WEBVTT - The AI Era of Mark Zuckerberg 0:00:01.200 --> 0:00:04.120 Hi everyone, I'm Emily Chang and this is the circuit. 0:00:05.280 --> 0:00:08.639 Mark Zuckerberg has fired his latest salvo in the AI wars, 0:00:08.880 --> 0:00:11.600 and he thinks it's a big one. Of course, you 0:00:11.640 --> 0:00:13.760 all know him as the face of Facebook, in the 0:00:13.760 --> 0:00:17.040 spotlight and in the hot seat. For twenty years, he's 0:00:17.040 --> 0:00:20.800 been on the offensive, buying Instagram, WhatsApp, and Oculus, and 0:00:20.880 --> 0:00:24.800 the defensive, battling a never ending cascade of criticisms about 0:00:24.800 --> 0:00:30.080 the social, political, and business impacts of his expanding empire. Lately, though, 0:00:30.280 --> 0:00:33.559 Zuckerberg seems to be having something of a zuck Aissance 0:00:34.200 --> 0:00:39.320 or a zuck Renaissance. He's rocking a different vibe. Gold chains, 0:00:39.600 --> 0:00:43.800 fuzzy jackets, martial arts on a barge in Lake Tahoe. 0:00:43.920 --> 0:00:48.640 Zuckerberg appears to be reinventing himself and his company, Metta's 0:00:48.680 --> 0:00:51.640 latest push is a major play to open source AI 0:00:51.920 --> 0:00:55.160 in contrast to closed source competitors like open AI and Google. 0:00:55.680 --> 0:00:59.280 Zuckerberg just unveiled his company's newest class of AI models, 0:00:59.600 --> 0:01:02.640 LAMA three point one, by the way, that stands for 0:01:02.800 --> 0:01:06.559 large Language Model Meta AI, and he believes this approach 0:01:06.600 --> 0:01:09.280 to AI will have a profound impact on the progress 0:01:09.319 --> 0:01:14.520 of tech, business, and maybe the world. I took a 0:01:14.560 --> 0:01:17.399 trip to Meta headquarters in Menlo Park, California to meet 0:01:17.480 --> 0:01:20.440 Zuck and hear about the AI future he says is 0:01:20.440 --> 0:01:23.800 the path forward. We also talked about what's next for 0:01:23.840 --> 0:01:27.080 social media, what's at stake in the US presidential election, 0:01:27.640 --> 0:01:32.039 and his famous side quests joining me now, Meta CEO 0:01:32.120 --> 0:01:34.120 and founder Mark Zuckerberg. 0:01:35.400 --> 0:01:36.280 Goodness, see you again. 0:01:36.360 --> 0:01:37.320 Thank you for having us. 0:01:37.480 --> 0:01:38.480 I'm excited to do this. 0:01:38.600 --> 0:01:41.240 I thought about sending my avatar to meet you, but 0:01:41.280 --> 0:01:43.000 I decided this was too good to pass up. 0:01:43.840 --> 0:01:44.960 So we're gonna walk slowly. 0:01:45.840 --> 0:01:47.560 Okay, you want to set the pace. I will set that. 0:01:47.920 --> 0:01:49.400 I will try it. Sometimes I walk too. 0:01:49.240 --> 0:01:51.800 Fast, so you will. This is your thing. 0:01:51.960 --> 0:01:56.080 So so in the future, will we be meeting as avatars? 0:01:56.120 --> 0:01:57.760 And will that feel totally normal? 0:01:58.960 --> 0:02:01.600 I think probably some of the time. Yeah. I mean, 0:02:01.680 --> 0:02:02.760 you know, I think you're going to be. 0:02:03.360 --> 0:02:06.080 You're gonna have meetings where instead of having to use 0:02:06.160 --> 0:02:09.480 zoom and look at a screen, you're gonna just have 0:02:10.120 --> 0:02:13.919 the people you're talking to as kind of three D avatars, 0:02:13.960 --> 0:02:17.079 just on your couch or around the table. And that's 0:02:17.080 --> 0:02:18.440 going to be pretty good. I mean we'll also be 0:02:18.480 --> 0:02:21.840 able to probably have AI coworkers who will be embodied 0:02:21.840 --> 0:02:24.280 as avatars sitting around the table with us too. So 0:02:24.360 --> 0:02:26.280 that's something that you're not gonna be able to do physically. 0:02:26.320 --> 0:02:29.080 But I don't know. I like it because I mean 0:02:29.120 --> 0:02:29.440 I'm a. 0:02:29.360 --> 0:02:34.440 Big believer in like kind of physical presence and interaction, 0:02:34.840 --> 0:02:37.880 and I mean I just think that like the ability 0:02:37.960 --> 0:02:40.079 to actually be in a physical space, like around a 0:02:40.120 --> 0:02:42.560 table or at a couch and kind of have the 0:02:42.600 --> 0:02:45.520 person there, even if they're not physically there is a 0:02:45.560 --> 0:02:47.720 lot better than just looking at a screen. 0:02:48.040 --> 0:02:50.640 How much of your day to day is meetings versus 0:02:50.680 --> 0:02:53.480 product reviews versus like big strategic thinking. 0:02:53.960 --> 0:02:55.959 It really depends on a day to day basis. I 0:02:56.000 --> 0:02:59.280 don't know that there's like a normal day, the ideal 0:02:59.360 --> 0:03:01.240 day I get to to work on all the long 0:03:01.320 --> 0:03:04.280 term stuff. But I mean, in this role, you sometimes 0:03:04.280 --> 0:03:07.079 get sucked into stuff that is not. 0:03:06.720 --> 0:03:09.440 What you want to be focused on within a given day. Yeah. 0:03:09.480 --> 0:03:12.720 So yeah, but I view and part of this is like, 0:03:13.520 --> 0:03:15.480 all right, how many days am I actually focused on 0:03:15.520 --> 0:03:17.440 the things that I want to be building stuff for 0:03:17.440 --> 0:03:18.000 the long term? 0:03:18.080 --> 0:03:19.280 How much are you still coding? 0:03:19.800 --> 0:03:22.640 Oh? I don't code for work at all anymore. I 0:03:22.720 --> 0:03:23.320 do it for fun. 0:03:23.720 --> 0:03:25.840 Yeah, I mean probably most of the coding that I 0:03:25.880 --> 0:03:28.560 do these days is with my daughters, teaching them. 0:03:28.680 --> 0:03:29.320 They love it. 0:03:29.760 --> 0:03:33.600 My daughter Aggie, she basically uses code as a medium 0:03:33.800 --> 0:03:36.360 for art and just to create somethings. She calls it 0:03:36.400 --> 0:03:38.920 code art, just like telling stories and different things by 0:03:39.000 --> 0:03:44.600 like making I don't know just's it's just coding in scratch. 0:03:44.640 --> 0:03:46.120 So I spent a bunch of time doing that with 0:03:46.160 --> 0:03:48.000 the kids, and that's my coding these days. 0:03:48.400 --> 0:03:52.240 So I found an article in the Harvard Crimson going 0:03:52.280 --> 0:03:55.000 back to two thousand and three where you were talking 0:03:55.040 --> 0:03:55.920 about open source. 0:03:56.120 --> 0:03:57.680 Okay, like over twenty years. 0:03:57.560 --> 0:04:00.400 Ago, really, so you've been thinking about this really long time. 0:04:01.120 --> 0:04:03.280 Well, yeah, it's a big part of the tech industry. 0:04:03.320 --> 0:04:04.760 I mean you wouldn't have been able to build the 0:04:04.800 --> 0:04:07.480 early version of Facebook without that. I mean all the 0:04:07.480 --> 0:04:11.920 stuff that we used, just the early versions of my 0:04:12.080 --> 0:04:16.560 sequel and PHP and Apache and Linux, I mean all 0:04:16.600 --> 0:04:18.320 that stuff. It's like, there's just no way I would 0:04:18.320 --> 0:04:20.560 have been able to build the first version without that 0:04:20.600 --> 0:04:21.119 whole stack. 0:04:21.400 --> 0:04:22.800 So it's like super critical. 0:04:23.440 --> 0:04:25.400 Oh yeah, yeah, I was a student. 0:04:25.440 --> 0:04:26.880 I didn't have access to a lot of capital I 0:04:26.920 --> 0:04:31.719 couldn't have gotten like a really expensive proprietary Unix system 0:04:31.839 --> 0:04:34.000 or something. So I mean there's the whole kind of 0:04:34.000 --> 0:04:38.400 hacker mentality. You just take the code, use it for 0:04:38.520 --> 0:04:40.839 the thing that you need to. It's more cost efficient, 0:04:40.960 --> 0:04:43.240 and I mean that's how you can start something like 0:04:43.279 --> 0:04:44.240 this in a dorm room. 0:04:44.279 --> 0:04:46.520 So speaking of that, some people see you as an 0:04:46.600 --> 0:04:50.440 unlikely champion of open source today you're laughing. 0:04:50.400 --> 0:04:53.000 Well, I don't know why. I mean, I actually think, well, 0:04:53.320 --> 0:04:55.080 I get it, but yeah. 0:04:55.000 --> 0:04:57.920 You understand, you understand the word unlikely. 0:04:58.080 --> 0:05:03.560 Yeah, Well, we've actually Meta have been pretty big proponents 0:05:03.560 --> 0:05:06.359 of open source for a while, and it maybe is 0:05:06.400 --> 0:05:10.280 a little bit of an accident of history that you know, 0:05:10.320 --> 0:05:15.080 we got started after companies like Google and built up infrastructure, 0:05:15.760 --> 0:05:18.359 but it was never a proprietary advantage for us because 0:05:18.400 --> 0:05:22.080 we grew up after them. So instead of kind of 0:05:22.120 --> 0:05:24.240 hoarding it and saying, Okay, this is going to be 0:05:24.240 --> 0:05:26.159 something that we keep as an advantage, because it actually 0:05:26.160 --> 0:05:28.240 wasn't an advantage relative to Google and other companies. 0:05:28.279 --> 0:05:30.359 It was just we were just catching up, like all right, 0:05:30.480 --> 0:05:31.040 let's create. 0:05:30.920 --> 0:05:33.640 An open ecosystem and make it so we can benefit 0:05:33.680 --> 0:05:36.480 from the innovation that other people bring and. 0:05:37.120 --> 0:05:38.800 It's been a really good formula for us. 0:05:39.040 --> 0:05:41.760 So dating back to how we design servers, how we 0:05:41.800 --> 0:05:44.039 design our data centers, we have this whole open compute 0:05:44.040 --> 0:05:47.600 project that kind of standardized a lot of the infrastructure 0:05:47.600 --> 0:05:49.680 for the industry. It was really good for us because 0:05:49.960 --> 0:05:52.040 now that all these other companies are using the same 0:05:52.080 --> 0:05:54.480 stuff as us, the supply chains got more built out 0:05:54.480 --> 0:05:56.440 around it, which meant that it was cheaper for us. 0:05:56.440 --> 0:05:58.479 So we've saved billions of dollars by kind of having 0:05:58.480 --> 0:06:00.839 it out there. And that's going to how with AI too. 0:06:01.160 --> 0:06:02.960 So you can look at our history, there's all these 0:06:02.960 --> 0:06:05.440 different projects that we've open sourced, and I think that 0:06:05.480 --> 0:06:07.480 the positive experience that we've had is one of the 0:06:07.520 --> 0:06:11.719 reasons why we're just a little more willing to put 0:06:11.720 --> 0:06:14.440 ourselves out there and open source AI models and generally 0:06:14.480 --> 0:06:17.600 believe that we're going to get the positive benefits from 0:06:17.600 --> 0:06:18.719 the ecosystem on that too. 0:06:18.960 --> 0:06:21.599 You're really putting a stake in the ground by open 0:06:21.600 --> 0:06:24.760 sourcing your AI in this attempt to build the AI 0:06:24.920 --> 0:06:28.960 rails for the future. How much of this is a 0:06:29.000 --> 0:06:32.800 strategic way to control or own the next technology wave. 0:06:33.440 --> 0:06:34.279 It's a few things. 0:06:35.080 --> 0:06:38.320 One is, if you just think about, like our kind 0:06:38.320 --> 0:06:41.440 of psychology and strategy on this, A lot of how 0:06:41.440 --> 0:06:43.680 we've grown up over the last ten or fifteen years 0:06:44.320 --> 0:06:49.320 was building our apps through phone platforms that our competitors controlled. 0:06:49.760 --> 0:06:51.880 And you know, there's all these analyzes that we've done 0:06:52.320 --> 0:06:54.560 where we would be like a lot more profitable, our 0:06:54.560 --> 0:06:56.800 business would be bigger if we hadn't gotten all these 0:06:56.880 --> 0:06:59.960 random taxes or rules that the mobile platforms. 0:06:59.400 --> 0:06:59.960 Had put on us. 0:07:00.360 --> 0:07:02.400 But honestly, that's not the big thing that bothered me. 0:07:02.800 --> 0:07:05.600 It was how it limited our creativity to build the 0:07:05.600 --> 0:07:09.800 best things that we could imagine. It's somewhat soul crushing 0:07:09.840 --> 0:07:12.560 to like go build something that you think is going 0:07:12.600 --> 0:07:14.520 to be good and then just get told by Apple 0:07:14.600 --> 0:07:16.320 that you can't ship it because they want to put 0:07:16.360 --> 0:07:18.480 us in a box because they view us as competitive. 0:07:19.000 --> 0:07:21.720 So you know, we're a big enough company now that 0:07:21.920 --> 0:07:23.920 one of the things that I've resolved is that for 0:07:24.040 --> 0:07:26.760 the next generation of technology, I want us to build 0:07:27.400 --> 0:07:30.200 and have more control over the next set of platforms 0:07:30.240 --> 0:07:31.640 that we're going to build. So I think AI is 0:07:31.640 --> 0:07:34.800 a critical one, and I think augmented and virtual reality 0:07:34.840 --> 0:07:37.000 is another critical one. But the thing is is that 0:07:37.240 --> 0:07:39.760 these platforms are not just It's not such a thing 0:07:39.800 --> 0:07:41.080 that you can go build in a lab. 0:07:41.480 --> 0:07:42.440 It's an ecosystem. 