Mark Zuckerberg on AI Glasses, Superintelligence, Neural Control, and More

0:00 Somewhere between 1 and 2 billion people

0:01 in the world to have glasses for vision correction today, within 5 to 7 years.

0:05 Is there like any world where those aren't all replaced with eyeglasses?

0:09 The neural pad is going to be personalized to you over time.

0:11 It's not actually picking up motion.

0:12 It's picking up your muscles, firing.

0:14 A lot of people who are designing technology,

0:16 they make something that's kind of bulky,

0:18 and they add a bunch of core functionality to and they're like,

0:21 why don't people want to wear this?

0:22 It's kind of a mantra internally that we have,

0:24 which is this idea that AI is going

0:25 to be the most important technology in our lives.

0:27 We want to make sure that AI is in service of people,

0:29 not just like something that sits in data centers and automates stuff.

0:33 There are rumors that you're,

0:34 like personally going founder mode calling researchers.

0:37 What drove the decision for this llama for was not

0:40 on the trajectory that they thought it needed to be on, like,

0:43 why am I going out and meeting all of the top AI researchers?

0:46 It's like, well, like, I want to know who the top AI researchers are,

0:48 and I want to, like, personally have relationships with them.

0:51 I moved everyone around me who sat with me, and now the lab is there.

0:55 So, I mean, I'm pretty hands on every time

0:57 I think of what a milestone would be in AI.

0:59 They all seem to get achieved sooner than we think.

1:05 Thanks so much for being here.

1:05 Yeah.

1:06 Good to see you.

1:06 Thanks for doing this.

1:07 So today we're talking everything connect 25.

1:12 Can you give us the rundown of everything

1:14 you're announcing and what you're personally most excited for?

1:17 Yeah.

1:17 So the main things that we announced at connect were

1:21 our fall 2025 line of glasses and, you have them here.

1:27 So, so we can just start and go through them.

1:29 I mean, the first one is the next generation of Ray-Ban meta.

1:32 So that's sort of the classic eye glasses that we've shipped so far.

1:37 They're some of the most, you know,

1:40 fastest growing consumer electronics of all time.

1:42 We're very happy with that.

1:44 And a lot of the improvements that we have for these are,

1:49 you know, we doubled the battery life.

1:51 We we have three K video resolution in them now for, for capture.

1:56 And we're introducing new AI features

1:59 like this, this thing conversation focus which,

2:02 allows you if you're in a loud place,

2:05 to basically turn up the volume on friends who you're talking to.

2:08 So if you're in a restaurant or something like

2:10 that, that I think is gonna be really neat.

2:12 Then we've got this guy, which is the Oakley Meta Vanguard.

2:16 So this is our second collab, with Oakley.

2:20 It's, it's more of kind of a performance glasses.

2:24 And, these I think are really cool for a number of things.

2:28 And they're designed for, for a number of things around sports.

2:31 You've got the camera that's centered.

2:33 So that's great for alignment.

2:34 It's wider field of view, louder speakers.

2:37 They're they're water resistant.

2:39 They can, you know, just just a bunch of that, they

2:43 can connect with your Garmin watch so you can,

2:45 you know, you can be running a marathon and, and basically, you know, like,

2:51 you can tell it every mile, capture a video and then at the end,

2:54 it'll produce a video for you where it stitches them all

2:57 together and puts your stats from Garmin on top of it.

2:59 So I think that that's that's a pretty neat thing.

3:01 But I think the most interesting thing by far that we announced is this one,

3:06 which we call ribbon, meta ribbon display.

3:11 And that is because it is the first eyeglasses that we

3:15 have shipped that have a high resolution display in them.

3:19 And the other big breakthrough is that they pair with, with this guy,

3:24 the, the meta neural band, which is the first mainstream neural interface,

3:30 that we're shipping as the way to control them.

3:33 And I think that that's pretty neat.

3:35 I mean, basically every computing platform has its own new input method, right?

3:41 So when you went from computers

3:42 with the keyboard and mouse to phones with touchscreens,

3:45 you kind of got a completely new input method.

3:48 And the same, I think is going to be true

3:50 for glasses and having a neural interface where you can just,

3:53 you know, send signals from your brain with micro muscle movements like this.

3:57 I mean, this is basically about as much movement as you need to make.

4:01 And you can, you know, enter text, neurally.

4:04 just do all the, all the fun stuff that you got a chance to try out.

4:07 But but I think that this is a pretty big breakthrough.

4:09 So I'm very excited about this, but but but overall,

4:11 I mean the whole lineup is good.

4:13 So so it was a it was a fun connect.

4:15 So you had the Ray-Ban metas for kind of every day.

4:19 The Oakleys for athletes.

4:20 And then the displays now for power users.

4:23 How do all these glasses kind

4:25 of tie into that personal superintelligence vision.

4:27 Yeah.

4:27 So I mean, our theory is that glasses

4:31 are the ideal form factor for personal superintelligence,

4:33 because it's the only real device that you can have that can see what you see.

4:39 You can hear what you hear, can talk to you throughout the day and can

4:44 generate a UI in your vision in real time.

4:48 And so there are other devices that people have.

4:51 I mean, obviously you can have some AI on your phone,

4:53 you can do some AI on your watch.

4:55 You could do some of it if you just have like AirPods type type thing.

5:00 But, I think glasses are going to be the only thing that can that can do

5:06 all of the pieces that I just said kind of visual and audio in and out.

5:13 And that that I think is going to be is going to be a really big deal.

5:19 So I got to try it for this interview.

5:21 Yeah.

5:21 And I have to say, there were way more many

5:24 use cases than I initially thought there was going to be.

5:28 Like the AirPods, for example,

5:29 have the live translation, but this is live translation.

5:32 Yeah.

