The Future Mark Zuckerberg Is Trying To Build

Show video

I'd love to start with these. 10 years of work  right there. Someone on your team called these   the real life Tony Stark glasses. Very hard  to make each one of these... That makes me feel   incredibly optimistic... In a world where AI  gets smarter and smarter... This is probably   going to be the next major platform after  phones... I miss hugging my mom. Yeah haptics  

is hard... How does generative AI change  how social media feels?... We haven't found   the end yet... The average American has fewer  friends now than they did 15 years ago. Why   do you think that's happening? I mean  there's a lot going on to to unpack there... I'm about to interview Meta CEO Mark Zuckerberg. There are not   that many people with more power over what our  future might look like. Nearly half the total human  

population now uses Meta products and I just  tested some of their new tech that feels like   science fiction. This is crazy! Mark Zuckerberg and  the team at Meta are imagining a future that billions   of other people might actually end up living in. So  my goal for this conversation is to try to figure   out what that future really looks like. To paint a  picture of the future Mark Zuckerberg is trying to   build so that you can decide for yourself what you  think of it. Welcome to the first episode of our   new series, Huge Conversations Hey, good to meet you! Thanks for doing this. Yeah looking forward   to it. Awesome. I'd love to tell you what my goal  is of this conversation. Go for it. We have a called  

huge if true which is this very optimistic about  science and technology and the potential futures   that we can build and in every episode we're sort  of exploring what does it look like if you play a   certain technological future out and so my goal  in this conversation is to try to help people   see the future that you're imagining when you're  building the products that you and the Meta team   are building. What are you imagining this looks  like in future? How are you imagining people use   this? All of that. Cool. All right awesome. So  I'd love to start with these. Let's do it. 10 years   of work right there! I got to demo them a little  bit earlier today. I heard someone on your team  

call these the real life Tony Stark glasses? We're  getting there. But I'd love to just hear in your   voice what are these? Well these are the first full  holographic augmented reality glasses I think that   exist in the world. We've made I think it's a  a few thousand or something right. Very hard to   make each one of these but this is the culmination  of 10 years of research and and development that   we've done to basically miniaturize all the  computing that you need to have glasses not a   headset but glasses that can put full holograms  into the world with a wide field of view. So you   can imagine sort of in the future we'd be having a  version of this conversation where you know maybe   I or you are not even here it's like one of us is  physically here and the other one is here as a as   kind of a full body hologram and it's not just  a video call you can actually interact you can   do things I mean in the the demo we had the you  know ping pong and games and things like that but   I mean you could you can interact you can work  together you can you know play poker play chests   whatever like the holographic cards holographic  board game. I just think it's going to be wild.

it's going to remake I think so many different  fields that we think about today from how we work   and productivity to a lot of things around science  a lot of things around education entertainment fun   gaming. But this is just the beginning you  know this is the first version, it's a   prototype version that we've made in order  to develop the next version which is hopefully   going to be the consumer one that we sell to  a lot of people. Why build these? Well I think   it's going to be the next major computing platform.  So if you look at like the grand arc of computing  

over time you've you've gone from like main  frames to computers that basically like live   on you know your desk or on a tower to phones  that you have in your hand that you basically   like you know can take with you everywhere that  you want but it's it's pretty unnatural right it   takes you away from the world around you and. I  think that the trend in computing is it gets more   ubiquitous it gets more natural and it just  gets more social right so you want to be able   to interact with people in the world around you  and I think that this is probably going to be   the next major platform after phones. I'll give  these to you. These are the clear ones that show   all the... The whole thing is a special edition and  this is like a really special edition. There's   not a single millimeter of of space. You know  everything in here from the micro projectors that   um basically shoot light into the wave guides  right it's a special type of display system. I   mean these aren't normal displays like you have  in a phone or a TV or computer like the type of   displays that people have been building for  decades. It's a waveguide system. The projector  

