Joe Rogan Experience #1342 - John Carmack

1. 0:00 – 15:00

   Three, two, one. (hands thud)…

   1. JRJoe Rogan0:01

      Three, two, one. (hands thud) Here we go. How are you, sir?

   2. JCJohn Carmack0:04

      I'm doing good.

   3. JRJoe Rogan0:05

      Um, I'm super happy to have you here. If there's a Mount Rushmore of video games, you're George Washington. You're up there.

   4. JCJohn Carmack0:12

      So, this is great. People have been kind of nudging us for years and years to get this done, so it's great that we're finally able to make it happen.

   5. JRJoe Rogan0:19

      Yeah. Um, w- I have been, from day one, a gigantic Quake junkie and a DOOM junkie, so for me to have you in here is a, a giant treat. And everybody ... You know, I've talked about your video games and your creations so many times on this podcast, so it's, it's very cool to have you here. And thank you very much for showing me ... Before the podcast got started, I should tell everybody, you showed me the latest and greatest version of Oculus Rift, which is amazing. It's so small. For people watching the YouTube, this is the entire unit. This, this thing that sits on your head, it's very light, and it's not attached to a computer. Like, you don't have to carry anything around, and there's no extra. Everything is in here.

   6. JCJohn Carmack0:57

      Yeah, so this is the Oculus Quest, the standalone device.

   7. JRJoe Rogan1:00

      (laughs)

   8. JCJohn Carmack1:00

      And it's been kind of the culmination of a bunch of different products that we've been working on. And it's the vision that we had, even six years ago, just the idea of not connected to anything. You put this magic hat on, and you're transported to these different worlds, and it's all coming together.

   9. JRJoe Rogan1:14

      And this is available right now. Anyone can buy this, right?

   10. JCJohn Carmack1:16

       Yeah.

   11. JRJoe Rogan1:16

       Yeah. Okay, so it's avai- ... Jamie just actually pulled up-

   12. JCJohn Carmack1:18

       Right.

   13. JRJoe Rogan1:19

       ... the, the website. Um, can I push this up to you? Pull that ... Yeah, there you go.

   14. JCJohn Carmack1:22

       There you go.

   15. JRJoe Rogan1:22

       Um, th- how ... What is the battery life on these things?

   16. JCJohn Carmack1:25

       So, it depends on what you're doing, where, uh, if you just sit there and watch Netflix, uh, you know, it'll last a little more than three hours. If you play some of the really hardcore games, it might only last two hours or so.

   17. JRJoe Rogan1:36

       Do you have numbers on how many people are watching Netflix on this thing?

   18. JCJohn Carmack1:39

       So, our previous one right before this, the Oculus Go, was a little bit more media-focused, and that's one of our more popular ac- applications. I mean, surprisingly, things ... Everybody thought that VR was going to be all about these just amazing gaming experiences, but some of the most popular experiences are doing reasonably conventional things, watching Netflix, watching YouTube, Amazon Prime, stuff like that. Where if you're, like ... If you're you and you've got a great home theater and everything, there's not this much benefit to having a theater like that in VR. But if you're in a situation like you're in a tiny room in Tokyo or something, the idea of being able to put on the VR headset and have this, like, lovely ski lodge atmosphere with a giant screen TV, it has some real benefits. So, in the end, VR should be a replacement for anything you do on screens today, whether it's your phone, your tablet, your TV, your laptop, your PC. All of these should eventually be superseded by just having more flexible screens in VR. I mean, we have lots of challenges now with resolution and comfort for long-term use, but this is the direction that everything's going. Not only do you have things in VR that you couldn't do anywhere else, just experiences that you can't have with that level of immersion, but it should pull along every other thing that people do with screens devices today.

   19. JRJoe Rogan2:55

       Uh, that's ... I didn't consider the television shows, but of, of course, people would be watching Netflix on this if it's possible. Have you done the, uh, Disney World ride, the Avatar ride, Flights of Passage?

   20. JCJohn Carmack3:07

       No, I haven't done that yet.

   21. JRJoe Rogan3:08

       It's amazing. You sit on this ... It's like a motorcycle-looking thing. It straps you in place, and it's supposed to represent one of those flying dragon things in Avatar. And then you have the headset, you put that on, and the vir- virtual reality experience is second to none. I mean, it's incredible. Super high resolution, and y- the, the motorcycle's moving around. You get wind and smells and all these, uh, sensory-

   22. JCJohn Carmack3:32

       Yeah, so that's one of the really interesting things. Like, I think about that whenever I am at amusement parks for things like the Harry Potter rides and stuff like that, where they're doing lots with screens and motion platforms, where ... I think about it from the VR perspective. And anything we're doing visually and audibly, we could go ahead and do a great job in the headset. So, it's cutting it down to these few physical things that you can't do. So, you've got things with motion platforms that actually jostle you around that you can't do in VR. You've got things like smell. And, like, The VOID, um, where they have the Star Wars, uh, experience there, we have a fantastic Star Wars experience on Quest, which in many ways has a lot of that magic, but in The VOID where they set it up and they blow hot air over what's the virtual lava, uh-

   23. JRJoe Rogan4:13

       Yeah.