0:07:42.720 --> 0:07:46.360 So we can build the best AI model, but over time, 0:07:46.480 --> 0:07:49.040 like Linux, the way that it becomes valuable is that 0:07:49.080 --> 0:07:51.280 there's going to be a whole ecosystem of companies that 0:07:51.320 --> 0:07:54.280 integrate with it and build out all these different capabilities, 0:07:54.360 --> 0:07:56.280 And if we were to keep that to ourselves, we 0:07:56.320 --> 0:07:57.360 actually wouldn't benefit. 0:07:57.520 --> 0:07:59.520 So it's kind of the way that we. 0:07:59.440 --> 0:08:01.640 Can control our own destiny on this and make sure 0:08:01.680 --> 0:08:04.560 that we have access to leading AI is by building 0:08:04.640 --> 0:08:07.840 it and having it become an industry standard. So it's 0:08:07.840 --> 0:08:09.960 somewhat counterintuitive, and I think a lot of people woul 0:08:09.960 --> 0:08:11.200 look at that and say, Okay, why don't you just 0:08:11.320 --> 0:08:12.800 if you build this thing, why don't you just keep 0:08:12.800 --> 0:08:16.120 it for yourselves? But it actually gets stronger by being 0:08:16.120 --> 0:08:18.680 able to share it and have the ecosystem around it. 0:08:18.720 --> 0:08:21.040 How much of an opportunity here do you see to 0:08:21.080 --> 0:08:22.160 take Apple out. 0:08:22.000 --> 0:08:22.440 Of the middle. 0:08:22.800 --> 0:08:26.600 Interestingly, Apple is not really the biggest player in AI, 0:08:26.720 --> 0:08:29.360 or at least major AI foundation models, so I'm sure 0:08:29.400 --> 0:08:31.400 that they'll do great stuff with their Apple Intelligence and 0:08:31.440 --> 0:08:33.800 the on device stuff, But I was more using that 0:08:34.120 --> 0:08:37.400 as an illustrative example that really kind of shaped me 0:08:37.800 --> 0:08:39.320 and I think has shaped the company over the last 0:08:39.320 --> 0:08:41.680 ten or fifteen years of one of the struggles that 0:08:41.679 --> 0:08:44.600 we've had building apps in their ecosystem. I mean, they're 0:08:44.640 --> 0:08:46.680 just not a neutral player in that right. I mean 0:08:46.720 --> 0:08:48.640 we're a competitor to them, and we have to deliver 0:08:48.679 --> 0:08:50.960 our services through a competitor, and that's a very difficult 0:08:51.000 --> 0:08:53.520 situation to be in. So going forward, I'm not sure 0:08:53.520 --> 0:08:55.400 that they are going to be the biggest challenge that 0:08:55.480 --> 0:08:57.600 we have, but I think the lessons from that If 0:08:57.640 --> 0:08:59.640 AI is going to be as important in the future 0:08:59.679 --> 0:09:02.199 as platforms are, then I just don't want to be 0:09:02.280 --> 0:09:05.839 in the position where we're accessing AI through I don't 0:09:05.880 --> 0:09:08.160 know whether it's Google or you know, whoever the other 0:09:08.240 --> 0:09:11.559 company is that also may be eventually down the road 0:09:11.559 --> 0:09:13.640 a competitive for us. It's just I think it's important 0:09:14.080 --> 0:09:16.160 if we want to build services that help people connect 0:09:16.160 --> 0:09:17.880 in all the ways that we want to do. It's 0:09:17.920 --> 0:09:20.079 just a thing where we're a technology company. We need 0:09:20.080 --> 0:09:22.240 to be able to build stuff not just at the 0:09:22.240 --> 0:09:25.280 app layer, but like all the way down, and it's 0:09:25.320 --> 0:09:27.200 worth it to us to make these massive investments to 0:09:27.240 --> 0:09:27.480 do that. 0:09:27.960 --> 0:09:31.840 You're continuing to improve meta AI across all of your products, 0:09:31.880 --> 0:09:34.720 but also as a standalone chatbot. Why should we use 0:09:34.760 --> 0:09:35.920 meta over chat GPT? 0:09:36.240 --> 0:09:38.120 Well, there's a bunch of things where it's better, And 0:09:38.240 --> 0:09:40.440 I mean one is there are all these tools for 0:09:40.600 --> 0:09:42.000 producing content. 0:09:42.400 --> 0:09:42.560 You know. 0:09:42.600 --> 0:09:45.200 One of the things that we're rolling out soon is 0:09:45.400 --> 0:09:46.920 the ability to just like imagine stuff. 0:09:46.960 --> 0:09:48.880 You're typing something in real time. I do this with 0:09:48.920 --> 0:09:49.719 my daughters all the. 0:09:49.679 --> 0:09:54.559 Time, and as you're typing and entering the query, it's 0:09:54.600 --> 0:09:56.960 just generating the images as you enter the keystrokes. 0:09:56.960 --> 0:09:57.840 It's just really cool. 0:09:58.320 --> 0:10:00.440 The other is just that it's kind of in integrated 0:10:00.480 --> 0:10:04.240 into the experiences that people use. So you can add 0:10:04.360 --> 0:10:07.560 medai to your chats with your friends and WhatsApp or 0:10:07.600 --> 0:10:10.840 Messenger or Instagram, so it can be there in group threads. 0:10:10.880 --> 0:10:12.880 I think that's really neat. But look, at the end 0:10:12.880 --> 0:10:15.920 of the day, I think the most important product for 0:10:16.200 --> 0:10:18.960 an AI assistant is going to be how smart it is. 0:10:19.440 --> 0:10:22.240 And the LAMA models that we're building are some of 0:10:22.280 --> 0:10:24.120 the most advanced in the world, and I think that 0:10:24.120 --> 0:10:27.240 that's why people want to use them that and obviously 0:10:27.240 --> 0:10:29.240 it's you know, we try to build our services and 0:10:29.240 --> 0:10:31.760 make them free. So a lot of other companies they 0:10:32.040 --> 0:10:34.280 take their best models and they charge for them, and 0:10:34.280 --> 0:10:35.800 we're going to try to take our best models and 0:10:35.840 --> 0:10:38.000 make them free so that way as many people as 0:10:38.040 --> 0:10:41.040 possible around the world can use them. And it's early, 0:10:41.080 --> 0:10:44.880 but it's basically working. My goal for the Medai launch, 0:10:45.000 --> 0:10:48.160 which I mean it's only really a few months old 0:10:48.360 --> 0:10:50.840 at this point, was to by the end of the year, 0:10:51.320 --> 0:10:54.040 have Metai be the most used AI assistant in the world. 0:10:54.320 --> 0:10:55.800 And I think we're basically on track for that. 0:10:55.840 --> 0:10:57.320 I mean, there's hundreds of millions of people who are 0:10:57.360 --> 0:10:59.120 at the end of this year, yeah, and I think 0:10:59.120 --> 0:11:00.600 we're going to be there before the end of the year. 0:11:00.600 --> 0:11:02.600 But there are already hundreds of millions of people who 0:11:02.600 --> 0:11:04.000 are using it. We're not even rolled out in all 0:11:04.000 --> 0:11:06.520 the languages yet. We have the big launch coming this 0:11:06.559 --> 0:11:09.360 week that's just like much better models across the board, 0:11:09.440 --> 0:11:11.280 so people are using that AI. It's just going to 0:11:11.280 --> 0:11:13.360 get smarter this week automatically. You don't have to pay 0:11:13.360 --> 0:11:15.240 for it. It's just going to keep on getting better 0:11:15.240 --> 0:11:15.600 and better. 0:11:15.720 --> 0:11:18.240 So you're releasing LAMA three point one, this family of 0:11:18.320 --> 0:11:21.439 models big and small, including the biggest open source model 0:11:21.520 --> 0:11:23.439 ever four hundred and five billion primers. 0:11:23.480 --> 0:11:25.440 Yeah, yeah, what does that jump unlock? 0:11:26.240 --> 0:11:30.080 In Layman's term, the bigger the model, the more intelligence 0:11:30.080 --> 0:11:33.880 can be encoded in it, but also the more expensive 0:11:33.920 --> 0:11:36.440 it is to operate and run. Right, So you don't 0:11:36.480 --> 0:11:38.839 always just want to use the biggest model. You want 0:11:38.840 --> 0:11:40.920 to use the most sophisticated model for what. 0:11:40.840 --> 0:11:41.360 You're trying to do. 0:11:41.440 --> 0:11:42.720 So if you're trying to run it on a phone 0:11:42.720 --> 0:11:45.640 where there's limited compute, you actually want a much smaller model, 0:11:45.720 --> 0:11:48.240 maybe like a two billion parameter model or three or 0:11:48.240 --> 0:11:48.840 something like that. 0:11:48.960 --> 0:11:50.000 People want to run stuff on. 0:11:50.000 --> 0:11:53.160 Their own laptops, and you can run like a seventy 0:11:53.200 --> 0:11:55.760 billion parameter model on a laptop. The four hundred and 0:11:55.760 --> 0:11:59.160 five billion parameter model is a really big, very sophisticated 0:11:59.200 --> 0:12:03.800 models that we're shipping, is basically competitive with all the 0:12:03.840 --> 0:12:06.360 state of the art models. People can run it directly 0:12:06.400 --> 0:12:08.280 if they want, But I actually think the main thing 0:12:08.320 --> 0:12:10.679 that people are going to do, especially because it's open source, 0:12:11.240 --> 0:12:14.320 is use it as a teacher to train smaller models 0:12:14.320 --> 0:12:17.480 that they use in different applications. We've actually done this ourselves, 0:12:17.520 --> 0:12:20.160 so we have smaller models, which actually the main ones 0:12:20.160 --> 0:12:22.280 that we use in our products because again they're more 0:12:22.280 --> 0:12:24.320 cost effective to run. And as soon as we finish 0:12:24.400 --> 0:12:27.319 training the four hundred and five billion parameter model, we 0:12:27.440 --> 0:12:29.280 used it to now train a better version of the 0:12:29.320 --> 0:12:32.920 small models. And I think one of the really powerful 0:12:32.920 --> 0:12:35.160 things is that if you just think about like all 0:12:35.160 --> 0:12:38.480 the startups out there, or all the enterprises or even 0:12:38.559 --> 0:12:42.080 governments that are trying to do different things, they probably 0:12:42.120 --> 0:12:44.760 all need to at some level build custom models for 0:12:44.800 --> 0:12:46.880 what they're doing. And it's really hard to do that 0:12:47.280 --> 0:12:50.560 with closed systems out there, whether that's open AI or Gemini, 0:12:50.720 --> 0:12:51.920 Google's thing or whatever. 0:12:52.440 --> 0:12:53.520 But with open source. 0:12:53.320 --> 0:12:54.920 That's really easy because you have the weights, so you 0:12:54.960 --> 0:12:58.080 can basically use the model to distill and train whatever 0:12:58.200 --> 0:13:01.080 size model you want. Gets to a pretty core part 0:13:01.080 --> 0:13:03.000 of our philosophy is we don't believe there's gonna be 0:13:03.040 --> 0:13:06.200 like one AI to rule them all. Our vision is 0:13:06.240 --> 0:13:09.320 that there's going to be millions or just billions of 0:13:09.320 --> 0:13:11.840 different models out there, and I think that's really what 0:13:11.880 --> 0:13:14.280 the LAMA three point one, four or five billion is 0:13:14.280 --> 0:13:15.760 going to allow. It's just going to be this teacher 0:13:15.800 --> 0:13:19.400 that allows so many different organizations to create their own 0:13:19.440 --> 0:13:22.040 models rather than having to rely on the kind of 0:13:22.040 --> 0:13:24.040 off the shelf ones that the other guys are selling. 0:13:24.120 --> 0:13:26.240 So not one god, but many, Is that the way 0:13:26.240 --> 0:13:26.880 to think about it? 0:13:27.080 --> 0:13:29.400 Well, I don't think they're gonna be gods, but I 0:13:29.480 --> 0:13:32.599 mean I think I think there's this weird semi theological 0:13:32.679 --> 0:13:35.439 thing that I think. I think you're referring to this 0:13:36.440 --> 0:13:39.120 conversation that I had on Instagram recently, where I do 0:13:39.200 --> 0:13:41.439 think that to some degree, if you're like an organization, 0:13:41.559 --> 0:13:43.400 you think you're going to create like this one super 0:13:43.400 --> 0:13:46.000 intelligence that does have this feel to me of like 0:13:46.200 --> 0:13:49.400 people trying to create a god. And that's just I 0:13:49.840 --> 0:13:51.720 find that both the wrong way to look at it, 0:13:51.760 --> 0:13:55.240 but just also very unappealing. I would rather create a world, 0:13:55.960 --> 0:13:57.720 or help people create a world where there can just 0:13:57.760 --> 0:14:02.440 be like a lot of different diferent people and services 0:14:02.440 --> 0:14:04.199 that you interact with. So, you know, one of the 0:14:04.280 --> 0:14:07.160 things that I'm excited about is making it so that, 0:14:07.240 --> 0:14:09.640 like you know, there's almost two hundred million creators on 0:14:09.679 --> 0:14:12.520 our platforms, they all are trying to build their community. 