that's shooting light basically goes into these  nano etchings across the wave guide that are what   catches and creates the holograms. In order to  synchronize that with your where you're looking   there's eye tracking and little cameras,  they illuminate your eyes and then of course   there's all the basic stuff that you need all the  computing, the batteries to power the whole thing,   microphones, the speakers because it needs to be  able to play audio and speak with you and the   cameras and sensors to see things around you in  the world so that way when it's placing holograms   in the world it can do that in the right place  and understand where you are so that probably   is still not covering everything because there's  a lot of things that need to go into syncing up   the holographic images between the two displays  because you don't just have a single display   like you have in a phone or TV you have two and  it moves around and you know physical things   are hard and need to be synced up. There's also  the radio that has to communicate with your other   computing devices to do heavier computing um and  the wrist based neural interface that you probably   got to try out. We kind of miniaturized all of this  and fit it into uh you know normal looking pair of   glasses which is... you know when I told the team  that we were going to do this 10 years ago you  

know people weren't sure if we were going to be  able to but I think you not only we're going   to be able to do this but I think we're going to  be able to get it cheaper and higher quality and   even even smaller and more stylish over time. So  I think this is going to be a pretty wild future.   There are so many versions of trying to get  a similar idea of digital objects in physical   space. I'm thinking of for example of glasses that  have heads up displays where it's headlocked and   it's moving with my eyes, glasses that are really  creating digital objects in physical space that   don't move as I move, I'm thinking of these, I'm  also thinking of the Snapchat Spectacles that they   just announced, then on the other hand there are  headsets like the Quest and also like the Apple   Vision Pro that seem to fall into a different  category. I'm curious how you would organize this   landscape for people and how you think about  people using these tools in their real lives   in the near future? Yeah so when we were getting  started on this about 10 years ago I thought that   something like this was going to be the ultimate  product for everyone. Right you get to you know  

normal looking pair of glasses and we'll continue  improving that that can have full holographic   images. I think it's super powerful  and it is sort of the science fiction future that   I think we all hope to get to. On the journey we  took a few other approaches as well um to help us   develop towards that including building glasses  that don't have displays to try to learn. Just   take a stylish pair of glasses today and put as  much technology into it as you can but really   focus on the form factor and that's the Ray Ban  Meta glasses and it's doing really well and   initially we thought that that was sort of intro  product for us to learn how to build this but one   of the things that's clear now is you're going to  be able to make that product a lot more affordable   than this probably permanently. So I actually think  that there are going to be a bunch of different of   these paths that we've taken are going to be  kind of permanent product lines that people   will choose. I think you'll see display-less glasses  like the Ray Ban Metas continue to get better and  

better, great for AI, no display but you can talk  to it, it can talk back. I think there's going to   be something in between these that's basically a  heads up display, so it's not a 70° field of view,   maybe it's a 20° or 30 degree field of view,  so that's not going to be what you want for   putting kind of a full hologram of a person or  interacting with the world around you but it's   going to be great for you know when you're talking  to AI, not just having voice but also being able   to see what it's saying or being able to text  someone with your wrist-based neural interface and   then have their text show up rather than having it  read to you, which is, we read faster than we   can listen or getting directions right or just  being able to search for information get all that.   So there's a lot of value for heads up display  that will be somewhat more expensive than the   display-less but somewhat cheaper than this.  Then I think you're going to get this. It's going   to be probably the most premium and and expensive  of glasses products but hopefully still something   that you know like a computer is generally  accessible to most people in the world but I think   that there are going to be all of those and I  I think people will like them. I also think that   the headsets that people are using around mixed  reality will continue to be a thing too because   no matter how good we get at miniaturizing  the tech for this you're just going to be   able to fit more compute into a full headset.  Fundamentally our mission is not you know build  

something that is advanced and only a few people  can use, we want to take it you the last mile and   do all the innovation to get it to everyone. We  you know just shipped or announced Quest 3S,   the new mixed reality headset where we basically are  delivering high quality mixed reality for $299.   I was really proud last year when we delivered  Quest 3, the first kind of really high quality   high resolution color mixed reality device for  $500, right it was like, it's like a fraction of   the cost of of what the competitors are doing  and I think it's actually higher quality in a   lot of ways, and now we've just doubled down on  that. So I think that they're all actually going   to end up being important long-term product lines:  display-less, heads up display, full holographic   AR, full headsets. I think that they're all going  to be important. Yeah. If you play out the future  