   24. JCJohn Carmack4:14

       ... towards you, that's something that you still don't get. But it's kind of like the, the age-old battle of what can you do differently in an arcade that you can't do as good in your, uh, your home system? And VR now takes ... Lets you do all of these amazing things there, but if you're willing to spend millions of dollars and build a theme park attraction, essentially, you can still throw some of these extra things in. People joke about, when is Smell-O-Vision coming to VR? And (laughs) there have actually been real companies that have spun up to say, it's like, "Oh, we want to do scent augmentation."

   25. JRJoe Rogan4:44

       Yeah.

   26. JCJohn Carmack4:44

       But it's not a great thing, uh, and those are still the last vestiges of things where you have to go someplace. But the promise of VR is to ... You know, the world as you want it, not having to go to some place to do something magical. And if you can get to 90-something percent of that experience staying in your own room, then that's great.

   27. JRJoe Rogan5:02

       What would they be able to do with Smell-O-Vision?

   28. JCJohn Carmack5:04

       (laughs)

   29. JRJoe Rogan5:04

       Would you have, like, a standalone unit that, like, has, uh, access to the program? So, like, if you were flying over orange fields, it would spray citrus in the air? Sort of, sort of like Soaring Over The World? Have you ever done that Disneyland ride?

   30. JCJohn Carmack5:17

       Yeah.
2. 15:00 – 30:00

   (laughs) So, I... Yeah,…

   1. JRJoe Rogan15:00

      extras and, and things you could download, and maps. So many different maps that people had created that were really interesting. Like, I remember one was a guy's apartment. Like, you could, you could play Quake in an apartment. Like, you could shoot... You could get to the top of the toilet and shoot at things off the toilet. It was really amazing. How did you feel about when people would play the game competitively and they would turn all the textures off?

   2. JCJohn Carmack15:22

      (laughs) So, I... Yeah, that's, uh, that especially is a sore point with the artists that have labored-

   3. JRJoe Rogan15:29

      Yeah.

   4. JCJohn Carmack15:29

      ... for years to build these glorious textures, and then you get the people that just turn them down. And there's two reasons to turn them down. You turn them down to help performance in some cases. In the early days, and especially the early graphics cards, you would get higher frame rate if you turned them down, so you'd have less latency in your response times. But there's also the even more nefarious thing about turning them all down to improve the contrast on your enemy acquisition.

   5. JRJoe Rogan15:52

      Yeah.

   6. JCJohn Carmack15:52

      I... So, people want this almost flat-shaded world, so that any moving set of pixels there, just turn and fire at that. And that's... Um, I never came to really great terms with that, where I always thought on... In the early days of esports and gaming, we did always insist that people have to play with at least plausible resolutions there because we want our game to look good.

   7. JRJoe Rogan16:14

      Yeah.

   8. JCJohn Carmack16:14

      We want people that are looking at it for the first time, seeing these professionals play it. I am... You know, we don't want them to look at that and say, "Well, this game looks like garbage. It's all flat-shaded or blurry." I am... And luckily, computers got fast enough that people could start playing at the frame rates that they wanted, even with the full textures, uh, running in it. I am... But the whole pace of doing the, um, kind of the esports and the competitive gaming was very interesting. We saw the dawn of that with Doom, but it's been pretty surprising. It's been amazing the state that it's taken, it's gotten to today, where I remember when we did the Quake Red Annihilation Tournament. I gave away my first Ferrari as grand prize, and I was thinking-

   9. JRJoe Rogan16:53

      Really?

   10. JCJohn Carmack16:53

       ... "This is just the most over-the-top thing. This is gonna be unmatched for years." And it was only a year later that there was some other tournament with $100,000 prize coming out. So, that went on, went much quicker than I expected. And then today, you have just the, the amazing celebrity of the top pro players. It's, uh, it's again, it's great to see the path that it's taken.

   11. JRJoe Rogan17:13

       It really is, and it's very interesting to see that they're now, like, legitimate sports stars, and they make a ton of money. Whereas if you were a kid 10, 15 years ago, your parents would tell you, "You're wasting your time. This is nonsense. Why are you playing these games?" But now, you have a legitimate opportunity to be a professional game player.

   12. JCJohn Carmack17:32

       Although, I do... I get... I- Hey, there's a hazard there, where-

   13. JRJoe Rogan17:36

       What's this, Jamie? What are you pulling up?

   14. JVJamie Vernon17:37

       Top earners this year-

   15. JCJohn Carmack17:39

       Yeah.

   16. JVJamie Vernon17:39

       ... uh, compared to the top, uh... Tiger Woods, how much he won in the Masters, and the top Indy 500 earners in winning.

   17. JRJoe Rogan17:44

       So, Tiger Woods won $2 million in the Masters. The top Quake players won $3,121,872. Oh, it's not Quake. What game is that?

   18. JVJamie Vernon17:52

       Dota.

   19. JRJoe Rogan17:53

       What does, what does that? What's Dota?

   20. JVJamie Vernon17:55

       Uh, it's a different-

   21. JRJoe Rogan17:56

       What's it stand for?

