Latest / NO FILTER / The Future Mark Zuckerberg Is Trying To Build(M4A_128K)

Transcript

The Future Mark Zuckerberg Is Trying To Build(M4A_128K) · NO FILTER

0:00
Speaker 1: I'd love to start with these ten years of work
0:02
Speaker 1: right there. Someone on your team call these the real
0:05
Speaker 1: life Tony Stark glasses.
0:06
Speaker 2: Very hard to make each one of these.
0:07
Speaker 1: That makes me feel incredibly optimistic in a world.
0:09
Speaker 2: Where AI gets smarter and smarter, this is probably going
0:12
Speaker 2: to be the next major platform after phones.
0:13
Speaker 1: I miss hugging my mom.
0:15
Speaker 2: The haptics is hard.
0:16
Speaker 1: How does generative AI change how social media feels?
0:20
Speaker 2: We haven't found the end yet.
0:22
Speaker 1: The average American has fewer friends now than they did
0:25
Speaker 1: fifteen years ago. Why do you think that's happening?
0:29
Speaker 2: I mean, there's a lot going on to unpack there.
0:36
Speaker 1: I'm about to interview Meta CEO Mark Sockerberg Marx the Soccerberg.
0:40
Speaker 1: There are not that many people with more power over
0:43
Speaker 1: what our future might look like. Nearly half the total
0:46
Speaker 1: human population now uses Meta products, and I just ested
0:50
Speaker 1: some of their new tech That feels like science fiction.
0:54
Speaker 1: This is crazy. Mark Sockerberg and a team at Meta
0:57
Speaker 1: are imagining a future that billyions of other people might
1:01
Speaker 1: actually end up living in. So my goal for this
1:03
Speaker 1: conversation is to try to figure out what that future
1:06
Speaker 1: really looks like, to paint a picture of the future
1:09
Speaker 1: Mark Zuckerberg is trying to build so that you can
1:11
Speaker 1: decide for yourself what you think of it. Welcome to
1:14
Speaker 1: the first episode of our new series, Huge Conversations.
1:23
Speaker 2: Okay, good to meet you. Yeah, looking forward to it.
1:27
Speaker 1: Awesome. I'd love to tell you what my goal is
1:31
Speaker 1: of this conversation. So we have a show called Huge
1:34
Speaker 1: If True, which is this very optimistic show about science
1:37
Speaker 1: and technology and the potential futures that we can build.
1:41
Speaker 1: And in every episode we're sort of exploring what does
1:43
Speaker 1: it look like if you play a certain technological future
1:47
Speaker 1: out and so my goal in this conversation is to
1:51
Speaker 1: try to help people see the future that you are
1:54
Speaker 1: imagining when you're building the products that you and the
1:56
Speaker 1: Meta team are building. What are you imagining this looks like, sture?
2:00
Speaker 1: How are you imagining people use this all that?
2:03
Speaker 2: Yeah?
2:04
Speaker 1: Cool? All rightwesome. So I'd love to start with these.
2:08
Speaker 2: Let's do it ten years of work right there.
2:10
Speaker 1: I got to demo them a little bit earlier today.
2:14
Speaker 1: I heard someone on your team call these the real
2:16
Speaker 1: life Tony Stark glasses.
2:18
Speaker 2: We're getting there, but I'd love.
2:20
Speaker 1: To just hear in your voice, what are these?
2:22
Speaker 2: Well, these are the first full holographic augmented reality glasses
2:28
Speaker 2: I think that exist in the world we've made. I
2:33
Speaker 2: think it's a few thousand or something right, very hard
2:37
Speaker 2: to make each one of these, but this is the
2:39
Speaker 2: culmination of ten years of research and development that we've
2:43
Speaker 2: done to basically miniaturize all the computing that you need
2:50
Speaker 2: to have glasses, not a headset, but glasses that can
2:55
Speaker 2: put full holograms into the world with a wide field
3:00
Speaker 2: of view. So you can imagine sort of in the future,
3:02
Speaker 2: we'd be having a version of this conversation where you know,
3:05
Speaker 2: maybe I or you or not even here. It's like
3:08
Speaker 2: one of us is physically here and the other one
3:09
Speaker 2: is here is as kind of a full body hologram,
3:12
Speaker 2: and it's not just a video call. You can actually interact.
3:14
Speaker 2: You can do things. I mean in the demo we
3:17
Speaker 2: had the you know, ping pong and games and things
3:19
Speaker 2: like that, but I mean you can interact, you can
3:21
Speaker 2: work together, you can play poker, play chess, whatever, like
3:24
Speaker 2: the holographic cards, holographic board game. I just think it's
3:28
Speaker 2: gonna be wild. It's gonna remake I think so many
3:31
Speaker 2: different fields that we think about today, from how we
3:36
Speaker 2: work in productivity to a lot of things around science,
3:39
Speaker 2: a lot of things around education, entertainment, fun, gaming. So
3:45
Speaker 2: but this is just the beginning, you know, this is
3:46
Speaker 2: the first version. It's like it's a prototype version that
3:49
Speaker 2: we've made in order to develop the next version which
3:54
Speaker 2: is hopefully going to be the consumer one that we
3:56
Speaker 2: sell to a lot of people. Why build these Well,
3:59
Speaker 2: I think it's going to be the next major computing platform.
4:02
Speaker 2: So if you look at like the Grand arc of computing,
4:04
Speaker 2: over time, you've you've gone from like mainframes to computers
4:09
Speaker 2: that basically like live on your desk or on a tower,
4:14
Speaker 2: to phones that you have in your hand that you
4:16
Speaker 2: basically like you know, can take with you everyworry that
4:19
Speaker 2: you want. But it's it's pretty unnatural, right. It takes
4:22
Speaker 2: you away from the world around you. And I think
4:26
Speaker 2: that the trend in computing is it gets more ubiquitous,
4:29
Speaker 2: it gets more natural, and it just gets more social, right,
4:34
Speaker 2: So you want to be able to interact with people
4:35
Speaker 2: in the world around you, And I think that this
4:38
Speaker 2: is probably going to be the next major platform after phones.
4:41
Speaker 1: Look at these two.
4:41
Speaker 2: These are the clear ones that show Yeah, the whole
4:44
Speaker 2: thing is a special edition and this is like a
4:46
Speaker 2: really special edition. There's not a single millimeter of space.
4:53
Speaker 2: It's you know, everything in here from the micro projectors
4:59
Speaker 2: that basically shoot light into the waveguides. Right, it's a
5:04
Speaker 2: special type of display system. I mean, these aren't normal
5:07
Speaker 2: displays like you have in a phone or a TV
5:10
Speaker 2: or computer, like the type of displays that people have
5:12
Speaker 2: been building for decades. It's a waveguide system. The projector
5:16
Speaker 2: that's shooting light basically goes into these nano etchings across
5:21
Speaker 2: the waveguide that are what catches and creates the holograms
5:25
Speaker 2: in order to synchronize that with your where you're looking,
5:30
Speaker 2: there's eye tracking and little cameras they illuminate your eyes.
5:34
Speaker 2: And then of course there's all the basic stuff that
5:37
Speaker 2: you need, all the computing, the batteries to power the
5:40
Speaker 2: whole thing, microphones, the speakers because it needs to be
5:44
Speaker 2: able to play audio and speak with you, and the
5:48
Speaker 2: cameras and sensors to see things around you in the world.