of not just the hardware that we've been talking  about so Meta Ray Bans, Quest, Orion, but also   the Llama models, if everything goes according to  you and the teams wildest dreams, I'd love for   you to just begin to describe what that feels like.  I mean I think that there are two primary values   that we're trying to bring. On the AR and kind of  mixed reality side, the main value we're trying to   bring is this feeling of presence .Right so there's  something that I think is just really deep about   being physically present with another person that  you don't get from any other technology today and   I think that's the thing when people have a very  visceral reaction to experiencing virtual or mixed   reality what they're really reacting to is that  they actually for the first time with technology   feel a sense of presence like they're in a place  with the person and that's super powerful. I  

focused on designing social apps and experiences  for 20 years that's sort of like the Holy Grail   of that is being able to build a technology  platform that delivers this like deep sense of   of social presence. The other big track is around  personalized AI and for that and that's sort of   where Llama and Meta AI and all those things are  going. There's all this development that's going   into making the models smarter and smarter over  time but I think where this is going to get   really compelling is when it's personalized for  you and in order for it to be personalized for you   it has to have context and understand what's going  on in your life both kind of at a global level and   like what's physically happening around you right  now and in order to do that I think that glasses   are going to be the ideal form factor because  they're positioned on your face in a way where   they can let them see what you see and hear what  you hear which are the two most important senses   that we use for for kind of taking information  and context about the world. I think that this is   all going to be kind of really deep and profound  stuff but it's basically those two things: It's   this feeling of presence and this capability  of really personalized intelligence that can   help you. I'd love to talk about each of those  two things. The first on presence, I owe a lot   to being able to connect with people online. Right  this job that I have is by definition that, also with   my family. My parents don't live anywhere close  to me. I video call them a lot and when I think  

about the progress of technology like this in a  timeline from the telegram to the telephone   to video call to some feeling of presence with  another person who's feels like they're right   there in front of me, that makes me feel incredibly  optimistic. I would love a future where like I can   lose in Scrabble to my mom and feel like she's  really there in front of me. Yeah and it feels like   we're not that far away from something - I agree! - that persuades my brain that that's happening. Yeah  

totally. And also I miss hugging my mom right like  that never goes away. Yeah haptics is hard. Yeah and   so my question is about that  it's about this this feeling of like it's hard   for me to imagine um a future where real physical  presence is not different and special in some way   where I don't miss literally hugging my  mom and I'm curious how you think about the   parts of human connection that are eye contact and  physical touch and things that our ape brains   value for connection with other people. Yeah well eye contact I think we're going to get to a lot before   the the touch part. For haptics I do think we'll  make progress on that but it's it's obviously   there's a spectrum there too from kind of hands  which is where if you you draw out the kind   of like homunculus version of a person in terms  of like what are what are our kind of sensory you   know what what's like the majority of what we're  sensing it's like yeah yeah so I think being   able to do that for your hands is probably the  most important place to start and you have a rough   version of that with controllers today. I think  that that'll get even more over time. We have this   demo playing pingpong where you have a controller  where as the digital ball hits the ping pong   paddle you feel it hit the as if it's hitting the  ping pong paddle wherever it is so you actually   have a sense of like where it's it's hitting  the the the paddle so I think that was that   was just a wild demo so I think we'll get some of  that the most extreme version of this is wanting   force feedback right so I mean like for doing a lot  of sports right it's it's like okay we can kind of   do a good approximation of like boxing today or  you get like good feedback on your hands but it   would be hard to do a virtual reality version of  Jiu-Jitsu where you're like grappling with someone and   you need like real kind of force feedback on  that so that's probably like the hardest thing   right to go do but I think we'll get there.  You know I think like most science fiction it's   not this binary thing that you just like wake up  one day and we're like oh we've realized all the   dreams but but I I do think that these platforms  are going to be the first time that I think that   there's a realistic sense of presence in all  the ways that that's special to people for   most things that people want to do which are not  the most physical ones and even some of the basic   physical ones I think we'll get. But then there's  a long tale of other stuff I mean smell is also  