   22. JVJamie Vernon17:56

       ... game than Quake, but it's, uh, it's very popular right now.

   23. JRJoe Rogan17:59

       What's it stand for? I've never even heard of it.

   24. JCJohn Carmack18:01

       It's a Valve game.

   25. JRJoe Rogan18:02

       Oh.

   26. JCJohn Carmack18:03

       You know, it's a, it's a MOBA, massively online battle arena game-

   27. JRJoe Rogan18:06

       Oh.

   28. JCJohn Carmack18:06

       ... where you're directing all the different little characters down there. And that, along with, especially League of Legends, even beyond that, uh, not as, uh, popular in the United States, but is just amazing overseas, where, uh, they... Like, if you look at the numbers for things, people think, "Oh, the Super Bowl, like the height of all competitive sports, whatever." And, uh, a lot of these esports games, especially in South Korea, are they dwarf those numbers.

   29. JRJoe Rogan18:30

       Really?

   30. JCJohn Carmack18:31

       And some people-
3. 30:00 – 45:00

   Mm. …

   1. JCJohn Carmack30:00

      up and kind of coast around, down things.

   2. JRJoe Rogan30:01

      Mm.

   3. JCJohn Carmack30:02

      So, sometimes you're making these trade-offs with your body on the VR experiences. Uh, we try to push people towards the no-trade-off games. You know, "Okay, you can sit here. You can do this amazing thing, and there's no downside to it." But for the adventurous, there are these other things you can do which might be exactly what hits the right buttons for you, but, uh, you may suffer some for it. (laughs)

   4. JRJoe Rogan30:22

      Do you guys take into account the possibility of people getting sick and suing? 'Cause then, I think you kinda have to, right?

   5. JCJohn Carmack30:28

      So, there's a lot of, uh... What we were more concerned about, health-and-safety-wise, is, on the new system where you can walk around-

   6. JRJoe Rogan30:35

      Oh.

   7. JCJohn Carmack30:35

      ... is, you know, people banging into walls, falling down steps, and so on. So, we spent a lot of effort building this, we call it the Guardian System. When you start up, you basically trace out what you say the safe area is. And that's on you. If you trace out the safe area over your staircase, um, you know, bad things can happen. But the idea is that, um, it has a known good area. It knows where you are. So, it will tell you. Uh, it'll bring up this little overlay to show you that, "Hey, you're approaching a wall. Do not be charging headlong." Now, people should not be sprinting in VR under almost any circumstances. But you do wind up getting into the action on things, and something jumps, it's behind you, startles you, and you jump around. So, you need a reasonable amount of space to do these, we call them room-scale games, where you're actually walking around. There's a lot of them that are stationary, that you're expected to pretty much stand still. But there's some amazing experiences about being able to clear out a room, walk around, and just not be on task as far as hyper-aware, but just being able to soak in a space and kind of crouch down and look at things. And I always like to say that modern games have so much artistry put into them that is largely wasted or lost on the players, where you have the fire extinguisher sitting in the corner that some artist lovingly crafted for a week, and you just glance at it, see the pixels, and move on. But in VR, you wind up saying... It's like, "Oh, I'm in this room. I'm looking at this." You re- you know, lean down, or even pick something up and turn it around and look at it. And, you know, and that's really great. But the, th- the larger spaces have these hazards, and it turns out that...It's hard to protect people from themselves in some cases, because we started saying, "Well, what if somebody's swinging their hand really hard?" I am... You know, we don't... We want them to be able to stop their hand. And even if we show when the hand is really close to it, they're not gonna have time, so we put in this predictive filter to say, "Based on the velocity of how fast you're moving your hand, maybe we'll start showing up the boundary a second earlier. If we think you're gonna hit it in a second, we'll show the boundary now." But then everybody says, "Oh, I don't want that flashing up. It's distracting me," and we need to put in parameters to turn it down. And it's a real balancing act there, where a lot of developers just kind of go in and say, "Turn off Guardian. Uh, I'll accept the risks for this," but I-

   8. JRJoe Rogan32:47

      Oh, wow.

   9. JCJohn Carmack32:47

      Yes, of course, the lawyers are very concerned for this. We have, uh... And we have some duty of care to the people that are our customers. Uh, there are some, uh, interesting YouTube videos of people playing VR, not setting things up right, and likes running into walls and so on.

   10. JRJoe Rogan33:01

       (laughs)

   11. JCJohn Carmack33:01

       And even internally, it's, uh... If you don't set it up right, if you, uh... If you extend it further than you should, there are people that have smacked their knuckles into a table or something, and kind of post the bloody knuckles picture from that-

   12. JRJoe Rogan33:14

       (laughs)

   13. JCJohn Carmack33:14

       ... but it's on you on how conservative you want to be with it.

   14. JRJoe Rogan33:17

       It was interesting when I stepped outside of the... What is it called, the guardian...

   15. JCJohn Carmack33:20

       Yeah.

   16. JRJoe Rogan33:21

       ... uh, whatever it is, uh, the, the, the matrix that shows up, and you, you walk through, then you see the regular world. Like, you step out of the virtual world, and you're into the regular world, but you're seeing it through this weird lens.