5:32 Yes.

5:33 The captions, there's so much more.

5:34 Yeah, the subtitles are awesome.

5:36 It's awesome.

5:37 So I'm curious, like,

5:38 what are your favorite kind of use cases that you've been using them for?

5:41 I mean, the thing that, that I really had in mind

5:46 when we were designing them for is just sending messages, right?

5:49 I mean, this is the thing that I think we do the most frequently on our phones.

5:53 I mean, I would guess that, you know,

5:54 we all are sending dozens of messages a day.

5:57 I don't really need to guess.

5:58 And we run a lot of the messaging apps that that people use.

6:01 And so I think we know that people spent sent dozens of messages a day

6:06 and, we wanted to make it so that the experience is just really great.

6:12 So now with the glasses, a friend texts you,

6:18 shows up in the corner of your eye for a few seconds.

6:21 It's off center, so it doesn't just block your view, it goes away quickly.

6:24 It's not distracting, but if you want,

6:28 you can just respond as easy as, you know, moving your finger like this.

6:33 And you know I can, right?

6:35 I mean, I'm up to like 30 words a minute typing with the neural text.

6:39 So I think this is I think this is great.

6:44 one of my kind of mental models for this is that, you know,

6:48 in-person conversations are awesome.

6:50 The one thing that I like about zoom is that you can multitask, right?

6:54 It's like if we're having a conversation and you say something,

6:57 I want to get some context,

6:59 on what you're saying or if you remind me that I should go do something or, oh,

7:04 you bring up a person and it's like, oh,

7:05 yeah, I've been meaning to reach out to that person.

7:07 Like if we're having a live conversation,

7:10 I'm not just going to do that right then and then the chances

7:13 that I'm probably actually going to forget by the end of the conversation.

7:15 I follow up on most those things on zoom now.

7:19 I kind of feel like you can just kind

7:22 of text someone off to the side and that's that's fine.

7:26 And this, I think is going to bring that to physical conversations,

7:31 to where it's like, you don't have to break your presence.

7:33 You don't have to like, you know, it's not it's not rude.

7:36 Right?

7:36 You're not like, breaking out a phone.

7:38 You're not really.

7:39 I mean, you can continue to really pay attention to the person

7:41 at just like a very quick gesture with your wrist.

7:44 And, I know I keep on having this, this, this phenomenon where there's,

7:49 like, some information I, I want to have in the middle of a conversation.

7:53 And normally otherwise,

7:53 I would have had to wait till the conversation and text or call the person

7:57 to get the information and go back to the person I was having the conversation.

8:00 Now I just get the information.

8:01 In the middle of the call.

8:02 I just text the person, information comes in and then it's like,

8:04 okay, you can have a much better and more informed conversation.

8:07 So texting I think is really going to be the big thing.

8:09 So how does a texting feature work?

8:11 I know there's like you can talk to me I but you're saying you can actually

8:15 write it down and it automatically goes

8:17 to the phone is that gets the glasses to okay.

8:20 So yeah.

8:21 Yeah.

8:21 With the neural band.

8:22 So like you just really just handwriting.

8:25 Yeah.

8:25 I mean it's you can kind of think about it

8:28 like as if you're holding a small pencil and then what,

8:32 like what motion you would make to write out certain letters.

8:37 But then what's going to happen is that the neural

8:39 band is going to be personalized to you over time.

8:42 So it's going to learn.

8:43 Okay.

8:43 How do you make W or J differently from someone else?

8:47 And it will learn and you will kind of co evolve with it to the K

8:55 to learn with the most minimal version of you making a J is or no or whatever.

9:00 And, and then the feedback loop around that will,

9:03 you'll just be able to make like increasingly subtle

9:05 movements over time where it's not actually picking up motion,

9:09 it's picking up, your muscles firing.

9:13 So over time, I think you're basically going to be

9:15 able to just move your muscles in opposition to each other.

9:19 Not even really move your hand and, and be able to send messages or control

9:24 a UI with your hand at your side in a jacket pocket behind your back.

9:28 It's not.

9:29 I mean, this isn't, it's not hand recognition.

9:33 It's not, it's not the cameras on the glasses

9:36 looking at what your hand is doing.

9:38 It's this band picking up micro signals that your brain

9:41 is sending to your muscles before they even move,

9:45 so you don't actually have to move if you just need

9:46 to, to be able to send the way to send the signals to it.

9:49 And it picks it up.

9:50 And yeah, that's so the wristband reads

9:52 the electrical signals before you even move.

9:55 Yeah.

9:55 Why is that important over.

9:57 Just like adding buttons to the glasses or like improving the voice commands.

10:01 Well, I mean, okay, for voice,

10:04 some of that, a lot of the time you're around other people.

10:08 Right?

10:08 So like, I just think that you want a way

10:11 to control your computing device that is private and discreet and subtle.

10:18 And one of the things that we thought about was,

10:20 okay, can we get whispering to work?

10:23 Because, by the way, voice does work too, right?

10:25 You can control the stuff with voice.

10:26 You can do voice dictation.

10:27 You talked about AI with voice,

10:29 but we just I just thought that there needed to be a way on top of that.

10:33 That was in voice, for for when you want to do

10:36 it more subtly and privately we could be having this conversation.

10:39 I could, you know, have the glasses on and, like, a message, it comes in and.

10:44 Okay, we're talking and I'm like, writing it and I've sent.

10:47 It's done.

10:48 I just I mean, in that time I could have

10:50 sent a ten word message and that would be fine.

10:52 That's so great.

10:53 Okay.

10:53 Then it's like, all right, now I got my response back in and okay,

10:56 here, like maybe I'm writing another response and whatever.

10:59 Yeah.

11:00 Okay.

11:00 Done.

11:01 So, I think that that's going to be I think it's gonna be awesome.

