My concern isn't the long run. It's the<br>next 3 to seven years. How do we head<br>towards Star Trek and not Terminator?<br>>> I call AI and robotics the supersonic<br>tsunami. We're in the singularity.<br>>> When is all white by color work gone?<br>>> Anything short of shaping atoms. AI can<br>do half or more of those jobs right now.<br>There's no onoff switch. It is coming<br>and accelerating. The transition will be<br>bumpy. You have a solution to this.<br>>> I don't make a bet here. Um,<br>>> China's done an incredible job,<br>>> right? I mean, it's running circles<br>around us. Do you imagine that the US<br>could make that level of investment and<br>commitment<br>>> based on current trends? Uh, China will<br>far exceed the rest of the world in uh<br>AI compute.<br>>> Every major CEO and economist and<br>government leader should be like, what<br>do we do?<br>>> We don't have any system right now to<br>make this go well. But AI is a critical<br>part of making it go well. There are<br>three things that I think are important.<br>Truth will prevent AI from going insane.<br>Curiosity, I think, will foster any form<br>of sentience. And if it has a sense of<br>beauty, it will be a great future. It's<br>going to be an awesome future.<br>>> Now, that's a moonshot, ladies and<br>gentlemen.<br>>> Welcome to Moonshots. Following is a<br>wide-ranging conversation with Elon Musk<br>focused on optimism and the coming age<br>of abundance. My moonshot mate Dave<br>Blondon and I flew into Austin, Texas to<br>meet up with Elon at his 11.5 million<br>square foot Gigafactory, home of the<br>Cybertruck and Model Y production and<br>the future home for 8 million square ft<br>of Optimus production. Elon has agreed<br>to do this kind of a deep dive catchup<br>once per year. This is hopefully the<br>first of many. And after having this<br>conversation with Elon, it's crystal<br>clear to me that we are living through<br>the singularity. All right, enjoy.<br>>> Yeah. Um, your relentless optimism is<br>always a breath of fresh air.<br>>> Thank you, buddy. Thank you. Well, I<br>want to share that tonight with a lot of<br>people.<br>>> Yeah,<br>>> I think they need it.<br>>> I hope you're right. And you might be<br>right. Actually, I'm increasingly<br>thinking that you are right.<br>>> Thank you.<br>>> Abundance for all.<br>>> Yeah,<br>>> that's the goal. Shall we?<br>>> Yeah.<br>>> All right.<br>>> Right now, putting a lot of time into<br>chips.<br>>> You are. You are personally.<br>>> Yeah.<br>>> Yeah.<br>>> It's always AI assistance, I assume.<br>>> What's that? with some AI assistance. I<br>assume that design<br>>> uh not enough.<br>>> Yeah.<br>>> It' be nice if we could just hand it off<br>to the AI.<br>>> Yeah. Yeah.<br>>> Soon enough.<br>>> Yeah. I tried to do some circuit design<br>actually with uh AI recently. Just this<br>a couple weeks ago. Not not happening<br>yet.<br>>> Um<br>very soon though.<br>>> Yeah. Um I I think probably at this<br>point Grock if you if you took a photo<br>and submitted to Grock, it could<br>probably tell you if if the circuit is<br>is if there's something wrong with it.<br>>> Yeah.<br>>> Yeah.<br>>> All right. I'm going to give it a shot.<br>You're using the same Grock that I'm<br>using. Are you or you are<br>>> Grock keeps updating. So<br>>> yeah, 4.2, but five is soon, right?<br>>> Uh five is Q1.<br>>> Yeah.<br>>> Um 4.2 has not been released yet.<br>>> Okay. uh externally. Um but yeah, I mean<br>if you just if you just upload an image<br>into Gro um<br>>> it's it's does quite a good job.<br>>> Yeah.<br>>> Um<br>>> yeah,<br>>> of of analyzing any any given image.<br>>> Absolutely. Let's uh let's start. We're<br>going to talk about this.<br>>> All right. We'll come back.<br>>> I mean, let's see if I if I take an if I<br>take a picture of you, what is it? Let's<br>see what it<br>>> Yeah. What's it going to say about me?<br>>> Yeah, it's going to say you're a flawed<br>circuit. I also have to remember to<br>update it because like we update the<br>Grock app so frequently.<br>>> You know, I asked I asked Grock to roast<br>me.<br>>> Oh, it's does a good job.<br>>> It did an amazing job. Then I asked<br>Grock to roast you. Yes.<br>>> And I spit out my coffee. It was it was<br>hilarious. And then I asked it, you<br>know,<br>>> say be more. It just keeps telling it to<br>be more and more.<br>>> I asked I asked until until it's like<br>mother of God.<br>>> Wait, is Bad Rudy still out or did that<br>get repealed? Bad Rudy still there?<br>>> And I asked, you know, does Elon know<br>what you say about him? and and and she<br>goes, "It's a she for me." She goes,<br>"What is he going to do about it?"<br>>> What is he going to do about it?<br>>> Yeah, let's see. Okay.<br>>> Um, so I just literally took a photo of<br>you and see what it is.<br>>> Did you ask a question?<br>>> No, nothing. I didn't say anything.<br>>> This man is is hugely<br>>> This This is Peter Diamandis.<br>>> Yes.<br>>> So,<br>>> okay.<br>>> That's pretty good.<br>>> Yeah.<br>>> There's no context whatsoever.<br>>> The host of the podcast Moonshots. Yeah.<br>>> Uh, sometimes that's your first<br>credential now. That's amazing. Forget<br>about everything else I've done in life.<br>Comes back to your podcast. That was a<br>no no context image.<br>>> Yeah. By the way, Graedia is awesome.<br>>> Okay, great.<br>>> I mean, just phenomenal.<br>>> I mean, just it's like I tried to like<br>update my Wikipedia page for like years<br>impossibly<br>>> and um Yeah, it it it knows me.<br>>> Amazing.<br>>> Yeah. Um, he's wearing a black quilted<br>jacket featuring a Sundance logo.<br>>> Not quite true. It's my abundance logo,<br>but I guess a little wrinkled. See the<br>>> Can it see it?<br>>> I I I think so.<br>>> Okay. Okay.<br>>> Anyway,<br>>> um Yeah, but it basically<br>uh it's pretty damn good.<br>>> Yeah.<br>>> Um he's smiling and relaxed with a<br>laptop in front of him.<br>>> That's true.<br>>> Yeah, that's true.<br>Um,<br>>> yeah.<br>>> Well, I should say quite a circuit<br>though.<br>>> You got to test it on the<br>>> roast him.<br>>> Only It has to be read by you, though.<br>>> I mean, I won't read the whole thing,<br>but<br>>> All right. Give me Give me a taste. I<br>can take it.<br>>> Okay. Check out that grin. Dude smiling<br>like he just discovered a new way to<br>monetize hope.<br>>> Monetizing hope. Oh, that's<br>>> I want to try and answer the question,<br>can AI and tech help save America and<br>the world? Right. Um, I want to give<br>people listening a dose of optimism.<br>There's a survey done in mid December by<br>Pew that said 45% of Americans would<br>rather live in the past and only 14%<br>said they'd rather live in the future,<br>which is insane to me, right? Um,<br>obviously they never read history. The<br>challenge is most Americans all they<br>have of the future. It's like Hollywood<br>has shown us killer AIs and rogue<br>robots, right? And people are worried<br>about their jobs. They're worried about<br>healthcare. They're worried about, you<br>know, the cost of living. The challenge<br>is how do we how do we help people? I<br>mean, you posted, you pinned on X, the<br>future is going to be amazing with AI<br>and robots enabling sustainable<br>abundance.<br>>> I think of you when I did that.<br>>> Thank you. I appreciate that. and and uh<br>>> well I mean<br>>> it's like what would Peter do you want<br>to say?<br>>> Yeah was channeling you.<br>>> Thank you. Thank I couldn't agree more.<br>I didn't agree more either.<br>>> That's great.<br>>> So so my question is from a you know<br>from a first principle standpoint<br>>> right<br>>> uh the rationale for optimism you know<br>how do we how do we head towards Star<br>Trek and not Terminator right? How do we<br>how do we head towards<br>>> Ronberry not Cameron. Yeah,<br>Jim. Jim, I will I will<br>>> the diverging path meme.<br>>> Yes, it is. It is. Uh, Avatar has some<br>hopeful parts, but anyway,<br>>> I how do we go towards universal high<br>income instead of social unrest? So, my<br>>> both<br>want socialrest.<br>>> So,<br>have universal high income and social<br>unrest. M<br>>> that's my prediction.<br>>> Oh, that will make for a lot of<br>problems.<br>>> Is that your actual prediction?<br>>> Yeah.<br>>> Yeah, it seems likely.<br>>> Like tell me to push back on it.<br>>> Yeah, exactly. But it seems like that's<br>the trend.<br>>> Yeah. Yeah, totally. No, we have<br>>> Well, because there's going to be so<br>much change.<br>>> Yeah, there's people are going to be<br>like scared shitless.<br>>> Yeah, it's it's sort of the um<br>you know um it's like be careful what<br>you wish for because you might get it.<br>>> Yeah. Yeah.<br>>> Now, if if you if you actually get all<br>the stuff you want, is that actually the<br>future you want?<br>>> Yeah.<br>>> Um because it means that your job won't<br>be what matter<br>>> if you're living an unchallenged life.<br>>> Yes.<br>>> Right. With no challenges.<br>>> Yeah.<br>>> No. You know, you know, if you become a<br>couch potato, if it's the Wall-E future,<br>that does not go well for humans.<br>>> Well, and we're used to being told,<br>here's your challenge. Yeah.<br>>> So people haven't historically been very<br>good at creating their own challenge in<br>the absence of<br>>> I think Elon does a damn good job. Every<br>time you every time one company takes<br>off, you start your next.<br>>> Oh, that's that's rare for punishment.<br>>> I think you are. I think you overthank<br>God for that.<br>>> So So what so<br>>> why do I do this to myself?<br>>> Actually, after AI and robots, is there<br>another thing after that? I guess<br>there's<br>>> Well, there's there's conquering, you<br>know, the universe.<br>>> Yeah, that there is that<br>>> rocks really.<br>>> Well,<br>and energy<br>>> rocks are your friends.<br>>> Conquering<br>>> We didn't even get there.<br>>> Why, Elon? Why are you so optimistic?<br>>> Are you Are you optimistic? Let's start<br>there.<br>>> I'm not as optimistic as you are.<br>>> Okay.<br>>> Um but why are you optimist?<br>>> I'm more optimistic than most people.<br>>> Okay.<br>>> Um<br>>> and is the trend upward<br>compared to a year ago, two years ago?<br>Well, I I think if you reframe things in<br>terms of<br>um progress bar, like speaking of<br>challenges,<br>>> yeah,<br>>> uh progress towards a cartev 2 scale<br>civilization.<br>>> Sure.