   17. JCJohn Carmack33:35

       Yeah, so it's... I... It's black and white because it's using the same cameras. Like, on the Quest, there are four cameras that, um... that are pointed off in different directions, and they have two jobs to do. They, they track where the headset is by looking at the world around it. They find stable little pieces of the world, and based on where they are as you move around, it can figure out where the headset is. And then they also look at the controllers, and the controllers have little invisible IR LEDs on it. Uh, but it takes this black-and-white image of the world from the four different views, and it can pass through the real world. It's pretty low-res, uh, doesn't have color, but it's good enough for important things like finding where you put your controllers and letting you orient yourself. Because depending on what you do in VR, if you spend a couple hours in an en- uh, environment where you're turning around and moving, it can be very shocking to people when they, uh, sort of come out of it, and they... they're in the opposite corner of the room they thought they should be, uh, they're pointed in a different direction. Uh, having the ability to kind of bring the world into place, especially in those cases where you're approaching the boundaries is, uh, it's nice.

   18. JRJoe Rogan34:39

       Is there anything that you guys have that can, or maybe in the future can, map out uneven terrain? Like, say if you were at a park, and there was a hill or in your backyard or something like that where you have var- various s- sort of surfaces. You have a sidewalk and then grass and then a hill. Is there ever going to-

   19. JCJohn Carmack34:57

       So there's, there's a few aspects to that. Uh, you can set up really large, uh, guardian boundaries. We do have an upper limit on it, uh, but again, some developers have disabled that to do even larger areas. And one of the coolest experiences that I had was at a convention last year. An artist had made a... basically a VR sculpture that was really big. It was like 50 feet long, uh, and you were inside this kind of aquarium-looking thing in a workshop. And we sketched out this giant boundary, and I walked around. Like, I walked literally from room to room in virtual reality. Again, very few people have the space to set something like this up in their house, but we were at a convention center. And the ability to walk through a door in virtual reality, get down on my hands and knees and crawl through a crawl space into another magical little area, that was really something. I mean, it's not clear how we can carry that over to other people, but the idea of doing it outside, uh... There's a few technical issues with it where bright sunlight overpowers, uh, the little tiny LEDs on the controllers. So while it's possible to sort of stop down the exposure on the cameras for tracking the headset, if it's reasonably bright, you're not gonna be able to resolve the controllers, which breaks some things. But people have found that if you get the right overcast day, uh, and you've got the right environment, you can go out to, uh, like a tennis court or, uh, or a big field or something, stretch out a large boundary, and explore some fairly sizable things. Now, really accurate, uh, determination of the world, Facebook Reality Labs has done a lot of research for almost what is the absolute limit of what we could possibly do with the sensors for building the most accurate representation of the world? And it's pretty damn good when you get to the point where they could take even a really cluttered room like this, scan the world for all of this, and they could make, uh, real 3D representations so you could then be good enough that I can go ahead and pick up individual things here. But it's still... Uh, it's pretty expensive to do that both in terms of they need some slightly different sensors, they have to project more structured light out at the world, takes a lot more calculation to do it. It's too much for this generation of products here, but certainly that's something that we're looking at in the future, where eventually we want to not have that one step of drawing out the guardian. You just want it to be both, um, sensible enough that it can tell what's going on in the environment and sort of smart enough to tell what's a hazard, because you want that magic of you just put it on, it does everything, and it just works. We're not there yet, but that is sort of on the roadmap for where we wanna go.

   20. JRJoe Rogan37:27

       So you think there will be a time where the, the, the technology will allow you to maybe possibly have several filter layers? Like, you can see the whole world, and it, it would be more of an augmented reality toward of a s- sort of a situation, or the whole world disappears, and then it could be virtual?

   21. JCJohn Carmack37:45

       Yeah, so there's... We have a lot of debates about, um, both the useful things that you might do there and then some real technical aspects where, in terms of augmented reality, this idea that we all buy into this future vision of a world where you've got something that's this... you know, that feels like sunglasses that you put on, and you can pull up all your information, and maybe it clouds over into a virtual experience. There are still, uh...... fundamentally unsolved problems in display technology to do the magic thing that we would really like from all of that. So you step back to saying that, "Well, what you saw in Quest when you saw the world through that, obviously it's low res, it's low frame rate, it's not good." But we could fix all of that. You could say, "Let's go high res color, high refresh rate," and you could make a, we would call that a passthrough rather than a see-through augmented reality system. And we could absolutely build that technology and we could make that pretty good. But then it comes down to what do we expect sort of the user story to be? If you had something like that, uh, would you be wearing this boxy thing, uh, you know, out into the world, riding the bus with it, doing different things? We have a little bit of a hard time seeing the kind of socially acceptable way that you're running around going about your life with sort of this shoebox-sized thing on your head.

   22. JRJoe Rogan38:54

       Mm.