11:06 But in terms of buttons, I mean, yeah, that works too.

11:10 I mean, you can you can take a photo by by tapping the button

11:12 on any of these, you can hold it down to take a video.

11:14 You can, you know, skip to the next track with the swipe or you can,

11:18 you know, start or stop music by tapping on it.

11:21 But, but I think the point is you don't want,

11:25 you don't necessarily want that to be the only way that you could do it.

11:28 Like now with the neural band, I can also just be like,

11:33 I don't know, start and stop music with it in my hand, on my side.

11:35 It's like, all right, like, it's like it's very subtle gesture.

11:38 Like you can't even tell, right?

11:40 Yeah.

11:40 We were joking in the demo room that it's

11:43 almost like right now when you go to classes,

11:45 all the teachers are like, oh, put your phones down, they can see you.

11:48 And like, students are like just checking their phones, distracted.

11:50 But now it's going to be like, hey, like, keep your thumbs out.

11:52 Almost, because it's like, I think it'll be take your glasses off

11:55 or take your glasses off and show your thumbs.

11:57 Maybe both.

11:57 Yeah.

11:58 But, yeah.

11:59 So looking ahead, I guess, what you're saying is,

12:02 like typing and speaking might not be enough to give you enough data to I.

12:06 What you've been typing and speaking might not be enough to give.

12:10 Well now we have electrical signals.

12:12 We can also give to, to the.

12:14 Yeah.

12:14 For the UI.

12:15 Yeah.

12:15 For controlling the UI.

12:16 Do you think like, the neural band is almost like a new interface to it?

12:21 I, yeah, I mean, I think it's it's AI and it's the glasses.

12:26 I mean, there's some parts of the glasses that are like, okay,

12:28 you just have a menu of your apps and you basically can just,

12:33 like, swipe through it very subtly with your hand and it's like, okay,

12:36 now I've selected okay, now I've picked up that, you know, it's I mean,

12:40 this is basically what the interaction looks like is like effectively it's like,

12:43 okay, swipe a couple of times, I brought up the camera.

12:45 Oh, now I have a viewfinder so I can see exactly what the what the image is.

12:49 I want to zoom in.

12:50 I just turn my finger a little bit.

12:51 Actually, that's one of the coolest is when you're playing music.

12:55 The way that you change the volume is you just pretend that there's

13:01 a volume dial and you just turn your wrist and it's like,

13:06 it just feels like magic.

13:07 It's like really fun.

13:09 The way you can bring up meta AI anywhere, you just tap twice.

13:15 It's like.

13:16 But but it's like.

13:17 But I mean tiny, right?

13:19 It's like the kick be your side.

13:21 Like no one would even notice.

13:22 So.

13:22 Yeah.

13:23 No, I think it's, I think this is just

13:25 going to be by far the most intuitive way.

13:28 But if you don't have it, yeah, you can control them with voice.

13:30 You can control them with, with, with, with the buttons.

13:33 I mean, different ways to do it.

13:34 Yeah.

13:35 So it's going to be both the AI

13:36 and the, the UI of glasses and potentially over time,

13:39 other things beyond glasses.

13:40 I think that you'll be able to control with the neural band.

13:43 That's really fascinating.

13:44 Yeah.

13:44 How long did it take you to pick up the controls?

13:47 For the neural band?

13:48 Yeah, for the neural band.

13:51 Not that long.

13:52 I mean, it's pretty intuitive.

13:53 I mean, the way that you kind of think about it is, navigating.

13:57 It's kind of like a D-pad.

13:58 You kind of swipe your thumb right, left, up, down.

14:02 You bring up the, the, the menu by kind of, you know, you can tap twice.

14:07 You bring up that AI by tapping twice, you type as if you have a there.

14:11 Enter text as if you have a pencil.

14:13 The trick is that when you're learning upfront,

14:16 the motions are a little bit bigger.

14:19 And then very quickly, within a few days of using it,

14:22 you get to like very subtle movements.

14:25 and the reality is, is that a lot of the optimization

14:29 isn't going to be on our side or your side,

14:31 it's going to be the machine learning system over time,

14:34 getting personalized to you and getting better.

14:36 And, you know, it's the the times when the the neural text entry is slow.

14:44 It's because I make a mistake and the autocorrect doesn't fix it.

14:47 So now I have to go back manually and re enter the text.

14:51 But yeah, I mean this is just like standard machine learning stuff

14:54 as this just gets better and better and learns your patterns better,

14:57 that just it's gonna be able to autocomplete more of what

14:59 you're doing so you won't even have to make all the motions.

15:01 So you've been calling classes kind of the next platform for years.

15:04 Yeah.

15:05 But the original or even met has really took off with just audio,

15:09 and I, what did you guys stumble on there?

15:13 So we have a few principles and building glasses,

15:16 and the three principles that we talk about

15:20 the most are they have to be great glasses first.

15:24 Right.

15:24 So you have to think about,

15:27 okay for any of the before you even get any of the technology.

15:31 You if you're going to wear these on your face all the time,

15:34 they need to be good looking.

15:35 They need to be light.

15:36 And these are right.

15:37 I think a lot of people that design, who are designing technology,

15:41 they make something that's kind of bulky and they

15:44 add a bunch of cool functionality to and they're like,

15:46 why don't people want to wear this?

15:47 It's like, well, most of the day

15:49 people aren't taking photos or something, right?

15:52 It's I mean, most of the day it's just there.

15:54 And the magic of these is that they're good looking,

15:56 comfortable glasses that you want to wear, and then you have them on your face.

15:59 So when you do want to listen to music or call I or whatever,

16:02 it's it's available to do that.

16:05 So good looking glasses first.

16:08 The other big principle is that the technology

16:11 needs to get out of the way, right.

