. And we're excited to go through— at Meet—

It's the same thing for me, though. I'm— I also run our business in Google and, and I, mostly play with Gemini to see— I just feel like in consumers' heads, I'm curious how you think this plays out because I feel like you've got OpenAI and Anthropic in this one world where ChatGPT is such a strong consumer brand. That is what people think of as AI. And then Claude has been super dominant in coding, in business. You guys, Google, you're sort of like asking people to run into AI as part of all these services that you have. And I don't know, I don't think we've, I don't know if we've ever seen a competition quite like this where, and then there's, there's X as well, where you have these, I always, I'm old, so I go back to like word processing days, you know what I mean? It's like, it's like, are you gonna, what are you gonna use? You're gonna use Microsoft Word. People settle on like a thing or in the browser wars it was like, there are only two: Internet Explorer and Netscape. And then that was, that was sort of like the rest of history for a long time. And so I'm— I don't know, do you see— I feel like these two groups have different challenges, and it's really not— I sort of feel like consumers, most people are still gonna pick— like, I do my AI on ChatGPT.

I just think we're so early. Um, like, I think that Um, you know, I think it's funny because I've reflected on this. Like, if you were sitting here, um, you know, a year ago, if you were to have this conversation, you would— we would always just say like, oh well, you know, OpenAI and ChatGPT have, you know, they've won on consumer already. They have the, they have, they have the biggest business, you know, they're just going to run away with the whole thing. And then fast forward a year. Anthropic has had this breakout success of Claude Code, which was, you know, somewhat foreseeable, but not super predictable at the time and, um, has overtaken them in revenue. And at the same time, Gemini has distributed quite a lot and actually has eaten a lot of, uh, consumer market share from the rest of the ecosystem, including ChatGPT. And so I think that we are in this phase of AI, which is just incredibly, incredibly dynamic. And I think it's very hard to say at any one moment that, you know, we're in the endgame because everything is just, you know, I think there's going to be so many new products built both for consumers and for developers and for businesses that haven't been invented yet that will each potentially be even bigger than the ones that we've had before. Like, I think it's pretty fascinating to me that ChatGPT obviously was this incredible hit and it was this like, you know, felt this, it was the fastest growing product and business that the world had seen till that point. And then Claude Code again is this incredible hit. It's the fastest growing business anyone has ever seen., until now. And I think that this is a really, there's something, this is a statement about something intrinsic about AI, which is as AI gets to new levels of intelligence and capability and, and overall performance, it just unlocks these new form factors that each will be this like, you know, incredible new wave of technology washing onto sort of like humanity's shores to some extent. So I think that Long story short, I think the next wave will be even bigger and the wave after that will be even bigger. And, you know, we're nowhere near the end. Like, there's going to be many more exciting new product paradigms that we'll see in the future.

I think the product overhang question is real. Like, we have these incredible models. What can we make that consumers actually want to use? But I'm also curious how you square this sort of sentiment of the average consumer and AI, because I'm in my 20s, I am not only in tech, and I see crazy stuff posted Instagram stories about how much people hate AI. The sentiment seems to be in the toilet. And then you guys have these billions of users and you're serving your AI as these buttons. And, you know, I'm curious how you square that sentiment, what you see on the consumer side of how they're, you know, receiving your technology that you guys are building.

Yeah, I mean, AI definitely has like a— sentiment is very low, to say the least. And I think this comes down to, um, on some fundamental level, we haven't yet demonstrated in, in a very real way how this is actually a tool for personal empowerment or personal agency, or how it just makes people's lives a lot better. Like, I think people's experience right now is that it, um, can be really helpful and it makes your life quite a bit better, but it's not this it's not overwhelmingly better versus I think for a lot of developers, I think their lives have actually totally changed. And I think most developers, they have very positive sentiment, maybe somewhat mixed, but very positive sentiment towards AI because they're now able to do things that they were just unable to do before and they can build so many more things faster and they can just build entire projects over a weekend. And it's just this incredible enabler of personal agency. And that moment hasn't happened for everyone else in the world yet. Like so far, we haven't yet given every person what the equivalent of Claude Code is that would enable them to, you know, do the projects they always wish they, they've, that they've always had in like, you know, the back of their mind or make their life way better or all of a sudden enable them to accomplish their goals. Like that hasn't happened yet. And same thing even for small businesses, like, you know, small business owners and entrepreneurs haven't yet had that full experience yet. So that's really what we're building towards at Meta is what does it look like to give very powerful agents to all of our consumers and all the small businesses in the world? And what does that look like if you're actually able to nail it in the form of like a huge increase in individual agency?