   23. JCJohn Carmack38:54

       Uh, it's an open question where everybody agrees if it came down to sunglasses, everybody in the world's gonna want something like that if it gives them these magical abilities and it can turn into virtual reality or augmented reality. It's an open question whether there's an in-between layer. If we get down to the point where it's something like swim goggles or very thin, um, sort of ski goggles, something that's half or a quarter of the volume of we- what we've got here, would that be something that people would wanna wear for long periods of the day? I lean towards no, but we haven't built it, so we don't know yet. Um, and then the question of what you wanna do with that in the augmented reality world, where people make these interesting little demos where, "All right, we've mapped the world in this incredible detail. Now we can flood it with water. We can do a simulation of all of this. Isn't this cool?" Or we can turn it into, uh, we can reskin your world as Bilbo Baggins' hobbit hole or something. And I am skeptical of the broad utility of a lot of these things where, like today, there's a ton of AR apps that you can get on your cellphone. You can hold your phone up and kind of look at things and interesting little things happen.

   24. JRJoe Rogan40:00

       Pokémon.

   25. JCJohn Carmack40:00

       Well, Pokémon Go is an interesting thing where that actually has more of a point for it, and the augmented reality side of it is very small. But the things that actually augment onto the world, I haven't seen anything that I, that I found really compelling there. Interesting technologies, but I think that the, um... I'm still betting more on the fully immersive experiences where you take over, and this idea of bringing part of it into the world, I'm a little more skeptical on, but I, I don't know how it's gonna play out.

   26. JRJoe Rogan40:29

       But that's the concept behind Magic Leap, right?

   27. JCJohn Carmack40:31

       Yeah.

   28. JRJoe Rogan40:32

       And Magic Leap was very, very hyped up a couple of years ago, but it hasn't really come to fruition yet in terms of like a consumer product that people could... I mean, remember they had the little girl and there was a ballerina that was dancing on her hand-

   29. JCJohn Carmack40:44

       Yeah.

   30. JRJoe Rogan40:44

       ... and it was like, "Is this gonna happen?"
4. 45:00 – 1:00:00

   Mm-hmm. …

   1. JCJohn Carmack45:00

      this for quite a while. It starts off with just one neuron. You stick in one sensor, uh, or one actuator, and in the Neuralink case now, they've got tens of thousands that they can put in, so much, much higher fidelity. Uh, even if it doesn't work forever, and they're still working out the different, uh, the codings, tactics, different installation procedures, uh, somebody's going to go in there and be, go from being profoundly disabled to probably being able to play a video game, you know, being able to sort of directly control things with your mind, where you start off being able to do maybe just very slowly driving a cursor. And people have done this, again, for years, where you can slowly move a mouse cursor or something and figure out a click thing. But here, when you have tens of thousands of neurons going in, you could go from this very slow moving something to doing this deep analog multidimensional, like playing a symphony with your brain output potentially.

   2. JRJoe Rogan45:54

      Mm-hmm.

   3. JCJohn Carmack45:54

      And then potentially feeding information in in a way that we can't right now, that you could have this sort of tactile feel to it.

   4. JRJoe Rogan46:01

      Do you remember Lawnmower Man?

   5. JCJohn Carmack46:03

      (laughs) Yeah, vaguely. I-

   6. JRJoe Rogan46:04

      That was the idea, was that the guy was-

   7. JCJohn Carmack46:06

      I've thought my- Yeah, I've thought to myself sometime I should go back and watch that, because I literally have not watched that since it came out when I was a teenager. And the VR is laughable at this point-

   8. JRJoe Rogan46:14

      Yeah.

   9. JCJohn Carmack46:14

      ... and everything, but it should be good for a chuckle.

   10. JRJoe Rogan46:16

       Well, um, it was a clun- I think it was a clunky movie. But the idea was based on a Stephen King book, and the Stephen King book where there was a guy who was sort of slow, and they did something to him, and all of a sudden he became some super genius, almost godlike character. Like, if this Neuralink stuff does work and you can take a person with profound disabilities-

   11. JCJohn Carmack46:39

       Mm-hmm.

   12. JRJoe Rogan46:39

       ... and all of a sudden, they become the smartest human beings on the planet, that would be really weird.

   13. JCJohn Carmack46:44

       So, I know that was one of the pitches that Elon was making early on, that, uh, Elon is very concerned about artificial intelligence, and-

   14. JRJoe Rogan46:51

       Very.

   15. JCJohn Carmack46:52

       ... part of Neuralink was this pitch that, "Well, maybe we can supercharge humans in a way that the AI won't run away so far, or we can at least interact with them on a more level playing field." I am less, uh, sanguine about that. I am, uh, that seems a little bit more of a reach to me, because I suspect that, okay, even if you do put a million neurons in, uh, when we're making artificial general intelligences, they're gonna have billions and billions of these different connections. And I think that it might be many steps above what a human could be, but if AI becomes possible and takes off, and I, I am a believer in artificial general intelligence. I think it's probably not as far off as many, uh, many people believe, that it is likely to be able to accelerate and advance faster than even a neural-linked human would be able to.

   16. JRJoe Rogan47:44

       How far away do you think we are from artificial general intelligence?

   17. JCJohn Carmack47:48

       So, by nature, I'm an optimist. I'm, you know, I tend to, uh, to underestimate how long things take. But on the other hand, as a programmer, I've usually been able to say, "Well, maybe I missed my estimate by 50% while everybody else blew it by 100%" or something. I think that we will have, we will potentially have, um, clear signs of AGI maybe as soon as a decade from now.