really important for people yeah right it's  I think it's disproportionately important for   memories and that's not really a thing that  I think in the next few years we're going to   have in any of these devices I mean that's a very  difficult and challenging thing on its own. What is   the piece of that that you feel most interested in,  that you keep coming back to in your mind? This has   the frustrating property to develop that the  sense of presence is almost like when you're   designing something that that's sort of trying  to artificially deliver it you're delivering   an illusion to a person and more than any one  thing that provides a sense of presence it's   actually more the case that any one thing done  wrong breaks the sense of presence. You kind   of know that you're interacting with technology  but it's so convincing that um that you just kind   of go along with it. You're like okay yeah no this  person feels like it feel like they're there right.   When I did that pingpong demo I like at the end  of it I dropped the pingpong paddle on the virtual   table and it shattered so that was not the best  for for our internal development but   like that's winning in our in our development  right it's like when when you feel like something   is is kind of so realistic that you you're just  convinced that um that it's there now and there   are a lot of things that can break that right so  I think a a field of view that's too low right so   something feels real but then you turn your head  and it's not there um latency read physics that   don't behave like realistic physics. It also is  interesting in some ways what people can accept   as physically real even though it's not right  so like we've done a ton of work on avatars we   we have this whole work stream on Kodak avatars  to do these photo avatars and it's I think it's   going to be incredbly compelling and people are  going to love it but one of the things I found   interesting is the ability to mix photorealistic  and expressive kind of the cartoony avatars with   photorealistic worlds and kind of more cartoony  computer game type worlds so you can have the   a Kodak kind of photorealistic avatar of a  person in what is clearly like a video game   or cartoon world and people are generally  pretty fine with that it's like okay that   that feels pretty good and similarly having  a photorealistic world but good increasingly   good kind of cartoon avatars as long as the  avatars move in a way that feels authentic to   the person you're interacting with it actually  feels pretty good you know it's when you look at   a 2d still frame of it some of the stuff can  look a little bit silly and and we've certainly   you know had had a our share of memes around  that but um but when you're in there you know and   you you've played around with lot of the stuff it  feels realistic because it's basically mimicking   the kind of authentic mannerisms of of a person  that you're interacting with and even if it's not   a Kodak photo realistic avatar if it's kind of  a more cartoony expressive one so I I think that   that's it's very interesting to see kind of  which pieces you need to unlock and what where you   just need to be like very technically excellent  and consistent but it's um this isn't a space   where it's like you deliver one thing and it's  good this is like there's a wide breth of things   that you need to nail and then have it all come  together and that's why these are you know 10 year projects. It seems like an interesting way to learn  about the human brain and what we actually care   about with respect to what feels real. I was  wondering about, there was this moment in an  

interview that you did with Lex Friedman, you quoted  research that says that the average American has   fewer friends now than they did 15 years ago  and I was so interested in that because   it seems like if we want to get to a world where  there's more human connection this is the trend   that we're going to have to grapple with and just  to give some data on this in the American Time Use  Survey over the last 20 years the amount of time  American adults spend socializing in person has   dropped by nearly 30%. For ages 15 to 24 according  to the Surgeon General it's nearly 70%. and I   look at that data and I think to myself well maybe  if we're all socializing digitally that doesn't   matter so much maybe there's a future where that's  actually fine but there's also data that suggest   that we're struggling somewhat. The number of  Americans who say that they don't have a single   close friend - yeah it's really sad - that share has jumped from 3% to 12% in the last 30 years. It feels to

me like with all the tools that we've built for human connection, we're struggling to connect and I'm curious   why do you think that's happening? I mean there's a lot  going on to to unpack there. A lot has changed   sort of economically and socially during that  period and a lot of those trends go back before   a lot of the modern technology. So I mean this  is something that a lot of academics and folks   have have studied but it is an interesting lens  to look at this though because I think whenever   you're talking about building digital types  of connection one of the first questions that   you get is is that going to replace the physical  connection and my answer to that especially in the   case of something like this is that no because  people already don't have as much connection   as they would like to have. It's not like this is  replacing some sort of better physical connection   that they would have otherwise had. It's that the  average person would like to have 10 friends and   they have two right or three and there's just  more demand to socialize than what people are   able to do given the current construct and giving  people the ability to be present with people who   are in other places physically just seems like  it will unlock more. It's not going to make it  