5:51
Speaker 2: So that way, when it's placing holograms in the world,
5:53
Speaker 2: it can do that in the right place and understand
5:55
Speaker 2: where you are. So that probably is still not covering everything.
6:00
Speaker 2: There's a lot of things that need to go into
6:02
Speaker 2: syncing up the holographic images between the two displays, because
6:07
Speaker 2: you don't just have a single display like you have
6:09
Speaker 2: in a phone or TV. You have two and it
6:12
Speaker 2: moves around, and you know, physical things are hard and
6:15
Speaker 2: need to be syncd up. There's also the radio that
6:18
Speaker 2: has to communicate with your other computing devices to do
6:21
Speaker 2: heavier computing, and the risk bace neural interface that you
6:25
Speaker 2: probably got to try out. We kind of miniaturized all
6:27
Speaker 2: of this and fit it into a normal looking pair
6:31
Speaker 2: of glasses, which is you know, when I told the
6:35
Speaker 2: team that we're going to do this ten years ago,
6:38
Speaker 2: you know, people weren't sure if who're going to be
6:39
Speaker 2: able to But I think, you know, not only we're
6:42
Speaker 2: going to be able to do this, but I think
6:43
Speaker 2: we're going to be able to get a cheaper and
6:44
Speaker 2: higher quality and even smaller and more stylish over time.
6:48
Speaker 2: So I think this is going to be a pretty
6:50
Speaker 2: wild future.
6:51
Speaker 1: There are so many versions of trying to get at
6:54
Speaker 1: a similar idea of digital objects in physical space. I'm
6:58
Speaker 1: thinking of, for example, glasses that have heads up displays
7:04
Speaker 1: where it's headlocked and it's moving with my eyes. Glasses
7:08
Speaker 1: that are really creating digital objects in physical space that
7:10
Speaker 1: don't move as I move. I'm thinking of these. I'm
7:14
Speaker 1: also thinking of the Snapchat spectacles that they just announced.
7:18
Speaker 1: Then on the other hand, there are headsets like the
7:20
Speaker 1: Quest and also like the Apple Vision Pro that seem
7:23
Speaker 1: to fall into a different category. I'm curious how you
7:25
Speaker 1: would organize this landscape for people and how you think
7:28
Speaker 1: about people using these tools in their real lives in
7:31
Speaker 1: the near future.
7:32
Speaker 2: Yeah, So when we were getting started on this about
7:34
Speaker 2: ten years ago, I thought that something like this was
7:37
Speaker 2: going to be the ultimate product for everyone. Right, you
7:39
Speaker 2: get to normal looking pair of glasses and we'll continue
7:43
Speaker 2: improving that that kind have full holographic images. I think
7:47
Speaker 2: it's super powerful and it is sort of the science
7:50
Speaker 2: fiction future that I think we all hope to get to.
7:53
Speaker 2: On the journey, we took a few other approaches as
7:57
Speaker 2: well to help us develop towards that, including building glasses
8:02
Speaker 2: that don't have displays, to try to learn just take
8:06
Speaker 2: a stylish pair of glasses today and put as much
8:09
Speaker 2: technology into it as you can, but really focus on
8:11
Speaker 2: the form factor. And that's the ray band metaglasses. And
8:14
Speaker 2: it's doing really well. And initially we thought that that
8:18
Speaker 2: was sort of intro product for us to learn how
8:21
Speaker 2: to build this. But one of the things that's clear
8:24
Speaker 2: now is you're going to be able to make that
8:27
Speaker 2: product a lot more affordable than this, probably permanently. So
8:33
Speaker 2: I actually think that there are going to be a
8:34
Speaker 2: bunch of different of these paths that we've taken are
8:38
Speaker 2: going to be kind of permanent product lines that people
8:40
Speaker 2: will choose. I think you'll see display those glasses like
8:43
Speaker 2: the ray band metas continue to get better and better.
8:46
Speaker 2: Great for AI. No display, but you can talk to it,
8:49
Speaker 2: it can talk back. I know there's going to be
8:51
Speaker 2: something between these that's basically a heads up display. So
8:55
Speaker 2: it's not a seventy degree field of view. Maybe it's
8:57
Speaker 2: a twenty degree or thirty degree field of you. So
9:00
Speaker 2: that's not going to be what you want for putting
9:04
Speaker 2: kind of a full hologram of a person or interacting
9:06
Speaker 2: with the world around you, but it's going to be
9:08
Speaker 2: great for you know, when you're talking to AI, not
9:11
Speaker 2: just having voice, but also being able to see what
9:13
Speaker 2: it's saying, right, or being able to text someone with
9:16
Speaker 2: your wrist based neural interface and then have their text
9:19
Speaker 2: show up rather than having it read to you, which
9:21
Speaker 2: is we read faster than we can listen, or getting
9:24
Speaker 2: directions right, or just being able to search for information
9:27
Speaker 2: and that. So there's a lot of value for a
9:29
Speaker 2: heads up display that will be somewhat more expensive than
9:32
Speaker 2: the displayless glasses, but somewhat cheaper than this. Then I
9:36
Speaker 2: think you're gonna get this. It's going to be probably
9:41
Speaker 2: the most premium and and expensive of the glasses products,
9:44
Speaker 2: but hopefully still something that you know, like a computer
9:47
Speaker 2: is generally accessible to most people in the world. But
9:51
Speaker 2: I think that they're going to be all of those,
9:52
Speaker 2: and I think people will like them. I also think
9:55
Speaker 2: that the headsets that people are using around mixed reality
9:59
Speaker 2: will continue to be a thing too, because no matter
10:01
Speaker 2: how good we get at miniaturizing the tech for this,
10:04
Speaker 2: you're just going to be able to fit more compute
10:06
Speaker 2: into a full headset. Fundamentally, our mission is not build
10:11
Speaker 2: something that is advanced and only a few people can use.
10:14
Speaker 2: We want to take it in the last mile and
10:17
Speaker 2: do all the innovation to get it to everyone. We
10:19
Speaker 2: know just shipped right now. It's Quest three S, the
10:23
Speaker 2: new mixed reality headset. We basically are delivering high quality
10:27
Speaker 2: mixed reality for two hundred and ninety nine dollars. I
10:30
Speaker 2: was really proud of last year when we delivered Quest three,
10:33
Speaker 2: the first kind of really high quality, high resolution color
10:37
Speaker 2: mixed reality device, for five hundred dollars. Right, it was like,
10:42
Speaker 2: it's like a fraction of the cost of what the
10:44
Speaker 2: competitors are doing, and I think it's actually higher quality
10:46
Speaker 2: in a lot of ways. And now we've just doubled
10:48
Speaker 2: down on that. So I think that they're all actually
10:52
Speaker 2: going to end up being important long term product lines,
10:56
Speaker 2: display lists, heads up display, full holographic ar are, full headsets.
11:02
Speaker 2: I think they're all going to be important.
11:03
Speaker 1: Yeah, if you play out the future of not just
11:07
Speaker 1: the hardware that we've been talking about so Metaway bands
11:12
Speaker 1: quest Orian, but also the Lama models, if everything goes
11:19
Speaker 1: according to you and the team's wildest dreams, I'd love
11:23
Speaker 1: you to just begin to describe what that feels like.