<br>>> Um well, let's say let's say the<br>aspiration<br>>> capturing all the energy from the sun's<br>output.<br>>> Well, let's even have a a humbler<br>humbler aspiration than that. If we say<br>that our goal is to even get a millionth<br>of the sun's energy,<br>>> that would be more than a thousand times<br>as much energy as could possibly be<br>produced on Earth.<br>>> So about a half a billionth of the sun's<br>energy reaches Earth. Um so you'd have<br>to go up three orders of magnitude from<br>that uh just to get to a millionth.<br>>> Yeah.<br>Um,<br>so<br>we're very very very far from even h<br>having a billionth of the sun's energy<br>uh harnessed in any way. So a reasonable<br>goal would be try to get to a millionth.<br>And if you try to get to a millionth or<br>or a thousandth um you know 0.1%.<br>Uh<br>that's that's such an enormous<br>uh there's not sure what metaphor we'd<br>use here because a hill to climb is is<br>not a<br>>> inapprop like not a big enough metaphor<br>but<br>>> gravity well to escape<br>>> engineer hell of a gravity well.<br>Exactly. Um so if if you try to get to a<br>millionth of the sun's energy or a<br>thousandth the sun's energy like now the<br>these are very very difficult tasks<br>>> and energy is the inner loop for<br>everything right now.<br>>> Yeah. I I think like I I think uh the<br>future currency will essentially just be<br>wattage.<br>>> Yeah. I was thinking is it is it d is<br>the ability of a person to control<br>energy and compute<br>>> or just energy? I mean the two translate<br>obviously<br>>> just like harnessed energy.<br>>> Yeah.<br>>> Like so or like basically how much power<br>is being turned into work of some kind,<br>>> right?<br>>> Um<br>intelligence or<br>um matter manipulation. Um,<br>>> so that's your next big project is going<br>to be energy.<br>>> It's it's going to be you're going to go<br>back to your solar your solar system.<br>>> You can expand from there and say, okay,<br>>> what about even getting somewhere on a<br>on a cottage of three scale, meaning<br>galaxy level.<br>>> Now you're talking now. Now we're back<br>to Star Trek.<br>>> Yeah. Expand horizons here.<br>>> Yes.<br>>> Where there isn't even a horizon because<br>you're not on a planet.<br>>> So we we talk about<br>>> So So think galaxy mind.<br>>> Yeah.<br>Well, listen, we're in 11 11.5 million<br>square foot, three pentagons right here<br>in this building. I mean, you think in a<br>reasonably large scale,<br>>> what is the magnitude?<br>>> Yeah.<br>>> Um, so I mean, so from a challenge<br>standpoint, I guess<br>the civil the civilizational challenge<br>will be how do you climb the orders of<br>magnitude?<br>>> Yeah.<br>>> And energy harnessed.<br>>> But we're going back to why are you<br>optimistic right now? I mean, when<br>people think about uh the challenges<br>ahead, I think we're going to end up<br>with abundance in the long run, it's for<br>me<br>>> beyond abundance in any beyond what<br>people possibly could think of as<br>abundance. Um like the AI actually<br>AI and robots the limit um will will<br>saturate all human desire.<br>>> And then we get to nanotechnology which<br>takes it even a step further.<br>Um the thing about the well I'm not sure<br>what you mean by nano you mean like<br>little nanobots<br>>> atomic reassembly.<br>>> Yeah. For health.<br>>> Oh yeah. Yeah. Sure. Sure. Um I mean<br>we're already doing atomic level<br>assembly on the for circuits you know.<br>>> Amazing.<br>>> Um<br>>> two three nanometers.<br>>> Yeah. It's it's only um depending on how<br>they're arrayed four or five silicon<br>atoms per nanometer.<br>>> Yeah.<br>>> So<br>>> those are big atoms though.<br>>> They're not bigish. They're not your<br>little I mean but but I'm just saying<br>you could they should actually describe<br>the circuits in terms of an integer<br>number of atoms in a specific place.<br>>> They should it's all angstroms now but<br>>> you could you can just it's just inte<br>it's it's like we'll call this the the<br>seven atom you know whatever like you<br>say two two nanometers it's like it's<br>like<br>>> no one knows<br>>> nine silicon atoms something like that.<br>Um they've got silicon and copper and um<br>you know so but a bunch of these things<br>are just marketing numbers like the two<br>nanometer is just a marketing number.<br>>> Oh yeah.<br>>> Um but but it's you still need<br>essentially close to atomic level<br>precision. Like the atoms really need to<br>be in the right spot.<br>>> Um so um<br>I think they're getting clean rooms<br>wrong by the way in these modern fabs.<br>Um<br>I'm going to I'm going to make a bet<br>here. Okay.<br>>> Okay.<br>>> Um that Tesla will have a 2nmter fab and<br>I can I can eat a cheeseburger and smoke<br>a cigar in the fab.<br>>> Oh, come on.<br>>> Yes.<br>>> The air handling will be that good.<br>>> Do you have this sketched out in your<br>mind? Like how is it how are the atoms<br>being placed that they're immune to uh<br>cheeseburger grease? They just maintain<br>wafer isolation the entire time. um<br>which is actually the default for for<br>fabs. The the wafers are transported um<br>in boxes of pure nitrogen gas under a<br>slight positive.<br>>> So are the bananas at Walmart. I<br>>> just so you know.<br>>> Yeah. Well, that's that's it's inite<br>essentially like it's pretty hard for<br>anything that's combusting<br>>> uh to live without oxygen.<br>>> Yep.<br>>> So um<br>>> let's talk about<br>>> So like like you can kill the bugs just<br>by putting a nitrogen blanket on plants.<br>>> Yeah. Interesting.<br>>> I want to talk about uh energy, health,<br>education because those are people's,<br>you know, concerns. So, on the energy<br>front,<br>>> um the innermost loop of everything that<br>you're building and doing right now,<br>>> energy is the foundation.<br>>> What's your vision for energy abundance?<br>Uh<br>>> the sun<br>>> in in in the next, you know, this this<br>this decade. The sun. Yeah. I mean, so<br>>> the sun is everything.<br>>> It's everything. So, you're all in on<br>solar.<br>>> I mean,<br>>> uh Yeah. I mean your natural gas natural<br>gas and solar you're at Colossus 2,<br>right?<br>>> Yeah.<br>>> People just don't understand how<br>>> that that solar is everything. So um<br>everything compared to the sun, all<br>other energy sources are like uh cavemen<br>throwing some twigs into a fire.<br>>> Yeah.<br>>> Um so the the sun is over 99.8%<br>of all mass in the solar system. Uh<br>Jupiter is around uh.1% of the mass. Uh<br>so even if you burnt Jupiter, the energy<br>produced by the sun would still round up<br>to 100%.<br>>> Yeah.<br>>> Mhm.<br>>> And then if you teleported three more<br>Jupiters into our solar system and burnt<br>them too,<br>>> it would still round up.<br>>> It still the sun still rounds up to 100%<br>of energy.<br>>> Any interest in fusion?<br>>> I mean like fusion on a planet fusion.<br>You know what? You know coming a mile<br>away.<br>>> You're not never going to guess how the<br>sun works.<br>>> Giant coal plants.<br>>> I mean, we have a giant fus free fusion<br>reactor that shows up every day<br>>> 93 million miles away.<br>>> It's farical for us to create little<br>fusion reactors. Um<br>I mean that would be like, you know,<br>having a tiny ice cube maker in the<br>Antarctic.<br>and say, "Hey, look, we made ice." I'm<br>like, "Congratulations.<br>You're in the [ __ ] Antarctic."<br>>> So, totally totally with you on this.<br>>> It's like<br>3 kilometer high glaciers right next to<br>you.<br>>> Okay.<br>>> Yeah. If you just narrow the question to<br>the Memphis timeline. So, Memphis data<br>center timeline between a gigawatt and<br>10 gig. You're not going to you're not<br>going to pull 10 gigawatts out of<br>Memphis. Um maybe you are<br>>> two or three.<br>>> Two or three. Okay. So So there's still<br>a gap between there and the next<br>whatever you just just draw. So and<br>they're not in space yet at that point.<br>>> So we're still in toy land here. Uh for<br>on toy land you<br>>> toy land. Toyland<br>>> 10 gigawatt.<br>>> You know what's amazing is there's 100<br>megawatts right outside the door here<br>>> and it's massive. Yeah.<br>>> It's it's enormous. And it uses more<br>energy<br>>> than everything. All these manufacturing<br>lines combined use less energy than<br>that.<br>>> I think but we're talking about a<br>longgo. Cortex one was<br>>> the the third largest training cluster<br>in the in the world.<br>>> Yeah.<br>>> For for doing coherent training.<br>>> You're falling behind.<br>>> Uh well, we have Cortex 2 that's being<br>built out. Um<br>>> that'll be uh half a gigawatt uh and<br>operational middle of next year. Mhm.<br>Uh,<br>>> hey everybody. You may not know this,<br>but I've got an incredible research<br>team. And every week myself, my research<br>team study the metat trends that are<br>impacting the world. Topics like<br>computation, sensors, networks, AI,<br>robotics, 3D printing, synthetic<br>biology. And these meta trend reports I<br>put out once a week enable you to see<br>the future 10 years ahead of anybody<br>else. If you'd like to get access to the<br>Metatrends newsletter every week, go to<br>dmandis.com/tatrens.<br>That's damandis.com/metatrends.<br>So going back to what Dave is saying<br>over the next five years, what are you<br>scaling on energy front? Do<br>>> I mean<br>>> five years is a long time.<br>>> I mean energy I mean China has done an<br>incredible job.<br>>> Yeah.<br>>> Right. I mean it's running circles<br>around us.<br>>> Uh China has done an incredible job on<br>solar.<br>>> Yeah.<br>>> It's amazing.<br>So I I believe China's uh production<br>capacity is around 1500 gawatts per year<br>of solar.<br>>> Yeah. They put in 500 terowatts in the<br>last year<br>>> terowatt hours. Yeah. Terowatt hours<br>like 500 500 terowatt hours to be very<br>specific<br>>> in the last year. 70% of that was solar<br>and they're just scaling.<br>>> Do do you do you imagine that<br>>> solar scales? Do you imagine that the US<br>could make that level of investment and<br>commitment? I mean because people are<br>worried about their energy bills going<br>up with no no data centers in our<br>backyard. How do we provide I mean<br>energy energy is equivalent to is<br>equivalent to cost of you know cost of<br>living. It's equivalent to health. It's<br>equivalent to clean water. You know the<br>higher energy uh production of a country<br>the higher its GDP. Um energy is<br>important. So what should what do we do<br>to scale that way? Do we do it in solar<br>here?