   18. JRJoe Rogan48:12

       Wow.

   19. JCJohn Carmack48:13

       Now, lots of people disagree. The majority of scientists, uh, working on it think it's like, "Oh, it's gonna be at least a few decades," and you still have a few holdouts that say, "Oh, it can't happen at all." But I'm a strict materialist. I think that, you know, our minds are just our body in action, and there's no reason why we can't wind up simulating that in some way. Um, so I don't think ... The question of how far off, there's a lot of numbers that you can play. Like, the, the brain has something like 85 billion neurons in it, and th- they have something like 10,000 connections between it. And you can multiply those out and compare them to what we have in computer memory and processing time, and you can say that, yeah, within 10 years, those curves should have crossed. But I would even go so far as to say most of those are probably not, uh, completely necessary. We know lots of biological systems. Like, we understand the processing that goes on a lot in the visual side, and we don't need nearly as many computer transistors as, uh, neurons that are used for processing some of those early layers. So, I suspect that even today, some of the government supercomputers, the, the biggest, the top 500 list that they have, those are remarkably, uh, probably useful for doing intel- artificial intelligence work. Where for a long time, for decades, I thought that was sort of just, uh, national chest thumping, the top 500 computers, because so many of them, they relied on replacing what used to be the old big iron Cray vector supercomputers, and they really weren't very easy to program. Most programs people wanna use, you can't run it on a supercomputer and just be a lot faster. One of the shocking things most people don't really appreciate is the fastest way to do most single threaded applications is an overclocked gaming computer today. You can't go spend a million dollars and buy a computer that will do many tasks faster than what you can just run on a gaming computer. And this is not at all the way things were for decades, where for a long time, you would go spend your millions of dollars on a Cray supercomputer and all of your code would run faster than anything you can get. But it turned out that the, uh, the processors that you wind up using in high-end gaming systems are, in many cases, the fastest, or in, in all cases, at least close to as fast for certain serial applications. So, the only thing else you can do is pile lots more of them together. And these big computers are football field sized, uh, systems that are just racks and racks of GPUs and CPUs and nowadays, for a long time I'd be like, "Well, what would you..." I would think, "How could I make a faster Quake map build or something on one of those?" Because we would sometimes have hours and hours spent processing this, and, um, at one point we had a computer that was almost in the top 500, um, at id Software, just for making our maps. Uh, but I looked at a lot of these supercomputers, I'm like, "Ah, these are terrible. Not very useful for what we want." Um, but now as I look at AI work and I think that, "Well, if you're just doing a whole bunch of these, uh, kind of general matrix multiplies, that computer right there is probably pretty good." So-I would suspect that you could do something, if we had the right algorithms, the right training schedule, and the right time to run through it. That it's probably possible on some systems today, and it'll just still take many years for the right algorithms to wind up being developed. Uh, the right training regimens to be run, and faster, cheaper hardware to wind up making it more economical to run all the experiments. Because in so many cases, the trick is not that the minimum requirements exist, but that 1,000 people have thrown themself at the wall of a problem, most of them have bounced off and failed, but eventually somebody gets through.

   20. JRJoe Rogan51:44

       Now, what about quantum computing? Is that something that could potentially break the bottleneck that we have with Moore's law?

   21. JCJohn Carmack51:51

       So, I don't... I am not an expert on quantum computing, and I, I think that I... Many times I beat myself up about it, where there are some simulators online where you can go and work on it, and I should work through the exercises of doing the basic factoring algorithms on quantum computing. But I... You know, my read on it right now is that it's probably not directly useful for most of the artificial intelligence tasks. The big things that people worry about that are, um, for things like breaking cryptography, uh, breaking the different hashes and encryption methods. That it's possible that... In many ways, that's almost a terrible technology, because it's a technology that doesn't solve so many of the problems that you'd like it to solve. And it does solve one of the problems you kind of wish nobody was able to solve (laughs) . I... So, I-

   22. JRJoe Rogan52:34

       What's that?

   23. JCJohn Carmack52:34

       By breaking all the encryption.

   24. JRJoe Rogan52:36

       Oh. Oh, okay.

   25. JCJohn Carmack52:36

       You know, like if somebody winds up, uh, with a quantum computer that they achieve quantum supremacy and it runs past all of our traditional computers, and all of a sudden they can break all of the, all of the secure socket layer stuff, break everybody's signatures, impersonate any public key sign stuff, there's no upside to this. That's all downside-

   26. JRJoe Rogan52:55

       Right.

   27. JCJohn Carmack52:55

       ... and all bad things that are gonna come from that. While it's not gonna make your video encoding go any faster, and it's probably not going to help artificial intelligence in many ways. So, I haven't found a whole lot to get me really excited about quantum, uh, computing. It may just be that... Uh, and with all these cases, why I beat myself up about not learning more about it, because in most cases when presented with some capability, there's some way to figure out how to apply it usefully to the things that you really want. In fact, I consider that almost the essence of engineering. Engineering is figuring out how to do what you want with what you've actually got (laughs) . And if somebody gives... Anytime somebody gives me new hardware, usually I can figure out some useful way to do things that I want with it, even if it's not immediately obvious. And maybe quantum computing plays out that way, but, um, it is still definitely the domain of big labs with cryogenic cooling and all that stuff. So...