11:26
Speaker 2: I mean, I think that there are two primary values
11:28
Speaker 2: that we're trying to bring on the AR and kind
11:31
Speaker 2: of mixed reality side. The main value we're trying to
11:34
Speaker 2: bring is this feeling of presence. Right, So there's something
11:36
Speaker 2: that I think is just really deep about being physically
11:41
Speaker 2: present with another person that you don't get from any
11:44
Speaker 2: other technology today. And I think that's the thing. When
11:47
Speaker 2: people have a very visceral reaction to experiencing virtual or
11:50
Speaker 2: mixed reality, what they're really reacting to is that they actually,
11:55
Speaker 2: for the first time with technology, feel a sense of presence,
11:58
Speaker 2: like they're in a place with the person, and that's
12:00
Speaker 2: super powerful. Like I've focused on designing social apps and
12:04
Speaker 2: experiences for twenty years, that's sort of like the holy
12:08
Speaker 2: grail of that is being able to build a technology
12:10
Speaker 2: platform that delivers this deep sense of social presence. The
12:15
Speaker 2: other big track is around personalized AI and for that,
12:19
Speaker 2: and that's sort of where Lama and medai and all
12:22
Speaker 2: those things are going. There's all this development that's going
12:25
Speaker 2: into making the models smarter and smarter over time. But
12:30
Speaker 2: I think where this is going to get really compelling
12:33
Speaker 2: is when it's personalized for you. And in order for
12:36
Speaker 2: it to be personalized for you, it has to have
12:38
Speaker 2: context and understand what's going on in your life, both
12:42
Speaker 2: kind of at a global level and like what's physically
12:45
Speaker 2: happening around you right now. And in order to do that,
12:48
Speaker 2: I think that glasses are going to be the ideal
12:50
Speaker 2: form factor because they're positioned on your face in a
12:53
Speaker 2: way where they can let them see what you see
12:55
Speaker 2: and hear what you hear, which are the two most
12:57
Speaker 2: important senses that we use for a kind of taking
13:01
Speaker 2: an information in context about the world. I think that
13:03
Speaker 2: this is all going to be kind of really deep
13:05
Speaker 2: and profound stuff, but it's basically those two things. It's
13:07
Speaker 2: this feeling of presence and this capability of really personalized
13:11
Speaker 2: intelligence that can help you.
13:13
Speaker 1: I'd love to talk about each of those two things.
13:17
Speaker 1: The first on presence. I owe a lot to being
13:20
Speaker 1: able to connect with people online. Right This job that
13:23
Speaker 1: I have is by definition also with my family. My
13:26
Speaker 1: parents don't live anywhere close to me. I video call
13:30
Speaker 1: them a lot. And when I think about the progress
13:33
Speaker 1: of technology like this in a timeline from like the
13:38
Speaker 1: telegram to the telephone to video call to some feeling
13:42
Speaker 1: of presence with another person who feels like they're right
13:45
Speaker 1: there in front of me, that makes me feel incredibly optimistic.
13:47
Speaker 1: I would love a future where like I can lose
13:52
Speaker 1: in scrabble to my mom and feel like she's really
13:55
Speaker 1: apparent in front of me, and it feels like we're
13:57
Speaker 1: not that far away from something persuades my brain.
14:00
Speaker 2: That that's happening. Yeah, totally.
14:03
Speaker 1: And also I miss hugging my mom right, like that
14:09
Speaker 1: never goes away.
14:10
Speaker 2: Yeah, Haptics is hard.
14:11
Speaker 1: Yeah, And so my question is my question is about that.
14:14
Speaker 1: It's about this feeling of like it's hard for me
14:16
Speaker 1: to imagine a future where real physical presence is not
14:22
Speaker 1: different and special in some way where I don't miss
14:27
Speaker 1: literally hugging my mom. And I'm curious how you think
14:32
Speaker 1: about the parts of human connection that are eye contact
14:38
Speaker 1: and physical touch and things that are eight brains value
14:42
Speaker 1: for connection with other people.
14:44
Speaker 2: Yeah, well, I contact, I think we're going to get
14:46
Speaker 2: to a lot before the touch part. For haptics. I
14:50
Speaker 2: do think we'll make progress on that. But it's obviously
14:53
Speaker 2: there's a spectrum there too from kind of hands, which
14:58
Speaker 2: is where if you dr out the kind of like
15:01
Speaker 2: homunculous version of a person in terms of like what
15:03
Speaker 2: are what are our kind of sensory you know, what's
15:06
Speaker 2: like the majority of what we're sensing, It's like it's yeah, yeah,
15:09
Speaker 2: So I think being able to do that for your
15:12
Speaker 2: hands is probably the most important place to start. And
15:15
Speaker 2: you have a rough version of that with controllers today.
15:19
Speaker 2: I that'll get even more over time. We have this
15:21
Speaker 2: demo playing ping pong where you have a controller whereas
15:25
Speaker 2: the digital ball hits the ping pong paddle, you feel
15:28
Speaker 2: it hit the as if it's hitting the ping pong
15:30
Speaker 2: paddle wherever it is, so you actually have a sense
15:32
Speaker 2: of like where it's it's hitting the paddle. So I
15:35
Speaker 2: think that was that was just a wild demo, So
15:38
Speaker 2: I think we'll get some of that. The most extreme
15:40
Speaker 2: version of this is wanting force feedback, right, So I
15:43
Speaker 2: mean like for doing a lot of sports, right, it's
15:46
Speaker 2: it's like, okay, we can kind of do a good
15:48
Speaker 2: approximation of like boxing today or you get like good
15:52
Speaker 2: feedback on your hands. But it would be hard to
15:54
Speaker 2: do a virtual reality version of jiu jitsu where you're
15:58
Speaker 2: like grappling with someone and you need like real kind
16:00
Speaker 2: of force feedback on that. So that's probably like the
16:03
Speaker 2: hardest thing to go do, But I think we'll get there.
16:08
Speaker 2: You know. I think, like most science fiction, it's not
16:11
Speaker 2: this binary thing that you just like wake up one
16:13
Speaker 2: day and we're like, oh, we've realized all the dreams.
16:15
Speaker 2: But I do think that these platforms are going to
16:18
Speaker 2: be the first time that I think that there's a
16:22
Speaker 2: realistic sense of presence in all the ways that that's
16:26
Speaker 2: special to people. For most things that people want to do,
16:31
Speaker 2: which are not the most physical ones, and even some
16:33
Speaker 2: of the basic physical ones I think we'll get, but
16:35
Speaker 2: then there's a long tail of other stuff. I mean,
16:37
Speaker 2: smell is also really important for people. Yeah, right, I
16:40
Speaker 2: think it's disproportionately important for memories, and that's not really
16:46
Speaker 2: a thing that I think in the next few years
16:48
Speaker 2: we're going to have in any of these devices. I mean,
16:50
Speaker 2: that's a very difficult and challenging thing on its own.
16:54
Speaker 1: What is the piece of that that you feel most
16:59
Speaker 1: interested in that you keep coming back to me in
17:00
Speaker 1: your mind?