0:14:12.559 --> 0:14:14.360 People want to interact with them, there aren't enough hours 0:14:14.360 --> 0:14:15.760 in the day, Like, I want to make it so 0:14:15.800 --> 0:14:18.120 that every single one of them can easily train like 0:14:18.160 --> 0:14:20.920 an AI version of themselves. 0:14:20.640 --> 0:14:22.040 That they can make it what they want. 0:14:22.080 --> 0:14:24.880 So it's almost like a kind of artistic artifact that 0:14:24.920 --> 0:14:27.240 they're putting out there for their community that allows their 0:14:27.280 --> 0:14:29.640 community and interact with them, but also gives them control 0:14:29.680 --> 0:14:31.400 over how that interaction happens. 0:14:31.920 --> 0:14:33.160 And I think that that's going to be great, and 0:14:33.200 --> 0:14:33.640 there's going to. 0:14:33.640 --> 0:14:35.720 Be millions there are, you know, eventually hundreds of millions 0:14:35.760 --> 0:14:39.400 of those simple businesses. There's hundreds of millions of small 0:14:39.400 --> 0:14:42.600 businesses on our platforms, and I want to make it 0:14:42.640 --> 0:14:46.080 so that any small business can just very easily pull 0:14:46.080 --> 0:14:48.440 in all their information from social media and their business 0:14:48.440 --> 0:14:51.800 catalog and get an agent that can help their customers 0:14:51.800 --> 0:14:54.520 with customer support and can help them with sales and 0:14:54.920 --> 0:14:57.800 can recommend new products. And I think that just like 0:14:58.000 --> 0:15:01.160 today every business has a web site and email address 0:15:01.160 --> 0:15:03.000 and social media accounts, I think in the future every 0:15:03.000 --> 0:15:05.000 business is going to have an AI business agent too, 0:15:05.040 --> 0:15:06.560 and I think that's going to be great. So I 0:15:06.560 --> 0:15:09.000 don't think that there's just like one AI use case, 0:15:09.040 --> 0:15:11.080 which is what we're trying to do. I think what 0:15:11.080 --> 0:15:13.840 we're trying to help enable the whole community to do 0:15:13.920 --> 0:15:16.320 is create all these different ais for all these things 0:15:16.320 --> 0:15:18.440 that people want to do. And that's kind of how 0:15:18.480 --> 0:15:19.840 I think the sense of being a good thing for 0:15:19.880 --> 0:15:20.240 the world. 0:15:20.720 --> 0:15:24.000 Some people don't see your AIS so open. So how 0:15:24.040 --> 0:15:24.960 open are we talking? 0:15:25.000 --> 0:15:25.280 Really? 0:15:25.520 --> 0:15:27.400 Will you show us how the model was built or 0:15:27.400 --> 0:15:28.680 what data you've trained on it? 0:15:29.320 --> 0:15:31.560 The model has open weights, so you can take the 0:15:31.560 --> 0:15:34.080 weights and you can modify them, and you can see 0:15:34.080 --> 0:15:36.280 how it's built, and you can see the architecture and 0:15:36.320 --> 0:15:39.000 all that. I'm not aware of anyone else doing more 0:15:39.080 --> 0:15:41.600 open work than this. The data question, I think is 0:15:41.640 --> 0:15:44.240 a kind of sensitive and important one. I mean, there 0:15:44.280 --> 0:15:46.040 is this question in science where people want to be 0:15:46.040 --> 0:15:48.640 able to reproduce science experiments, and I think some people 0:15:48.680 --> 0:15:50.080 would like to be able to do that here. The 0:15:50.120 --> 0:15:52.840 reality is, you know, it also costs for the LAMA 0:15:52.880 --> 0:15:55.800 three model at least hundreds of millions of dollars of compute, 0:15:55.800 --> 0:15:57.800 and going forward it's going to be billions and many 0:15:57.800 --> 0:16:00.680 billions of dollars of compute. So is it reason or 0:16:00.720 --> 0:16:02.240 is anyone actually going to be able to reproduce it. 0:16:02.280 --> 0:16:02.840 I'm not sure. 0:16:03.040 --> 0:16:05.200 I don't think that that's like as important of a 0:16:05.240 --> 0:16:08.480 thing to push on and for data, I think that 0:16:08.920 --> 0:16:11.680 even though it's open, but we are designing this also 0:16:11.800 --> 0:16:14.200 for ourselves and we want to make sure that this 0:16:14.320 --> 0:16:17.040 works well. And we do work with different organizations to 0:16:17.120 --> 0:16:19.440 license data in different ways, and some. 0:16:19.400 --> 0:16:21.080 Of those sources are proprietary. 0:16:21.120 --> 0:16:23.040 Even if the data is public, you know, you don't 0:16:23.040 --> 0:16:25.880 necessarily have the kind of right out of the gate 0:16:26.120 --> 0:16:27.800 to be able to train on it. You sometimes have 0:16:27.840 --> 0:16:31.160 to go make deals, and so we can't just go 0:16:31.280 --> 0:16:34.360 ship all the data even if it's public data that 0:16:34.480 --> 0:16:38.280 we've used to train on. And sometimes it's not even 0:16:38.360 --> 0:16:41.520 clear that it's the right thing to disclose or tell 0:16:41.560 --> 0:16:44.520 people what data we've trained on. So yeah, I mean 0:16:44.720 --> 0:16:46.280 I get the point on that. You know that you 0:16:46.320 --> 0:16:49.520 can always take what you're doing and make it more 0:16:49.560 --> 0:16:52.080 open and even more extreme way. But look, I mean, 0:16:52.080 --> 0:16:54.440 I think what we're doing is state of the art 0:16:54.640 --> 0:16:56.960 and kind of open source a eye models. I think 0:16:57.000 --> 0:16:58.680 for people who want to use this to build stuff, 0:16:58.680 --> 0:16:59.800 I think that they're pretty happy with. 0:16:59.800 --> 0:17:00.160 What we do. 0:17:00.440 --> 0:17:03.160 You are using Facebook and Instagram data right as a 0:17:03.200 --> 0:17:06.000 public data, public data as I understand it, and is 0:17:06.040 --> 0:17:09.720 that does that give you an advantage versus other models, 0:17:09.760 --> 0:17:12.679 and should those users get something in return for their 0:17:12.720 --> 0:17:13.600 data being used. 0:17:14.200 --> 0:17:16.320 I actually think I'm not sure how much it's an 0:17:16.359 --> 0:17:18.160 advantage because a lot of the public data on those 0:17:18.200 --> 0:17:20.840 services we allow to be indexed and search engines, So 0:17:20.880 --> 0:17:24.159 I think Google and others actually have the ability to 0:17:24.280 --> 0:17:24.800 use a lot of. 0:17:24.800 --> 0:17:26.520 That data too. In some way. 0:17:26.560 --> 0:17:28.960 If we didn't use the data, we might actually be 0:17:29.000 --> 0:17:31.960 at a disadvantage because if it were in like Google Search, 0:17:32.000 --> 0:17:34.560 and we somehow couldn't use it because we were saying, Okay, 0:17:34.560 --> 0:17:36.840 we're not going to use data from our services even 0:17:36.880 --> 0:17:38.720 if it's public, but Google can use it because it's 0:17:38.720 --> 0:17:39.560 not from their services. 0:17:39.560 --> 0:17:40.800 It like kind of doesn't make sense. 0:17:41.440 --> 0:17:45.480 But I do think that overall, there's code and technology 0:17:45.480 --> 0:17:48.399 that you develop, there's compute to train the models on, 0:17:48.840 --> 0:17:50.720 and there's what your data mix is, and those are 0:17:50.800 --> 0:17:53.480 probably some of the biggest pieces that contribute to the 0:17:53.640 --> 0:17:56.240 end quality of the model. So we try to innovate 0:17:56.280 --> 0:17:58.359 on all of those, and we want to build the 0:17:58.400 --> 0:18:00.720 best thing because it's the foundation that we use for 0:18:00.760 --> 0:18:02.760 our products. But we also want to build the best 0:18:02.760 --> 0:18:04.920 thing because we want it to become the open source 0:18:05.000 --> 0:18:07.160 standard that then other people adopt and then it makes 0:18:07.160 --> 0:18:09.560 it even better, which helps us serve the people that 0:18:09.600 --> 0:18:10.840 we're trying to build for as well. 0:18:11.240 --> 0:18:13.480 So let's get into the broader strategy and how this 0:18:13.640 --> 0:18:16.399 is all going to work, Like, will we have AI 0:18:16.520 --> 0:18:22.000 generated influencers with AI generated captions and avatars talking to avatars? 0:18:22.119 --> 0:18:23.440 Yeah, I think we'll have all of it. 0:18:23.560 --> 0:18:27.399 What and like, how do you want to create the 0:18:27.440 --> 0:18:29.480 first AI generated social network? 0:18:30.040 --> 0:18:31.840 Well, I think that that will be part of this, 0:18:31.920 --> 0:18:33.360 but it's not going to be the only thing. I mean, 0:18:33.400 --> 0:18:35.399 it's people come to our services because they want to 0:18:35.400 --> 0:18:37.800 connect with people. But you know, actually, one of the 0:18:37.840 --> 0:18:40.840 most interesting use cases that people have for medai today, 0:18:40.840 --> 0:18:43.560 it's like in the top four use cases is role 0:18:43.560 --> 0:18:46.439 playing difficult social interactions that they're going to have. 0:18:46.760 --> 0:18:47.920 So whether it's in a. 0:18:47.800 --> 0:18:50.720 Professional context like okay, you want to ask your manager 0:18:50.800 --> 0:18:53.280 for a raise, or like I'm having this hard conversation 0:18:53.320 --> 0:18:54.880 with my girlfriend or my friend, they. 0:18:54.840 --> 0:18:57.119 Could have role played this conversation with you today. 0:18:57.400 --> 0:18:59.679 I mean, like, hopefully it's not stressful for you, but 0:18:59.720 --> 0:19:04.919 it's But yeah, I mean I think that that's like 0:19:04.960 --> 0:19:08.000 it's a clear social tool where there's there's no judgment. Right, 0:19:08.040 --> 0:19:10.920 the AI isn't sitting there, there's no like social repercussion 0:19:11.000 --> 0:19:11.800 for what you ask it. 0:19:12.320 --> 0:19:14.480 But what evidence do you have that people want to 0:19:14.520 --> 0:19:17.480 live in this virtual world and socialize with avatars or 0:19:17.520 --> 0:19:19.000 that it's actually good for us? 0:19:19.320 --> 0:19:21.280 Well, I think that people want to connect with each other. 0:19:21.520 --> 0:19:23.800 I actually think like all the other stuff is generally noise, 0:19:24.240 --> 0:19:26.960 but all the technology allows you to do that in 0:19:27.000 --> 0:19:28.800 a better and better way. Right when I got started 0:19:28.800 --> 0:19:32.280 with Facebook, it was mostly text, and then this whole 0:19:32.320 --> 0:19:35.160 there's been this whole kind of like technological evolution where 0:19:35.200 --> 0:19:40.240 we got smartphones that we could primarily be taking photos, 0:19:40.320 --> 0:19:42.720 and then the mobile network's got good, so then you 0:19:42.760 --> 0:19:45.080 could be kind of sharing video and videos a lot 0:19:45.160 --> 0:19:47.800 richer and consuming videos a better experience. So I just 0:19:47.840 --> 0:19:49.160 don't think that's the end of the line. I think 0:19:49.160 --> 0:19:52.160 it gets more immersive. And that's one of the reasons 0:19:52.200 --> 0:19:55.080 why I'm so convinced that ar glasses by the time 0:19:55.080 --> 0:19:57.320 you just get to like normal stylish glasses where you 0:19:57.359 --> 0:19:58.840 can have a sense of presence. 0:19:58.840 --> 0:19:59.960 I mean, I think within. 0:20:00.000 --> 0:20:01.639 Five years we're going to be at a point where 0:20:01.800 --> 0:20:04.760 we could be having this conversation and like I could 0:20:04.840 --> 0:20:07.439 just be like a full kind of three D avatar here, 0:20:07.440 --> 0:20:09.520 and it would feel like a much more realistic sense 0:20:09.560 --> 0:20:12.720 of presence than just talking to each other over a 0:20:12.800 --> 0:20:13.560 zoom screen or. 0:20:13.560 --> 0:20:15.440 Something like that. That's what it's about, though. 0:20:15.440 --> 0:20:19.040 I think people want to connect, and we're basically building 0:20:19.320 --> 0:20:21.879 all the different technology that allows people to do that. 0:20:22.440 --> 0:20:24.960 You renamed your company meta, You're still pouring billions of 0:20:25.040 --> 0:20:26.160 dollars into the metaverse. 0:20:26.440 --> 0:20:28.280 Are we as far along as you thought we'd be 0:20:28.920 --> 0:20:31.639 a few years down the line? Are there any lessons 0:20:31.680 --> 0:20:34.000 in the urgency of the pivot? 0:20:34.440 --> 0:20:34.720 Well? 0:20:35.280 --> 0:20:37.199 A lot of the reason why I did that is 0:20:37.240 --> 0:20:41.840 because I think we were getting pigeonholed as just this 0:20:41.960 --> 0:20:44.399 like social media app company, right, And that was like 0:20:44.480 --> 0:20:45.359 so ingrained, right. 0:20:45.480 --> 0:20:46.560 So we were called Facebook. 