   28. JRJoe Rogan53:46

       Mm-hmm.

   29. JCJohn Carmack53:46

       It just hasn't been at the top of my radar.

   30. JRJoe Rogan53:48

       Now, when we talk about technology and you talk about the exponential increase in the powers of technology, is it possible that we could come to a point in time, somewhere in the future, where there's no way to encrypt anything? Where it's not possible to hide things? Where we won't be able to do banking online, we won't be able to have digital currency, because virtually everyone will have access to all the information? Because essentially, digital currency, or anything that's encrypted, it's just information. Right? It's just ones and zeros. Is it possible that technology will reach a point in time where borders and boundaries are impossible?
5. 1:00:00 – 1:09:34

   Mm-hmm. …

   1. JCJohn Carmack1:00:00

      DOOM, and Quake, we had at least one VR entrepreneur that wanted to work with us, uh, because finally here was content. Because they were like, everybody had this vision of VR, like it's this hazy vision of the future where cool stuff happens when you put the helmet on, but they didn't have this concrete, uh, instantiation of like, well, what do you actually do? And then people saw the 3D games, like, "Oh, that's what you wanna do inside the virtual reality helmet." So, they would come up, and they'd wanna basically work with us, license the technology, and every time I looked at these, I'm like, "Oh, this is just, isn't gonna work out." I, in many cases, they were people that were very, they were, you know, high on enthusiasm but a little low on the raw technical necessities to make something like this happen, because really it was too early. I mean, there were systems that if you had unlimited money you could get a big SGI refrigerator-sized infinite reality system, uh, that could render images that could be interesting. But people talk about the current resolution limits here or, like, the original, uh, the original DK1 kind of the views for there. Those headsets back then, they were the resolution of, remember the original pixel DOOM? Uh, like 320 by 240-

   2. JRJoe Rogan1:01:07

      Mm-hmm.

   3. JCJohn Carmack1:01:07

      ... sort of resolutions there, but wrapped around your head. Somebody called them football-sized pixels. You know, like you've-

   4. JRJoe Rogan1:01:13

      (laughs)

   5. JCJohn Carmack1:01:13

      ... got this big blurry blob off to the side. So, it was never gonna happen in those early days. So, uh, we always made the point of saying, "Well, okay, we'll license the stuff to you, but, uh, (laughs) you know, I wouldn't put a dime of my investment money into something like that," because I just thought it was too early. And then here we are a couple decades later, and the future has arrived.

   6. JRJoe Rogan1:01:32

      Yeah, it seems like it was about a decade ago that people started really taking it seriously again, b- where the technology had caught up to the vision. Is that-

   7. JCJohn Carmack1:01:40

      So, it's, uh, I think that, I think we can date all of it to the demo that I gave at E3, uh, which had the DOOM 3 as the kind of running there, because it was amazing talking to people. I got into it, uh, several months before that, and I had this thought. Okay, I had just finished, finished Rage, uh, the last game that I had worked on there, and, and I was like, all right, between each project I would do research. I would take a time to go ahead and explore some new technology, whether it was in the game rendering or something related to it. And I thought, well, virtual reality. Back in, I remember dealing with that in the '90s. It's been 20 years. Surely they've sorted this out by now. You know, it was a matter of the technology was terrible back then, but here we are closing in on a million times faster processors. Surely somebody's just sorted this out. And I was shocked when I went and I looked through, I surveyed everything, and it really wasn't. There was a cottage industry of people that would serve basically Department of Defense, uh, contracts that would make these very expensive systems that were, you know, 10s to, in some cases like $150,000 for some of these big things, and they weren't even very good. It was, it was offensive to me as an engineer, where I look at the capabilities of what's possible. You know, I, I say, "This is our display, this is our processing, these are our sensors. What is the limit?" I always talk about the speed of light calculations, like if everything was perfect, how good could it be? And I could look at what we had and say, "It could be a whole hell of a lot better than what we have here, what people are shipping and that are charging these very high prices for it." And so that was when I started cobbling things together myself. Uh, you know, and that's what led to working with, uh, you know, Palmer at the start of Oculus and led to where we are today.

   8. JRJoe Rogan1:03:22

      What year was that?

   9. JCJohn Carmack1:03:23

      So, 2011 in the first work, I think.

   10. JRJoe Rogan1:03:25

       Mm, that's so recent.

   11. JCJohn Carmack1:03:27

       Yeah. Mm-hmm.

   12. JRJoe Rogan1:03:27

       That's crazy. So yeah, less than a decade, and, um, now we have it down to a one wearable headset that sits on your head. My, my friend Duncan's first unit was connected to a computer-

   13. JCJohn Carmack1:03:41

       Mm-hmm.

   14. JRJoe Rogan1:03:42

       ... and you had cables that you would trip over, and I think there was a backpack involved as well?