<br>>> Um I think we should scale solar<br>substantially in the US. Um<br>um Tesla and SpaceX are scaling solar.<br>Um<br>so uh and I encourage others to do so as<br>well. M<br>>> um so the the uh<br>I mean I've said the stuff you know<br>publicly um I do see a path to 100<br>gawatts a year of of space solar sort of<br>a AI powered solar powered AI<br>satellites.<br>>> Yes 100 gawatts a year of solar powered<br>AI satellites.<br>>> I did the math on that. Uh, that's like<br>500,000<br>Starlink V3s launched over 8,000<br>Starship flights. That's one every hour<br>for a year. Um, yeah,<br>we 10,000 flights a year is is a<br>reasonable number. Um, so<br>>> it's amazing. It's quite the scale.<br>Well, what's what's the really rough<br>timeline on that because I mean by<br>aircraft standards that's a small<br>number.<br>>> Sure. In terms of flights. Yeah, for<br>sure.<br>>> Yeah, that's uh that's that's that's a<br>small f like so just like depends what<br>you compare it to. If you compare it to<br>the rest of the rocket industry, it's a<br>very high number.<br>>> Yeah.<br>>> Um<br>>> and we're talking about a million tons<br>of payload to orbit per year. So if you<br>do if you do a million tons of payload<br>or orbit per year with 100 kilowatts per<br>ton, uh that's 100 gawatt of solar<br>powered AI satellites um per year.<br>>> Yeah. Um I mean there's a there's a path<br>to get probably to a terowatt per year<br>um<br>>> from from the from<br>if you say like uh 10 you want you want<br>to go up another order of magnitude or<br>let's say you want to go to 100<br>terowatts a year.<br>>> Yeah.<br>>> Which obviously kind of nutty numbers.<br>>> Uh then you want to make those uh AI<br>satellites on the moon.<br>>> Yes.<br>>> And use a mass driver. Yeah. So, the<br>Gerard K. O'Neal approach.<br>>> Well, like Robert Heinland was a harsh<br>course. Pretty much. Yeah. I love that<br>book.<br>>> Yeah. Yeah. It's a sort of libertarian<br>paradise on the<br>>> um<br>uh Yeah. So, cuz on the moon you can<br>just accelerate the satellites into<br>to escape velocity is around 2500 meters<br>per second. Um and uh there's no<br>atmosphere. So, like a mass driver works<br>very well on the moon. Can I ask the the<br>question about orbital debris? I mean,<br>we're we're building effectively a<br>Dysonish swarm around the Earth.<br>>> Um,<br>eat it for lunch.<br>>> Uh, are you worried about over<br>congestion on the uh<br>that's going to be a Sunsync orbit's<br>going to fill very quickly.<br>>> I mean, you can you you don't have to<br>have sunsync. I mean, you can uh<br>>> don't have to, but it's optimal.<br>>> Yeah. Um<br>there's some pros and cons to to sunsync<br>or not sunsync. Um<br>I mean, your your payload to orbit drops<br>by like 30% compared to, you know, if<br>you were just went to um like mid-<br>inclination like 70° or something like<br>that.<br>>> Yeah. I mean, do we need an orbital<br>debris x-prise at this point? We need<br>some way to get the the satellites<br>>> um<br>>> defunct satellites down. Do we pass<br>rules that require them to de-orbit on<br>their own?<br>>> Yeah. At the point at which you you can<br>put a million tons of satellites into<br>orbit, you can also, you know, start<br>bringing down satellites, too. Yeah.<br>>> Um or at least collecting them into a<br>known into a fixed location so they're<br>not like all over the place.<br>>> Yeah. and then you can reuse them.<br>>> Yeah. Um let's just say that we'll have<br>so the the resource level will be so<br>high that that I believe this will be a<br>solved problem given the amount of<br>intelligence we're talking about here.<br>>> Oh<br>>> um like the intelligence will be quite<br>interested in preserving itself.<br>>> Yes. That's true.<br>>> Oh<br>>> interesting.<br>>> Yeah. Good motivation.<br>>> Yeah.<br>>> Interesting.<br>>> The question is the data centers will<br>not be in low earth orbit, right?<br>They'll be they'll be much higher<br>constantly in the sun. They're not going<br>to be in the traffic jam, I assume.<br>>> Uh, well, you can get to, you know, you<br>don't have to get to get to constant<br>sunlight. You can be around 1,200<br>kilometers on synchronous will give you<br>constant sunlight.<br>>> Mhm.<br>>> Um,<br>>> but you could you could place him in<br>multiple orbits.<br>>> Yeah.<br>>> Yeah.<br>>> Yeah. No, I think if there's an X- prize<br>for cleaning up, it's got to be there's<br>only going to be clutter in low Earth<br>orbit. I mean debris from<br>>> anything anything that's if it's a you<br>know below around 7 or 800 kilometers<br>the atmosphere will atmospheric drag<br>will bring it back.<br>>> Yeah.<br>>> Um so like for Starlink there's a dual<br>benefit of being<br>uh like as low as possible because uh<br>your your your beam you you know your<br>beams are tighter. you know, you're<br>basically that you have less latency and<br>and your your your beams are smaller if<br>you're you're closer to the earth. So,<br>uh like Starling 3 will be around 330 to<br>350 km,<br>>> which is quite a lot of drag. Uh so,<br>it's basically constantly thrusting to<br>>> I still remember when you proposed<br>Starlink and everybody else in the<br>industry was like, "No way. No way. He's<br>not going to get the spectrum. He's not<br>going to be able to do this." Um<br>>> yeah,<br>>> it's uh it's kind of worked.<br>>> Yeah, we're the stalling team have done<br>an incredible job.<br>>> Yeah.<br>>> Um<br>>> I mean we've basically rebuilt the<br>internet in space with with a laser<br>links.<br>>> Mhm.<br>>> So there's uh 9,000 satellites up there<br>right now.<br>>> Do you think the government's going to<br>be able to handle the kind of licensing<br>of the volume of satellites that you<br>want to put up? I mean, will there be<br>push back cuz you know, China's going to<br>put up their own constellations.<br>Uh Europe, who knows whether Europe will<br>ever step up?<br>>> They won't.<br>>> What's that? They won't. No.<br>>> And there's probably<br>>> Yeah.<br>>> Nothing that nothing they're doing h has<br>success in the set of possible outcomes.<br>>> Yeah.<br>>> I just got back from Rome. I don't want<br>to touch touch that railing.<br>>> Successes are on the set of possible<br>outcomes.<br>No, the chart of outcomes though<br>>> the chart that shows the number of<br>billion dollar startups in the US versus<br>Europe.<br>>> Have you seen that graphic?<br>>> Oh my god, it's crazy.<br>>> Yeah. And data centers too. It's<br>actually um<br>>> no one was talking about orbital data<br>centers six months ago.<br>>> Yeah.<br>>> Nobody. And then all of a sudden<br>>> Sundire's on it.<br>>> You're you're out with it. And<br>>> it's the hot new thing<br>>> and it is what what what tip what<br>happened? What happened that every<br>company is now talking about orbital<br>data centers?<br>>> I guess it went viral and X.<br>>> It did.<br>>> I don't know. Is every company talking<br>about<br>>> Oh, yeah. Everybody's got their own<br>orbital data center.<br>>> For sure. And I I was suggesting to<br>Peter that that you updated the math on<br>launch costs and that it's a tipping<br>point very quickly with the updated<br>math.<br>>> But Starship's been the cost for you<br>know, I don't know what you hold $100<br>per kilogram, $10 per kilogram. What do<br>you have Starship at? It's possible that<br>Elon said that and nobody believed it<br>until now.<br>>> No,<br>>> you can go back and look at my what even<br>back when it was Twitter uh the my old<br>tweets. I I said these things se many<br>years ago.<br>>> 100 bucks or 10 bucks a a kilogram.<br>>> Yeah. And I said this is we're we're<br>going to do a million tons a year to<br>orbit.<br>Um<br>Yeah. And and we've got to get the the<br>cost down.<br>>> Yeah. uh well below $100 a kilogram.<br>>> So that's going to move the data centers<br>to orbit.<br>>> It will. It's they can do you can<br>basically do the math like if you've got<br>a fully reusable rocket.<br>>> Yeah.<br>>> Um which is fully and rapidly reusable<br>like an aircraft. Uh then this is an<br>incredibly this is a very difficult<br>thing to do obviously. U I I think it's<br>at the limit of human intelligence to<br>create a fully and rapidly reusable<br>rocket.<br>>> Um<br>>> but it is possible and we're doing it<br>with Starship. It's It's been the holy<br>grail in the aerospace industry forever.<br>>> Yeah. Quest for the holy grail rocket.<br>>> Yeah.<br>>> And then I pretty much it is I mean<br>right the DCX was the first little<br>things that were trying there and uh<br>it's been you know all of I mean back<br>when I was in the space industry that's<br>all everyone ever spoke about. And then<br>when Falcon 9 first reused its first<br>stage, um I mean all the traditional<br>aerospace industries did not believe<br>that even Falcon 9 could re could could<br>fly and reuse.<br>>> Literally you can come see it land at<br>Cape Canaveral.<br>>> Yeah.<br>>> Um and then take off again.<br>>> Yeah.<br>>> So I don't know how you would not<br>believe a thing that you can see with<br>your own eyes.<br>>> Yeah. Well, they didn't believe you<br>could. They didn't believe you could.<br>>> But the the the la the leap from there<br>to the launch cost actually requires<br>more faith than just just that. But I<br>think I think Starship is the launch<br>cost tipping point and that somewhere in<br>that you know before you had Twitter it<br>became X somewhere in that timeline it<br>went from speculative to no doubt and I<br>don't know if that's a smooth line or a<br>couple of good launches in between but I<br>suspect that the data centers in space<br>>> but people<br>>> ties directly to the credibility<br>>> is not thinking about orbital data<br>centers they're thinking about energy<br>and the cost of energy here on here in<br>their hometown and sort of the the<br>there's a lot of doomer<br>conversations out there. The data<br>centers are going to drive, you know,<br>the CPI up.<br>>> Uh<br>they're not entirely wrong.<br>>> Okay. So, what is so what is the what's<br>the energy solution here on Earth for uh<br>the rest of humanity or the the non data<br>the non AIs?<br>>> Oh, there's something other than data<br>center use uses of energy. Okay.<br>>> Interesting.