16:13 So I think a lot of people, when they're designing technology, they, you know,

16:18 they have this impulse to kind of add a lot

16:22 of flourishes and, make it so that the thing that they've built,

16:28 is kind of center stage.

16:30 And, one of the things that I keep pushing back on with our team is no,

16:36 actually, look, especially if you're building glasses,

16:38 it is extremely important that whatever technology you have,

16:42 it is there when you want to use it and then it gets out of the way.

16:44 Right?

16:44 You don't want random stuff in your vision, right?

16:46 It's like, I don't care how cool

16:47 of an animation you made on the design team, right?

16:50 It's like like people, we want this to be super minimal, right?

16:53 If message comes in, you see the message, it goes away when you don't want it.

16:58 that I think the, the classic design just distill those principles.

17:05 So yeah, I think it was kind of a hit for those reasons.

17:08 The third principle, for what it's worth, is take superintelligence seriously.

17:12 So it's a it's kind of a mantra internally that we have, which is,

17:16 this idea that AI is, is going to be the most important technology in our lives.

17:21 And, we want to make sure that AI is in service of people,

17:24 not just like something that sits in data centers and like automate stuff.

17:29 And, so we design these, even the early version,

17:34 the first version of this, to be able to just have easy software

17:40 updates to have improvements like the conversation

17:43 focus thing that we talked about.

17:45 So that way you can get like,

17:46 can I have this great accessibility feature, you're in a loud place.

17:48 You can turn up the volume on a person, right?

17:50 It's like kind of crazy, but it's like, okay,

17:52 so you build in the sensors first that you think

17:55 that the AI is going to need over the next couple of years,

17:58 and then you make it so as you get the software working,

18:00 you can just update it so that we

18:02 can bring the latest AI technology to two people,

18:04 make them smarter with a software update.

18:07 Well, I'm a big fan of personal question.

18:11 Yeah.

18:11 How far away do you think we are from these glasses being so good?

18:14 You just replace your phone?

18:17 So I think about it a little bit differently.

18:19 I think about what is your main computing device right now?

18:23 Phones are your main computing device.

18:25 It's my main computing device for sure.

18:29 But I didn't get rid of my computer, right.

18:32 I just I have my computer, but I just use it less.

18:35 Even a lot of the time.

18:36 I'm even at my desk, and, my computer's right there.

18:41 But if I want to go do something,

18:42 I just take out my phone because that's like my primary thing that I do.

18:44 So what I think is going to happen here.

18:48 Is that I don't think we're going to get rid of our phones anytime soon.

18:51 I just think our phone is going to stay in our pocket more or in our bag.

18:56 And, even very subtle things like, I don't know how many times a day I used

19:02 to look at my phone to see what time it was.

19:05 Now, I don't do that when I have the glasses.

19:07 You just tap quickly.

19:09 The clock comes up done right.

19:12 It's like, okay, I don't need to take out my phone to look at the time, okay?

19:15 Now, like I don't need to take out my phone to look at messages, right?

19:19 It's like, okay, now with these, the camera's great.

19:22 But now with the viewfinder, I know, like, exactly what I'm capturing.

19:26 So, so just, I think you just go kind of use case by use case and all

19:30 these things I would have used to have taken out my phone for, I now don't,

19:34 I think that's sort of the way it's going to go, but I would be surprised if.

19:40 If we kind of get rid of our phones anytime in the next five years,

19:44 I just think it's going to be less and less use slowly using the phone less.

19:49 Yeah.

19:49 I mean, the GPS feature is another one, right?

19:52 Instead of having to look at my phone constantly,

19:53 I'm going to use this like all the time.

19:55 Yeah.

19:55 Just have it in my glasses.

19:57 Yeah.

19:57 Navigating me where I'm wearing to go with the lines.

20:00 Translation even was useful.

20:01 Yeah, there's a lot I'm curious to see

20:04 how other people are going to be using it, but, I think and I think people are

20:08 going to be really curious about is the metaverse.

20:11 How much of the vision is still distilled in these products?

20:14 Well, I think we're getting closer to it.

20:16 Right.

20:16 So the metaverse is obviously it's a very visual experience presence.

20:21 There's some parts of it like stereo audio, which, I mean,

20:24 you can 3D audio which you can, you can get that as great.

20:27 But, you know, starting to get the display, and then, you know,

20:35 the Orion prototype that we showed last year, it's a wider field of view.

20:40 It'll be a more expensive product when it's, when it's ready,

20:43 but I think you're basically going to get to, you know,

20:46 this is one where you can you can have a hologram.

20:49 You can you can kind of have context.

20:51 It's enough to kind of watch a video or have a message thread,

20:54 but this product, especially because it's monocular,

20:59 it's not putting 3D objects in the world.

21:02 Right.

21:02 So there still are things around delivering the sense of presence,

21:06 like you're there with another person that I

21:10 think is going to be really magical.

21:11 And when we get the consumer version of Orion,

21:14 you'll you'll really be able to, to start doing that.

21:17 So I think we're kind of inching our way there.

21:20 But this is a big breakthrough.

21:23 I think the first kind of high resolution display with the neural band,

21:27 and it's going to be a good step for learning how that goes.

21:31 But yeah, no, I think the vision is that all

21:34 of the kind of immersive software around presents that we have for VR,

21:37 we would like a version of that to run on glasses, too.

21:40 I want to switch gears to the the Vanguards.

21:42 Yeah.

21:42 The other fun, really stylish.

21:44 I love to design, a recreation.

21:47 We go on runs and on cycle.

21:49 So I am using those.

21:50 Well, how did you build these with athletes in mind?

21:55 Well, I think a lot of people on the team,

21:58 are pretty intense athletes, and and some of it is, you know,

22:03 we just use the Ray-Bans, and, you know,

22:05 I can't tell you how many pairs of Ray-Bans I fried taking them surfing.