0:20:46.600 --> 0:20:48.639 Facebook is one of our apps, right, and the others 0:20:48.640 --> 0:20:51.600 are just as important now. And I've never thought about 0:20:51.600 --> 0:20:54.280 the company as a social media app company. I mean, 0:20:54.320 --> 0:20:56.359 we didn't start as that. We weren't you know, there 0:20:56.359 --> 0:20:58.879 weren't apps when we got started. We were a technology 0:20:58.920 --> 0:21:02.240 company that is to help people connect and kind of 0:21:02.240 --> 0:21:06.560 to build the future of human connection. And I thought 0:21:06.560 --> 0:21:11.320 that the rebrand around meta both was healthy because. 0:21:11.040 --> 0:21:12.679 We just have more products at this point than just 0:21:12.720 --> 0:21:14.840 Facebook and two. 0:21:14.920 --> 0:21:18.359 It kind of re anchored people's thinking about the company 0:21:18.359 --> 0:21:20.440 in terms of, oh, this is a company that's developing 0:21:20.680 --> 0:21:24.040 kind of longer term technology around people connecting, which is 0:21:24.080 --> 0:21:26.040 really how I think about what we should be doing. 0:21:26.440 --> 0:21:28.359 And when you're trying to push something in a direction, 0:21:28.400 --> 0:21:30.960 it's helpful for people both inside the company and outside 0:21:30.960 --> 0:21:32.480 to think about it in terms of what you're actually 0:21:32.480 --> 0:21:34.479 trying to do. So I've been very happy with how 0:21:34.520 --> 0:21:36.920 that's gone. The metaverse thing was always going to be 0:21:36.960 --> 0:21:38.560 a very long term thing. I think some things have 0:21:38.640 --> 0:21:42.879 gone better than I thought. Some have gone slower. The glasses, 0:21:43.000 --> 0:21:44.919 I think is probably the best example of something that 0:21:44.960 --> 0:21:47.720 is going better. I would have thought a few years 0:21:47.760 --> 0:21:51.720 ago that ar glasses wouldn't really be a mainstream thing 0:21:52.280 --> 0:21:55.199 until we got kind of these full holographic displays, like 0:21:55.320 --> 0:21:57.600 I think the type of stuff that you've been able 0:21:57.600 --> 0:22:00.520 to see. It's been one of the most positive surprises 0:22:00.600 --> 0:22:02.920 that the collaboration that we have with ray Band, it's 0:22:02.960 --> 0:22:05.720 going extremely well, and I think part of it is 0:22:06.320 --> 0:22:09.159 they're stylish, they're good glasses. Part of it is that 0:22:09.200 --> 0:22:11.359 it's a great form factor for AI. We didn't know 0:22:11.400 --> 0:22:12.600 that AI was going to be a thing when we 0:22:12.640 --> 0:22:14.720 started working on that project, or I mean, we thought 0:22:14.760 --> 0:22:15.960 it was going to be a thing like ten years 0:22:15.960 --> 0:22:17.800 from now. But if you asked me five years ago, 0:22:18.240 --> 0:22:20.440 I would have guessed that AR would come before AI, 0:22:20.640 --> 0:22:22.920 not the other direction. So some of this is just 0:22:22.960 --> 0:22:24.879 about kind of setting yourselves up to ride the different 0:22:24.920 --> 0:22:25.840 waves when they come in. 0:22:26.400 --> 0:22:28.639 I've gotten to try out a bunch of your future 0:22:28.680 --> 0:22:32.320 facing technology here with Quest with the ray band, metas 0:22:32.400 --> 0:22:35.560 with O Ryan. How big an opportunity do you see 0:22:35.600 --> 0:22:38.640 to own your own operating system? And down the line 0:22:38.640 --> 0:22:40.879 do you see your reshuffling of the power relationships in 0:22:40.920 --> 0:22:41.960 the tech industry Now. 0:22:42.560 --> 0:22:45.720 I don't always think about things primarily from a business perspective. 0:22:45.760 --> 0:22:48.119 I mean, I like building stuff. What I've found is 0:22:48.160 --> 0:22:52.239 that if you build good things, then eventually you were 0:22:52.280 --> 0:22:55.280 able to build successful products and businesses around it, even 0:22:55.320 --> 0:22:59.159 if it takes a while. And one of my lessons 0:22:59.200 --> 0:23:01.560 for the last ten years is just building apps that 0:23:01.600 --> 0:23:05.280 we ship through our competitors' mobile platforms. I just think 0:23:05.320 --> 0:23:07.760 we're going to be freed up to be more creative 0:23:07.880 --> 0:23:10.280 and build better stuff if we can control more of 0:23:10.320 --> 0:23:12.600 the core technology. So I mean, look, we're not going 0:23:12.680 --> 0:23:15.600 to be the only kind of operating system and augmented 0:23:15.600 --> 0:23:18.040 in virtual reality. There will be others. We're not going 0:23:18.080 --> 0:23:20.080 to be the only AI system. There will be others. 0:23:20.520 --> 0:23:22.760 But I don't necessarily think about it as trying to 0:23:22.800 --> 0:23:25.359 shuffle the industry. I just want to be able to 0:23:25.400 --> 0:23:28.400 build good things, and I think that like not being 0:23:28.440 --> 0:23:30.639 dependent on competitors that are trying to put you in 0:23:30.640 --> 0:23:33.239 a box, is an important part of doing that. 0:23:33.320 --> 0:23:34.640 But it's also a luxury, right. 0:23:34.640 --> 0:23:36.280 It's like when I was a startup, where we were 0:23:36.280 --> 0:23:38.679 a startup, like you don't necessarily have the resources to 0:23:38.720 --> 0:23:41.479 go invest tens of billions of dollars in building out 0:23:41.520 --> 0:23:44.879 these technologies. But you know, now we're here and we 0:23:44.920 --> 0:23:47.960 have a successful business and we can do this innovation. 0:23:48.160 --> 0:23:49.760 And it's just one of the things that I want 0:23:49.760 --> 0:23:51.040 to do. I don't want us to just like rest 0:23:51.040 --> 0:23:54.200 on our laurels and maximize our profits in the near term. 0:23:54.200 --> 0:23:56.919 I want to pour all the profits and success that 0:23:56.960 --> 0:24:00.240 we have into building the next generation of things, to 0:24:00.240 --> 0:24:03.000 do it in an open way. I think one analogy 0:24:03.400 --> 0:24:07.239 that people misread from history is there have always been 0:24:07.280 --> 0:24:10.719 multiple operating systems. Right with PC, it was Windows and Apple, 0:24:10.960 --> 0:24:14.760 and Windows was the open one back then, and it won, 0:24:15.400 --> 0:24:16.960 right So I think that a lot of people now 0:24:17.240 --> 0:24:20.040 have this massive recency bias and they think you just 0:24:20.080 --> 0:24:22.720 have the iPhone Android analogy in your head and you're like, oh, like, 0:24:22.760 --> 0:24:25.080 well Apple won that. It's like, yeah, they won this 0:24:25.200 --> 0:24:29.280 generation of computing. But it's not always that way. For PCs, 0:24:29.320 --> 0:24:32.840 the open one was actually the primary platform. And part 0:24:32.840 --> 0:24:35.199 of my goal is to make it so that for 0:24:35.440 --> 0:24:38.600 the next generation of platforms, whether it's AR glasses or 0:24:38.640 --> 0:24:42.399 mixed reality headsets or AI systems, it's not just that 0:24:42.440 --> 0:24:44.680 there should be an open platform. I think the open 0:24:44.680 --> 0:24:47.439 platform can actually be the best one, and I think 0:24:47.480 --> 0:24:49.560 that that's kind of a cool thing for the industry. 0:24:50.119 --> 0:24:53.360 Sakoya calls AI the six hundred billion dollar question. There's 0:24:53.359 --> 0:24:56.560 allays investment in chips and the infrastructure and the data centers, 0:24:57.040 --> 0:24:58.640 but when does it start paying off? 0:24:58.880 --> 0:25:02.919 Like is it a bubble? And well, if not, like, 0:25:03.000 --> 0:25:04.439 when do you start seeing the money. 0:25:04.680 --> 0:25:07.199 I think bubbles are interesting because a lot of the 0:25:07.200 --> 0:25:10.360 bubbles ended up being things that were very valuable over time, 0:25:10.400 --> 0:25:12.000 and it's just more of a question of timing, like 0:25:12.040 --> 0:25:14.480 you're asking, right. Even the dot com bubble, it's like 0:25:14.520 --> 0:25:16.520 there's all this fiber laid and it ended up being 0:25:16.560 --> 0:25:19.480 super valuable, but it just wasn't as valuable as quickly 0:25:19.480 --> 0:25:20.200 as people thought. 0:25:20.280 --> 0:25:22.399 So is that going to happen here? I don't know. 0:25:22.440 --> 0:25:24.400 I mean, it's hard to predict what's going to happen 0:25:24.440 --> 0:25:26.080 in the next few years. I think AI is going 0:25:26.119 --> 0:25:29.159 to be very fundamental. If the products were able to 0:25:29.240 --> 0:25:32.000 grow massively over the next few years, which I think 0:25:32.000 --> 0:25:34.960 there's a very good chance of, then I'd much rather 0:25:35.240 --> 0:25:39.000 over invest in play for that outcome then just try 0:25:39.040 --> 0:25:41.560 to say, Okay, maybe we'll save some money by developing 0:25:41.600 --> 0:25:43.879 it a little more slowly. I think that there's a 0:25:44.040 --> 0:25:46.520 meaningful chance that a lot of the companies are overbuilding 0:25:46.520 --> 0:25:48.920 now and that you look back and you're like, oh, 0:25:48.960 --> 0:25:51.840 we maybe all spent some number of billions of dollars 0:25:51.920 --> 0:25:54.320 more than we had to. But on the flip side, 0:25:54.440 --> 0:25:56.320 I actually think all the companies that are investing are 0:25:56.320 --> 0:26:00.439 making a rational decision, because the downside of being is 0:26:00.440 --> 0:26:03.080 that you're out of position for the most important technology 0:26:03.119 --> 0:26:07.720 for the next ten to fifteen years, whereas if you overinvest, 0:26:07.760 --> 0:26:11.200 then you're probably just losing some amount of money that's 0:26:11.560 --> 0:26:14.320 for these companies generally an affordable amount of money that 0:26:14.359 --> 0:26:17.480 they can lose for something that's just a really important prize. 0:26:17.520 --> 0:26:21.760 So with this economy, does the year of efficiency continue 0:26:21.840 --> 0:26:24.560 into the AI era? Are we talking years of efficiency 0:26:24.880 --> 0:26:26.080 or has it belt loosened? 0:26:26.840 --> 0:26:30.000 Well, I'm really glad that we did all the efficiency 0:26:30.040 --> 0:26:33.320 push because I think that that basically created enough capital 0:26:33.359 --> 0:26:36.840 for us to go invest in massive amounts of infrastructure. 0:26:36.840 --> 0:26:39.160 And I think going forward, I would guess that most 0:26:39.200 --> 0:26:41.560 of the investment that we make is going to be 0:26:41.920 --> 0:26:45.919 building out AI compute rather than massively growing the number of. 0:26:45.880 --> 0:26:46.679 People at the company. 0:26:46.720 --> 0:26:48.919 We are growing, We're going to hire more people, but 0:26:49.000 --> 0:26:51.720 I think at this point the biggest part of the 0:26:52.080 --> 0:26:55.160 new investment that we're making is in building these kind 0:26:55.160 --> 0:26:58.800 of giant AI superclusters to train the future AIS. 0:26:59.480 --> 0:27:00.359 So it was. 0:27:00.359 --> 0:27:02.480 An interesting thing for like the first almost twenty years 0:27:02.480 --> 0:27:04.919 of the company that it was just like growing quickly 0:27:04.960 --> 0:27:08.439 in people every year, and I think it's pretty healthy 0:27:08.480 --> 0:27:11.359 to focus on efficiency around that. But yeah, I'm definitely 0:27:11.359 --> 0:27:14.440 glad that we kind of gave ourselves the flexibility. 0:27:13.840 --> 0:27:14.520 To build this. 0:27:14.840 --> 0:27:17.800 You said your goal is getting to artificial general intelligence 0:27:18.359 --> 0:27:21.960 or AGI. How do you define AGI and do you 0:27:22.000 --> 0:27:22.600 get there first? 0:27:22.760 --> 0:27:23.480 That's a good question. 0:27:23.880 --> 0:27:25.960 I'm not sure we can answer the second question about 0:27:26.000 --> 0:27:29.000 who gets there first, but actually maybe maybe we'll start there. 0:27:29.040 --> 0:27:31.840 I do think open sources gaining ground pretty quickly. If 0:27:31.880 --> 0:27:34.159 you look at LAMA two last year, we were like 0:27:34.200 --> 0:27:38.040 a whole generation behind the frontier, and for LAMA three, 0:27:38.480 --> 0:27:40.720 we're basically competitive with the state of the art models 0:27:40.760 --> 0:27:42.280 LAMA three and the LAMA three. 