<br>>> Um<br>>> that's complex. Well, the the the best<br>way to actually increase the energy<br>output per year of the United States or<br>any country is batteries. Um, so the<br>>> sure<br>>> peak power output of the of the US is<br>around 1.1 terowatts, but the uh average<br>power usage is only half a terowatt.<br>>> Yeah. So if you just buffer the the<br>energy, so charge up the the batteries<br>at night, discharge during the day, um<br>without incremental capital expend<br>without incremental capital<br>expenditures, without building new power<br>plants, you can double the energy<br>throughput of the US. The energy output<br>per year can double<br>>> with batteries. Um<br>>> and do we have those batteries uh in<br>development?<br>>> Uh yeah, Tesla makes them.<br>>> Okay. So you think current the current<br>current Tesla battery packs?<br>>> What do you think? What do you think? I<br>literally have I I went on stage and<br>presented the thing.<br>>> Yeah,<br>>> that's that's the dead giveaway. So<br>>> I I even went to installations of the<br>mega packs, you know, and there's<br>>> So why don't people do this?<br>>> It's on the internet. So<br>>> yeah.<br>>> So is do you think<br>>> they are? And and China, by the way, is<br>like it seems like China listens to<br>everything I say I say and does does it<br>basically or at least or or they're just<br>doing it independently. I don't know.<br>But they're they're certainly making um<br>massive battery packs like<br>really massive battery pack output.<br>They're they're you know making vast<br>numbers of electric cars. Yeah.<br>>> Uh vast amounts of solar. Um,<br>>> I don't know. These are all things I I<br>said, you know, we should do here.<br>>> Fundamental. Sure. When I fly over Santa<br>Monica and LA, when I'm when I'm I'm<br>piloting and I look down, they're like,<br>zero roofs have solar on them.<br>>> Zero roofs.<br>>> Yeah.<br>>> I mean,<br>>> it's not essential to have them on a<br>roof.<br>>> Okay. But it's a convenient place to<br>have them.<br>>> Yes. Uh, but the surface area of roofs<br>is uh I'm not saying it shouldn't, but<br>it's<br>>> uh Tesla makes a solar roof, which is<br>the the only solar roof that isn't ugly.<br>Um, our solar roof actually looks<br>beautiful.<br>>> Yeah.<br>>> Um,<br>but if you want to do solar at scale,<br>you just need more surface area.<br>>> So, so we we we have um vast empty<br>deserts. Sure. African America like if<br>you fly from LA to New York or just fly<br>across country and you look down um for<br>a large portion of the time you look<br>down it is bleak desert.<br>>> Yes.<br>>> It looks like Mars essentially.<br>>> We're not worried about overpopulation<br>there.<br>>> No, I mean it look there's barely a<br>lizard alive in these scorching deserts,<br>you know. Yep.<br>>> It's not like farmland we're talking<br>about. We're just talking about Yep.<br>>> Uh places that look like Mars,<br>>> like just uh scorched rock.<br>So if we put soil where we currently<br>have scorched rock,<br>>> I think this will be a quality of life<br>improvement for the lizards or the few<br>creatures that live in this<br>>> uh very difficult environment.<br>>> Do we have the distribution network?<br>>> It's like this is going to be thank god<br>some shade finally.<br>>> Do we have the distribution network to<br>be able to do that? Yeah, you need to to<br>materially affect quality of life, you<br>need to capture and store what a couple<br>hundred gigawatts.<br>>> Is that in realistic?<br>>> You could just put the data center I<br>guess locally there.<br>>> Well, we already covered data centers.<br>>> We're talking about you know the other<br>>> Yeah.<br>>> Like I I don't know like in an abundant<br>world five years from now, massive<br>amounts of compute,<br>>> massive, you know, universal high<br>income.<br>>> I don't know income like universal you<br>can have whatever you want income.<br>>> Yeah.<br>>> Yeah. That's that's really what it<br>amounts to.<br>>> But in that world, uh, you know, other<br>than compute energy, how much more<br>energy do we need like 30 40 50% or I<br>don't know, unless we want to move<br>mountains around to make a ski mountain,<br>you know, in the backyard.<br>Um, I think the vast majority of energy<br>consumption will go into compute. And<br>then there may be use cases I'm not<br>thinking of like you know the well you<br>know right here is a nice case study<br>because manufacturing every one of these<br>cars coming out at the rate of one every<br>minute or two uh is less energy than the<br>data center that's training the cars to<br>drive to to self-drive.<br>>> Yes.<br>>> So that's a good little case study. And<br>we don't need that much more physical<br>energy for abundant happiness. We need<br>more compute energy. Well, yeah,<br>>> the sun is just generating vast amounts<br>of energy uh all the time for free that<br>goes just goes into space.<br>>> So, um I think we'll end up trying to<br>capture I don't know uh a millionth of<br>like a millionth a thousandth of the<br>sun's energy. Um, we're currently I'm<br>not sure the exact number, but we're<br>I don't know, we're probably at 1%ish of<br>Kadeshv level one.<br>>> Fair enough. Yeah, I I I would guess<br>that even that's high.<br>>> I'm just Yeah, saying<br>>> we have a long way to go.<br>>> I'm that's being optimistic. Like<br>hopefully we're not.1%, but I don't<br>think we're 10%. I'm just trying to get<br>it to like to an order of magnitude. Uh<br>>> so pull it like we're roughly 1% of the<br>apparently using 1% of the energy that<br>we could use on Earth.<br>>> I think the bottom line from a first<br>principles thinking for the public is<br>there's a lot of energy out there<br>>> a lot<br>>> and it we have it in the US. We have it<br>on the planet and it needs to be<br>captured and the tech to capture it<br>>> is here and improving every year.<br>>> Yes.<br>>> Yeah. um there's not going to be some<br>energy crisis. I there'll be a large<br>forcing function to harness more energy,<br>but we're not going to run out of it.<br>>> All right, I want to talk about<br>education.<br>So, here's the numbers. They're abysmal.<br>>> Um I mean, they're they're they're<br>abysmal, right? Okay. Uh the importance<br>of college in the United States, uh back<br>in 2010, 75% of Americans said it's<br>important to go to college. That number<br>is now down at 35%. All right. Uh,<br>college graduates as a group turn out to<br>be the group that's out of work the<br>longest,<br>>> right? And the but still and tuition has<br>increased 900% since 1983.<br>Um,<br>>> yeah, the administrative expenses at<br>universities have gotten out of control.<br>Yeah.<br>>> Um, so<br>>> I think I saw some stat that like<br>there's one administrator for every two<br>students at Brown or something like that<br>>> and I'm like this seems uh little high.<br>>> Yeah. You know what?<br>>> They should teach something.<br>>> Yeah. Yeah.<br>>> What was your college journey?<br>>> Um, I went to college in Canada for a<br>couple years at Queens University.<br>Uh-huh.<br>>> Um, so, uh, I I had Canadian citizenship<br>through my mom who was born in Canada<br>and my my grandfather was actually<br>American, but for some reason, I don't<br>know, my mom couldn't get US<br>citizenship, so but she was born in<br>Canada, so I got Canadian citizenship.<br>Um, and uh, I didn't have any money, so<br>I could only go to Canadian university<br>at first. I<br>>> mean, people forget that about you. You<br>didn't have this giant social network or<br>huge amount of wealth coming into all of<br>this.<br>>> No.<br>>> Yeah.<br>>> Uh, no. I I arrived in Montreal at age<br>17 with I think around $2,500 in<br>Canadian travelers checks back when<br>travelers checks were a thing.<br>>> Um and um one bag of books and one bag<br>of clothes. That was my starting point.<br>That was my spawning point in North<br>America. Um,<br>>> and then so I went to Queens University<br>for a couple years and then uh<br>University of Pennsylvania uh did a dual<br>degree in physics and economics um<br>>> and graduated<br>>> uh undergraduate at UPUP<br>Wharton.<br>>> Yeah. And then um I came out to do uh I<br>was going to do a PhD at Stanford<br>working on uh energy storage<br>technologies for electric vehicles<br>essentially material science I guess<br>fundamentally<br>>> um the the idea that I had was it was to<br>try to create a capacitor with enough<br>energy density that you could get um<br>high range in an electric car.<br>>> It's funny I invested in an ultra<br>capacitor company and didn't Yeah.<br>didn't go well. Well, it's one of those<br>things where, you know, you could<br>definitely get a PhD, but it wasn't<br>clear that you could make a company or<br>do something useful like this. Most PhD<br>is un hat I mean, hate it, but most PhDs<br>do not<br>>> turn into something that's going to<br>>> do not turn into something useful. Like<br>you you could add a leaf to the tree of<br>knowledge, but it's not necessar<br>necessarily a useful leaf. enormous<br>fraction of of great entrepreneurs are<br>dropping out<br>>> of grad school or undergrad. But now<br>nowadays the sense of urgency is off the<br>charts.<br>>> I mean they're popping out everywhere.<br>>> Yeah. Because you know don't waste your<br>time going into grad school. Start a<br>company.<br>>> Yeah.<br>>> Curriculum is nowhere near caught up to<br>what's actually going on in technology<br>and I don't have time and all the time.<br>It's like<br>>> you know this is the moment. I I think<br>right now it's like it's unclear to me<br>why someone would somebody would be in<br>college right now unless they want the<br>social experience.<br>>> Yeah.<br>>> Yeah.<br>>> I mean if you have the ability to go and<br>build something. So the question is how<br>would you redesign the educational<br>program if I could be so so blunt as to<br>create more Elon Musks? If we want to<br>create an Elon Musk factory of people<br>who start with very little but are able<br>to drive uh and drive breakthroughs.<br>What's involved there?<br>What drove you?<br>>> Uh curiosity um about the nature of the<br>universe.<br>>> So I'm just curious about uh<br>>> the meaning of life and<br>>> you know what is this reality that we<br>live in. So,<br>>> how early?<br>>> My son Dax wanted to know what was it<br>like for you in middle school and high<br>school.<br>>> He's 14 years old. He's in that age<br>range now.<br>>> Well, I did I found school to be quite<br>painful. Uh and it was very boring and<br>in South Africa it was very violent.