That would be a crazy thing to nail because if you go to a small town in anywhere in America and go to that restaurant's website. I mean, it hasn't been updated since 2002. So giving everyone, you know, multi-agent architecture products sounds like a huge leap.

And I mean, to Kylie's, like, earlier question, I mean, look, I— there's things that I like that Meta does and there's things I don't like. I think there's huge swaths of the public that view the company quite cynically. It just feels like you guys do have a— it's like you said, AI in general, not always most beloved thing at the moment. I do feel like the bar is higher for you guys to like, to get people to trust you.

Yeah, 100%. But I think that, um, like, again, I think if we think about what is the best thing that we can do, it's really we should build the best possible products that we think are genuinely amazing for those who use them. Like, I think, I think we can build products that can transform the lives of most small business owners. And we have, again, hundreds of millions of small businesses all around the world that are on Meta. You know, a bunch of them use WhatsApp to run their businesses like you do. A bunch of them have Facebook pages or Instagram pages. A bunch of them, you know, use our advertising solution. So I think there's this, there's an opportunity that exists there that, like, on some level only we have, because only we have, again, billions of users, billions of people around the world who use our products, and hundreds of millions of small businesses. And one of the ideas that gets me personally really excited is, you know, if you're able to build agents for both sides of this, like, ecosystem, you know, for all the consumers as well as all the small businesses, then what does that look like when you enable the mechanism for those agents to work together and collaborate with one another? And so, you know, Dario always talks about a country of geniuses in a data center. I think we're excited about building an economy of agents in a data center. Like, what does that actually— if you fundamentally change how supply and demand work in the economy and it's like mediated by agents, I think that could be like, there's like very, very exciting things that we can build towards. And you're totally right that that has to be done in lockstep with ensuring that we have social permission and that people really care about— people see that we care about how these things are deployed and that we're genuinely making people's lives better as a result.

I mean, one place you guys had won clear hearts and minds was by making these things open source. And I'm an old open source fan and I kind of like believe in it philosophically. And yeah, I mean, so where, where are we going with, with that since Muse Spark?

Yeah, yeah, we're, yeah. So, so, um, you know, models are a lot more powerful than they were even back in the Lomini, even though it's so, it's so recent. And so, um, one of the, and one of the things that is very important to me is safety, um, for these models. And so, um, one of the things that we instituted as part of our Advanced AI Scaling framework was, you know, we have to take very seriously when the models that we develop trigger various safety guardrails, especially around bio, chem, cyber capabilities and loss of control. So MuseSpark, in our testing, did trigger some of those safety checks that we did. And we detailed all this in the preparedness report of MuseSpark that we published. And so, as a result, MuseSpark in its current form is not suitable for open sourcing, but we are working on developing versions of the model that are suitable to be open sourced. Literally, a meeting I had earlier today was actually to review the progress on this. So we're excited to actually continue supporting the open source ecosystem and developing open source models. And I expect that You know, again, we'll have more to share on that, you know, in the coming months. But that's an exciting milestone for us as well.

So, okay, you're really going to stick with it because, you know, you did the— I always appreciated that you guys did the Open Compute Project. You're in Sun Microsystems' old building. Again, I'm just a history nerd. They were such a champion of open source software and were always kind of like this foil to what Microsoft had built in the world. And I kind of think it's important and I mean, so it sounds like you're saying for you guys are committing that that's like still going to be something that Meta does that's quite different to most of your competitors. Yeah.

I mean, I think I've said this a bunch of times, like we're— we, we will continue open sourcing models, but we also have to take safety seriously. And so we will— our most powerful models, we have to consider whether or not they're safe enough to be open sourced.

Okay. What's— and then, I mean, there's another If you read the stories about your tenure at Meta, I mean, one thing that drops out is this, you know, I think it was the New York Times or someone did this story on Alex and Zuck see the world one way. They're very research forward and want the best model in the world. And Boz and Chris Cox are more focused on products. And Meta is this company that has to serve billions of users and do so as cheaply as possible. You could argue that, you know, and doesn't charge for its models today. I'm sure you probably expected we would ask some question along these lines, but, but so where are all you guys philosophically? And, and are you— is there all this division about what direction to take your AI strategy?

Yeah, I mean, first off, the one thing this job has taught me is the bar for journalistic reporting at major outlets is, you know, the line between gossip and reporting is like, is remarkably thin.