so, if I have glasses, it's not going to make  it that I spend less time with my wife, it's going   to make it so that I spend more time with you  know my sister who lives across the country. And   that's, I think that's good. I  think people need that. As for the rest, I   I think we could probably spend a multi-hour  podcast just going into all of the different   kind of socioeconomic political dynamics that are  going on but none of the trends that I've seen   does it seem like the primary thing that's going  on is that because people are interacting online   they're now not interacting with their with people physically. Now certainly I think you   you I do interact with people online who I also  like to interact with physically but and I think   that that's kind of like a combination um like  more combined richer relationship that you have   overall but I think that there's a lot going  on with the loss of of kind of social capital and   connections that really predates a lot of the  modern technology. The goal of what, I'm what I'm   trying most to learn about is how we can structure  the technologies that we use in the future to get   toward this future I think you're imagining of  more human connection in more ways. I'm curious, you  

brought up the other big pillar of AI and in some  of your conversations, I'm thinking of a conversation   with Tim Ferris in particular, you talked about  a lot of different use cases of AI and they seem   to me to fall on somewhat of a spectrum. Like  for example you mentioned automatic real-time   translation, like basically the Star Trek  Universal translator. We're pretty   much there! Yeah and that's one example on one  end of the spectrum where some people might argue   that there is a chance that someone is less likely  for example to learn a language because we can all   speak to each other in real time in different  languages. I think nobody would really argue  

that therefore we shouldn't have that kind of  universal translator. People still learn Latin and   Greek. Right exactly and so I think that end  of the spectrum is something like um technologies   that really measurably unlock our humanity because  they remove a struggle between people and then on   the other end of the spectrum there are a lot of  educational things for example where the struggle   is kind of the point right? Like it's like building  a muscle. I can think of so many times  

in my life where like the reason why I was doing  something was not the output it was the fact that   I was trying so hard to do it. There's one example  in the Tim Ferris interview where you talked   about your kids struggling to articulate  themselves emotionally and adults very much had   the same problem and you talked about AI as a way  to help them articulate those emotions. Yeah and   I thought about all of the many times in my life  where I have struggled to articulate my emotions   and how I really could have used some help in  those moments and I also found myself thinking   about the times when that was really building  a muscle where like the act of struggling to   communicate with someone and understand what they  wanted from me was was important to my development.   And so my question is if you think about that  as a spectrum between things that are really   important to our humanity where and the struggle  being removed is helpful versus things where the   struggle is the point and it unlocks  something about our humanity and is important   to preserve like building a muscle, how do you  draw the line between those things and how do   we ensure that the muscles that we're building for  this future are stronger and not weaker? Yeah it's   interesting I mean I think we're always going to  find new things to struggle with and I mean it's   you can always get better at communicating with  other people and kind of expressing yourself and   understanding other people so having a tool that  can help you do that better isn't going to mean   that like oh now we perfectly understand every you  know it's I mean I think the maybe one   of the most functional aspects of this you're  already seeing a lot of these AI models really   help people with coding right like a generation  ago um before I was getting started a lot of   coding was like really low-level system software  and you know then by the time that I got into   it there was a little bit of that but um you you  can make websites pretty easily make apps pretty   easily and I think in 20 years or a lot sooner  than that you're going to basically be in a   world where kids will be able to just describe the  things that they want and build incredibly complex   pieces of software so it's um in that world  are kids going to be not struggling I I don't   think so I think that they're going to be just  expressing their creativity and and it'll it'll   be this kind of constant iterative feedback loop  around like okay like yeah I you know took a few   minutes to describe this thing and like yeah this  whole like amazing virtual world was created that   I can have see on my glasses or whatever but like  these things are not exactly what I want them to   be so now I need to like go back and edit them it  just I don't know I think that there's always   more. Another way to get this - it's one of the things  that I think makes makes people so good. It just   there's there's always more to do. We'll always  find the struggle? Yeah. Another way to get at this   is if you if you play this out to make the  tools even better in like 10 years let's say   your kids are in high school are there ways that  you would want them using AI because you think it   would accelerate them intellectually and ways that  you would advocate for them not to use it or   things that you would have concerns about? I mean  I think that there's some things that you need   to be able to do yourself. I think that's a lot of  the basic fear that people have around this is 