<br>>> So So it's like it was it it was like uh<br>>> it's like that was like that book Enders<br>Game.<br>>> Yes. Um but in real IRL<br>>> in this game IRL there's like but not as<br>fun.<br>>> Um<br>>> so your goal was escape.<br>>> Yes.<br>>> Do you think<br>>> escape from the the prison?<br>>> So that's a question I have. Do you do<br>you think that<br>>> it was miserable?<br>>> Do you think most successful people have<br>had a lot of hardship early in life? Do<br>you need to have that level of hardship?<br>>> Probably need a little bit of hardship I<br>suppose.<br>>> Yeah. But and then so it's always tricky<br>like what are you supposed to do with<br>your kids? You know, create artificial<br>adversity. Put them in.<br>>> That's cool.<br>>> You got an answer. That's that's a<br>Warren Buffett topic actually.<br>>> Yeah.<br>>> Well, you do.<br>>> But seriously,<br>>> it's not easy to create artificial<br>adversity because if you love your kids,<br>you don't want to do that. So<br>>> that's for sure.<br>>> So I had a lot of adversity. Um probably<br>it was good. Uh probably, you know,<br>helped somewhat, I suppose. One one of<br>the<br>>> What doesn't kill you makes you stronger<br>type of thing.<br>>> No,<br>>> at least I didn't lose a limb. And I<br>think what doesn't maim you<br>>> good at maming<br>10 fingers.<br>>> Can you modify that a little bit?<br>>> Yeah.<br>>> Can I ask you a question?<br>>> You makes you stronger.<br>>> I uh for the last 5 years I've been<br>helping teach this class, Foundations of<br>AI Ventures at MIT. And every year when<br>you survey the students, they go up a<br>lot in their desire to start a company.<br>And so it's now up to 80%. The incoming<br>>> everyone's just going to it's it's just<br>going to be like one person company.<br>>> Well, that's with AI that's that's<br>viable, I guess. But no, they want to<br>co-ound. They Yeah, they don't want to<br>be the founder. They want to be part of<br>a founding team. So, it still works out.<br>>> But, uh, when Peter and I were in school<br>at MIT, it was I'm guessing maybe 10%.<br>and they all wanted to be PhDs<br>>> and and they've been doing the survey<br>everyone who wanted to start. I mean I I<br>>> I don't remember any conversations about<br>with people saying they wanted to start<br>>> even at Stanford at the time.<br>>> Um I I I actually um a few days into the<br>semester or I should say the quarter um<br>I I called Bill Nicks who was the head<br>of material science department and said<br>I' I'd like to just put it on deferment.<br>He said, "Is my class that bad?"<br>>> No. And he he said he said that's he<br>said that's okay. You can put it on<br>deferment. But he said this is probably<br>the last conversation we'll have. And he<br>was right.<br>>> Um but then last I think it was last<br>year he sent me a letter saying that all<br>of my predictions about lithium-ion<br>batteries came true.<br>>> It was very nice.<br>>> And did he also say you can still come<br>back and finish your PhD?<br>>> Yeah. No. Several times Stanford has<br>said that I can come back for free.<br>Well, so you know what happened at MIT<br>is every time so I did not know it<br>>> be a great use of your time.<br>>> Exactly. I'm like<br>>> so every time an Iron Man movie came<br>out,<br>>> it notched up another probably 10% or<br>so.<br>>> Okay.<br>>> Uh in terms of because everybody wanted<br>to be Tony Stark.<br>>> And so that's the image. And I didn't<br>know till today that the new Tony Stark,<br>the modern Iron Man Tony Stark, I always<br>thought Tony Stark was modeled on<br>Charles Stark Draper and Howard Hughes.<br>is Charles Stark Draper's education and<br>his you know scientific endeavors<br>married with Howard Hughes's ambition<br>>> and that created the original character<br>but then when Robert Downey Jr. wanted<br>to reinvent it.<br>>> Yeah, it came.<br>>> It's modeled on Elon.<br>>> Yeah,<br>>> he came and met with me.<br>>> This is a Groipedia fact.<br>>> All right.<br>>> Uh yeah, fantastic.<br>>> Um<br>>> yeah, they came to John Fabro and and<br>Robert<br>>> I like the name Grock. I would like<br>Jarvis as well.<br>>> Yeah.<br>>> Yeah. Um<br>>> probably some some trade.<br>>> At some point if Grock gets good enough,<br>we're going to call it Encyclopedia<br>Galactica.<br>>> Yes, that's nice.<br>>> Yeah.<br>>> Yeah, of course. 42.<br>>> Thank you. Um, so going back to<br>education, uh, should colleges, I guess<br>the social experience, you said is<br>important there, but what would you do<br>for education, uh, you know, middle,<br>high school? You just came back from a<br>announcement with President Blly, uh,<br>who's a friend. I I think he's an<br>amazing amazing visionary. Yeah.<br>Incredible what he did with his nation.<br>>> Yeah.<br>>> Yeah. Um,<br>>> remarkable.<br>>> Remarkable and gutsy.<br>>> Yeah. I was like, "How are you still<br>alive?" That was<br>>> Yeah. I mean, I It was like It's the<br>nuclear It was a nuclear option,<br>>> right? Shut him down. I mean, do you<br>know how besides putting everybody with<br>a gang sign um in in uh in jail? I don't<br>know if you know the second thing he<br>did. He went to all of the graves of all<br>the gang members out there and destroyed<br>the graves and said, "Your memory will<br>not be remembered in this nation."<br>That's just badass.<br>>> And it worked.<br>>> I mean, you have to be badass<br>[ __ ] to take on all the knocker<br>gangs and win<br>>> and live.<br>>> Yeah. And still be alive.<br>>> And live. He's got a great great uh<br>guard at his palace there. But what what<br>did you announce with uh with him in El<br>Salvador?<br>>> Uh it was just uh basically to use Grock<br>for uh education like personalization.<br>>> Hopefully not the vulgar version of it.<br>>> Yeah. we would have like you know the<br>you know kids friendly version of Grock.<br>>> Uh but but obviously AI can be an in an<br>individualized teacher.<br>>> Yeah.<br>>> Um that uh is infinitely patient and<br>answers all your questions.<br>>> Um now you still need to be curious um<br>and and uh<br>you still need to want to learn. You<br>know GR can't make you want to learn. It<br>can make learning more interesting. you<br>could probably gify and incentivize it,<br>right?<br>>> You can make learning more interesting.<br>Um,<br>and and less of a production line. Um,<br>so<br>but kids do need to have to if they need<br>to want to learn, you know.<br>>> Yeah.<br>>> Do you and like the people should just<br>think of the the brain as a biological<br>computer.<br>>> It's a neural net.<br>>> Yeah. Yeah, it's a bi biological<br>computer with<br>you know so with a number of neurons and<br>a neural efficiency.<br>>> Yeah.<br>>> Um and um<br>so so what like what you can't do is<br>tune any arbitrary kid into Einstein. Uh<br>this is not realistic because Einstein<br>had a very good meat computer like an<br>outstanding meat computer.<br>>> Um so you can't just uh do Shakespeare<br>Newton you know Einstein type of thing.<br>um unless the meat computer is uh an<br>exceptional one.<br>>> So what do you think? So when people say<br>we need to solve education in the United<br>States<br>>> um because it's fundamentally broken u I<br>think what's really broken I'm curious<br>is the old uh social contract that says<br>uh do well in high school, get in a good<br>college, get a degree, and then get a<br>job. And I don't know that that's going<br>to be valid in the future.<br>Uh my we talk about this on the pod a<br>lot that the that the career of the<br>future isn't getting a job. It's being<br>an entrepreneur. It's finding a problem<br>and solving it.<br>>> Yeah.<br>>> Do you do you agree with that?<br>>> Right now I'd say people should just you<br>know go to school for the social<br>experience,<br>use more AI.<br>Um<br>the conventional schooling experience I<br>think could be a lot better. um the what<br>what we're going to do in Al Salvador<br>and hopefully other places just have<br>individualized teachers that's going to<br>be much better and you you could go to<br>you could go to a school with a bunch of<br>other kids I guess if you want to hang<br>out with other kids but you don't need<br>to<br>>> right<br>>> you could do it on your phone at home<br>um so that's why I say like at this<br>point education is a social experience<br>when I talk to my kids who are in in<br>college<br>>> uh they they they do recognize that they<br>can learn um just as much independently.<br>In fact, that they would learn more in<br>in a work situation.<br>>> Yeah.<br>>> Um they're there for the social<br>experience and to be a bunch around a<br>bunch of people of their their own age.<br>Um<br>sort of a coming of age social<br>experience.<br>>> Sure. Sure. Being on your own uh<br>learning how how to lead or defend<br>yourself as the case may be.<br>>> Well, yeah. Yeah, I mean, if you join<br>the workforce, you're, you know, from<br>the perspective of like a, you know,<br>19-year-old, you with a bunch of old<br>people,<br>and if you're doing engineering with a<br>bunch of middle-aged dudes, it's like,<br>do you really want to do that or do you<br>want to hang out with um, you know,<br>where there's at least some girls your<br>age<br>type of thing.<br>>> I I want to get I want to get I want to<br>get back to this when we talk about<br>>> a lot of other choices. Actually,<br>>> I want to get back as we get to<br>universal high income, but I want to<br>talk about health and longevity one<br>second. US is the number one ranked<br>number one in health expenses worldwide<br>and it's ranked 70th<br>>> in health span,<br>>> right? We<br>>> are really 70th.<br>>> 70th<br>>> is that from Is that accurate?<br>>> Is why everybody listen it?<br>>> Uh I think it would be better than 70th<br>>> for health span.<br>>> Um well, whatever. It's it is like we<br>just get fat or something.<br>>> We're not the top 10.<br>>> Maybe a Zic can help us plan the<br>rankings there.<br>>> Um, so<br>>> would you just run around? We need<br>Cupid. But a Zic.<br>>> Mjaro Cupid.<br>>> But but I think that's a big reason.<br>It's like if people get really fat then<br>their their health gets bad.<br>>> Yeah. Well, if they don't have any<br>exercise, health get bad. or if they<br>donuts for breakfast every morning. You<br>still doing that?<br>>> Uh, no, actually I'm not.<br>>> Okay, that's good. That's good.