But you guys weren't fighting like crazy?

No, I don't think so. I mean, I think in general, like, yeah, no, I think, I think we're very— we're all very aligned on what, what is important. Like, we all know we need to have very advanced models, both to support our core business and to build the existing apps and products and services that we, we have for our users and our small businesses to be the best version they can be in the world. Like we were, we were working on, we've been working on business agents since long before I got to Meta and those require the best models possible. But we also know that, so we all know we need to build the best possible models and we all know that We need to then integrate those into our business and utilize these models to build products and services that are, um, that are incredible for our consumers and for, um, the businesses on our platform. So I think that there's, yeah, there's not, there's no real disagreement. I mean, like, I think like any company, you know, there's always like, you know, we debate the points deeply, like we talk about them and we, think through the implications and we talk about— we want to make sure everyone has like the ability to chime in on these issues. But there's no like major beef, as it were.

So you think that was just total bullshit?

I think so. Yeah, I really do think so.

On the Meta stuff, I mean, right before you got— you made this transition from Scale to Meta, you were out doing, you were doing a lot of stuff in DC. You were flagging the danger of China in, in the whole AI race. When I saw you guys do that deal, I was trying to square that in my head because, you know, I know they were, they were putting offices in Singapore and creating some of this distance. I don't know. It just, it, it seemed like the type of situation where, um, getting much closer with a Chinese startup and, and a, company with the resources like Meta, you know, it seemed a little different to me than what you'd been saying rhetorically. Does that make sense?

Yeah. I mean, obviously the whole Meta situation is pretty complicated. It's hard to—

super complicated.

And I unfortunately can't really go into any, any real detail. But I think what I will say is like, I think the The menace, I think in general, when you think about these questions of geopolitics and, and, and whatnot, you always have to separate the sort of the people from the state in some sense. So I think that like, you know, my parents are from China. I think there's lots of very incredible, very talented people who are Chinese. And, you know, many of them Some of them move to Singapore, some of them move to the US, some of them move elsewhere in the world. And I think there are— many of them are incredibly talented, and I feel lucky when I get to work with them. And that is like separate from my overall beliefs on, you know, the Chinese Communist Party and the, the sort of like actions they're taking and what that means for the— what the— how the United States should be thinking about our overall strategy as a country. So I think that, yeah, I think it's important to draw a distinction between these two. And I think there's sometimes a real pull inside Silicon Valley tech to be somewhat unnuanced about this and to just really like, like anything that involves China, you know, we, we sort of like lump together. And we sort of like, you know, Twitter or X in particular is like particularly unnuanced about this.

So nobody's nuanced about anything.

Not nuanced about anything. But I think that like, yeah, I think to me this, you know, whether or not there's amazing people who happen to have been born in China, who we would love to work with, is like totally independent from what I believe about like US versus China overall geopolitics.

And you can't comment because, I mean, if you— it looks like China's shut the deal down. If you can't comment on it, that means there's still machinations at play, like something, something could still happen.

Um, I just can't comment on it. Yeah.

Touching on that sentiment, what was that newspaper ad you put out about AI war? Was that in the New York Times, that full-page ad about AI and war and we need to take this quite seriously. Do you remember while you were at SCaLE?

Yeah. And this goes back to, I mean, I think, yeah, I mean, zooming way out, I think, I think that was at a moment that felt very critical, which was that I felt it was very important at that time for the United States government to understand that AI was going to enable a large step change in what it meant for national security and defending our country and defending our citizens. And I think in some ways what we've seen since then, which is Mythos and other pretty meaningful events in terms of the importance of AI for national security, have proven that to be very correct. And that was really at a moment where it was important for— I think there's pretty clear evidence that the Chinese Communist Party and the PLA have always taken AI extremely seriously as a technology that has very far-reaching implications for national security. And, um, and that was at a moment where it was very important for us in the United States to take that as seriously. And I think the US government today is taking AI very, very seriously as it pertains to national security.. And I think a lot of what we're seeing, um, is a demonstration that, like, the plea that I had and that many other people, um, in the tech ecosystem and in DC had have been really internalized, and that we are really, um, thinking quite deeply about this today.

So you don't think Anthropic are overdoomers?