17:01
Speaker 2: This has the frustrating property to develop that. The sense
17:04
Speaker 2: of presence is almost like when you're designing something that
17:08
Speaker 2: that's sort of trying to artificially deliver it. You're delivering
17:12
Speaker 2: an allusion to a person, and more than any one
17:16
Speaker 2: thing that provides a sense of presence, it's actually more
17:20
Speaker 2: the case that any one thing done wrong breaks the
17:23
Speaker 2: sense of presence. You kind of know that you're interacting
17:27
Speaker 2: with technology, but it's so convincing that that you just
17:33
Speaker 2: kind of go along with it and you're like, Okay, yeah, no,
17:35
Speaker 2: this person feels like it feel like they're there. Right.
17:37
Speaker 2: When I did that ping Pong demo, I like, at
17:40
Speaker 2: the end of it, I dropped the ping pong paddle
17:43
Speaker 2: on the virtual table and it shattered. So that was
17:46
Speaker 2: not the best for our internal development. But but like
17:51
Speaker 2: that's winning in our in our development. Right. It's like
17:53
Speaker 2: when you when you feel like something is is kind
17:55
Speaker 2: of so realistic that you're just convinced that that it's
18:00
Speaker 2: there now, and there are a lot of things that
18:02
Speaker 2: can break that, right, So I think a field of
18:05
Speaker 2: view that's too low, right, so something feels real, but
18:07
Speaker 2: then you turn your head and it's not there. Latency right,
18:12
Speaker 2: Physics that don't behave like realistic physics. It also is
18:15
Speaker 2: interesting in some ways what people can accept as physically
18:20
Speaker 2: real even though it's not right. So we've done a
18:23
Speaker 2: ton of work on avatars. We have this whole workstream
18:26
Speaker 2: on codec avatars to do these photo realistic avatars, and
18:29
Speaker 2: it's I think it's going to be incredibly compelling and
18:32
Speaker 2: people are gonna love it. But one of the things
18:35
Speaker 2: that I found interesting is the ability to mix photorealistic
18:39
Speaker 2: and expressive kind of the cartoony avatars with photo realistic
18:43
Speaker 2: worlds and kind of more cartoony computer game type worlds.
18:48
Speaker 2: So you can have the a codec kind of photo
18:51
Speaker 2: realistic avatar of a person in what is clearly like
18:54
Speaker 2: a video game or cartoon world, and people are generally
18:57
Speaker 2: pretty fine with that. It's like, Okay, that feels pretty good.
19:00
Speaker 2: And similarly having a photo realistic world but good increasingly
19:06
Speaker 2: good kind of cartoon avatars. As long as the avatars
19:10
Speaker 2: move in a way that feels authentic to the person
19:13
Speaker 2: you're interacting with, it actually feels pretty good.
19:16
Speaker 1: You know.
19:16
Speaker 2: It's when you look at a two D still frame
19:18
Speaker 2: of it, some of the stuff can look a little
19:20
Speaker 2: bit silly, and we've certainly, you know, had had our
19:23
Speaker 2: share of memes around that. But but when you're in there,
19:27
Speaker 2: you know, and you've you've played around with a lot
19:28
Speaker 2: of the stuff, it feels realistic because it's basically mimicking
19:32
Speaker 2: the kind of authentic mannerisms of a person that you're
19:35
Speaker 2: interacting with. And even if it's not a codec photo
19:38
Speaker 2: realistic avatar, if it's kind of a more cartoony expressive one.
19:41
Speaker 2: So I think that that's it's it's very interesting to
19:44
Speaker 2: see kind of which pieces you need to unlock and
19:47
Speaker 2: what where you just need to be like very technically
19:49
Speaker 2: excellent and consistent. But it's this isn't a space where
19:53
Speaker 2: it's like you deliver one thing and it's good. This
19:55
Speaker 2: is like there's a wide breadth of things that you
19:59
Speaker 2: need to nail and then have it all come together.
20:01
Speaker 2: And that's why these are ten year projects.
20:05
Speaker 1: It seems like an interesting way to learn about the
20:08
Speaker 1: human brain and what we actually care about with respect
20:12
Speaker 1: to what feels real. I was wondering about There was
20:16
Speaker 1: this moment in an interview that you did with Lex Friedman.
20:20
Speaker 1: You quoted research that says that the average American has
20:25
Speaker 1: fewer friends now than they did fifteen years ago, and
20:28
Speaker 1: I was so interested in that because it seems like
20:31
Speaker 1: if we want to get to a world where there's
20:33
Speaker 1: more human connection, this is the trend that we're going
20:37
Speaker 1: to have to grapple with. And just to give some
20:41
Speaker 1: data on this, in the American Time you survey over
20:45
Speaker 1: the last twenty years, the amount of time American adults
20:47
Speaker 1: spent socializing in person has dropped by nearly thirty percent
20:52
Speaker 1: for ages fifteen to twenty four according to the Surgeon
20:55
Speaker 1: in general, it's nearly seventy percent. And I look at
20:59
Speaker 1: that data and I think to myself, well, maybe if
21:01
Speaker 1: we're all socializing digitally, that doesn't matter so much. Maybe
21:04
Speaker 1: there's a future where that's actually fine. But there's also
21:07
Speaker 1: data that suggests that we're struggling somewhat. The number of
21:11
Speaker 1: Americans who say that they don't have a single close friend. Yeah,
21:17
Speaker 1: that share has done three percent to twelve percent in
21:20
Speaker 1: the last thirty years. It feels to me like with
21:24
Speaker 1: all the tools that we've built for human connection, were
21:28
Speaker 1: struggling to connect. And I'm curious, why do you think
21:33
Speaker 1: that's happening.
21:37
Speaker 2: I mean, there's a lot going on to unpack there.
21:42
Speaker 2: A lot has changed sort of economically and socially during
21:46
Speaker 2: that period, and a lot of those trends go back
21:48
Speaker 2: before a lot of the modern technology. So I mean,
21:52
Speaker 2: this is something that a lot of academics and folks
21:54
Speaker 2: have studied. But it is an interesting lens to look
21:59
Speaker 2: at this though, because I think whenever you're talking about
22:02
Speaker 2: building digital types of connection, one of the first questions
22:07
Speaker 2: that you get is is that going to replace the
22:11
Speaker 2: physical connection? And my answer to that, especially in the
22:16
Speaker 2: case of something like this, is that no, because people
22:23
Speaker 2: already don't have as much connection as they would like
22:26
Speaker 2: to have. It's not like this is replacing some sort
22:28
Speaker 2: of better physical connection that they would have otherwise had.
22:32
Speaker 2: It's that the average person would like to have ten
22:35
Speaker 2: friends and they have two, right or three, and there's
22:42
Speaker 2: just more demand to socialize than what people are able
22:46
Speaker 2: to do given the current construct and giving people the
22:50
Speaker 2: ability to be present with people who are in other
22:52
Speaker 2: places physically just seems like it will unlock more. It's
22:55
Speaker 2: not going to make it so if I have glasses,
22:57
Speaker 2: it's not going to make it that I spend less
22:59
Speaker 2: time with my wife. It's going to make it so
23:01
Speaker 2: that I spend more time with, you know, my sister
23:04
Speaker 2: who lives across the country, and that's I think that
23:08
Speaker 2: that's that's good and I think people need that. As
23:12
Speaker 2: for the rest, I think we could probably spend a
23:15
Speaker 2: multi hour podcast just going into all of the different
23:19
Speaker 2: kind of socio economic political dynamics that are going on.