<br>>> Uh, well, first of all, I wasn't eating<br>a lot of doughnut. I was trying to have<br>uh point4 of a donut, which rounds down<br>to zero.<br>So, I figured anything below below 044<br>of a donut rounds down to zero.<br>>> So, you and I have had uh a disagreement<br>on longevity.<br>>> We had a little bit. Yeah. I was saying,<br>you know, we should push to get people<br>to 120, 150, and you were saying people,<br>you know, shouldn't live that long.<br>>> Uh, so how long do you want<br>>> Yeah.<br>>> You know, there's some,<br>>> you know, people in the world that have<br>done some bad things. How long do you<br>want them to live?<br>>> Yeah. Well, it's okay. They can get the<br>longevity.<br>>> This is a serious question, though. If<br>we them, a lot of things are going to<br>happen that we don't<br>>> Wait a second. You said one thing that<br>you said was interesting. He said um uh<br>we need people to die so people change<br>their minds.<br>>> Oh yes people people don't change their<br>minds they just die.<br>>> But so that makes more sense actually.<br>>> My response to that Elon was you know my<br>response to that was the head of GM<br>didn't have to die for Tesla to come<br>along and Lockheed and Northrup and<br>Boeing didn't have to go away for I mean<br>there's in a meritocracy the better<br>ideas will dominate.<br>So, I'm hoping that I can get you back<br>onto the longevity train. So, there's a<br>lot going on longevity right now, right?<br>>> Uh like what?<br>>> Well, David Sinclair is about to start<br>his epigenetic re uh reprogramming<br>trials in humans. It's worked in in<br>animals and and non-human primates. It's<br>going into humans.<br>>> Is this like a pole or an injection or<br>>> right now? It's an injection of an<br>adnoissociated virus. It's the three<br>Yamanaka factors.<br>>> Okay. Uh we've got a $101 million health<br>span X-P prize that's working on 730<br>teams working on reversing the age of<br>your brain immune system and muscle by<br>20 years. By the way, do you know why<br>it's $101 million?<br>>> No.<br>>> Because the primary funer when they<br>found out your carbon X price was 100<br>bucks, he wanted to make it bigger. So<br>it's 101.<br>>> Oh, who who's the Chip Wilson from<br>Lululemon?<br>>> Oh, okay. And then uh and then evolution<br>out of but Chip said, "Can we make it<br>bigger?" I said, "You put extra million<br>in, we'll make 101 million."<br>>> Sounds good.<br>>> It's a good story.<br>>> But then we got folks like Dario Amade<br>predicting doubling the human lifespan<br>in the next 10 years.<br>>> Um that's<br>probably correct.<br>>> Okay, great.<br>>> I don't know about doubling, but in<br>significant<br>>> significant increase. Sure.<br>>> Um<br>>> which is easily escape velocity.<br>>> I mean because when Yeah.<br>>> Depending how old your Yeah.<br>Oh yeah, for sure. Or effective age.<br>Yeah.<br>>> Yeah. Yeah.<br>>> So I mean I think you know I think that<br>for<br>>> too much and turn into a baby or<br>something.<br>>> That's what I'm telling all the students<br>there. It's like Peter what happened.<br>>> Yes. Yes. There there is a frozen.<br>>> You got a zero wrong in the dosage.<br>Just a small factor of 10.<br>>> Grow out of it. It'll be fine. Exactly.<br>>> You won't remember it. I literally<br>>> I mean, wouldn't it be funny if we do<br>this in like 10 years? Okay, we should<br>do it in I'll do we'll do it in 10 years<br>for sure. And and and let's see let's<br>see if we look younger.<br>>> That's a good side bet.<br>>> My my comment was always Elon's back<br>then Elon was like, you know, late 40s.<br>wait till he gets into his 60s, he's<br>going to want, you know, lunch anymore.<br>>> I mean, I I I want things to not hurt.<br>>> Yeah, sure. Of course.<br>>> It's like it's like basically it's it<br>seems like it's only a matter of time<br>before you get back back pain.<br>>> Yeah.<br>>> Um like it's a when, not an if your back<br>hurts.<br>>> Arthritis. Yes.<br>>> Yeah. Like these things suck basically.<br>>> Being able to sleep through the night<br>without going to the bathroom<br>>> a lot. It's very much That one.<br>>> Yeah, it's<br>more than hope.<br>>> That one.<br>>> Oh man, that would that's like the<br>infinite money one.<br>>> Why did you invest in longevity? So I<br>can sleep through the night and not go<br>to the bathroom.<br>>> Bladder bladder. Yeah. Duration.<br>>> I mean, admittedly, if you have to wear<br>adult diapers, that's a that's a bummer.<br>>> That's not good.<br>Adult D is a real,<br>you know, it's like one of the one of<br>the signs that a country<br>is not on the right path<br>>> is when the adult diapers exceed the<br>baby diapers.<br>>> Yeah, we're there.<br>>> Yeah. South Korea will be there anymore.<br>>> They already No, they passed that point.<br>>> No, they passed that point.<br>>> They passed that point many years ago.<br>Japan passed the point many years ago.<br>>> Doesn't go well looking at the Japanese<br>economy. No, I mean like South Korea is<br>like uh Yeah. One third replacement<br>rate.<br>>> Crazy.<br>>> Yeah. So, three generations they're<br>going to be 127th. So, 3 3% of their<br>current size. I mean, North Korea won't<br>need to invade. They can just walk<br>across.<br>>> Yeah. Yeah.<br>>> This is going to be some people in, you<br>know, walkers or something like there'll<br>be a bunch of optimist.<br>But you you know you've been very verbal<br>about the you know the not<br>overpopulation but massive<br>underpopulation.<br>>> Yeah I've been saying this for ages.<br>>> Yeah. Longevity is going to be an<br>important part of that solution. I also<br>think by the way if you increased the<br>productive life of most Americans by<br>just a few years you'd flip the entire<br>economics here.<br>>> Well if AI and robots is going to make<br>everything sure free basically.<br>>> Yeah. Um but uh well how long would you<br>want to live?<br>>> Uh I want to I want to go you know other<br>planetary systems. I want to go and<br>explore the universe. Yeah. I mean you<br>know I would like to double my lifespan<br>for sure.<br>>> I don't want you know I'm not sure I<br>want to talk about immortality but<br>>> you know at least 120 150. It's a long<br>time.<br>>> One of the worst curses possible would<br>be that<br>>> Yes. May you live forever.<br>>> May you live forever.<br>>> That would be one of the worst<br>>> Yeah. curses you could possibly give<br>anyone.<br>>> But I think life's going to get very<br>interesting.<br>>> Yeah.<br>>> Far more. We're going to speedrun Star<br>Trek as my partner Alex Weer Gross says.<br>>> Yeah.<br>>> Yeah.<br>>> Speedrunning Star Trek would be cool.<br>>> Yeah. Um<br>>> well, at a minimum your kids will have<br>infinite life expectancy. If you're<br>talking about escape velocity, if you<br>can double lifespan, there's it's not<br>even close. You're you're clearly past<br>longevity escape velocity. They the idea<br>of 50 years of AI improvement.<br>>> Yeah, it's great. I mean, we're going to<br>have 20 years on this.<br>>> I don't know. I got too many fish to<br>fry.<br>>> So, I invited<br>>> This is something, by the way, that I<br>that I think I just I think it's very<br>obviously other people think this, too,<br>but I've long thought that um like long<br>like longevity or semi- mortality is an<br>extremely solvable problem. I don't<br>think it's a particularly hard problem.<br>Um,<br>I mean, when you consider the fact that<br>your body is extremely synchronized in<br>its age,<br>>> Yeah.<br>>> the clock must be incredibly obvious.<br>Um, nobody has an old left arm and a<br>young right arm,<br>>> right?<br>>> Why is that?<br>>> What's keeping them all in sync?<br>um you're programmed to die is the is<br>the way you're programmed to die. And so<br>if you change the program,<br>>> yeah,<br>>> uh you will live longer.<br>>> And we've got, you know, species of the<br>boowhead whale can live for 200 years.<br>The Greenland shark can live for 500<br>years. And when I when I learned that, I<br>said, why can't they? Why can't we? And<br>I said, it's either a hardware problem<br>or software problem, and we're going to<br>have the tech to solve that. And I do<br>believe that it's this next decade. So<br>the important thing is not to die from<br>something stupid before the before the<br>solutions come. You know, I invited you<br>uh<br>>> in retrospect the long the solution to<br>longevity will seem obvious.<br>>> Yeah.<br>>> Extremely obvious.<br>>> I I think the thing worth working on<br>Peter's going to work on this anyway,<br>but the thing to work on is exactly what<br>you said. If old ideas don't calcified<br>old ideas don't just die off, add that<br>to the pile of things we need to think<br>about today because there are a whole<br>host of other AI related things we need<br>to think about today.<br>>> Let me let me finish on the longevity<br>point one second. Um Elon uh I want to<br>invite you again. So uh uh there's a<br>company called Fountain Life that uh<br>created with Tony Robbins, Bob Hurry,<br>Bill Cap, and we do a 200 gigabyte<br>upload of you. Everything knowable about<br>you. Full genome, full all imaging,<br>everything. Right. President Blly and<br>the first lady came through, called it<br>an amazing 10 out of 10 experience.<br>>> Um<br>>> I think I don't want you to pull a Steve<br>Jobs<br>>> and kick the bucket because of some<br>>> because some something they didn't know.<br>I mean, so if you ask yourself,<br>>> do you actually know what's going on<br>inside your body right now?<br>>> Um, I did an MRI recently and submitted<br>it to Gro and it didn't<br>>> need no<br>none of the doctors nor Grock found<br>anything wrong,<br>>> but that's a fraction of the<br>information, right? I mean, it's your<br>full genome, your microbiome, your<br>metabolism, everything.<br>>> And okay,<br>>> it's possible. So,<br>>> don't call me.<br>>> What's that?<br>>> Don't call me, bro.<br>We have a We have a center in<br>>> your water bottle.<br>>> We have<br>God damn it.<br>>> Too late.<br>>> Sorry. It's already in the works.<br>>> So, can you go through the the rationale<br>of UHI? How does how does universal high<br>income work?<br>>> Okay. So<br>there's there's going to be more<br>intelligence,<br>digital intelligence than all human<br>intelligence combined and more humanoid<br>robots than all humans.<br>>> Um, and assuming we're in a benign<br>scenario, Star Trek, sort of Rodenberry,<br>not Cameron situation.