that while we're building these amazing tools we  get away from this self-confidence and ability of   being able to do like this basic stuff yourself so  it's like all right you have a calculator but it's   still good to be able to do kind of basic math in  your head because there are a lot of things that   come up throughout the day that you just want to  have a general numeracy around right that often   they're not expressed in numerical terms but just  in terms of understanding trends or understanding   arguments that people are making, you you kind of  need to understand the shape of how numbers come   together and so I think one of the big debates  is like should we still teach our kids to program   computers even though you're going to have these tools in the future that are just so much   more powerful than anything that we have now to  produce incredibly complicated pieces of software.   I think the answer to that is probably yes  because I think teaching someone how to code   is teaching them a way to think rigorously and  that even if they're not doing most of the code   production I think it's important that you kind of  have the ability to think in that way and I think   it's going to just make you generally a better  thinker and better person so yeah maybe that's   like this generation's version of calculators  it's like so you you want to you want to use the   calculator but you'll also want to be able to  generally do without it. Other ones like language   I don't know I mean different people can come  out I think this is one of the interesting   questions about parenting these days is like is  is just kind of like what what's important to   teach your your kids and in an era where so much  is going to change over the the time that they're   even in school. Language I think you can make  similar arguments. I think there's a lot of it's   like it's probably going to be less functionally  relevant in the future to learn multiple languages   but it sort of helps you think in different ways, you know I found from the languages that I've   studied that a lot of it you learn about  the structure of your own language, you can   you know you also learn about the culture right  because so much of how things are expressed in   different places is tied to the nuance and the  history of kind of what how so I think   like you that's all valuable and interesting  stuff to get into but then I don't know at the   same time we only have so many hours in the day  so people need to prioritize what they're going   to learn and it may be that okay in a world with  perfect translation which by the way we basically   just announced on the Ray Ban Metas that now  you're going to be able to just like you go to   countries yeah we're starting out with just a few  languages but we'll roll it out to more and you   know you'll be you could be traveling anywhere and  you have your glasses and they just translate in   real time in your ear. So it's wild, yeah so  I think people are going to need to choose what  

what what they want to focus on going forward.  How do the developments that we've been talking   about in AI intersect with social media and the  platforms that most people use today? There's a  future where there's images and generated text  and maybe AI influencers. How does generative   AI change how social media feels in the future?  Yeah I mean I think that that's a really   deep one. You know there's already been one  big shift which is that social media started  

out as people primarily interacting with their  friends and now it is you know at least half of   the content is basically people interacting  with creators or content that's not created   by people who they kind of personally know so  we sort sort of already have that paradigm and   I think AI is probably going to accelerate that. It  will give all these people additional tools right   so your friends will create kind of funnier memes  and more interesting content um that'll come from   a lot of different ways. I think some of it will  be okay your friends have glasses and they capture   a bunch of stuff and before they might have not  been able bble to edit it to make it interesting   or maybe it was just too much work or they didn't  even realize that they captured something amazing   but now the AI is like hey I like made this thing  for you out of your content um it's like okay   that's awesome like people will enjoy that. Creators  obviously kind of much more specialized skills   are going to be able to use even more advanced AI  tools to make more compelling content but then I   think that there will be a bunch of kind of green  field type stuff where maybe in the future there   will be content that is purely generated by AI  by the system personalized for you maybe it's   summarizing things that are out there that that  are going to be interesting maybe it's um just   producing something funny that makes you laugh  this is going to be like a very kind of deep zone   that there's a lot to to experiment with.  I think there are going to be AI creators as well, 

as creators building AI versions of themselves,  I mean that's a thing that we just showed too   at Connect is basically I mean if you're a Creator one of the big challenges is   like all right there are only so many hours  in the day and your community probably has a   nearly unlimited demand to interact with you and  you want to interact with them because you're   trying to grow your community. I mean that's both  socially and from a business perspective that's   sort of you know growing the community is an  important part of what every creator does so   okay if we can make it so that each creator  can basically make an like an AI artifact   that their community can interact with people be  clear it's not the actual creator themselves but   it's almost like a piece of digital art that  you're producing like an interactive sculpture   or something that it's like it's like you train  it to here's the context that I wanted to have   here's the topics I wanted to communicate  on here's stuff that I wanted to stay away   from you're giving your community something to  interact with when you can't be there to to kind   of answer all the questions and I think that's  going to be super compelling so there's like   these interesting things but I think it's I AI  it's kind of like the internet in a way where   it's probably going to change almost every field  and almost every feature of every application that   we use um it seems sort of hyperbolic to say that  but I do think that's true and it's just hard to   sort of enumerate all the different things up  front but I think that over the next 5 to 10   years we're just going to explore the impacts  in each of these areas and it's going to be   like an amazing amount of innovation and really  exciting. I feel two things simultaneously when   you say that. I feel both like I really want  to be optimistic about the future of these  