0:27:42.160 --> 0:27:44.040 Point one release that we're putting out now. 0:27:44.240 --> 0:27:47.200 We're basically already starting to work on LAMA four, and 0:27:47.640 --> 0:27:49.840 our goal is to completely close the gap with all 0:27:49.840 --> 0:27:51.399 the others on that. So I don't know, I mean, 0:27:51.480 --> 0:27:52.919 do we get to AGI first? I mean, I think 0:27:52.920 --> 0:27:54.800 that there will probably be some breakthroughs between now and then. 0:27:54.840 --> 0:27:57.399 It's hard to just predict in a straight line, but 0:27:57.520 --> 0:28:00.520 we're certainly putting together a world class effort and we're 0:28:00.600 --> 0:28:01.280 focused on it. 0:28:01.560 --> 0:28:03.159 Then you get to the more complicated question, which is 0:28:03.160 --> 0:28:03.679 like what is it. 0:28:04.040 --> 0:28:07.040 I don't know that there's one specific definition for this 0:28:07.560 --> 0:28:10.840 because I think intelligence is multivariant, right, It's not like 0:28:11.040 --> 0:28:14.879 there's one number that is your intelligence. I think of 0:28:14.920 --> 0:28:19.560 AGI is intelligence has different kind of capabilities, Right, So 0:28:20.040 --> 0:28:23.520 the first set of models could reason over text, and 0:28:23.560 --> 0:28:25.960 then you added in the ability to do photos, and 0:28:26.000 --> 0:28:29.240 now you're adding in the ability to do videos, to 0:28:29.320 --> 0:28:32.520 understand videos and produce videos, and then you're making that 0:28:32.600 --> 0:28:34.720 it works well with audio. I think the ability to 0:28:34.760 --> 0:28:37.919 reason and be able to produce three D worlds and 0:28:37.960 --> 0:28:40.680 three D content is going to be really important. I 0:28:40.760 --> 0:28:42.600 care a lot, and our company cares a lot about 0:28:42.600 --> 0:28:44.600 people interacting with each other. So like when you think 0:28:44.600 --> 0:28:46.800 about the human brain, there's like whole parts of it 0:28:47.040 --> 0:28:50.080 that are just focused on basically like reading people's expressions. 0:28:50.400 --> 0:28:52.640 It's like if you move your eyebrow a millimeter, it 0:28:52.920 --> 0:28:54.640 means something. It's like I could pick that up, whereas 0:28:54.680 --> 0:28:57.040 if something moves over there in the corner a millimeter, 0:28:57.080 --> 0:28:59.880 I'm not going to notice it. There's probably a specific 0:29:00.200 --> 0:29:03.080 aspect of intelligence or modality which is like reading people's 0:29:03.120 --> 0:29:05.640 faces and emotions, and that's something that I care about, 0:29:05.640 --> 0:29:07.000 so I think we'll probably try to build that in 0:29:07.040 --> 0:29:09.040 at some point. So I don't know if it's just 0:29:09.080 --> 0:29:13.560 like increasing the amount of knowledge or some like IQ score. 0:29:13.560 --> 0:29:16.160 It's kind of layering in all these different things, which 0:29:16.200 --> 0:29:18.880 is why it's a little bit hard to say who 0:29:18.880 --> 0:29:21.840 gets their first because I actually think over time, different 0:29:21.880 --> 0:29:23.680 companies might optimize for different things. 0:29:23.840 --> 0:29:26.920 I know, you've always been fascinated by China and you 0:29:27.000 --> 0:29:29.920 learn to speak Mandarin, and what do you know about 0:29:29.960 --> 0:29:32.640 where China is on AI and AGI? 0:29:33.160 --> 0:29:36.000 I don't personally know a ton. I know on open 0:29:36.200 --> 0:29:39.400 source they're leading models. You know, LAMA three point one 0:29:39.840 --> 0:29:43.000 is well ahead of that. The kind of leading Chinese 0:29:43.080 --> 0:29:47.920 model before was right at the level of the previous 0:29:47.960 --> 0:29:50.960 seventy billion parameter model. But there's been no Chinese version 0:29:51.040 --> 0:29:53.840 that's anywhere near the frontier, like a four hundred billion 0:29:53.840 --> 0:29:56.400 parameter model, and the LAMA three point one seventy billion 0:29:56.640 --> 0:29:57.640 is ahead of where they are. 0:29:57.720 --> 0:30:00.280 So I feel good about that. I don't know. 0:30:00.320 --> 0:30:02.120 I think that's going to be one of the interesting 0:30:02.360 --> 0:30:05.880 questions over time. You know, geopolitically, when you think about 0:30:05.920 --> 0:30:09.000 the rivalry, there's this question which is how should the 0:30:09.080 --> 0:30:13.000 US approach kind of AI competition with China. And there's 0:30:13.040 --> 0:30:14.560 one strain of thought which is like, okay, well, we 0:30:14.600 --> 0:30:17.000 need to like lock it all down and I just 0:30:17.040 --> 0:30:19.840 happen to think that that's really wrong because the US 0:30:19.880 --> 0:30:23.440 thrives on open and decentralized innovation. I mean, that's the 0:30:23.440 --> 0:30:26.520 way our economy works. That's like how we build awesome stuff. 0:30:27.040 --> 0:30:30.000 So I think that locking everything down would hamstring us 0:30:30.040 --> 0:30:31.920 and make us more likely to not be the leaders. 0:30:32.520 --> 0:30:35.440 And the other question is if we do lead, what's 0:30:35.480 --> 0:30:38.120 the chance that we're actually going to prevent them from 0:30:38.160 --> 0:30:39.680 being able to steal it anyway. 0:30:39.680 --> 0:30:41.520 I mean, it fits on like a thumb. 0:30:41.360 --> 0:30:44.360 Drive or so it's I just think that that's not 0:30:44.480 --> 0:30:47.520 a realistic way to approach it. I personally think that 0:30:47.600 --> 0:30:50.520 the right configuration of this is open source. 0:30:50.560 --> 0:30:52.200 I think is going to be the leading ecosystem. 0:30:52.240 --> 0:30:54.960 It will be very robust, it will eventually be available 0:30:55.000 --> 0:30:58.080 to everyone in the world, including China. But I think 0:30:58.080 --> 0:31:00.920 the leading companies should work with the US government and 0:31:01.520 --> 0:31:04.240 make sure that our national defense and things like that 0:31:04.320 --> 0:31:07.640 have sort of a perpetual first mover advantage on the 0:31:07.720 --> 0:31:08.920 leading technology in the world. 0:31:08.960 --> 0:31:11.320 What do you say to the skeptics who think, you know, 0:31:11.520 --> 0:31:15.200 this could be exploited by our rivals, criminals could get 0:31:15.240 --> 0:31:17.240 their hands on it, Like, how do you think about 0:31:17.280 --> 0:31:17.960 those risks. 0:31:18.360 --> 0:31:20.000 Well, I think that there are different kinds of risks, 0:31:20.000 --> 0:31:22.240 so you have to take them one by one. The 0:31:22.360 --> 0:31:27.200 risk of China or a very large sophisticated state that 0:31:27.240 --> 0:31:29.640 has a lot of resources, I think is different from 0:31:29.640 --> 0:31:32.240 like an individual criminal who might just try to do 0:31:32.280 --> 0:31:35.080 something bad. I think to neutralize a criminal, I mean, look, 0:31:35.760 --> 0:31:37.160 you want to do a lot of testing up front. 0:31:37.560 --> 0:31:40.040 We do all this work before we release every single 0:31:40.120 --> 0:31:42.320 one of the models to make sure that we understand 0:31:42.360 --> 0:31:43.840 what the risks are and that we mitigate them as 0:31:43.920 --> 0:31:45.240 much as possible. 0:31:45.200 --> 0:31:45.960 At the end of the day. 0:31:46.480 --> 0:31:48.920 The way that we've approached this with social media, and 0:31:49.000 --> 0:31:51.840 the way that I think society approaches this overall is 0:31:52.040 --> 0:31:55.280 you have kind of better AIS or more resources to 0:31:55.320 --> 0:31:57.840 go fight people who have less resources. So there are 0:31:57.880 --> 0:32:01.000 all these sophisticated folks who try to do bad stuff 0:32:01.000 --> 0:32:03.480 on the social networks, but we invest like billions of 0:32:03.480 --> 0:32:05.720 dollars a year and have these various sophisticated systems. 0:32:05.760 --> 0:32:07.320 So I do think that will play out with AI. 0:32:07.160 --> 0:32:11.040 Too, which is that more sophisticated AIS with more compute 0:32:11.400 --> 0:32:14.960 will be able to generally check less sophisticated folks, whether 0:32:14.960 --> 0:32:17.480 they're just kind of everyday criminals or whatever trying to 0:32:17.560 --> 0:32:20.160 use something but they have less compute. 0:32:19.800 --> 0:32:21.840 So we win the AI wars this way. 0:32:21.960 --> 0:32:25.000 Well, I think we're checking less sophisticated actors with less 0:32:25.040 --> 0:32:27.760 compute that will work most of the time. Now that 0:32:27.760 --> 0:32:29.920 doesn't mean it automatically works for everything. I think a 0:32:29.960 --> 0:32:32.120 lot of attention needs to be paid to it. So 0:32:32.160 --> 0:32:33.960 it's a thing that we need to focus on. But 0:32:34.160 --> 0:32:36.480 I do think that that's how society has been stable 0:32:36.520 --> 0:32:39.720 for hundreds of years, is you have larger forces, whether 0:32:39.760 --> 0:32:42.800 it's polices or armies or whatever, checking kind of smaller 0:32:42.800 --> 0:32:45.640 groups that are trying to create harm. How you deal 0:32:45.720 --> 0:32:50.680 with the geopolitical rivals is a more complicated question, but 0:32:50.760 --> 0:32:52.800 I think there there's the question of what can you 0:32:52.840 --> 0:32:55.280 hope to achieve. If you're trying to say, okay, should 0:32:55.280 --> 0:32:57.600 the US try to be five or ten years ahead 0:32:57.600 --> 0:33:00.520 of China. I just don't know if that's a reasonable goal, 0:33:00.960 --> 0:33:03.080 because I think China has a lot of resources and 0:33:03.120 --> 0:33:06.200 a lot of great scientists, and they're great at espionage, 0:33:06.280 --> 0:33:07.880 right it's like all the stuff. So I'm not sure 0:33:07.920 --> 0:33:10.360 if you can maintain that. But what I do think 0:33:10.440 --> 0:33:14.360 is a reasonable goal is maintaining a perpetual six month 0:33:14.640 --> 0:33:18.080 to eight month lead by making sure that the American 0:33:18.320 --> 0:33:21.600 company is in the American folks working on this continue 0:33:21.640 --> 0:33:24.960 producing the best AI systems, and then having a direct 0:33:25.000 --> 0:33:27.800 effort to try to make it so that those efforts 0:33:27.800 --> 0:33:31.400 are integrating with the National Security establishment, so that way, 0:33:31.520 --> 0:33:34.440 the US government has a kind of perpetual six. 0:33:34.280 --> 0:33:37.080 Month eight month advantage on what everyone else in the 0:33:37.080 --> 0:33:37.680 world gets. 0:33:37.720 --> 0:33:39.760 And at some level six months may not seem like 0:33:39.760 --> 0:33:42.040 a lot, but part of how I think about it 0:33:42.120 --> 0:33:45.040 is like you're using an iPhone, you know all the competitors, 0:33:45.440 --> 0:33:48.360 and it's like, is the iPhone from three years ago 0:33:48.440 --> 0:33:50.280 better than the best Samsung phone today? 0:33:50.760 --> 0:33:51.200 No way. 0:33:51.600 --> 0:33:53.880 But you know, the fact that Apple has just generally, 0:33:53.880 --> 0:33:56.000 in the eyes of most consumers, been a little bit 0:33:56.000 --> 0:33:59.080 ahead at each point has meant that over fifteen years 0:33:59.160 --> 0:34:01.880 or however long, in almost twenty years, they've just had 0:34:01.880 --> 0:34:04.360 this big compounding lead, And I think if the US 0:34:04.440 --> 0:34:07.280 can maintain that advantage over time, that's just a very 0:34:07.280 --> 0:34:08.040 big advantage. 0:34:08.080 --> 0:34:09.440 So that's my philosophy. 0:34:09.560 --> 0:34:12.680 I think it's what I think is the most reasonable 0:34:13.000 --> 0:34:16.520 thing to shoot for, and that optimizes for making sure 0:34:16.520 --> 0:34:18.160 that Americans continue to lead in this. 0:34:19.040 --> 0:34:21.359 Well, look, you're not an AI dooomer, but some really 0:34:21.400 --> 0:34:24.959 influential people are like, how can you be so sure 0:34:25.040 --> 0:34:28.040 there won't be a roboc ellipse or THATAI will lead 0:34:28.120 --> 0:34:29.160 us to human extinction? 0:34:29.480 --> 0:34:31.240 Well, I don't think you can ever be one hundred 0:34:31.280 --> 0:34:33.680 percent sure. There are a lot of bad things that 0:34:33.680 --> 0:34:37.000 could happen. Climate change is terrible, nuclear war is terrible. 0:34:37.480 --> 0:34:41.560 But I generally think you can manage this in a 0:34:41.640 --> 0:34:45.960 responsible way where you just maximize the chance of the 0:34:46.040 --> 0:34:51.120 net prosperity that's created in society being really good, and where. 0:34:50.960 --> 0:34:52.200 You can manage the risk. 0:34:52.280 --> 0:34:56.040 And it's like I think that for the AI risks, 0:34:56.640 --> 0:34:59.480 especially like the existential type stuff that you're talking about. 0:35:00.160 --> 0:35:02.680 I just personally think that we will have more of 0:35:02.719 --> 0:35:05.200 a sense of the different capabilities of these systems with 0:35:05.239 --> 0:35:07.040 each model that we build, and they're getting better, right, 0:35:07.080 --> 0:35:09.080 So LAMA three is better than Lama too, Lama four 0:35:09.080 --> 0:35:11.319 will be better than LAMA three. And this is why 0:35:11.400 --> 0:35:13.800 we do all the safety testing up front. We study safety. 