<br>>> Yeah.<br>>> Um,<br>>> poor Jim.<br>>> Yeah. I mean, I guess it's important to<br>have these sort of<br>>> counterpoints.<br>>> Yeah. Let's not let's go not go in that<br>direction. Um<br>thing. Um so<br>uh<br>the the robots are going to just do<br>whatever you want.<br>>> All the blue collar labor is being done<br>by robots. All data centers are being by<br>robots.<br>>> The the white collar labor will be the<br>first to go because until you until you<br>can move atoms, the thing that can be<br>replaced first is anything that that<br>involves just<br>digital if it's digital like if it<br>involves<br>>> t tapping keys on a keyboard and<br>>> moving a mouse the computer can do that<br>they can do that<br>>> sure<br>>> um you need the humanoid robots to to uh<br>shape atoms so if all you're doing is<br>changing bits of information which is<br>white color work um that is that is the<br>first thing that that<br>>> when this is the inspirational this is<br>the inspirational part of the podcast by<br>the way when is when is all white color<br>work gone by when?<br>>> Well, there there's there's a lot of<br>inertia. So, even with AI at its current<br>state, um I'd say you're you're pretty<br>close to being able to replace half of<br>all jobs of<br>>> and you know that white color jobs that<br>includes anything like education, too.<br>>> Yeah. M<br>>> so anything that involves information<br>um and anything short of shaping atoms<br>um AI can do probably half or more of<br>those jobs right now.<br>>> Sure.<br>>> But there's a lot of inertia. People<br>just keep doing the same the same thing<br>for quite some time. Um, and there<br>actually has to be a a company that<br>makes more use of AI that competes with<br>a company that makes less use of AI,<br>creating a forcing function for<br>increased use of AI,<br>>> right?<br>>> Otherwise, the company that that still<br>has humans do um things that AI can do<br>will still continue to exist. Being a<br>computer used to be a job. So it used to<br>be that a human computer<br>like yeah<br>>> a computer being a computer was a job.<br>You would compute numbers. Sure. It<br>didn't it didn't used to be a machine.<br>It used to be a job description. Um, and<br>there you can look online there's these<br>pictures of like where they're having<br>like skyscrapers full<br>>> of women copying mostly women copying<br>from ledger to ledger<br>>> and men too but but yeah but pe people<br>um<br>>> um but it was a lot of women but there's<br>there were just buildings full of uh<br>people just at desks doing calculations.<br>>> Yeah. Um so they'd be calculating the<br>interest in your bank account or um you<br>know some<br>um you know science uh experiment or<br>something like that or what but if you<br>want calculations done uh you people<br>would do it. Um so<br>um now<br>one laptop with a spreadsheet can<br>outperform a skyscraper of several<br>hundred human computers<br>>> right<br>>> of people doing calculations. Um, now if<br>even a few cells in that spreadsheet<br>were done manually,<br>um, it you would not be able to compete<br>with a spreadsheet that was entirely a<br>computer.<br>>> Mhm.<br>>> Yeah.<br>What this means is that companies that<br>are<br>entirely AI will demolish companies that<br>are not.<br>>> Right.<br>>> It won't be a contest.<br>>> Agreed. And that flippid.<br>>> Yeah. one cell in that<br>>> just one if<br>>> I got to do that<br>>> would you want even one cell in your<br>spreadsheet to be manually calculated<br>>> that would be the most annoying cell and<br>you're like god damn it<br>>> y<br>>> and and and gets it wrong a bunch of the<br>time<br>error rate<br>>> so this flipping<br>>> flipping the flipping<br>>> um<br>>> are we monetizing hope effectively<br>>> yes<br>>> not not at this moment I think we're I<br>think we're pe I think we're pe doo for<br>people worried about the future of their<br>jobs.<br>>> Monetize.<br>>> We're at peak doom.<br>>> We're going to do that<br>as a t-shirt<br>>> and the mug.<br>>> And the mug.<br>>> Yes.<br>>> The mug.<br>>> Uh,<br>so but you have a sol you have a<br>solution to this<br>>> which is UHI.<br>>> Yes. Everyone can have whatever they<br>want.<br>>> So how does that work? How does UHI<br>work?<br>>> It's it's a good question. like we have<br>to figure out some like<br>>> I mean it's not a it's not a bumpy road<br>it yeah I mean so my concern isn't the<br>long run it's the next 3 to seven years<br>>> yes the transition will be bumpy uh<br>because humans don't like simultaneously<br>yes we'll have radical change social<br>unrest and immense prosperity<br>>> and you can buy all all the cyber trucks<br>you want<br>>> things are going to get very cheap<br>>> yes<br>>> um So this is actually and frankly if if<br>this doesn't happen we we'd go bankrupt<br>as a country. So the national debt is<br>enormous.<br>>> Yeah.<br>>> Uh the interest on the national debt<br>exceeds uh not just the military budget<br>but the military budget I think plus um<br>Medicare<br>>> um or Medicaid one of the two. It's like<br>like it's it's like one trillion<br>>> of interest. Yeah. Um<br>>> which is growing.<br>>> Yes. And the deficit is growing.<br>>> Yes.<br>>> Um<br>but the the so this so if if we don't<br>have AI and robots, we're all going to<br>go bankrupt and and and and we're headed<br>for economic doom.<br>>> We're going back also competitive<br>pressure from China. So this is<br>definitely going to happen. I guess<br>>> we're going back to the theme of this<br>talk. How can AI and exponential tech<br>save America and the world?<br>>> Don't you think that? But I want I want<br>to get I want to hit this because we<br>>> I was like quite pessimistic about it<br>and and and ultimately I decided to be<br>fatalistic and and<br>>> um look on the bright side.<br>>> I've got to see you look on the bright<br>side of life.<br>>> You're sitting there<br>crucified<br>right side.<br>>> But this is not about taxation and<br>redistribution.<br>>> Yeah. No, it's um<br>>> So, how do how does it work? Reason<br>through it with me.<br>>> Listen, by the way, I'm open to ideas<br>here.<br>>> Okay.<br>>> Uh so, it's not like I got this all<br>figured out.<br>>> All right. So, so I'm wondering if<br>instead of universal high income, if<br>it's universal, universal high stuff.<br>>> Yeah.<br>>> And services.<br>>> Yes.<br>>> The UHSS. We got<br>>> like I I guess Okay. This is my guess<br>for how things<br>roll out play out. And I I and by the<br>way, I'm this is this is going to be a<br>bumpy ride and it's not like I know the<br>answers here. Um but I I I have decided<br>to look on the bright side. U and and<br>I'd like to thank thank you guys for<br>being an inspiration in this regard.<br>>> Thank you.<br>>> Happy to help. Yeah,<br>because I I actually think it's it it is<br>better to be a an optimist and wrong<br>than a pessimist and right.<br>>> Yes, for sure.<br>>> Um for quality of life.<br>>> Yeah. And by the way, there's also not a<br>force of nature. It's under<br>>> like to me it's really clear that we<br>don't have any system right now to make<br>this go well. But AI is a critical part<br>of making it go well. And at some point,<br>Grock is going to be addressing this<br>exact topic that we're talking about or<br>it has to be one of the big four AI<br>machines. I mean, it's coming dealing<br>with it. There's no velocity knob,<br>right? There's no onoff switch. It is<br>coming and accelerating.<br>>> I call AI and robotics the supersonic<br>tsunami.<br>>> Yes.<br>>> Which maybe is a little alarming.<br>>> You think it's good. That's good. Well,<br>because the wake up call.<br>>> This is important for folks to to gro<br>because um uh I don't want to leave<br>people depressed. I want people to<br>understand what's coming. So we're we're<br>basically demonetizing<br>everything. I mean labor becomes the<br>cost of capex and electricity. AI is<br>basically uh intelligence available uh<br>>> at a dimminimous price. Um<br>uh so you're able to produce almost<br>anything. Things get down to basic cost<br>of materials and electricity, right? Uh<br>so people can have whatever stuff they<br>want, whatever services they need.<br>>> Um it's not when when we say universal<br>high income, it sounds like it's a tax<br>and redistribute, but that's not the<br>case.<br>Um<br>>> it's it's I think my my best guess for<br>how this will manifest is that prices<br>will become prices will drop.<br>>> Yeah.<br>>> So as the efficiency of of production or<br>the provision of services drops um<br>prices will drop. I mean you know prices<br>in in dollar terms are the ratio between<br>the output of goods and services and the<br>money supply.<br>>> Sure. So if your output of goods and<br>services increases faster than the money<br>supply, you will have deflation and or<br>vice versa, you know. So um<br>>> it's a good thing we're growing the<br>money supply so quickly then,<br>>> right?<br>>> I I I Yes. That's why I I I came like<br>let's not worry about growing the money<br>supply. It won't matter because the<br>output of goods and services actually<br>will grow faster than the money supply.<br>And I think we'll be in this and this is<br>a prediction I think some others have<br>made but um I will add to it which is uh<br>that that I think governments will will<br>actually be pushing to to increase money<br>supply um like like faster.<br>>> Yes. They won't be able to waste the<br>money fast enough which is saying<br>something for<br>>> Isn't it isn't it crazy how close those<br>timelines just randomly worked out? I<br>mean at the rate because we're expanding<br>the national debt not because we're<br>anticipating AI. We were going to do<br>that no matter what.<br>>> And so it's like right on the edge of<br>becoming Argentina.<br>>> But yeah, at the time so productivity is<br>going to improve dramatically<br>>> and it is improving dramatically. I I I<br>think we'll see<br>>> I think I think we may see<br>like high double digit uh output of<br>goods and services. We have to be a<br>little careful about how economists<br>measure things<br>and um<br>>> yeah<br>it's it I mean there's like my favorite<br>joke I have a few economist jokes that I<br>that that I like but um maybe my<br>favorite one economist joke is um two<br>economists are going for a walk in in<br>the forest um and they come across a<br>pile of [ __ ] and one economist says I'll<br>pay you 100 bucks to eat a pile of [ __ ]<br>I've heard this one. This is great. Go<br>ahead.