23:24
Speaker 2: But none of the trends that I've seen does it
23:28
Speaker 2: seem like the primary thing that's going on is that
23:31
Speaker 2: because people are interacting online, they're now not interacting with
23:35
Speaker 2: there with people physically. It's now certainly I think you you,
23:44
Speaker 2: I do interact with people online who I also like
23:47
Speaker 2: to interact with physically. But and I think that that's
23:48
Speaker 2: kind of like a combination, like more combined richer relationship
23:53
Speaker 2: that you have overall. But I think that there's a
23:56
Speaker 2: lot going on with the loss of of kind of
23:59
Speaker 2: social capital and connections that really predates a lot of
24:03
Speaker 2: the modern technology.
24:05
Speaker 1: The goal of what I'm trying most to learn about
24:10
Speaker 1: is how we can structure the technologies that we use
24:14
Speaker 1: in the future to get toward this future. I think
24:17
Speaker 1: you're imagining of more human connection in ways. I'm curious
24:21
Speaker 1: you brought up the other big pillar of AI and
24:26
Speaker 1: in some of your conversations I'm thinking of a conversation
24:29
Speaker 1: with Tim Ferris in particular, you talked about a lot
24:31
Speaker 1: of different use cases of AI, and they seem to
24:35
Speaker 1: me to fall on somewhat of a spectrum. Like, for example,
24:40
Speaker 1: you mentioned automatic real time translation. It's like basically the
24:44
Speaker 1: Star Trek translator, Yeah, Verstal.
24:46
Speaker 2: Translator pretty much there.
24:48
Speaker 1: Yeah, And that's one example on one end of the
24:51
Speaker 1: spectrum where some people might argue that there is a
24:56
Speaker 1: chance that someone is less likely, for example, to learn
24:58
Speaker 1: a language, because we can all speak to each other
25:00
Speaker 1: in real time in different languages. I think nobody would
25:03
Speaker 1: really argue that therefore we shouldn't have that kind of
25:06
Speaker 1: universal translator, right right exactly. And so I think that
25:13
Speaker 1: end of the spectrum is something like technologies that really
25:17
Speaker 1: measurably unlock our humanity because they remove a struggle between people.
25:23
Speaker 1: And then on the other end of the spectrum, there
25:25
Speaker 1: are a lot of educational things, for example, where the
25:27
Speaker 1: struggle is kind of the point, right, Like it's like
25:29
Speaker 1: building a muscle, like you, I can think of so
25:32
Speaker 1: many times in my life where like, the reason why
25:34
Speaker 1: I was doing something was not the output, it was
25:36
Speaker 1: the fact that I was trying so hard to do it.
25:38
Speaker 1: There's one example in the Tim Ferrisson interview where you
25:41
Speaker 1: talked about your kids struggling to articulate themselves emotionally, and
25:47
Speaker 1: adults very much have the same problem, and you talked
25:50
Speaker 1: about AI as a way to help them articulate those emotions.
25:53
Speaker 1: And I thought about all of the many times in
25:56
Speaker 1: my life where I have struggled to articulate my emotion
26:00
Speaker 1: and how I really could have used some help in
26:01
Speaker 1: those moments, and I also find myself thinking about the
26:05
Speaker 1: times when that was really building a muscle, where like
26:08
Speaker 1: the act of struggling to communicate with someone and understand
26:10
Speaker 1: what they wanted from me was important to my development.
26:14
Speaker 1: And so my question is if you think about that
26:16
Speaker 1: as a spectrum between things that are really important to
26:20
Speaker 1: our humanity, where and the struggle being removed is helpful,
26:24
Speaker 1: versus things where the struggle is the point and it
26:28
Speaker 1: unlocks something about our humanity and is important to preserve,
26:32
Speaker 1: like building a muscle. How do you draw the line
26:35
Speaker 1: between those things, and how do we ensure that the
26:38
Speaker 1: muscles that we're building for this future are stronger and
26:41
Speaker 1: not weaker.
26:42
Speaker 2: That's interesting. I mean, I think we're always going to
26:44
Speaker 2: find new things to struggle with, And I mean it's
26:47
Speaker 2: you can always get better at communicating with other people
26:50
Speaker 2: and kind of expressing yourself and understanding other people. So
26:53
Speaker 2: having a tool that can help you do that better
26:56
Speaker 2: isn't going to mean that like, oh, now we perfectly
26:58
Speaker 2: understand every you know, it's I mean, and I mean
27:03
Speaker 2: I think the maybe one of the most functional aspects
27:07
Speaker 2: of this. You're already seeing a lot of these AI
27:08
Speaker 2: models really help people with coding. Right, Like a generation ago,
27:13
Speaker 2: before I was getting started, a lot of coding was
27:15
Speaker 2: like really low level system software, and you know, then
27:19
Speaker 2: by the time that I got into it, there's a
27:21
Speaker 2: little bit of that. But you know, you can make
27:24
Speaker 2: websites pretty easily, make apps pretty easily. And I think
27:29
Speaker 2: in twenty years or a lot sooner than that, you're
27:32
Speaker 2: going to basically be in a world where kids will
27:35
Speaker 2: be able to just describe the things that they want
27:37
Speaker 2: to build incredibly complex piece of software. So it's in
27:43
Speaker 2: that world are kids going to be not struggling. I
27:46
Speaker 2: don't think so. I think that they're going to be
27:48
Speaker 2: just expressing their creativity and it'll it'll be this kind
27:52
Speaker 2: of constant iterative feedback loop around like okay, like yeah,
27:57
Speaker 2: I took a few minutes to describe this thing, and like, yeah,
28:02
Speaker 2: this whole like amazing virtual world was created that I
28:05
Speaker 2: can now see on my glasses or whatever. But like,
28:08
Speaker 2: these things are not exactly what I want them to be,
28:10
Speaker 2: so now I need to go back and edit them.
28:12
Speaker 2: Just I don't know. I think that there's always more
28:16
Speaker 2: another way to get at this. The things that they
28:18
Speaker 2: mix people so good. It's just there's always more to do.
28:22
Speaker 1: We'll always find the struggle. Yeah, another way to get
28:24
Speaker 1: at this is if you play this out to make
28:27
Speaker 1: the tools even better. And like ten years, let's say
28:31
Speaker 1: your kids are in high school. Are there ways that
28:33
Speaker 1: you would want them using AI because you think it
28:35
Speaker 1: would accelerate them intellectually and ways that you would advocate
28:40
Speaker 1: for them not to use it, or things that you
28:43
Speaker 1: would have concerns about.
28:44
Speaker 2: I mean, I think that there's some things that you
28:46
Speaker 2: need to be able to do yourself. I think that
28:49
Speaker 2: that's a lot of the basic fear that people have
28:51
Speaker 2: around this is that while we're building these amazing tools,
28:57
Speaker 2: we get away from this self confidentdence and ability of
29:01
Speaker 2: being able to do like this basic stuff yourself. So
29:04
Speaker 2: it's like, all right, you have a calculator, but it's
29:08
Speaker 2: still good to be able to do kind of basic
29:10
Speaker 2: math in your head because there are a lot of
29:12
Speaker 2: things that come up throughout the day that you just
29:16
Speaker 2: want to have a general numeracy around, right that often
29:20
Speaker 2: they're not expressed in numerical terms, but just in terms
29:23
Speaker 2: of understanding trends or understanding arguments that people are making.