Um, I mean, that's That's a complicated question. I think it depends on which part. I think that Anthropic are, well, yeah, it depends on which part, but I think on the whole, I think that whenever you listen to people in the AI industry talk about AI, it's important to, I think, separate maybe the exact things that they're saying versus what's the sort of core message they're trying to get across. And I think that some of the overall message from Anthropic, which I think is quite fair, is that these models already are very, very capable and very, very powerful, and they're only going to be more capable and more powerful going into the future. And we obviously think that this could be this incredible boon for human society. Like, I would not be working on this if I didn't believe that this could be so, so positive for humanity. Um, you know, some of the areas that we care a lot about are scientific discovery and health. One of the things that we have a whole effort on is health superintelligence. Like, I think this can be an incredibly positive technology, but it's also very important to factor in what are the risks of the technology and make sure that we're taking those seriously.

I want to jump into Ashlee Vance's favorite topic with the time we have left, which is you guys just bought a humanoid robotics startup. Can you tell us more about those ambitions and like whatever you can tell us about what you're hoping to build and use? I imagine these models to bring into the real world.

Yeah, 100%. I mean, I think, um, what was it called? It was Assured Robot Intelligence.

Okay. And they made hardware?

Uh, no, they do not make hardware. They made, um, AI for various hardware targets.

Okay.

Yeah.

Um, Yeah, I think that, again, taking a step back, if you take superintelligence seriously and you take very seriously this premise that we will have very, very powerful intelligent systems, then you kind of realize we're going to have digital superintelligence. We're going to have the current form of superintelligence that we're targeting, but then not long thereafter, physical superintelligence becomes really, really important and very critical. And so if you have short timelines, which we do, that, you know, very powerful AI capabilities are coming, it just means that you have to take robotics capabilities and physical intelligence very seriously as something that you need to be building towards in the span of years. And so that's kind of the overall core premise is that, you know, this is Physical intelligence and robotic capabilities are very much so on the natural continuum of what your roadmap has to be if you want to build superintelligence as a company. And I think there will be all sorts of ways that we apply this technology over time. I think that we will use the technology to accelerate scientific discovery. I think we'll use the technology to figure out how to make accelerate, you know, goods manufacturing. I think we'll also use it to figure out how to make people's lives better in sort of a more, like, local sense. Like, what does it look like for robots to make all of our lives way easier? So I think there's, there's, there's obviously, like, a near infinite number of applications of robotic technology. But, but the other key part here is, we really think that In the same way that digital superintelligence benefits from scaling, so does robotic intelligence. And so given that we are building the compute infrastructure to enable just massive scaling of these systems and these models, it would almost be a waste if we didn't integrate that with efforts in world modeling and physical intelligence.

It feels like something you guys are really trying to own, the hardware, bringing the models into the real world. But this whole time I am unfortunately thinking about the metaverse no-legs situation and what critics might think about bringing humanoids from Meta into the world. Like, what makes you guys right to do this? What have you learned that makes you feel like we can, we can do this and change that sort of reputation?

I think that like ultimately, you know, there's a world where we could be so scarred by what has happened in the past that we just like didn't get out of bed in the morning and we just sort of, you know, stayed home. But I think that what we like, we are so excited and incredibly inspired by the potential of the technology and also just building amazing products. And I generally subscribe to the belief like if we build great products very thoughtfully and are very, take a lot of care in how we deploy them and how we roll them out to the world, I think that, you know, I think that people will be excited about those.

All right. I'm just looking at the time. We're going to lose you in a second. Can we go rapid fire real quick? Mango model, alive, dead?

The mangoes are alive and kicking. They're—

they're always fruit themed.

I know. I'm wondering how do mangoes grow? Do they grow on trees or— I was going to say on the vine, but anyway, alive and well.

Okay, because my nerds in AI land were telling me there's things afoot with the Mango model.

Another AI app?

This is what I'm talking about. There's just so much— there's so many like spurious rumors that are not grounded in any reality. But okay, I mean, I feel like— I mean, as much as, you know, we are self-important, like we get a fraction that the other labs get. So, um, you know, I have a lot of, I think, empathy for The drama of it all, the like rumor mill and what that feels like.

So Nat Friedman and Daniel were two of the biggest investors in John Carmack's AI effort. He's been very quiet. He obviously used to work at Meta. Do you talk to him? Is there any chance of getting the band back together? Um, uh, do you know what he's doing?

I actually don't really know what he's doing. I don't know if anyone really knows what he's doing. He's obviously like one of the GOAT programmers, so I respect him a huge amount.