22:09 But it was worth it.

22:10 But they're not water resistant, so.

22:12 I mean, it's pretty obvious, like, you were like,

22:15 okay, we want something that's water resistant.

22:17 The extra, sound, the kind of power in the speakers.

22:21 These are six decibels louder is really

22:25 valuable and helpful when you're doing loud things.

22:28 So, if you're cycling at 30 miles an hour and it's windy.

22:32 The other day, a few weeks back, I was taking a call on a jet ski,

22:36 and I could hear the person completely fine over the engine of the jet ski.

22:40 And, and we have this advanced, like, noise reduction in the background.

22:46 So we've designed these.

22:49 It's got the microphone in the nose pad,

22:51 which basically means that you could be in a wind

22:55 tunnel and the other person on the other

22:57 end of a phone call would not be able to hear any of the background noise.

22:59 You just come in crystal clear.

23:01 So, I think stuff like that is awesome.

23:04 Then, you know, pairing it with, and then there's all

23:07 this other stuff that I think you want for sports.

23:10 The video stabilization is good.

23:11 The pairing with Garmin and Strava, I think is good.

23:13 We put an LED in the corner, so that you can have it light up when you're,

23:18 to give you a reminder if you're to, to make sure you

23:21 stay on pace that you want or in your target heart rate zone.

23:26 So yeah, I mean, there's a lot of there's a lot of things like

23:29 that that I think are going to be are going to be pretty neat for athletes.

23:32 Yeah.

23:33 It's fascinating.

23:34 Why do you think it was important

23:36 to design so many different glasses for different lifestyles?

23:40 And how much broader do you really expect to go from here?

23:43 Well, I mean, I think people have different styles, right?

23:45 So, this isn't like a phone where everyone's going

23:49 to be okay having the same thing and just like,

23:51 can maybe have a slightly different color case or like

23:54 put a different sticker on my case and then I'm good.

23:56 Like, I think people have different face shapes.

23:59 They have different esthetics.

24:00 Like it's I mean, people wear different kinds of clothes.

24:02 I think this is like an important part of our identity and sense of style.

24:07 And some of it is functional too.

24:09 I mean, some of it is a little more lifestyle, some of it is more active.

24:12 And, you know, part of this is,

24:18 some people prefer thinner frames, some people prefer thicker frames.

24:22 You will always be able to fit more technology and thicker frames.

24:26 So as we get to more display, like more advanced holograms,

24:31 it will always be a little bit thicker of a frame

24:35 than what you can do for the thinner ones.

24:37 So I think that there's a choice that people will make around okay.

24:41 Like, do I prefer the esthetics and the kind

24:43 of the simplicity of the thinner ones,

24:45 or do I like the esthetics of the of the thick ones?

24:49 It's unfortunately, fortunately,

24:53 kind of big glasses are in style so that that kind of work.

24:55 So there's that, there's the whole, spectrum.

24:59 And then there's also the price point in affordability.

25:01 Right.

25:01 And less technology.

25:02 You're always gonna be able to offer for, for, for a more affordable price.

25:08 And that's going to be great too, because we want to make it so that, you know,

25:10 there's like more than a billion people.

25:12 It's somewhere between 1 and 2 billion people

25:13 in the world to have glasses for vision correction today.

25:17 And I don't know, I think within, you know, whatever it is, 5 to 7 years.

25:22 Is there like any world where those aren't all replaced with eyeglasses.

25:25 And it feels to me like it's kind of like

25:27 the iPhone came out and everyone had flip phones,

25:29 and it's like just a matter of time before they're all smartphones, right?

25:32 So I think that that's going to be the case with glasses too.

25:34 But people have different glasses and they want different glasses

25:37 and they want different glasses at different price points and style.

25:40 So our goal is to just,

25:42 work with a lot of iconic brands to do some, some great work.

25:45 So on the Vanguards, you mentioned the Garmin integration.

25:48 I thought this was fascinating.

25:49 Yeah.

25:50 Because this basically allows AI to kind of see your heart rate pace,

25:55 location, everything kind of all hands free while you're running or cycling.

25:59 How long do you think it will be

26:01 before we have like a personalized coach that's there,

26:03 like proactively in your ear?

26:04 Yeah.

26:05 I mean, it can do that a bit.

26:07 So I think, for running, for example, if you want to ask it what your heart rate

26:14 is or what your pace is, it can answer now.

26:17 I mean, I don't know how hard you run, but when I'm running,

26:20 like for performance, I don't really want to talk to something.

26:24 Right.

26:24 It's like there's you're probably above that heart rate zone.

26:29 But there are times when I'm like skiing really fast and I

26:32 want to know how fast I'm going because that's kind of fun.

26:34 And, you know, you're not necessarily out of breath doing that.

26:37 So, yeah, I mean, part of it is being able to talk

26:39 to it and it can give you the stats in real time.

26:42 That's one of the reasons why we built the LED in is,

26:46 we want to make it so that, you

26:49 just have a very simple visual indicator of, okay,

26:52 am I on my my kind of pace target?

26:55 Am I on my heart rate target?

26:56 But I know you can kind of get a sense of, like, over time,

27:00 it's going to be pretty good to have a display in those, too.

27:03 Yeah.

27:03 I mean, the real time coaching is.

27:05 Yeah.

27:05 So, like, I guess for an example,

27:07 if I'm running and I want to say in zone two, which is,

27:10 yeah, below a certain heart rate,

27:12 I don't wanna go to too hard because I want to stay in the zone.

27:15 I guess my question is, will I be able to have a coach that's telling me,

27:19 oh, hey, slow down a little bit.

27:21 You're all too fast.