platforms and I obviously have gained so much from  an enormous pace of change right like everything   that we're doing now and what I actually feel is  worried. I feel some specific concerns around the   way that you know I might communicate with an  audience and the way that they might respond to   that or the way that human communication might  change but also more generalized just sort of   fear of the pace of change and and worry and I  don't think I'm alone in that feeling. Yeah and   you're supposed to be the optimist! I know! And I'm  curious like how you talk to people who feel that   way. What concerns do you feel are most legitimate  and what do you feel most misunderstood? I think   the pace of change is always a concerning  thing right it's there is a lot of uncertainty   about how how things will go in the future and  we're all going to get really amazing new tools   to do both our hobbies and our jobs and they'll make it so we can do better work and   have better lives but at least on the professional  side it's going to be our responsibility to keep   up with that or else it's going to be difficult  for us to compete with other people who are   doing a good job of kind of keeping up with  the new trends. So I get it. I mean I think   you know especially in the you know line of of  work of being a creator and it's a very sort of   competitive space, I don't think that like creators  necessarily think about it as competitive but it   is right it's like it's you know and um and so I  get it. I think that this is going to make it so   that like the quality of work that people produce  and how interesting it is and how much they can   communicate and like really efficiently is is  just going to kind of go through the roof but   but when you're staring down a set of changes like  you know that there's some big change coming and   you don't know what it is that's always a time of  anxiety so I get it. If I take my creator hat  

off and I'm just a person who is youngish starting  out my career-ish, starting out building a family,   how would you advise someone like me to prepare  well for the future that we're headed toward   to be able to learn new skills now or just think  about this future in an educated way? Yeah I mean   I just think maintaining curiosity about things is  is important. I do think we can overstate to what   extent the next 10 years is going to be sort of  different from the last 10 or 15. I mean a ton of   stuff changed over the last 10 or 15 years too.  It's not like this is the only time in history   where there's some technology it's going to make  it so there's new opportunities and things change   the internet coming into maturity and everyone  having smartphones has already rewired things   dramatically and I mean maybe the next period will  be a somewhat bigger change or maybe it won't I   think it'll feel different to different people but I don't think this is like going from zero to   one it's not like okay everything's just kind of  been normal and now like now it's about to change   it's like the technology of evolves over time and  and like the opportunities that we have evolve and   improve and I think that's like the people who  do well I think are are people who are generally   curious about it and and dig in and and try  to use it to live better lives rather than the   people who who basically you know try to fight it  in in some way. One thing that I really want to ask   you about is open source. Yeah. I think imagine that  we're talking to an audience that has maybe heard   that term but doesn't have any real idea of how  that might impact them in the development of AI. 

How would you explain the reasonable debate  that people in your field are having about this   right now? Well I think there are two pieces. I mean  so what does open source mean? It means that people   can build a lot of different things right so at  a high level I look at the vision that a bunch of   companies have right so Open AI, Google, they're  building an AI right like one AI that I think   in general they're like okay this is going to be  it's like you're going to use they think you're   going to use Gemini or ChatGPT for like all the  different things that you want to interact with   and at a high level that's just not how I think  the world is going to go. I think we're going to   have a lot of different AI systems just like we're  going to have we have a lot of different apps.   I think in the future every business just like  they have a website and a phone number and an   email address and a social media account is also  going to have an AI that can interact with with   their customers to help them sell things to help  them do support. I think a lot of creators will   have their own AIs right I think like a lot  of people will interact with with a bunch of   different things. There's a question of okay do you  want a future that's fundamentally kind of very  

concentrated and where you're interacting with  kind of one system for everything or do you want   one where a lot of different people are building a lot of different AIs and systems just kind   of like you probably didn't want there to be you  know just one app or just one website. It's like a   richer world when there's a diversity of different  things so that's one piece is is just giving   people the ability to build it themselves and  what open source does it makes it that everyone   can take and modify the model and build stuff on  top of it which is different from the kind   of closed and centralized approach. The safety  debate is a specific part of this which is in a   world where AI gets smarter and smarter, what's the  way that we have the highest chance of of having a   a a kind of positive future and and not having  a lot of the safety concerns? And I think some   people think that if we keep the model closed  and don't give it to a lot of developers that   should make it safer because then you don't get  bad developers doing bad things with the model.