29:27
Speaker 2: You kind of need to understand the shape of how
29:29
Speaker 2: numbers come together. And so I think one of the
29:32
Speaker 2: big debates is like, should we still teach our kids
29:34
Speaker 2: to program computers even though you're going to have these
29:39
Speaker 2: these tools in the future that are just so much
29:42
Speaker 2: more powerful than anything that we have now to produce
29:45
Speaker 2: incredibly complicated piece of software. I think the answer to
29:49
Speaker 2: that is probably yes, because I think teaching someone how
29:54
Speaker 2: to code is teaching them a way to think rigorously,
29:58
Speaker 2: and that even if they're not doing most of the
30:01
Speaker 2: code production, I think it's important that you kind of
30:03
Speaker 2: have the ability to think in that way, and I
30:05
Speaker 2: think it's going to just make you generally a better
30:08
Speaker 2: thinker and better person. So yeah, maybe that's like this
30:14
Speaker 2: generation's version of calculators. It's like, so you'll you'll want
30:17
Speaker 2: to you want to use the calculator, but you'll also
30:20
Speaker 2: want to be able to generally do without it. Other
30:23
Speaker 2: ones like language. I don't know, I mean, different people
30:26
Speaker 2: can come out. I mean, I think this is one
30:27
Speaker 2: of the interesting questions about parenting these days, is like,
30:32
Speaker 2: is just kind of like what's important to teach your
30:35
Speaker 2: kids in an era where so much is going to
30:38
Speaker 2: change over the time that they're even in school language.
30:41
Speaker 2: I think you can make similar arguments. I think there's
30:43
Speaker 2: a lot of It's like it's probably going to be
30:46
Speaker 2: less functionally relevant in the future to learn multiple languages,
30:49
Speaker 2: but it sort of helps you think in different ways.
30:53
Speaker 2: You know, I've found from the languages that I've studied
30:55
Speaker 2: that a lot of it. Yeah, you learn about the
30:57
Speaker 2: structure of your own language, you can, and you know,
31:00
Speaker 2: you also learn about the culture, right, because so much
31:03
Speaker 2: of how things are expressed in different places is tied
31:05
Speaker 2: to the nuance and the history of kind of what
31:08
Speaker 2: how So I think, like you, that's all valuable and
31:11
Speaker 2: interesting stuff to get into, But then I don't know.
31:15
Speaker 2: At the same time, we only have so many hours
31:17
Speaker 2: in the day, so people need to prioritize what they're
31:18
Speaker 2: going to learn, and it may be that okay, in
31:20
Speaker 2: a world with perfect translation, which by the way, we
31:23
Speaker 2: basically just announced on the ray Bound metas that now
31:25
Speaker 2: you're going to be able to just like you go
31:27
Speaker 2: to different countries. We're starting out, Yeah, we're starting out
31:30
Speaker 2: with just a few languages, but we'll roll it out
31:32
Speaker 2: to more and you know, you'll you could be traveling
31:34
Speaker 2: anywhere and you have your glasses and they just translate
31:36
Speaker 2: in real time in your ear. So it's wild. Yeah,
31:41
Speaker 2: So I think people are gonna need to choose what
31:43
Speaker 2: they want to focus on going forward.
31:45
Speaker 1: How do the developments that we've been talking about in
31:48
Speaker 1: AI intersect with social media and the platforms that most
31:52
Speaker 1: people use today. There's a future where there's generated images
31:58
Speaker 1: and generated text and maybe AI influencers. How does generative
32:03
Speaker 1: AI change how social media feels in the future.
32:08
Speaker 2: Yeah, I mean, I think that's that's a really deep one.
32:12
Speaker 2: You know, there's already been one big shift, which is
32:14
Speaker 2: that social media started out as people primarily interacting with
32:19
Speaker 2: their friends, and now it is you know, at least
32:22
Speaker 2: half of the content is basically people interacting with creators
32:26
Speaker 2: or content that's not created by people who they kind
32:29
Speaker 2: of personally know. So we sortaly sort of already have
32:33
Speaker 2: that paradigm, and I think AI is probably going to
32:38
Speaker 2: accelerate that. It will give all these people additional tools. Right, so,
32:42
Speaker 2: your friends will create kind of funnier memes and more
32:45
Speaker 2: interesting content that'll come from a lot of different ways.
32:51
Speaker 2: I think some of it will be Okay, your friends
32:52
Speaker 2: have glasses and they capture a bunch of stuff, and
32:56
Speaker 2: before they might have not been able to edit it
32:58
Speaker 2: to make it interesting, or maybe it's just too much,
33:00
Speaker 2: or they didn't even realize that they captured something amazing.
33:02
Speaker 2: But now the AI is like, hey, I like made
33:04
Speaker 2: this thing for you out of your content, and it's like, okay,
33:07
Speaker 2: that's awesome, Like people will enjoy that. Creators obviously kind
33:12
Speaker 2: of much more specialized skills are going to be able
33:16
Speaker 2: to use even more advanced AI tools to make more
33:19
Speaker 2: compelling content. But then I think that there will be
33:21
Speaker 2: a bunch of kind of green Field type stuff where
33:24
Speaker 2: maybe in the future there will be content that is
33:29
Speaker 2: purely generated by AI, by the system personalized for you.
33:35
Speaker 2: Maybe it's summarizing things that are out there that are
33:38
Speaker 2: going to be interesting. Maybe it's just producing something funny
33:42
Speaker 2: that makes you laugh. This is going to be like
33:44
Speaker 2: a very kind of deep zone that there's a lot
33:48
Speaker 2: to experiment with. I know there are going to be
33:50
Speaker 2: AI creators as well as creators building AI versions of themselves.
33:54
Speaker 2: I mean, that's a thing that we just showed too.
33:57
Speaker 2: It connect is basically I mean, if you're a creator,
34:01
Speaker 2: one of the big challenges is like, all right, there
34:03
Speaker 2: are only so many hours in the day, and your
34:06
Speaker 2: community probably has a nearly unlimited demand to interact with you,
34:10
Speaker 2: and you want to interact with them because you're trying
34:12
Speaker 2: to grow your community. I mean that's both socially and
34:14
Speaker 2: from a business perspective. That's sort of growing the community
34:17
Speaker 2: is an important part of what every creator does. So okay,
34:21
Speaker 2: if we can make it so that each creator can
34:24
Speaker 2: basically make like an AI artifact that their community can
34:29
Speaker 2: interact with, people will be clear that it's not the
34:30
Speaker 2: actual creator themselves, but it's almost like a piece of
34:34
Speaker 2: digital art that you're producing, like an interactive sculpture or
34:37
Speaker 2: something that it's like, it's like you train it to
34:39
Speaker 2: Here's the context that I wanted to have, Here's the
34:41
Speaker 2: topics I wanted to communicate on. Here's stuff that I
34:44
Speaker 2: wanted to stay away from. You're giving your community something
34:48
Speaker 2: to interact with when you can't be there to kind
34:51
Speaker 2: of answer all the questions, and I think that's going
34:55
Speaker 2: to be super compelling. So there's like all these interesting things.