I interviewed Priscilla Chan. CZI is investing billions and billions of dollars into science and biotech. I don't know, it seems you guys were scoring really high on these like health benchmarks and, and SAC obviously has an interest there as well. So it just seems like, man, you guys would have resources to tap into that other folks wouldn't. Is that in the cards? Is that— I don't know if those things have to be separate or—

No, no, yeah, we're going to be collaborating closely with CZI to build the best. As I mentioned, health superintelligence is so important for us. We think that there's just so much potential in enabling equal access all around the world to very powerful health AI systems. And That's one of the things I think is, you know, we uniquely can deliver actually to sort of billions of people, billions, billions of people all around the world because they use a lot of our products already every day. So yeah, it's, it's a very exciting and, and important initiative for us.

Tell us, like, I mean, you're being coy on some of the technical stuff. I know you don't want to speak about the new models, but tell us, like, tell us, like, one thing that you guys feel like you're really doing different or ahead. Of everybody else on something that, that you think you've figured out?

Well, you know, I, you know, we never, never want to— you never want to like— you always want to show, not tell, you know. So, but, but I think that, you know, we are really excited about the models that are cooking right now. We were, I think, We're really excited about the results that we're seeing from scaling our models. And we, we think everyone's going to be pretty excited and we expect them to be state of the art in, in some of the areas that we're really, really focused on.

And then, I mean, just kind of the last one, I mean, philosophically, do you feel like you have a different approach to all the other frontier labs or even you? I feel like you're a bit of a mystery. I don't know. It's like I kind of know where Dario stands. Definitely know where Elon stands. Feel like I have a handle from time to time on Sam. Dennis is very science-focused. You're running this massive, massive lab, and I'm not sure I, like, really know, you know, what you think about this, this technology that's being unleashed on the world.

Yeah.

A few things worth, worth saying. I think one is Well, first, I do, I'm a huge believer in the technology in the sense that I do believe we're going to have very, very powerful AI systems. And we're building towards that, but so is everyone else that you mentioned. We're all building towards true superintelligence. And I think first off, table stakes, that we have to take safety incredibly seriously as a topic. There is, I think there's no such thing as building superintelligence without being very, very thoughtful and thinking very seriously about what are all the safety risks associated with developing and deploying this technology and ensuring that you are able to mitigate as many of those as possible and have strategies and research methods to be able to develop those in a thoughtful way, develop the models in a thoughtful way. So I think this is an area where I agree with, you know, some, a subset of the people that, that you have mentioned, which is that safety is an incredibly, incredibly important effort. And, you know, you've seen this in terms of MSL, you know, we published a very detailed preparedness report for Muse Spark, more detailed than Meta has historically. And that's due to a commitment that we have towards that. I think that where we specifically as Meta want to, what we want to build towards is this world of personal superintelligence. It is deployed very, very widely and broadly. Billions and billions of people all around the world have access to it. It's this, in many ways, this democratized technology and capability that everyone has equal access to. And then that enables this era of just incredible human abundance. Like, we all have tools to, to, you know, of great agency. We have the ability to accomplish so much more than any human has ever been able to accomplish in the past. And we're sort of augmented by this incredible sort of like agent economy that has making incredible progress on scientific discovery and making great advancements in health and I think that one of the things that I always think to myself is, how can we build paradise on Earth? And I think that superintelligence is a key milestone to get there. And then one last thing that, you know, some people may kill me for mentioning this, but I do think one topic that is increasingly important that I think a lot about, and maybe it does express some of core philosophically what I believe is, you know, there's this kind of hot topic these days of model welfare, which is—

I love talking about this.

Which is, you know, is it important for us to treat models well?

You guys got a philosopher, right?

Yeah, yeah, exactly. Yeah. And yeah, is it important for us to treat models well and to think about, you know, whether or not models have moral weight and these sort of more, I think in some ways they feel heady, but also I think they do change your actions on a day-to-day basis, given that so many of us are using AI so much. And I think it's very important. I mean, I think in a world where obviously we care, most humans care about how we treat many other living things like plants or animals or certainly other humans. I think in that world, it really does make sense for us to be thoughtful about how we treat the models. And, you know, one of the things that we really care about is how can we develop the models and deploy the models in a way that is thoughtful about their subjective feeling through it. And, you know, it's interesting, there's been research. You are able to measure a lot of this on, you know, there are ways to measure the sort of subjective experience of the models. And—

Elios does that.