27:22 Yeah, that's the goal of the LED to start.

27:25 And if you're trying to stay in the zone two, you probably can't talk to it.

27:27 Right.

27:28 Because I mean that's kind of the definition of zone two, right.

27:30 It's like you should you should be able to.

27:31 Right.

27:32 So it's I don't know if it's the definition, but it's a proper answer.

27:36 Yeah.

27:36 Yeah.

27:37 So, so you can, you can, you can basically you can ask at different stats

27:42 and, and it knows if you're connected to Garmin.

27:46 Yeah.

27:46 So I know you've been training a lot.

27:48 Have you like found any good use of it too

27:51 in your training as, like almost like a coach or.

27:53 Well, I take them surfing and they don't break, which is nice.

27:56 There we go.

27:56 Yeah.

27:57 So, there's the ravens are good.

27:59 But like I said before, I've, fried a few pairs of those, okay?

28:02 And they captured moments, like, in real time too, right?

28:06 Yeah.

28:06 I mean, one of the nice things about this is because the camera is centered.

28:11 It's like, really perfectly aligned.

28:13 I mean, the image really in the video

28:15 feels like it's coming exactly from, from your perspective.

28:19 It's nice even when the camera's off to the side.

28:22 But it's like these things about delivering a sense of presence.

28:28 They kind of have this feeling where it's like when you get it to be perfect,

28:31 it really feels like it clicks.

28:32 And if you're like two degrees away, it's still pretty valuable.

28:35 But it's not.

28:36 But there's something about having it in the middle

28:38 that I think is really quite compelling.

28:40 Okay.

28:40 And wide field of view.

28:42 I'm excited to try them out.

28:43 Yeah.

28:43 So you're raising three kids while building superintelligence.

28:49 I'm curious what conversations you're having right now about what?

28:53 What was the about to grow up in, Well, I mean,

28:59 it it comes up in a bunch of different things.

29:01 But, in our family, we build a lot of robots.

29:07 I, I'm really into 3D printing.

29:10 I think that's like a really fun DIY hobby.

29:14 My daughters are really into it.

29:16 They're like making stuff.

29:18 Kind of got to the point where I'm like,

29:20 all right, you need to design your own stuff.

29:22 I'm not just going to print stuff that we find on the internet.

29:24 So now they're using all the different tools

29:26 to to do that, and I work with them on that.

29:28 It's like a fun hobby, but in a bunch of ways.

29:31 I kind of think of robotics as like the thing after you get I mean,

29:39 obviously they'll they'll kind of develop contemporaneously to some degree.

29:43 But, but I think we're gonna have a really smart,

29:48 kind of just kind of intellectual eye.

29:51 And then you get the robotics,

29:52 which is going to be the kind of fully embodied version of it.

29:55 So, so, yeah, I mean, there's there's not that much that, you know,

30:00 the kids could do with, you know,

30:03 developing AI at their age, but you can actually,

30:07 they can and they just enjoy making stuff.

30:10 So, I mean, it's like kids love 3D printing and it's like, not very complicated.

30:16 And it's just a fun hobby.

30:18 And so yeah, so we have this project going on right now.

30:21 First I just ordered a bunch of robots

30:23 from the internet that you could assemble.

30:25 That because there's like a bunch of starter

30:27 packs and things like that and all of that.

30:29 And then, the next one is basically designing

30:33 our own and basically just like, you know,

30:35 okay, 3D print the shell is you have the, the, you know, whatever the, you know,

30:40 you do a version with the Raspberry Pi or the, the, the Jetson,

30:44 you know, and just like, get that to, to to work pretty well

30:47 for, for running language models and doing interesting stuff with it.

30:50 Yeah.

30:51 That's, that's a step up over Legos.

30:53 That's fun.

30:53 Yeah.

30:54 Yeah, yeah.

30:55 No.

30:56 That's true.

30:58 It's nice because Legos are, You run out of blocks.

31:05 Yeah, right.

31:06 But true.

31:07 But with, like, the internet, you don't you don't run out.

31:10 Are there any, like, specific, I guess,

31:13 traits or skills you're trying to teach your kids?

31:16 So, for example, Dennis Hassabis, he said learning how to learn is a really

31:21 important trait is I think you're thinking of teaching,

31:24 I mean, I mostly try to teach our kids to be good people.

31:32 I mean, I guess there's like the there's

31:34 all the intellectual stuff they're talking about,

31:35 like, what's interesting, robots or whatever.

31:37 But, I don't know.

31:40 I mean, I think, like being caring and kind

31:45 and things like that, I think are really important.

31:47 So just a, a bunch of values type things like that.

31:53 But yeah, I mean, I, I agree with Dennis's answer too.

31:57 It's less about like learning a specific thing and more

32:03 about kind of learning how to go deep on a thing.

32:07 So I'm I'm a big believer that.

32:13 So a lot of people say some variant of like, oh yeah,

32:16 you want to learn how to learn but okay, but how do you do that?

32:18 Right.

32:19 And I guess my theory on how you

32:21 do that is basically a very depth first approach.

32:24 So rather than just like trying to have some conceptual framework,

32:31 you complete a project, right?

32:32 So like you will learn a lot by building a robot, like,

32:37 whether you, actually care about robotics

32:41 after that is completely incidental to, like,

32:45 you took a problem with decomposed into all

32:48 these things we needed to figure out, like the design for the 3D printing.

32:51 And she needed to get the printers to work and then like,

32:54 and then we needed to, like, get the Raspberry Pi to work, and it's like, okay,

32:58 we're like learning a lot about wheels right now, right?

33:00 It's like, okay, we're going to do treads.

33:01 Are we going to do wheels?

33:02 Like, how many wheels?

33:02 Like, how do you want to, you know.