34:58
Speaker 2: But I think it's AI. It's kind of like the
35:02
Speaker 2: Internet in a way where it's probably going to change
35:06
Speaker 2: almost every field and almost every feature of every application
35:11
Speaker 2: that we use. It seems sort of hyperbolic to say that,
35:16
Speaker 2: but I do think that's true, and it's just hard
35:18
Speaker 2: to sort of enumerate all the different things upfront, but
35:22
Speaker 2: that over the next five to ten years, we're just
35:23
Speaker 2: going to explore the impacts in each of these areas
35:26
Speaker 2: and it's just going to be an amazing amount of
35:28
Speaker 2: innovation and really exciting.
35:30
Speaker 1: I feel two things simultaneously when you say that. I
35:33
Speaker 1: feel both like, I really want to be optimistic about
35:37
Speaker 1: the future of these platforms, and I obviously have gained
35:41
Speaker 1: so much from an enormous pace of change, right like
35:45
Speaker 1: everything that we're doing now, And what I actually feel
35:48
Speaker 1: is worried. I feel some specific concerns around the way that,
35:54
Speaker 1: you know, I might communicate with an audience, in the
35:56
Speaker 1: way that they might respond to that, or the way
35:58
Speaker 1: that human communication might change, but also more generalized just
36:02
Speaker 1: sort of fear of the pace of change and worry.
36:05
Speaker 1: And I don't think I'm alone in that feeling. Yeah,
36:07
Speaker 1: And you're supposed to be the optimist I know, And
36:09
Speaker 1: I'm curious, like how you talk to people who feel
36:13
Speaker 1: that way. What concerns do you feel our most legitimate
36:17
Speaker 1: and what do you feel most misunderstood.
36:21
Speaker 2: I think the pace of change is always a concerning thing, right,
36:26
Speaker 2: It's there is a lot of uncertainty about how how
36:30
Speaker 2: things will go in the future. And we're all going
36:33
Speaker 2: to get really amazing new tools to do both our
36:37
Speaker 2: hobbies and our jobs, and they'll make it so we
36:43
Speaker 2: can do better work and have better lives. But at
36:45
Speaker 2: least on the professional side, it's going to be our
36:47
Speaker 2: responsibility to keep up with that or else it's going
36:50
Speaker 2: to be difficult for us to compete with other people
36:53
Speaker 2: who are doing a good job of kind of keeping
36:54
Speaker 2: up with a new trend. So I get it. I mean,
36:56
Speaker 2: I think, you know, especially in the you know, line
37:01
Speaker 2: of work of being a creator, and it's a very
37:03
Speaker 2: sort of competitive space. I don't think that like creators
37:06
Speaker 2: necessarily think about it as competitive, but it is, right,
37:08
Speaker 2: It's like it's you know, and and so I get it.
37:14
Speaker 2: I think that this is going to make it so
37:16
Speaker 2: that like the quality of work that people produce and
37:19
Speaker 2: how interesting it is and how much they can communicate
37:21
Speaker 2: and like really efficiently is just going to kind of
37:25
Speaker 2: go through the roof. But when you're staring down a
37:29
Speaker 2: set of changes, like you know that there's some big
37:33
Speaker 2: change coming and you don't know what it is. That's
37:37
Speaker 2: always a time of anxiety. So I get it.
37:40
Speaker 1: If I take my creator hat off and I'm just
37:42
Speaker 1: a person who is young ish, starting out my career, ish,
37:48
Speaker 1: starting out building a family, how would you advise someone
37:53
Speaker 1: like me to prepare well for the future that we're
38:00
Speaker 1: added toward, to be able to learn new skills now,
38:05
Speaker 1: or just think about this future in an educated way.
38:08
Speaker 2: Yeah, I mean, I just think maintaining curiosity about things
38:12
Speaker 2: is important. I do think we can overstate to what
38:16
Speaker 2: extent the next ten years is going to be sort
38:19
Speaker 2: of different from the last ten or fifteen. I mean
38:21
Speaker 2: a ton of stuff changed over the last ten or
38:22
Speaker 2: fifteen years too. It's not like this is the only
38:24
Speaker 2: time in history where there's some technology that's going to
38:27
Speaker 2: make it so that there's new opportunities and things change.
38:31
Speaker 2: The Internet coming into maturity and everyone having smartphones has
38:35
Speaker 2: already rewired things dramatically, And I mean, maybe the next
38:39
Speaker 2: period will be a somewhat bigger change, or maybe it won't.
38:44
Speaker 2: I think it'll feel different to different people, But I
38:49
Speaker 2: don't think this is like going from zero to one.
38:51
Speaker 2: It's not like Okay, everything's just kind of been normal
38:53
Speaker 2: and now like now it's about to change. It's like
38:55
Speaker 2: the technology evolves over time, and like the opportunities that
38:59
Speaker 2: we have evolve and improve. And I think that's like
39:03
Speaker 2: the people who do well, I think are people who
39:05
Speaker 2: are generally curious about it and dig in and try
39:11
Speaker 2: to use it to live better lives, rather than the
39:14
Speaker 2: people who who basically, you know, try to fight it
39:20
Speaker 2: in some way.
39:21
Speaker 1: One thing that I really want to ask you about
39:24
Speaker 1: is open sourced.
39:25
Speaker 2: Yeah.
39:26
Speaker 1: I think imagine that we're talking to an audience that
39:29
Speaker 1: has maybe heard that term, but doesn't have any real
39:34
Speaker 1: idea of how that might impact them in the development
39:36
Speaker 1: of AI. How would you explain the reasonable debate that
39:40
Speaker 1: people in your field are having about this right now?
39:42
Speaker 2: Well, I think there are two pieces. I mean, so,
39:45
Speaker 2: what does open source mean. It means that people can
39:47
Speaker 2: build a lot of different things. Right, So, at a
39:49
Speaker 2: high level, I look at the vision that a bunch
39:52
Speaker 2: of companies have, right so open AI, Google, they're building
39:56
Speaker 2: an AI right like one AI that I think in general,
40:00
Speaker 2: they're like, Okay, this is going to be It's like
40:01
Speaker 2: you're going to use They think you're going to use
40:03
Speaker 2: Gemini or chat GPT for like all of the different
40:06
Speaker 2: things they want to interact with, and at a high level,
40:09
Speaker 2: that's just not how I think the world is going
40:10
Speaker 2: to go. I think we're going to have a lot
40:12
Speaker 2: of different AI systems, just like we're going to have
40:15
Speaker 2: we have a lot of different apps. I think in
40:17
Speaker 2: the future, every business, just like they have a website
40:21
Speaker 2: and a phone number and an email address and a
40:23
Speaker 2: social media account, is also going to have an AI
40:26
Speaker 2: that can interact with with their customers to help them
40:29
Speaker 2: sell things, to help them do support. I think a
40:33
Speaker 2: lot of creators will have their own AIS, right, I
40:36
Speaker 2: think like a lot of people will will interact with
40:38
Speaker 2: with a bunch of different things. There's a question of, Okay,
40:40
Speaker 2: do you want a future that's fundamentally kind of very
40:45
Speaker 2: concentrated and where you're interacting with kind of one system
40:48
Speaker 2: for everything, or do you want one where a lot
40:51
Speaker 2: of different people are building a lot of different AIS
40:55
Speaker 2: and systems just kind of like you probably didn't want
40:57
Speaker 2: there to be just one app or just one webs
41:00
Speaker 2: It's like a richer world when there's a diversity of
41:02
Speaker 2: different things. So that's one piece is just giving people
41:06
Speaker 2: the ability to build it themselves. And what open source
41:08
Speaker 2: does it makes it so that everyone can take and
41:09
Speaker 2: modify the model and build stuff on top of it,
41:12
Speaker 2: which is different from the kind of closed and centralized approach.