0:35:14.280 --> 0:35:15.799 We try to have as much of a sense of 0:35:15.920 --> 0:35:18.520 here are the different capabilities that would be bad. Right, 0:35:18.600 --> 0:35:20.600 It's like we don't want it to be able to 0:35:20.760 --> 0:35:24.560 lie or self replicate, or like, just is it intentionally 0:35:24.560 --> 0:35:26.960 deceiving people or stuff? And I just don't know that 0:35:27.000 --> 0:35:31.200 we're seeing those things yet. So I think there are 0:35:31.239 --> 0:35:33.680 different questions that happen once you start getting to a 0:35:33.920 --> 0:35:36.279 level of intelligence that's a lot greater than where we 0:35:36.320 --> 0:35:39.359 are now. But there is this myth that I think 0:35:39.360 --> 0:35:43.960 people kind of anthropomorphize and assume that intelligence is going 0:35:44.000 --> 0:35:48.640 to take the form of something that's conscious or sentient 0:35:48.920 --> 0:35:50.239 or physical. 0:35:51.080 --> 0:35:52.360 And I actually think in some ways. 0:35:52.239 --> 0:35:53.880 This has been one of the most surprising things over 0:35:53.880 --> 0:35:56.480 the last few years, is you can have something that 0:35:56.560 --> 0:36:00.840 is pretty reasonably intelligent that is actually completely separate from 0:36:00.880 --> 0:36:04.160 a consciousness. I do think that there's this thing through 0:36:04.200 --> 0:36:09.200 the history of science where people keep thinking that they're 0:36:09.360 --> 0:36:11.920 special or that they're like the center of the universe 0:36:11.960 --> 0:36:15.640 in different ways. And I think people are very special 0:36:16.239 --> 0:36:20.600 and like life and kind of our consciousness and the 0:36:20.600 --> 0:36:23.240 connections we make. I think all that stuff really matters 0:36:23.280 --> 0:36:26.319 in like a deep way. But I'm not sure if 0:36:26.360 --> 0:36:30.680 the specific like intelligence of productivity is actually the thing 0:36:30.719 --> 0:36:33.680 that makes our lives meaningful. It's obviously it's one of 0:36:33.680 --> 0:36:36.480 the defining characteristics of being a person. 0:36:36.640 --> 0:36:37.480 But I don't know. 0:36:37.520 --> 0:36:39.239 I think it's like as much of it as like 0:36:39.719 --> 0:36:43.200 the love and connection and camaraderie that you feel with 0:36:43.239 --> 0:36:45.680 other people. And I don't know, I think that that 0:36:46.480 --> 0:36:49.200 stuff exists, even if you could like completely separate out 0:36:49.239 --> 0:36:51.520 intelligence from kind of the rest of the human experience. 0:36:51.640 --> 0:36:54.719 Yeah, there are obviously a lot of players, you know, 0:36:54.880 --> 0:36:57.560 trying to do, you know, some similar things. What do 0:36:57.600 --> 0:37:00.319 you make of Sam Maltman's leadership. Do you trust him 0:37:00.320 --> 0:37:03.640 and Opening Eye to get to AGI responsibly. 0:37:04.000 --> 0:37:05.400 I know Sam pretty well. 0:37:05.640 --> 0:37:08.520 He's actually on the board of the Biohub with me 0:37:08.719 --> 0:37:11.200 at the chan Zuckerberg Initiative with me and Priscilla, So 0:37:11.680 --> 0:37:14.000 we kind of sit in there and we grill all 0:37:14.040 --> 0:37:16.719 the scientists and push them on using AI in more 0:37:16.719 --> 0:37:19.360 effective ways. And that's kind of a fun way that 0:37:19.400 --> 0:37:21.839 we get to work together. I mean, look, he caught 0:37:21.840 --> 0:37:24.120 a bunch of stuff early on what was going to 0:37:24.200 --> 0:37:26.960 scale with large language models that a lot of other 0:37:27.000 --> 0:37:30.759 people had written off, and I think he deserves a 0:37:30.800 --> 0:37:36.600 lot of credit for how that organization has developed, also 0:37:36.680 --> 0:37:39.760 having gotten a lot of public scrutiny. Myself, I think 0:37:40.080 --> 0:37:41.600 it's like, look, when you're going through it for the 0:37:41.600 --> 0:37:44.759 first time, you don't handle it as as perfectly as 0:37:44.760 --> 0:37:46.839 you would like. But I think he's handling it very 0:37:46.920 --> 0:37:50.720 gracefully and is generally doing I think he's doing better 0:37:50.760 --> 0:37:52.360 than I did, and I. 0:37:52.120 --> 0:37:53.040 Respect him for that. 0:37:53.840 --> 0:37:55.680 Over the long term, there's just the question of which 0:37:55.719 --> 0:37:58.560 model ends up being the right one. And it's somewhat 0:37:58.560 --> 0:38:01.560 ironic thing to have an organization that's named open AI 0:38:01.920 --> 0:38:04.680 but as sort of the leader in building closed AYE models. 0:38:04.760 --> 0:38:07.400 And it's not necessarily bad, but it's kind of a 0:38:07.400 --> 0:38:10.719 little funny. But I'm not like such an absolutist on 0:38:10.800 --> 0:38:13.719 open sources. I think anything closed is bad. I mean, 0:38:14.080 --> 0:38:15.839 at Meta we do a lot of open source stuff. 0:38:15.840 --> 0:38:17.200 We do a lot of closed source stuff, right, So 0:38:17.239 --> 0:38:18.759 it's like I do both. I think I get the 0:38:18.840 --> 0:38:21.560 value of it. I would imagine that open ai will 0:38:21.600 --> 0:38:24.080 continue being an important company for a while to come, 0:38:24.560 --> 0:38:28.800 But personally I am more optimistic about a more positive 0:38:28.800 --> 0:38:31.960 AI future where open source is the industry standard. 0:38:32.000 --> 0:38:33.600 And that's that's just my view. 0:38:33.880 --> 0:38:36.880 I want to talk about the twenty twenty four presidential election. Okay, 0:38:38.080 --> 0:38:41.680 Facebook has been a flashpoint in many elections around the world, 0:38:42.200 --> 0:38:45.319 and you personally have been called out, most recently by 0:38:45.480 --> 0:38:46.600 former President Trump. 0:38:47.760 --> 0:38:50.719 This is a big election. What do you think is 0:38:50.760 --> 0:38:51.200 at stake? 0:38:52.160 --> 0:38:55.640 Well, I mean, look, it's it's obviously a very important 0:38:55.640 --> 0:38:58.480 and it'll be a historic election. And look, I mean 0:38:58.640 --> 0:39:01.399 the main thing that I hear from people is that 0:39:02.000 --> 0:39:04.239 they actually want to see less political content on our 0:39:04.280 --> 0:39:07.080 services because they come to our services to connect with people. 0:39:07.600 --> 0:39:10.320 So you know, that's what we're going to do, where 0:39:10.840 --> 0:39:13.520 we give people control over this, but we're generally trying 0:39:13.600 --> 0:39:16.560 to recommend less political content. So I think you're going 0:39:16.640 --> 0:39:18.520 to see our services play less of a role in 0:39:18.560 --> 0:39:21.239 this election than they have in the past. And personally, 0:39:22.040 --> 0:39:25.120 I'm also planning on not playing a significant role in 0:39:25.120 --> 0:39:27.680 the election. I've done some stuff personally in the past. 0:39:27.719 --> 0:39:30.600 I'm not planning on doing that this time. And that includes, 0:39:30.960 --> 0:39:34.000 you know, not endorsing either of the candidates. Now, Like 0:39:34.080 --> 0:39:35.600 I mean, there's obviously a lot of crazy stuff going 0:39:35.640 --> 0:39:37.319 on in the world. I mean the historic events over 0:39:37.320 --> 0:39:41.160 the last like over the weekend, and I mean a 0:39:41.200 --> 0:39:44.680 personal note, it's you, I mean seeing Donald Trump get 0:39:45.000 --> 0:39:48.920 get up after getting shot in the face and pump 0:39:49.000 --> 0:39:51.920 is fist in the air with the American flag is 0:39:52.640 --> 0:39:54.120 one of the most badass things I've. 0:39:53.960 --> 0:39:54.920 Ever seen in my life. 0:39:55.400 --> 0:39:58.040 At some level as an American, it's like hard to 0:39:58.080 --> 0:40:01.640 not get kind of emotional about that spirit in that fight, 0:40:01.680 --> 0:40:03.080 and I think that that's why a lot of people 0:40:03.800 --> 0:40:07.880 like the guy. But look, I mean, we're living in 0:40:07.920 --> 0:40:10.600 a pretty crazy time, and I view our role here 0:40:11.280 --> 0:40:14.480 is to make it so everyone can express their views 0:40:14.560 --> 0:40:16.560 on this stuff, but we're going to try to manage 0:40:16.560 --> 0:40:18.799 it so that way the politics doesn't drown out the 0:40:18.880 --> 0:40:21.080 human connection and the community, which is I think the 0:40:21.080 --> 0:40:24.200 main thing that people come to our services for. And 0:40:24.840 --> 0:40:26.319 you know, we're not always going to get that right, 0:40:26.480 --> 0:40:28.640 but that's what we're going to try to do, and 0:40:28.680 --> 0:40:30.520 I think that that's probably the best role that. 0:40:30.520 --> 0:40:31.040 We can play. 0:40:32.640 --> 0:40:36.239 President Biden signed a bill to ban TikTok into law, 0:40:36.280 --> 0:40:39.600 but whether it happens is an open question. Former President 0:40:39.719 --> 0:40:43.120 Trump has given a reason to not ban TikTok is 0:40:43.120 --> 0:40:44.560 that he would see the market. 0:40:44.560 --> 0:40:46.680 To you, what do you make of that logic. 0:40:47.800 --> 0:40:48.239 I don't know. 0:40:48.600 --> 0:40:52.080 The national security questions around whether TikTok should be allowed 0:40:52.200 --> 0:40:54.600 is obviously you know, above my pay grid as something 0:40:54.600 --> 0:40:56.800 that the folks in our government and Congress need to 0:40:57.080 --> 0:40:57.880 go figure out. 0:40:58.160 --> 0:40:59.800 From what I see on a day to day basis, 0:41:00.320 --> 0:41:02.640 I think the competition is focusing. 0:41:02.840 --> 0:41:05.440 It's good. I like competing with different companies. I think 0:41:05.440 --> 0:41:08.560 we're doing pretty well here. We're gaining market share. So 0:41:09.280 --> 0:41:11.200 I don't know they'll go do what they need to do, 0:41:11.320 --> 0:41:15.000 but I think you know where we're going to be 0:41:15.000 --> 0:41:18.200 fine and we're going to continue doing well in this 0:41:18.239 --> 0:41:19.160 space either, right, So. 0:41:19.080 --> 0:41:20.680 Any thoughts on whether or not it should be. 0:41:20.640 --> 0:41:23.440 Banned, I really think that's that's above my BIY grade. 0:41:23.520 --> 0:41:27.640 So between looking at the broader social media ecosystem, these 0:41:27.680 --> 0:41:31.120 little apps, between TikTok and Snap and Instagram and threads 0:41:31.160 --> 0:41:34.760 and Facebook, how do you see the social media ecosystem 0:41:35.160 --> 0:41:36.560 changing in an AI world? 0:41:36.640 --> 0:41:39.719 Do new players emerge? Is their consolidation? How do you 0:41:40.120 --> 0:41:41.760 win the battle for the young people? 0:41:42.640 --> 0:41:44.160 And I think it's going to be all of this stuff. 0:41:44.239 --> 0:41:47.120 There's definitely going to be an opportunity for people to 0:41:47.120 --> 0:41:49.000 build new apps. I mean, I think we've seen that 0:41:49.080 --> 0:41:52.920 it's very competitive TikTok guru from being very small to 0:41:53.080 --> 0:41:54.760 now having more than a billion people. 0:41:55.080 --> 0:41:57.560 I'm sure there will be more more apps. There are 0:41:57.640 --> 0:41:58.319 also part of. 0:41:58.280 --> 0:41:59.880 The reason why we want to innovate and push on 0:41:59.880 --> 0:42:02.120 that is that we also want to integrate that stuff 0:42:02.120 --> 0:42:04.319 into our experiences, right. We want to make it so 0:42:04.320 --> 0:42:07.440 that creators can create more interesting contents, that there are 0:42:07.480 --> 0:42:09.120 new ways for people to interact with the. 0:42:09.040 --> 0:42:10.120 People that they care about. 0:42:10.840 --> 0:42:13.640 AI has already been so important for just recommending people 0:42:13.719 --> 0:42:17.919 great content in these platforms. I think that will continue, 0:42:18.239 --> 0:42:20.920 but also there will be another wave where AI now 0:42:21.000 --> 0:42:23.680 is helping people not just get good recommendations, but also 0:42:23.760 --> 0:42:28.640 create new content. So I'm pretty optimistic about that. I 0:42:28.640 --> 0:42:30.040 hope it's not just going to be like one new 0:42:30.040 --> 0:42:32.799 format like video or photos. I think that there's going 0:42:32.840 --> 0:42:35.160 to be I would guess dozens or hundreds of new 0:42:35.360 --> 0:42:39.320 types of content formats, and some will be made by startups, 0:42:39.600 --> 0:42:42.120 and there'll be new apps built around them. Some hopefully 0:42:42.160 --> 0:42:45.360 will pioneer and popularize. But it's going to be a 0:42:45.440 --> 0:42:46.360 very dynamic space. 