33:03 So and that stuff, I think you basically

33:07 it's like you're learning how to decompose problems, how to debug things.

33:13 And that's like you just learn by doing.

33:16 So that's, that's kind of my, my philosophy on it.

33:18 Yeah, I like that.

33:20 And I think just doing stuff we have common interests.

33:22 So yeah, you know, it's like sometimes we, build a robot,

33:25 sometimes we watch K-pop doing hunters, you know, it's like you got to have fun.

33:28 Okay.

33:29 Okay, so when we do achieve superintelligence,

33:35 these glasses are going to see everything we see here.

33:38 Everything we.

33:38 And now even some of our muscle impulses.

33:42 What human traits should we fight to preserve?

33:44 And what should we just let go?

33:47 Well.

33:50 I'm not sure I agree that they're going to see everything.

33:52 Or maybe there will be like some sensor that does,

33:55 but they're not necessarily going to retain it.

33:56 I'm not sure that you I mean, do you want them to, but so that's a whole

34:02 separate direction that we can that we could go into.

34:05 But I think just kind of like being

34:07 able to pick out the salient bits are important,

34:09 and giving people control over that is important.

34:11 But, but that's not really the core of what you're asking.

34:14 So the, I think like to me creativity is very important.

34:21 And, like having a sense of here's like what you want to make in the world,

34:30 And like, having an idea of how you,

34:32 I mean, it's, it's some of the themes that we just talked about,

34:35 like, how do you decompose that?

34:36 What are the steps to do that?

34:38 What are the tools that are going to be most helpful in doing that?

34:40 I think is part of the job of a creator or a builder is to, like,

34:45 be a kind of master of the tools that are available to them, right?

34:49 If you're not, then you're not going to kind of be at the edge of, of your work.

34:53 You know, it's a that's a competitive world.

34:57 So, yeah, I kind of think.

35:06 I don't know, I think the, the, all of the AI progress,

35:09 I think has been very fascinating because people have sort of conflated

35:14 intelligence with, like an intent or like a desire to do something.

35:20 And I think part of what we've seen so far from the AI

35:23 systems is that, you can actually separate those two things, right?

35:27 Like the AI system has no impulse or desire to create something.

35:31 It just it's kind of it sits there and waits for you to give it directions,

35:36 and then it can go off and do a bunch of work.

35:38 And so I think the human piece at the end is going to be,

35:44 well, what do we want to do to make the world better?

35:47 And I think some of that will be around like

35:50 a personal kind of creative manifestation of what you want to build.

35:56 But I do think some of it then,

35:57 and I think in our industry we probably underplay this a bit,

36:00 is just like caring about other people and, you know,

36:04 taking care of people and spreading kindness and like

36:07 so I think that that stuff is really important too.

36:08 And I think I will help with that as well.

36:10 So this conversation wouldn't be without obviously,

36:14 the superintelligence labs that you've you've been on a spree

36:17 over the last couple past months and all over the news.

36:19 Yeah, it's been busy.

36:21 Yeah, yeah.

36:21 When did you decide to kind of make that change from the outside?

36:26 There are rumors that you were, like,

36:28 personally going founder mode, calling researchers, emailing all that stuff.

36:32 What drove the decision for this?

36:36 Yeah.

36:36 I think I mean, it was largely lama for was not

36:42 on the trajectory that I thought it needed to be on what I mean,

36:46 Lama for was in many ways a big improvement over Lama three.

36:49 But like, we weren't trying to be a bit better than Lama three, right?

36:54 We were basically trying we want to build like we're a frontier lab, right?

36:59 That's who wants to be doing leading work, right?

37:02 That, so I kind of felt like.

37:08 I learned enough about how to set up a lab

37:12 that, that I wanted to kind of reformulate the work that we're doing,

37:19 with a few basic principles.

37:20 So one is huge focus on talent density, right?

37:24 So you don't need many, many hundreds of people.

37:27 You need, you know, 50 to 100 people.

37:31 It's a group science project who can basically

37:33 keep the whole thing in their head at once.

37:36 Like, seats on the boat are precious, right?

37:39 It's like, if someone is not, is not,

37:42 isn't pulling their weight on that, then that's like it has this huge,

37:49 negative effect in the way that it doesn't

37:51 for a lot of other parts of the company.

37:53 So I think like just this absolute focus

37:56 on, like having the highest talent density in the industry,

38:00 drove a lot of what you heard about.

38:03 Okay.

38:04 Well, like, why am I going out and meeting all of the top AI researchers?

38:07 It's like, well, like, I want to know who the top AI researchers are.

38:10 And I want to, like, personally have relationships with them.

38:12 And, and I want to build the strongest team that we can.

38:17 So that was that was a big part of the focus.

38:22 There are other parts to like, I think when you're doing long term research,

38:25 one of the principles that we have for the lab

38:28 is we actually don't have deadlines that are top down, right?

38:31 It's like it's research, right?

38:32 You don't know how long the thing is going to take.

38:33 All the researchers, you know, everyone's competitive.

38:36 They all want to kind of be at the frontier and doing leading work.

38:38 So, you know, me setting a deadline for them isn't going to help.

38:43 It's not it doesn't help them at all.

38:45 Right?

38:45 It's like they're moving as fast as they can to try to understand,

38:50 the, the kind of what we need to solve the problem.

38:53 So, but that's, that's kind of.

38:56 Yeah.

38:57 So we organize the lab to be very flat.

39:00 We don't want layers of management that are not technical.

39:04 Right.

39:04 It's a the issue is people start off technical and then they go into management

39:07 and then six months or a year later you like still think you're technical,

39:11 but you're actually like haven't been doing the stuff for a while.

39:14 So you're the, the knowledge just kind of slowly decays or quickly

39:18 decays in a, in an environment that's moving as quickly as that.