41:20
Speaker 2: The safety debate is a specific part of this, which is,
41:26
Speaker 2: in a world where AI gets smarter and smarter, what's
41:29
Speaker 2: the way that we have the highest chance of having
41:35
Speaker 2: a kind of positive future and not having a lot
41:37
Speaker 2: of the safety concerns. And I think some people think
41:41
Speaker 2: that if we keep the model closed and don't give
41:46
Speaker 2: it to a lot of developers, that should make it safer,
41:51
Speaker 2: because then you don't get bad developers doing bad things
41:55
Speaker 2: with the model. Historically, I think what we've seen with
42:01
Speaker 2: open source is actually the opposite, which is that this
42:05
Speaker 2: is not the first open source project, right, I mean,
42:08
Speaker 2: this is obviously this has been a thing in the
42:09
Speaker 2: industry for decades, and I think what we've traditionally seen
42:12
Speaker 2: is that open source software is safer and more secure
42:16
Speaker 2: largely because you put it out there, more people can
42:20
Speaker 2: scrutinize it because they can see all parts of the system.
42:23
Speaker 2: And then there are inevitably issues with any software, there
42:27
Speaker 2: are bugs, there are security issues, and initially with open source,
42:31
Speaker 2: people thought, hey, if you're putting the software out there
42:34
Speaker 2: and there are holes in it, isn't everyone just going
42:36
Speaker 2: to go exploit those holes, and especially the bad guys.
42:39
Speaker 2: But it turned out that it's sort of in this
42:41
Speaker 2: counterintuitive way that by making by adding more scrutiny to
42:46
Speaker 2: the systems, the holes became apparent quicker, and then we're fixed,
42:50
Speaker 2: and then people roll out a new version, just like
42:52
Speaker 2: we roll out a new version of our models, right,
42:54
Speaker 2: Lama three, Lama three point one, Lama three point two,
42:56
Speaker 2: everyone upgrades. So I think the same thing is going
43:00
Speaker 2: to happen here. I think it's sort of this counterintuitive
43:02
Speaker 2: thing where even though I think there's some concern around, all, right,
43:09
Speaker 2: are bad guy is going to do bad things with
43:12
Speaker 2: these models, I actually think you just get a kind
43:15
Speaker 2: of smarter and safer model for everyone the more it's
43:18
Speaker 2: rolled out and the more kind of scrutiny is on it.
43:22
Speaker 2: And then part of that is we get feedback and
43:24
Speaker 2: we make the model saferst that as we roll it
43:26
Speaker 2: out to more people, it's safer for more people you use.
43:28
Speaker 2: So I think that the history of open source and
43:31
Speaker 2: the software industry generally would suggest that open source is
43:35
Speaker 2: going to lead to a more prosperous and safer future.
43:39
Speaker 1: Our show is called huge if True, and what I
43:41
Speaker 1: mean by that is kind of testing the most optimistic,
43:45
Speaker 1: non obvious thing. And so my question to you is
43:48
Speaker 1: what is the biggest, open, genuine question on your mind
43:51
Speaker 1: right now?
43:56
Speaker 2: In which field.
43:58
Speaker 1: You're in? So many I am particularly curious about the
44:05
Speaker 1: combination of AI and hardware, but I realized that we've
44:09
Speaker 1: covered a lot, so I'm curious the direction you'd take
44:12
Speaker 1: this on a question that occupies you right now.
44:14
Speaker 2: Gosh, I mean, I think maybe one that's a little
44:16
Speaker 2: more AI specific. Is there a current set of methods
44:20
Speaker 2: that seem to be scaling very well? Right? So, with
44:25
Speaker 2: past AI architectures, you could kind of feed an AI
44:29
Speaker 2: system a certain amount of data and use a certain
44:32
Speaker 2: amount of compute, but eventually it hit a plateau. And
44:36
Speaker 2: one of the interesting things about these new transformer based
44:39
Speaker 2: architectures over the last five to ten years is that
44:44
Speaker 2: we haven't found the end yet. So that leads to
44:46
Speaker 2: this dynamic where LAMA three we could train on ten
44:53
Speaker 2: to twenty thousand GPUs, LAMA four we could train on
44:56
Speaker 2: more than one hundred thousand GPUs Lama five. We can
44:59
Speaker 2: plan and to scale even further, and there's just an
45:03
Speaker 2: interesting question of how far that goes. It's totally possible
45:08
Speaker 2: that at some point we just like hit a limit
45:10
Speaker 2: and just like previous systems, there's an asymptote and it
45:12
Speaker 2: doesn't keep on growing. But it's also possible that that
45:16
Speaker 2: limit is not going to happen anytime soon, and that
45:21
Speaker 2: we're going to be able to keep on just building
45:23
Speaker 2: more clusters and generating more you know, synthetic data to
45:27
Speaker 2: train the systems, and that they're just going to keep
45:30
Speaker 2: on getting more and more useful for people for quite
45:32
Speaker 2: a while to come. And it's a really big and
45:36
Speaker 2: high stakes question, I think, for for the companies because
45:40
Speaker 2: we're basically making these bets on how much infrastructure to
45:43
Speaker 2: build out for the future, and this is like hundreds
45:47
Speaker 2: of billions of dollars of infrastructure. So like, I'm clearly
45:50
Speaker 2: betting that this is gonna keep scaling for a while.
45:54
Speaker 2: But it's one of the big questions I think in
45:55
Speaker 2: the field because it is possible that it doesn't, you know,
45:59
Speaker 2: that it would lead to a very different world where
46:03
Speaker 2: I mean, I'm sure people still figure out eventually just
46:06
Speaker 2: need to make some new fundamental improvements to the architecture
46:09
Speaker 2: in some way, but that might be a somewhat longer
46:12
Speaker 2: trajectory for Okay, maybe you know, the kind of fundamental
46:16
Speaker 2: AI advances slow down for a bit, we just take
46:19
Speaker 2: some time to build new products around this, or it
46:22
Speaker 2: could be the case, And that's what I'm betting on,
46:25
Speaker 2: that the fundamental AI will just continue advancing for quite
46:28
Speaker 2: a while, and that we're going to get both a
46:30
Speaker 2: new set of products that are just really compelling in
46:33
Speaker 2: all these ways, and that the technology landscape and what's
46:35
Speaker 2: possible will just continue being dynamic over like a twenty
46:38
Speaker 2: year period. And that's probably what I guess is going
46:42
Speaker 2: to happen. But I think it's one of the bigger
46:45
Speaker 2: questions in the industry and kind of for technology across
46:48
Speaker 2: the world today.
46:49
Speaker 1: Is there anything else that you want to say? I
46:52
Speaker 1: don't know, for good amazing, Yeah, thank you so much
46:56
Speaker 1: for doing this.
46:56
Speaker 2: Yeah, no, thank you