Historically I think what we've seen with open  source is actually the opposite which is that   this is not the first open source project right  I mean this is obviously this has been a thing in   the industry for decades and I think what we've  traditionally seen is that open source software   is safer and more secure largely because you put  it out there more people can scrutinize it because   they can see all parts of the system and then  there are inevitably issues with any software   there are bugs there are security issues and  initially with open source people thought hey if   you're putting the software out there and there  are holes in it isn't everyone just going to go   exploit those holes and especially the bad  guys but it turned out that it sort of in this   counterintuitive way that by making by adding more  scrutiny to the systems the holes became apparent   quicker and then were fixed and then people  roll out a new version just like we roll out   a new version of our models right Llama 3, Llama  3.1, Llama 3.2 everyone upgrades, so I think the   same thing is going to happen here I think it's  sort of this counterintuitive thing where even   though I I think there's some concern around all  right are bad guys going to do bad things with   these models. I actually think you just get a kind  of smarter and safer model for everyone the more   it's rolled out and the more kind of scrutiny  is on it and then part of that is we get   feedback and we make the model safer so that is  we roll it out to to more people it's safer   for more people to use. So I think that the history  of open source in the software industry generally  

would suggest that open source is going to lead  to a more prosperous and safer future. Our show   is called Huge If True and what I mean by that is  kind of testing the most optimistic non-obvious   thing and so my question to you is what is the  biggest open genuine question on your mind right now? In which field? You're in so many! I am  particularly curious about the combination of   AI and hardware but I realize that we've covered  a lot so I'm curious the direction you'd take this   on a question that occupies you right now. Gosh  I mean I think maybe one that's a little more   AI specific is there a current set of methods  that seem to be scaling very well right so with   past AI architecture you could kind of feed an  AI system a certain amount of data and and use   a certain amount of compute but eventually it  hit a plateau and one of the interesting things   about these new transformer based architectures  over the last you know 5 to 10 years is that we   haven't found the end yet. So that leads to this  dynamic where Llama 3 you know we could train on  

you know 10 to 20,000 gpus, Llama 4 we could train  on you know more more than 100,000 gpus, Llama 5   we can plan to scale even further and there's just  an interesting question of how far that goes. It's   totally possible that at some point we just like  hit a limit and just like previous systems there's   an asymptote and it doesn't keep on growing but  it's also possible that that limit is not going   to happen anytime soon and that we're going to be  able to keep on just building more clusters and   generating more you know synthetic data train the  systems and that they're just going to keep on   getting more and more useful for people for quite  a while to come and it's a really big and high   stakes question I think for for the company is  because we're basically making these bets on how   much infrastructure to build out for the future  and this is like hundreds of billions of dollars   of infrastructure so like I'm clearly betting  that this is going to keep scaling for a while   but it's one of the big questions I think in the  field because it is possible that it doesn't. You   know that obviously would lead to a very different  world where it's I mean I'm sure people still   figure it out eventually just need to make some  new fundamental improvements to the architecture   in some way but that might be a somewhat longer  trajectory for okay maybe you know the the kind   of fundamental AI advances slow down for a bit  and we just take some time to build new products   around this or it could be the case and that's  what I'm betting on that the fundamental AI will   just continue advancing for quite a while and that  we're going to get both a new set of products that   are just really compelling in all these ways  and that the technology landscape and what's   possible will just continue being dynamic over  like a 20-year period and that's probably what   I'd guess is going to happen but it I think it's  one of the bigger questions in the industry and   kind of for technology across the world today.  Is there anything else that you want to say? I  

don't know! Awesome. We're good. Amazing yeah thank  you so much for doing this. Yeah no thank you...

2024-10-02

Show video