39:21 So anyway, those are the basic principles.

39:25 But I think that overall, taking a step back, I think that AI is going to be

39:29 the most important technology in our lives in our lifetimes.

39:32 Building the capacity to build absolutely leading models, I think,

39:38 is going to be a really critical

39:40 and very valuable thing for, for unlocking creativity,

39:42 building a lot of awesome things.

39:47 So, yeah, I'm absolutely just going to focus and do the things that we need

39:50 to do in order to, to to make sure that we can continue to do that.

39:53 How hands on are you still with the lab?

39:57 Well, I'm in the lab.

39:59 I moved everyone around me who sat with me, and now the lab is there, so.

40:03 I mean, I'm pretty hands on.

40:05 That's like.

40:05 I mean, I mean, you the chief scientist, you know, sits right next to me.

40:10 So.

40:10 And, you know, a lot of the, the other teams sit within 15ft of me.

40:14 So, so, I mean, what would I say?

40:17 I, I mean, look, I'm not an AI researcher, right?

40:20 So it's not like I'm, like, in there, like,

40:23 telling them what research ideas to to do the things

40:27 that I can do is as CEO of the company,

40:31 I can make sure that we have the very best people and talent density.

40:35 I can make sure that we have by far the highest compute per researcher,

40:39 and that we do whatever it takes to go build out all that capacity.

40:43 I mean, you probably saw the announcement

40:45 that we had we're building this, Prometheus cluster,

40:47 which is I think it's going to be the the first kind of gigawatt

40:53 plus single cluster that's like a contiguous

40:55 cluster for training coming online next year.

40:58 And, we're building a, cluster in Louisiana that is,

41:05 it's going to be five gigawatts,

41:08 and we're building several more clusters like that that are

41:11 that are just going to be multiple gigawatts.

41:13 So, yeah.

41:14 So I mean, that's that obviously takes some conviction.

41:16 I mean, we're talking about many hundreds of billions of dollars of capital.

41:20 So you both need to have like a good business model

41:22 that can support it and you need to believe in it.

41:24 So I, that's that's one thing.

41:27 And then like, there's all these other parts of running a company where,

41:30 I don't know, I mean, like big companies have positives and negatives,

41:33 and it's sort of my job to make sure that we

41:36 channel all the positives that, that we can towards this effort.

41:40 And then I can clear as many of the kind of things

41:42 that would otherwise be annoying about running a company out of the way.

41:44 So, you know, when we build something awesome,

41:47 we'll get in our apps and get it to billions of people quickly.

41:49 That's a good thing that you can do inside a big company.

41:52 And then like whatever the other kind of things, that would be more annoying.

41:55 It's sort of my job to clear them out of the way.

41:56 So how do I do that?

41:58 Well, I want to make sure that I know the researcher as well.

42:01 Right.

42:01 And that they, you know, that, like when they have an issue that they feel

42:06 comfortable texting me or just like walking over in my desk.

42:09 And, so I think that that's actually part

42:11 of running the whole thing well is like,

42:14 is that the kind of like I have that rapport with them?

42:17 And, in that way I can be in the information flow to understand what

42:21 the issues are that they're having so I can go sell them for them.

42:25 So it's like building the culture of everything.

42:27 I was open, everything moves fast.

42:29 Yeah.

42:29 And you're personally talking with all these researchers.

42:32 Yeah.

42:33 I mean, I'm doing other stuff for the company too, obviously.

42:35 But yeah.

42:36 And but yes, I'm just trying to make sure

42:39 that we move as quickly as, as, as, as we

42:42 can to get the researchers what they need

42:44 to do the best research in the industry.

42:46 Last question.

42:47 If we do achieve superintelligence, then whenever that is,

42:51 what are the plans to integrate it into existing products?

42:54 Yeah.

42:54 I mean, I think when you have superintelligence,

42:57 the nature of what a product is will change pretty fundamentally.

43:00 I mean, you think about our products today, a lot of the big ones,

43:06 or if you look at Instagram or Facebook or our business model around ads,

43:11 they're basically already these very large

43:16 transformer systems or the recommendation systems,

43:19 not language models today, but they're basically already these big,

43:25 massive scale machine learning systems.

43:30 So, I mean, in the future.

43:34 Do you think we'll, we'll get to a point where, you know,

43:37 instead of having it be a recommendation system that, you know,

43:40 has some kind of basic understanding of what you might care about

43:43 and therefore can kind of spit out an idea for a story.

43:47 And then you have the app that renders it.

43:49 I just think you'll have models that you'll

43:52 interact with directly that, that, that it'll, it'll recommend content.

43:58 It'll generate content.

43:59 You'll be able to talk to it.

44:00 Right.

44:00 It's like just so I think it's this stuff will be much more integrated.

44:04 And then, I think that that sort of culminates in the glasses version where I

44:12 think what you're going to get is

44:15 eventually this always on experience that, I mean,

44:17 you can control when it's on and off,

44:19 but it can be always on if you want where it can,

44:22 you can just let it see what you see here,

44:24 where you hear it can go off and think about the context of your conversations

44:28 and come back with, with more context

44:30 or knowledge that it thinks you should have.

44:33 When you need an app, it can just generate the UI from scratch for you.

44:38 In your vision.

44:41 I, I'm not sure how long it's going to take to get to that.

44:44 I don't think this is like five years.

44:47 I think it's going to be quicker.

44:50 So, two, three.

44:52 It's hard to exactly know, but, but I don't know, I would guess it's,

45:00 every time I think of what a milestone would be in AI,

45:05 they all seem to get achieved sooner than we think.

45:09 So I think like my optimism about AI

45:15 has generally only increased as time has gone on.

45:18 In terms of both the timeline for achieving it and how awesome it's going to be.