   15. JCJohn Carmack1:03:46

       Yeah (laughs) . So there's still some reason to want to use the computer, where one of the points that I like to make is that, uh, while cellphone technology, which is definitely what these standalones run on, is astounding in how much power we've gotten out of these. But your high-end gaming PC rig, I...It's a difference of, it's 50 times more powerful.

   16. JRJoe Rogan1:04:05

       Hmm.

   17. JCJohn Carmack1:04:05

       It is just way, way, way more powerful. And if you just wanna make something happen quickly and easily, it's easy on the PC, where you have to sweat pretty hard to make some intense things happen on the cellular systems. I, and one of the interesting things is, I make the point to people that we are so used to computers just getting faster and faster, and they have for, you know, for decades and decades, our entire life basically, but we are approaching the end of that. People talk about the end of Moore's law, like you mentioned, and I've had to tell people that while we've still got a lot more power to come, uh, the next decade's still gonna be good, it is very likely, barring some magical new technology which, fingers crossed, maybe we get, but you will probably never get the cell phone technology, uh, to the point that a, a modern gaming PC is. Which, on the one hand, sounds like obviously that's a different thing, but on the other hand, if you've been here through this million X increase in performance, it's a little bit of a downer to think that no, we just can't wait another 10 years and we'll have everything we could do there available in the systems. So it does mean a little bit of a cultural change to, to start thinking more about performance rather than just say throw everything at it and then... And that used to be what I did in the early days at id. We would make these bleeding edge things where only a few people at the start had good enough computers, or people would get it and they were running at a low frame rate and it made them wanna go upgrade their computer and wanted to go get the latest thing or buy a GPU and, you know, do the thing to make the game better. But we are, we are approaching the limits of what's going to be happening with that. I am... So you have to be better, you know. You have to be a, a more conscientious developer, you have to, uh, start paying attention to all of the different aspects of performance that on the PC you can still largely gloss over. Uh, and I like that. I mean, as an old school optimizer where I always appreciated the challenge, that's why my entire time at Oculus I've been focused on these mobile systems where in many ways it's easier to do spectacular things on the PC, but mobile is super important and it's more of a challenge. You know, it winds up fitting a little bit better for me, but there are generations of game developers especially now that have been, uh, have grown up making PC titles where it's easy and they have to educate themselves quite a bit now to go ahead and make the step down to something with less than a 10th or, you know, less than a 50th in some cases of the raw power.

   18. JRJoe Rogan1:06:23

       When you're talking about the end of Moore's law, what- what is the limitation that we're facing technologically? Like why is it going to, why are there- there- why is there going to be a point where they can't get any more powerful?

   19. JCJohn Carmack1:06:35

       So the way, uh, the chips work is you have these, uh... You know, they wind up sketching out, uh, basically wires onto the- the silicon chips and they have gotten so small that the wire that the current's flowing through is a handful of atoms wide, which is just astounding if you think about it. You know, these are these fundamental, uh, elements of matter and the wire is this small integer number of atoms wide. Now in theory, you can keep going down and say, "Well, maybe we can make a one atom wide, uh, electrical path," but you wind up running into eventually all these quantum effects where if you make a very narrow wire and you pack them very close together, you have two wires there, an electron won't necessarily stay on that one wire of conductor that you want it to be on. Because of the way, uh, quantum mechanics work, it is going to wind up jumping, they call it quantum tunneling. There is a percentage chance, and quantum is all about randomness like that, but an electron flowing here, there's going to be this chance that it just teleports essentially to a nearby wire. It takes this discrete quantum jump to another wire. And this is reality. It's shocking, it's not intuitive, uh, people have a hard time kind of grasping a lot of this, but quantum tunneling's a real thing and we are bumping into quantum limits. They can still shrink more than we are right now. We're- we're down at seven nanometers in the latest, uh, stuff, although there's all sorts of issues with marketing speak about exactly how they measure it, but they're still getting smaller and there's still room to get smaller still, but the end is in sight. It can't go too much. And one of the things that becomes, uh, an issue is just the economics of it. Each generation has gotten more and more expensive. If you went back, um, 30 years, there were a whole bunch of semiconductor places that could fab, um, you know, different chips. You could go ahead and have a design and you could shop it out to a whole bunch of different places, find the one that worked best for you, but it's come down to the point now where it costs billions of dollars to make a new fab and at the high-end processes you're left with just TSMC, Samsung, and Intel, um, very few companies. You know, AMD held on for a while until they spun theirs out, uh, and it's ex- it's so expensive and it, that's one of the challenges where they will... I have full confidence that we'll see a couple more node shrinks, um, so it'll still make chips cheaper, uh, somewhat faster, more cores on them, but it is going to get an end of the line. But I hold out hope for potential other things. You know, there are directions that, you know, maybe you have your carbon nanotube wires or you're starting able, to be able to do some things with photonic processing in different ways. There are possible outs for it, but I don't know that any of them are a sure enough thing to really be counting on at this point.

   20. JRJoe Rogan1:09:21

       It's so hard for a dummy like me to wrap my head around that, but when you're talking about these wires, so if these wires, it's size dependent when they get too small then this quantum tunneling becomes completely unpredictable? Is that what it is?

Episode duration: 2:36:46