0:42:46.960 --> 0:42:50.520 We're facing a crisis in mental health, especially a teenagers. 0:42:51.120 --> 0:42:53.440 The Surgeon General is now calling for a warning. 0:42:53.239 --> 0:42:56.800 Label on social media, saying that it's partially to blame 0:42:57.239 --> 0:42:58.480 with everything that you know? 0:42:58.640 --> 0:43:02.560 Now, does you have point? How are you thinking about this? 0:43:03.160 --> 0:43:05.440 Yeah? So I guess I come at this from a 0:43:05.440 --> 0:43:06.920 few directions. 0:43:08.160 --> 0:43:11.879 One is that there's clearly an issue with mental health 0:43:11.920 --> 0:43:14.919 in the country, So I think that's like a really 0:43:14.920 --> 0:43:18.880 important thing, and for kids and teens it's especially important. 0:43:18.920 --> 0:43:20.799 And I think the focus on that is right. 0:43:21.360 --> 0:43:24.960 You know, I have three young girls and being a 0:43:25.040 --> 0:43:28.239 parent is hard, and you just want to make sure 0:43:28.239 --> 0:43:31.719 that they have like good lives. And from that perspective, 0:43:31.960 --> 0:43:34.239 what I aim for us to do is build our 0:43:34.280 --> 0:43:37.799 services in a way that's aligned with parents, giving them 0:43:37.840 --> 0:43:40.360 the controls that they need to basically oversee how the 0:43:40.400 --> 0:43:42.520 services work for their kids. I mean, I think different 0:43:42.560 --> 0:43:44.640 families are going to have different rules for how they 0:43:44.680 --> 0:43:47.279 want this stuff to work, and there's probably not a 0:43:47.320 --> 0:43:49.440 one size fits all thing that's right. But I think 0:43:49.480 --> 0:43:52.720 we have a role in making sure that we study 0:43:52.719 --> 0:43:56.160 the stuff, understand what's good what's not broadly, but then 0:43:56.280 --> 0:43:58.560 for individual families, giving the parents the. 0:43:58.440 --> 0:43:59.400 Controls that they need. 0:44:00.120 --> 0:44:02.760 What the data says today is a little bit different 0:44:02.800 --> 0:44:04.960 from what the basic meme is that's out there. I 0:44:04.960 --> 0:44:06.960 think a lot of people kind of act as if 0:44:07.040 --> 0:44:10.560 there's been this proven connection between these and I just 0:44:10.640 --> 0:44:13.200 don't think that the science supports that. Today It's obviously 0:44:13.280 --> 0:44:16.239 something that will continue to be studied over time. There's nuances. 0:44:16.320 --> 0:44:19.200 I think social media is different from phones overall. It 0:44:19.239 --> 0:44:21.480 might be that phones have an issue, right if you're 0:44:21.480 --> 0:44:23.879 getting notified for something or it's buzzing you and it's 0:44:23.880 --> 0:44:26.480 like preventing you from sleeping. I mean that is different 0:44:26.520 --> 0:44:28.239 from social media. But I think there's all these other 0:44:28.280 --> 0:44:31.680 issues too. But I just haven't seen of all the 0:44:31.680 --> 0:44:35.480 most rigorous studies that have been published, including in the 0:44:35.480 --> 0:44:39.799 most prestigious scientific journals, the link between this on kind 0:44:39.800 --> 0:44:42.200 of a causal basis has not been well established. Now 0:44:42.280 --> 0:44:43.680 that doesn't mean that there isn't work for us to do. 0:44:43.719 --> 0:44:45.200 I mean, we want to make sure we do a 0:44:45.200 --> 0:44:49.560 good job on this, but I think for whether it's 0:44:49.560 --> 0:44:52.560 the Surgeon General or other folks, kind of jumping to 0:44:52.600 --> 0:44:55.560 a conclusion that the science doesn't yet support is not 0:44:55.600 --> 0:44:57.719 the most helpful thing. But look, I mean there are 0:44:57.760 --> 0:45:00.799 clearly issues here across society, and want to make sure 0:45:00.840 --> 0:45:02.400 we're part of the solution to that. 0:45:02.880 --> 0:45:05.400 Facebook, and you have been blamed for a lot of things, 0:45:05.880 --> 0:45:08.040 whether you like it or not, whether you agree or not, 0:45:08.640 --> 0:45:10.160 why should we trust you with AI? 0:45:11.120 --> 0:45:14.440 Well, the loaded question we have gotten blamed for a 0:45:14.440 --> 0:45:16.799 lot of things, and I mean, look, I take our 0:45:16.880 --> 0:45:18.560 role in all this stuff seriously, and I think we've 0:45:18.600 --> 0:45:21.319 tried to handle all this as well as possible. I'm 0:45:21.320 --> 0:45:24.160 not sure that it's all been fair, but look, I 0:45:24.160 --> 0:45:26.440 mean we're in I'd like to think that we're an 0:45:26.440 --> 0:45:28.759 important and relevant company. So I think the scrutiny is 0:45:28.800 --> 0:45:32.360 generally healthy. I mean for AI, I mean look compared 0:45:32.360 --> 0:45:34.480 to the other companies. I think doing this in an 0:45:34.480 --> 0:45:37.520 open way actually increases the chance that it ends up 0:45:37.560 --> 0:45:39.640 being done safely. If you look at the history of 0:45:39.680 --> 0:45:43.840 open source, open source software counterintuitively ends up being more secure. 0:45:43.960 --> 0:45:45.840 At the beginning of the open source movement, a. 0:45:45.880 --> 0:45:47.400 Lot of people thought, hey, it's not going to be 0:45:47.440 --> 0:45:49.200 secure because anyone can just see who the bugs are. 0:45:49.320 --> 0:45:51.000 It's like, yeah, well, anyone can see where the bugs 0:45:51.040 --> 0:45:53.200 are is then you fix them. Whereas in closed systems 0:45:53.560 --> 0:45:55.400 you have these holes that are just out there that 0:45:55.520 --> 0:45:58.160 like governments or hackers or whatever exploiting for a long 0:45:58.200 --> 0:46:00.399 time because there's less transparency on them. 0:46:00.480 --> 0:46:02.520 So I think that that's. 0:46:02.320 --> 0:46:05.200 Going to be one of the defining things around open 0:46:05.239 --> 0:46:09.120 source is that anyone can scrutinize the work, and because 0:46:09.120 --> 0:46:10.600 of that, I think it just puts a lot of 0:46:10.640 --> 0:46:13.080 pressure to make sure that the quality of the work 0:46:13.080 --> 0:46:14.760 that you're doing gets better really quickly. 0:46:15.280 --> 0:46:19.160 And so I guess that's just a big difference. 0:46:18.800 --> 0:46:22.439 In our approach here compared to the approach that others 0:46:22.480 --> 0:46:26.640 are taking, and I think that people should take confidence 0:46:26.680 --> 0:46:29.879 in that, and look, when we get stuff wrong, then 0:46:30.040 --> 0:46:32.279 we're going to get called out on it, and I 0:46:32.320 --> 0:46:34.560 think it'll just generally create these systems that are more 0:46:34.560 --> 0:46:37.000 hardened and more secure and safer. But I know that 0:46:37.000 --> 0:46:38.680 there's the whole debate around isn't going to be safe 0:46:38.680 --> 0:46:41.160 to do open source AI. My personal view is that 0:46:41.200 --> 0:46:44.040 open source AI is going to be safer than closed 0:46:44.040 --> 0:46:47.799 for the exact same reasons that open source historically has 0:46:47.840 --> 0:46:50.440 been safer and more secure than closed source software. 0:46:50.560 --> 0:46:53.000 Let's talk about regulation, because in the shape of AI, 0:46:53.120 --> 0:46:55.600 regulation is going to be an issue around the world 0:46:55.920 --> 0:47:00.279 for the next several years. What should this regulation look like? 0:47:00.600 --> 0:47:03.839 Do you support something like an independent agency that would 0:47:03.880 --> 0:47:07.120 oversee AI like we have for nuclear energy. 0:47:08.160 --> 0:47:10.520 I think it's pretty early and thinking through this stuff, 0:47:10.760 --> 0:47:13.920 and I haven't seen any proposal yet that I look 0:47:13.960 --> 0:47:15.880 at and I'm like, oh, that's exactly what we should do. 0:47:16.200 --> 0:47:18.440 I think there's a time question for this, right. I 0:47:18.440 --> 0:47:20.840 think in a lot of things, if you regulate it 0:47:20.840 --> 0:47:22.920 too late, there could be too much harm. If you 0:47:23.000 --> 0:47:25.680 regulate it too early, you could stop innovation. I do 0:47:25.760 --> 0:47:27.839 think we're seeing this issue in Europe now, where they 0:47:27.920 --> 0:47:29.960 just like put in place a ton of regulations and 0:47:30.000 --> 0:47:32.120 a lot of companies are just not launching stuff there. 0:47:32.160 --> 0:47:35.919 So I mean that's an issue. So I think there's 0:47:36.000 --> 0:47:39.880 clearly going to need to be this interaction between governments 0:47:40.000 --> 0:47:42.280 and companies to go oversee the work. 0:47:42.440 --> 0:47:43.799 I'm not sure what the right model is. 0:47:44.280 --> 0:47:48.239 The thing that I've focused on most is the viability 0:47:48.280 --> 0:47:51.880 of open source, because I obviously believe in this deeply 0:47:52.120 --> 0:47:55.120 and it's both kind of our strategy and approach. 0:47:55.239 --> 0:47:57.120 But I think it's the best. 0:47:56.920 --> 0:48:00.840 Chance for creating an AI future that is positive, where 0:48:01.280 --> 0:48:04.360 the prosperity that is created can be shared by the 0:48:04.360 --> 0:48:07.000 most people, and that we end up with the safest outcome. 0:48:07.040 --> 0:48:08.799 I think that open source will do that. 0:48:09.280 --> 0:48:11.360 But there's a big intellectual debate around this, and I 0:48:11.400 --> 0:48:13.680 think that there's a bunch of people who think, okay, well, 0:48:14.600 --> 0:48:18.520 open sources inherently maybe like a little less controllable because 0:48:18.680 --> 0:48:21.160 there aren't just like a small number of companies that 0:48:21.200 --> 0:48:22.759 you can go choke when you want to, like go 0:48:22.840 --> 0:48:25.279 get them to do something right. So it is like 0:48:25.640 --> 0:48:28.239 a little more forward leaning on. Okay, we're going to 0:48:28.280 --> 0:48:32.200 try to enable this kind of decentralized innovation, and that's 0:48:32.320 --> 0:48:34.879 what is typically worked on the Internet, and I think 0:48:34.880 --> 0:48:37.839 it's what gives the best shot and the scrutiny around 0:48:37.840 --> 0:48:39.360 that I think is the thing that will be the 0:48:39.360 --> 0:48:42.160 most likely to produce a safe outcome. But you know, 0:48:42.200 --> 0:48:44.440 there's a real debate around this today or is like 0:48:44.480 --> 0:48:46.120 should that be the way that it goes now? I 0:48:46.160 --> 0:48:49.600 think that open sources gaining popularity with every day, and 0:48:49.680 --> 0:48:52.000 I think the LAMA three point one release, with the 0:48:52.000 --> 0:48:54.600 four or five billion model that is basically going to 0:48:54.640 --> 0:48:56.759 be kind of this teacher model where like all these 0:48:56.800 --> 0:48:59.920 different companies or institutions, universities, academics are going to use 0:48:59.920 --> 0:49:02.120 it to train different models. I think it's just going 0:49:02.160 --> 0:49:04.520 to keep on gaining more and more popularity and I 0:49:04.560 --> 0:49:07.200 would guess will eventually become sort of the industry's standard 0:49:07.200 --> 0:49:09.440 for how this stuff works. But I just think we 0:49:09.480 --> 0:49:12.000 need to be careful about not doing things that are 0:49:12.040 --> 0:49:14.960 going to prevent what is actually the best outcome over time, 0:49:15.080 --> 0:49:19.120 just because it's maybe like a little less deterministic or 0:49:19.360 --> 0:49:20.040 something like that. 0:49:20.760 --> 0:49:26.360 You've been at this now for twenty years years. Yeah, 0:49:26.640 --> 0:49:27.560 it's been a journey. 0:49:28.719 --> 0:49:33.000 How would you describe your leadership style before versus now? 0:49:34.560 --> 0:49:36.640 I don't know, I probably need to think more about that. 0:49:37.719 --> 0:49:39.800 I mean, the roles have changed so much at the company. 0:49:39.840 --> 0:49:42.600 And when I started early on, I really knew nothing 0:49:42.640 --> 0:49:46.640 about building a company. I was a kid, and I 0:49:46.760 --> 0:49:49.360 was an engineer, and I like coded the first version 0:49:49.560 --> 0:49:52.560 myself and did most of the coding for maybe like 0:49:52.600 --> 0:49:54.960 the first couple of the years of the company. And 0:49:55.000 --> 0:49:59.080 then we got all these amazing people, right. So, I mean, Cheryl, 0:49:59.160 --> 0:50:01.680 I often joke that she raised me as a manager 0:50:02.000 --> 0:50:05.399 like a parent, and I think that that's like really true, right, 0:50:05.400 --> 0:50:07.399 It's like, I mean, I literally I think it's hard 0:50:07.440 --> 0:50:10.080 to overstate how little I knew about running a company. 