Yeah. So anyway, so this is, I think, a very important topic. I think actually nobody is talking about it enough from my perspective, given how much we are now, especially in tech, all using these models and they are really our work partners in a very deep way. And I think it's quite important.

I remember talking to Richard Sutton about this. He seems pretty serious. I mean, you're kind of a sci-fi had then. I've listened to a couple of your other interviews where you're talking about, um, you're just really dialed in on Neuralink and, and what BCIs could mean to the future of humanity. And then, I mean, so I'm just getting the sense you're, you're kind of—

yeah, my favorite things to do are read sci-fi and walk in the woods.

So, yeah, well, this is what always threw me off. This is why I would text you about country music, because if I'm honest, like, you did not strike me as like the country, country music type. I had a different picture of you in my, in my head. So, okay, so you're mixing— you're of these two worlds, nature and, and like our transhumanist future.

Well, I do think— yeah, I mean, on the topic, I do think like, uh, if you were to think about which technologies are like critical path for humanity, BCI is definitely one of them. It's like, um, superintelligence, obviously, robotics for sure, and brain-computer interfaces. Like, these are the critical path areas. And if you think about what are the, what are the things that we work on today that will scale to literally infinity far into the future, it's, you know, energy, compute, and robots.

So there's like one guy who's betting on those bigger than everyone else. That would be Elon. And then I feel like China. And then I feel like Meta on some of these fronts taking some of like, especially around BCI, like the motor neuron stuff and everything, you know, very like more bets than I see some of the other AI companies. But yeah, I mean, don't you, so if that's what you believe, I would say Elon's a little more all in on robotics, energy, BCI than anyone else. Don't you have to, does that mean you guys, or Meta's ratcheting?

I think the details really matter here. Like you do, I think you have to build these in stages. You do have to build superintelligence. That is a very important prerequisite to being able to be in a position to build the rest. And one area in which my opinions differ from Elon's, I think a lot of people's opinions differ from Elon's, is I do think research is incredibly important and that building superintelligence is fundamentally a research activity. On some level, We are in the fog of war of knowledge, and we are trying— you do experiments to poke and prod in this fog of war to understand what it would mean to build superintelligence. And that is research. And so I think sequencing really matters. I think how you approach it over time really matters. I think being thoughtful about the milestones over time matters. But yeah, I mean, we're doing One of the research areas in FAIR is called Tribe. We had a milestone in the past year, Tribe V2, around building foundation models for brain prediction. One of the cool results that we found was good zero-shot generalization. So without even knowing who you are or having any data about your brain, we can do a reasonable job of predicting how your brain would respond to various images or videos or audio. Yeah, I think, I think we're making important bets in many of the key areas.

Okay.

I think we're going to— we'll set you free. You haven't talked like this since you took this job. We dragged you around. I never really do this, but open floor. If there's, you know, if there's something we didn't hit that you— I just feel like you haven't, like, kind of had a chance. I mean, I guess you could do it whenever you want it, but to tell the world or whatever you want coming out of this experience so far, or maybe we hit hit everything. I don't know.

Yeah, no, I think, I think we talked about, um, a lot of the key things. I mean, ultimately what we really are building towards at Meta is like, how do you build a world that has, um, massive amount of personal empowerment? So each individual person or small business or entrepreneur has just incredible tools to empower them to build more than any human has been able to ever build in the literally in the history of humanity. That's like this incredibly exciting concept for us. And then how do you do so in a way that also, alongside all the humans in the economy, you empower this economy of agents that is there to sort of facilitate and optimize and enable incredible progress alongside the humans. And the economy of agents in a data center is just as clear exciting outcome that we're excited to be able to create in the world. And I think it's actually quite tractable for us to develop. And then all along the way, drive incredible scientific progress, dramatically improve health outcomes through health superintelligence. There's a lot of things that we're really fundamentally excited about on this journey.

Okay, well, thank you. Thanks so much for making time.

Thank you.

It's nice to see you again.

See you again in another year.

Out of the world. No, it is cool to see you again. So thanks. Thank you for coming by.

Yeah.

Good to see you guys.

The Core Memory Podcast is hosted by me, Ashlee Vance, and/or Kylee Robison, or both of us together. It is produced by me and David Mickelson. Our theme song is by James Mercer and John Sortland, and the show is edited always by the John Sortland. Thank you so much to Brex and Send Cut Send for all your support, and thank you most of all to everybody for listening or watching. We love you. Please leave us a like, a review, a subscribe, all those tremendous things. Thank you, and we'll see you again.