0:50:10.480 --> 0:50:13.280 And she's just like a really special person has played 0:50:13.320 --> 0:50:15.800 this hugely important role in the history of the company 0:50:15.800 --> 0:50:19.399 and like training me and so many of the other 0:50:19.520 --> 0:50:21.880 leaders of the company. And then you know this funny 0:50:21.880 --> 0:50:24.480 thing happened over the last fifteen years where there were 0:50:24.480 --> 0:50:27.719 all these kids who just kind of grew up and 0:50:28.160 --> 0:50:31.160 have had this context and these relationships with each other, 0:50:31.239 --> 0:50:33.600 and you know, now I think we're all a little 0:50:33.640 --> 0:50:37.320 bit more sophisticated, hopefully about managing something like this well 0:50:37.360 --> 0:50:39.600 and building stuff for the longer term and doing it 0:50:39.640 --> 0:50:42.880 more responsibly. And I don't know, there's like a lot 0:50:42.920 --> 0:50:45.120 of lessons that you learn along the way, you know, 0:50:45.120 --> 0:50:47.399 when I talk to other founders or CEOs and ask 0:50:47.440 --> 0:50:50.160 how I built the team here, that I think is 0:50:50.200 --> 0:50:52.120 really kind of like the key to how we get 0:50:52.120 --> 0:50:54.000 everything done. I mean, there's always a lot of attention 0:50:54.200 --> 0:50:56.840 placed on like the CEO or the top person, but 0:50:56.920 --> 0:50:59.720 it's never one person, right, It's always a group of people, 0:50:59.840 --> 0:51:02.200 and I don't know, it's hard to replace some of that. 0:51:02.239 --> 0:51:03.839 I think a lot of this is like you. 0:51:03.880 --> 0:51:06.400 Just grow up and you learn together, and you bond together, 0:51:06.520 --> 0:51:10.279 and it's like a really kind of interesting and special group. 0:51:10.440 --> 0:51:13.959 Looking forward, will the next Mark Zuckerberg start the next 0:51:13.960 --> 0:51:16.239 Facebook or will Ai do it for him? 0:51:16.360 --> 0:51:17.760 Yeah? 0:51:17.800 --> 0:51:20.520 Well, I think we use technology tools for stuff. I 0:51:20.560 --> 0:51:22.000 think one of the things that's going to be interesting 0:51:22.040 --> 0:51:24.960 in the future is you know, now to build meta 0:51:24.960 --> 0:51:27.480 and the services that we have. We have tens of 0:51:27.520 --> 0:51:29.880 thousands of people working at this company. And one of 0:51:29.920 --> 0:51:31.239 the things that I think is going to be really 0:51:31.280 --> 0:51:34.719 powerful in the future is the next entrepreneur, you know, 0:51:34.760 --> 0:51:37.640 sitting in their dorm room or high school or whatever 0:51:37.680 --> 0:51:39.560 it is, is going to now have all these tools 0:51:39.600 --> 0:51:42.319 to be able to have the productivity of like a 0:51:42.400 --> 0:51:45.200 large company, but maybe it's just them doing it, or 0:51:45.239 --> 0:51:46.240 them and like a small. 0:51:46.040 --> 0:51:47.440 Group of friends working together. 0:51:47.920 --> 0:51:50.160 Part of why I was able to start this is 0:51:50.200 --> 0:51:52.320 because there's the saying in science that you stand on 0:51:52.360 --> 0:51:54.920 the shoulders of giants, and I think that's true with 0:51:54.960 --> 0:51:57.480 open source and technology too. And you know, like I 0:51:57.480 --> 0:51:59.480 couldn't build this myself, I was able to build it 0:51:59.480 --> 0:52:02.200 because there's all this other technology that I can build on, 0:52:02.280 --> 0:52:04.320 and I just think that that's going to keep getting 0:52:04.360 --> 0:52:04.960 better and better. 0:52:05.160 --> 0:52:07.279 So, you know, it's like I watch my kids. 0:52:07.040 --> 0:52:09.880 Be creative and the stuff that they can do today, 0:52:10.440 --> 0:52:11.640 like I wouldn't have been able to do when I 0:52:11.640 --> 0:52:13.600 was a kid because it just didn't exist. So I 0:52:13.640 --> 0:52:15.880 think we're going to live in a profoundly more creative 0:52:15.920 --> 0:52:19.040 future where more people can have the capacity to do 0:52:19.560 --> 0:52:22.239 like pretty amazing things than has been the case in 0:52:22.239 --> 0:52:22.640 the past. 0:52:23.080 --> 0:52:25.720 What about Elon? You're kind of like the anti Elon 0:52:25.840 --> 0:52:29.239 now and it's working for you. Where do you see 0:52:29.280 --> 0:52:30.080 your differences? 0:52:30.560 --> 0:52:31.080 I don't know. 0:52:33.200 --> 0:52:35.759 I know him less well, to be honest, I think 0:52:35.800 --> 0:52:39.040 he's obviously an amazing entrepreneur and has done a lot 0:52:39.040 --> 0:52:41.640 of really great stuff, you know. I mean, it takes 0:52:41.640 --> 0:52:43.919 some courage to go out there and speak your mind 0:52:43.960 --> 0:52:45.600 as much as he does. I don't think like I 0:52:45.600 --> 0:52:47.920 would be comfortable doing that as much. Maybe I'm just 0:52:47.920 --> 0:52:50.440 like a somewhat more reserved person, but I feel like 0:52:50.680 --> 0:52:53.279 all my career people are just telling me like, oh, 0:52:53.320 --> 0:52:55.800 go out like be more yourself. And so there is 0:52:55.840 --> 0:52:57.640 something that I think you have to admire about someone 0:52:57.880 --> 0:53:00.000 who does that and maybe even takes it to an extreme. 0:53:00.760 --> 0:53:03.239 But I don't know what that version of me would be. 0:53:03.320 --> 0:53:04.440 Like, We're going to go. 0:53:04.520 --> 0:53:08.640 Visit you in Tahoe and follow you on your side quests. 0:53:08.600 --> 0:53:10.640 And you're going to surf too, well, I'm going to try. 0:53:11.160 --> 0:53:11.839 I'm going to try. 0:53:12.440 --> 0:53:14.319 You're going to teach me, right, I think that was 0:53:14.320 --> 0:53:16.359 what I was told, but I'm not really I don't 0:53:16.360 --> 0:53:17.400 know if a good teacher. 0:53:17.160 --> 0:53:21.200 But either way, you're always one upping yourself. And I'm 0:53:21.239 --> 0:53:24.400 sure there's a metaphor in there for the company. You also, 0:53:25.160 --> 0:53:26.880 you know, you're doing all these things to make us 0:53:26.920 --> 0:53:29.239 more virtual, but you also love so many things about 0:53:29.239 --> 0:53:32.520 the whole world. How do you wrestle with those two ambitions. 0:53:32.680 --> 0:53:34.000 Well, I don't know that there are odds. 0:53:34.200 --> 0:53:38.000 I just think that people are very physical beings, and 0:53:38.400 --> 0:53:41.680 I think that we sort of mythologize intelligence. I don't know, 0:53:41.680 --> 0:53:43.200 there's like a bunch of people in the tech community 0:53:43.239 --> 0:53:45.160 you think like, oh, we'll just like separate out our 0:53:45.160 --> 0:53:47.640 consciousness and intelligence and like upload it to the cloud. 0:53:47.640 --> 0:53:50.440 And I'm like, that just sounds ridiculous to me. I 0:53:50.440 --> 0:53:52.319 mean to me, it's like part of what makes you 0:53:52.360 --> 0:53:55.440 a person is you like are active and you have energy. 0:53:55.719 --> 0:53:57.560 We're not just minds. 0:53:58.080 --> 0:54:00.640 Like I think the energy and like the love and 0:54:00.719 --> 0:54:04.000 all those things are probably as foundational to what makes 0:54:04.000 --> 0:54:08.560 you a person. And so I don't know, I think 0:54:08.560 --> 0:54:10.520 my life has gotten a lot better. You know, in 0:54:10.520 --> 0:54:12.040 the early days of the company, I just didn't have 0:54:12.080 --> 0:54:14.479 time to do anything else. We were like always about 0:54:14.520 --> 0:54:17.920 to die, and you know, just and it's like being 0:54:17.920 --> 0:54:21.880 a startup is really stressful, right, and obviously there's stress 0:54:21.960 --> 0:54:24.200 now too, but it's just it's a little more managed 0:54:24.200 --> 0:54:26.080 and we have like more good leaders at the company 0:54:26.080 --> 0:54:28.920 and all that. And my life has just gotten so 0:54:29.000 --> 0:54:31.799 much better since I took the time to go make 0:54:31.840 --> 0:54:34.240 sure that I can go do physical things all the time, 0:54:34.520 --> 0:54:36.920 and I think it's made me a better person. I 0:54:37.400 --> 0:54:40.880 injured my knee last year fighting, and I thought, you know, 0:54:40.880 --> 0:54:42.839 Priscilla was gonna give me a hard time about it. 0:54:42.960 --> 0:54:44.719 Like I kind of thought like she was just gonna like, 0:54:44.760 --> 0:54:46.800 you're an idiot. Why are you fighting. You're like running 0:54:46.800 --> 0:54:49.080 this company. It's like you shouldn't be doing this. But 0:54:49.480 --> 0:54:52.239 she actually was like, you know, I know it's a 0:54:52.280 --> 0:54:55.080 long recovery, but when you're done, you better go fight again. 0:54:55.239 --> 0:54:57.320 It's like just because you're like a much better person 0:54:57.680 --> 0:55:01.520 when you're like going and and doing all this physical stuff. 0:55:01.520 --> 0:55:03.800 And it's like it's just kind of like a family value. 0:55:03.880 --> 0:55:06.040 I mean, we do it. I mean I do it. 0:55:06.080 --> 0:55:09.360 Priscilla does it with me. She actually she hits pads. 0:55:09.200 --> 0:55:11.759 That's what I was referring to, tell us, serve with you. 0:55:11.960 --> 0:55:13.880 But you know when when she's sitting pads. You can 0:55:13.920 --> 0:55:16.000 hear it from like down the streets. So I mean she, 0:55:16.239 --> 0:55:19.080 I mean, she's like I think maybe one day we'll 0:55:19.080 --> 0:55:21.960 talk her into fighting. But but she's quite good at surfing. 0:55:22.280 --> 0:55:23.719 And we teach the kids too, and it's just like 0:55:23.760 --> 0:55:26.480 a fun thing that we do together. And I don't know, 0:55:26.600 --> 0:55:28.120 I really believe in that. I think that that, like, 0:55:28.160 --> 0:55:30.120 it just makes you like a better person. 0:55:30.440 --> 0:55:31.480 Tell me about the necklace. 0:55:32.040 --> 0:55:37.120 Oh, this this is something that I worked with a 0:55:37.160 --> 0:55:40.920 designer to make that has engraved on it. The prayer 0:55:41.040 --> 0:55:43.680 that I sing to my daughters every night when I 0:55:43.680 --> 0:55:46.200 put them to bed. It's a Jewish prayer called mesche 0:55:46.239 --> 0:55:52.080 Birach and it's basically a prayer for health and courage, 0:55:52.160 --> 0:55:55.719 and it says, may we have the courage to make 0:55:55.760 --> 0:55:59.279 our lives a blessing. And I just think that's like 0:55:59.400 --> 0:56:01.520 and I've sung it to them basically every night of 0:56:01.560 --> 0:56:02.560 their lives since. 0:56:02.320 --> 0:56:03.480 They were born. 0:56:03.760 --> 0:56:06.560 And I unless I'm out of traveling or something, but 0:56:06.600 --> 0:56:08.960 I try to be around for bedtime. That's kind of 0:56:09.000 --> 0:56:10.520 my thing and when I hang out with the kids, 0:56:11.280 --> 0:56:15.799 and yeah, I don't know, it's just it's meaningful for me. 0:56:16.040 --> 0:56:16.680 In our family. 0:56:16.960 --> 0:56:20.520 That's beautiful. Thank you. Thank you for sharing your time 0:56:20.520 --> 0:56:20.879 with us. 0:56:20.880 --> 0:56:23.719 Thank you for letting us into your home, and you know, 0:56:24.239 --> 0:56:25.680 taking the time to explain all of. 0:56:25.680 --> 0:56:27.520 This, like I think it's really important. It helps us 0:56:27.520 --> 0:56:29.160 to get to know you cool. 0:56:32.200 --> 0:56:34.480 Thanks so much for listening to this edition of the Circuit, 0:56:34.800 --> 0:56:37.600 and please watch our video episode with Mark Zuckerberg on 0:56:37.640 --> 0:56:41.200 Bloomberg Originals. I visit his retreat in Tahoe, we hang 0:56:41.239 --> 0:56:44.560 with his wife Priscilla, and yes we go wait surfing. 0:56:44.800 --> 0:56:47.560 You'll see how that works out. I'm Emily Chang. 0:56:47.600 --> 0:56:49.880 You can follow me on Twitter and Instagram at Emily 0:56:49.960 --> 0:56:53.080 Chang tv and watch new episodes of the Circuit on 0:56:53.120 --> 0:56:56.960 Bloomberg Television, streaming on the Bloomberg app or on YouTube. 0:56:57.400 --> 0:57:00.360 And check out other Bloomberg podcasts on Apple podcast, the 0:57:00.400 --> 0:57:03.880 iHeartMedia app, or wherever. You listen to your shows and 0:57:03.960 --> 0:57:05.680 let us know what you think by leaving a review. 0:57:05.760 --> 0:57:09.160 Those extra reviews make a big difference. I'm your host 0:57:09.160 --> 0:57:12.680 and executive producer. Our showrunner is Lauren Ellis, our editor 0:57:12.760 --> 0:57:13.680 is Alis and Casey. 0:57:14.120 --> 0:57:14.959 Catch you next time.