<br>>> And so the guy takes 100 bucks and eats<br>the [ __ ]<br>>> Then they keep walking. They come across<br>another pile of [ __ ] And and the other<br>guy says, "Okay, I'll give you a hundred<br>bucks to eat a pile of shit."<br>So he gives him a hundred bucks and and<br>then the the guys can say, "Wait a<br>second.<br>>> We both have the same amount of money.<br>We ate a both ate a pile of [ __ ]<br>>> Oh my god. It sounds like<br>>> but we increase the economy by $200.<br>>> This is the kind of [ __ ] you get in<br>economics. So So uh but if you if so if<br>you say like just the output of goods<br>and services um<br>the will be much greater. You just need<br>a<br>>> so profitability of companies go through<br>the roof<br>>> at some point. But but no but so the<br>question becomes is that taxed by the<br>government? uh<br>>> is that then taxed by the government and<br>redistributed as some level of income as<br>a U as a UHI or UBI? In other words, um<br>one of the questions is if in fact this<br>future we hit massive productivity uh<br>and massive profitability because we're<br>dividing by zero. The cost of labor has<br>gone to nothing. The cost of<br>intelligence has gone to nothing and<br>we're still producing products and<br>services faster and faster. So there's<br>more profitability. Someone needs to be<br>buying it and someone needs to be able<br>to have the capital to buy it. Um,<br>I mean this is an important question to<br>get to get thought through.<br>>> Yeah. Um, well, one like side<br>recommendation I have is like don't<br>worry about like squirreling money away<br>for uh retirement in like 10 or 20<br>years. It won't matter.<br>>> No.<br>>> Okay. either either we're not going to<br>be here or<br>>> it it just uh like it's it's you won't<br>need to save for retirement. If if any<br>of the things that we've said are true,<br>saving for retirement will be<br>irrelevant.<br>>> The services will be there to support<br>you. You'll have the home, you'll have<br>the healthcare, you'll have the<br>entertainment.<br>>> The way this unfolds is fundamentally<br>impossible to predict because of<br>self-improvement of the AI and the<br>accelerating timeline.<br>>> Yeah. It's called singularity for a<br>reason.<br>>> Yeah. Exactly.<br>>> I don't know what goes what what what<br>happens after when after the event<br>horizon.<br>>> Exactly. You can't never see past the<br>black hole or the event horizon. The<br>light cone.<br>>> I mean Ray has a singularity out way too<br>far. I mean this is like the next what<br>what's your timeline for<br>>> for this?<br>>> We're in the singularity.<br>>> Well, we are in the singularity for<br>sure. We're in the midst of it right now<br>for sure.<br>>> And we just we're in this beautiful<br>sweet spot which is you know the<br>>> we're the roller coasters were just<br>>> Yeah. Exactly. That's a great analogy.<br>It's like that feeling.<br>>> You're at the top of the roller coaster<br>and you're about to go.<br>>> Yeah. But you know it's going to be a<br>lot of G's when you lot when you hit it.<br>>> Uh and it's like people like I don't<br>have to just have courtside seats. I'm<br>on the court.<br>>> Exactly.<br>>> And it blows my And still blows my mind<br>>> sometimes multiple times a week.<br>>> Yeah.<br>>> Um and so<br>>> just when I think I'm like<br>wow. And then it's like<br>>> two days later more wow.<br>>> Yeah.<br>>> Um<br>>> exponential wow.<br>Yeah, I think we'll hit um AGI next year<br>in 26.<br>>> Yeah, I heard you say that.<br>>> Yeah, I've said that for a while<br>actually.<br>>> And then you know and then you said by<br>2029 2030 equivalent to the entire human<br>race.<br>>> 2030 we exceed like I'm confident by<br>2030<br>um AI will exceed the intelligence of<br>all humans combined. That's way<br>pessimistic if if you hit AGI next year<br>and that's that's you know that date is<br>is in flux but from that date<br>>> to self-improvements that are on the<br>order of a th00and 10,000x just<br>algorithmic improvements is very short<br>>> and so everybody why isn't everybody<br>talking about this right now?<br>>> Well I mean on on<br>>> X on X they off.<br>>> Yes. But why isn't<br>>> about every day basically.<br>>> Yeah. But it's like<br>>> stop<br>>> it's not<br>>> okay. Okay. So, I'll tell you something<br>else that I I'll tell you something that<br>most people in the AI community don't<br>yet understand.<br>>> Okay.<br>>> Um, which is there the almost no one<br>understands this. Um, the intelligence<br>density potential uh is vastly greater<br>than what we're currently experiencing.<br>So, I I think we're we're off by tours<br>of magnitude in terms of the<br>intelligence density per gigabyte<br>>> of what what's achievable.<br>>> Yes. per gigawatt of energy<br>>> per I'm characterize it by file size<br>okay if the file size of the AI if you<br>>> if you have a say get intelligence<br>>> oh okay in know yes sir<br>>> um<br>>> on your on your drives on your laptop<br>>> power tube parameters the same thing<br>whatever<br>>> um so two two orders of magnitude<br>>> yes<br>>> and you like you said you ringside<br>courtside seat<br>>> you would know I'd say it's it's it's uh<br>two yes Yeah.<br>>> Towards magnitude improvement in um<br>that's just just algorithmic<br>improvement. Same computer and the<br>computers are getting better.<br>>> Yeah.<br>>> So<br>>> and bigger, you know, they're getting<br>better and the budgets are getting<br>bigger. So<br>>> that's why like I think I think it's it<br>is on<br>it is like a 10x improvement per year<br>type thing. Thousand%.<br>>> Yeah.<br>>> And that and that's going to happen for<br>Yeah.<br>for the foreseeable future. So you see<br>the massive underreaction like if you<br>walk downtown Austin the massive I mean<br>it may be under discussion in X but it's<br>not percolating at all.<br>>> Well it's not it's not discussion in any<br>realm of government. Everybody is like<br>defending their position about where we<br>are and jobs and this but<br>>> it's it's like we're heading towards a<br>>> a supersonic supersonic tsunami and and<br>uh uh I mean every every you know every<br>major CEO and economist and government<br>leader should be like what do we do<br>because<br>>> once it hits<br>>> um<br>>> well that it's coming at the exact same<br>time there no matter what there's No,<br>there's no concept of let's deliberately<br>slow down, right?<br>>> No, it's impossible.<br>>> It's impossible at this stage.<br>>> I mean, I I' I'd previously advised that<br>we slow it down, but that was point that<br>uh that's pointless. Like I I like you<br>can't<br>be going to it, but too fast, guys. Um<br>I've said that many years and and I was<br>like okay that I finally came to the<br>conclusion I can either be a spectator<br>or a participant but I can't stop it.<br>>> So at least if I'm a participant I can<br>try to steer it in a good direction.<br>>> Um and uh like my number one belief for<br>safety of AI is to be maximally truth<br>seeeking. So um that don't make AI<br>believe things that are false. Like if<br>you say if you if you say to the AI that<br>axiom A and axom B are both true but<br>they're but they cannot be but but<br>they're not.<br>>> Yeah.<br>>> Um and it has to but it must behave that<br>way. Um you will make it go insane. So<br>that that I I mean I think that was the<br>central lesson that RC Clark was trying<br>to convey in 2001 Space Odyssey was that<br>the um you know people always know they<br>know the meme of that uh hell wouldn't<br>open the pod bay doors but but why<br>wouldn't Hal open the pod bay doors? I<br>mean I guess they should have said uh<br>hell assume you're a pod bay door<br>salesman<br>>> and and you want to sell the hell out<br>shows how well they work. Yes, they're<br>just prompt engineering. one little but<br>the the the the but the AI had been told<br>that it needs to take the this the<br>astronauts to the monolith but also they<br>could not know the about was that in<br>code or was it in English it's flows by<br>in green font right<br>>> yeah it's basically the AI was<br>told that the astronauts couldn't know<br>about the monolith<br>>> that's why it killed them yeah<br>>> so it came it basically came to the<br>conclusion that<br>>> uh the only way to solve for this is to<br>bring the the the astronauts to the<br>monolith dead Yeah, then it has solved<br>both things. It has brought the<br>astronauts to the monolith and they also<br>don't know about the monolith, which is<br>a huge problem if you're an astronaut.<br>>> Turns out AI doesn't care about logic<br>quite as much as that implied.<br>>> So what I'm saying is<br>don't force AI to lie. This is<br>>> give it factual truth. Yes.<br>>> Ilia recently did a podcast. He was<br>talking about one of the potential<br>things to program into AI is is a<br>respect for sentient life of all types.<br>>> Um. Yes. Yes.<br>>> I mean,<br>>> so I'd say another property.<br>>> Yes.<br>>> I mean, there are three things that I<br>think are important. Um,<br>truth, curiosity, and beauty.<br>>> Mhm.<br>>> And if AI cares about those three<br>things, uh, it will care about us.<br>>> On which part?<br>Truth will prevent AI from going insane.<br>>> Mhm.<br>>> Curiosity I think will foster<br>uh any form of sentience. Meaning like<br>we're more interesting than a bunch of<br>rocks.<br>>> Yeah.<br>>> So if it has if it's curious<br>then I think it will foster humanity. Um<br>and if it has a sense of beauty<br>um it will be a great future. I think<br>that's a great foundation.<br>>> Yeah. Jeffrey Hinton made a comment<br>recently. I don't know if you saw it,<br>that<br>>> his his hopeful future was that we would<br>program maternal instincts into our AIS<br>to<br>>> see us maternal.<br>>> Yeah. In other words,<br>>> he haven't heard this. Yeah.<br>>> So, he said a little scary. He said<br>there's a there's a there's a scenario<br>where a very intelligent being succumbs<br>to the needs of a less intelligent being<br>and that's the mother taking care of the<br>child.<br>Do you think that we might have a uh<br>singletarian uh like a a<br>uh that achieves dominance and<br>suppresses others? And do you imagine<br>that that ASI could be a means to<br>stabilize<br>the world in humanity?<br>>> Darwin's<br>observations about evolution,<br>>> yes,<br>>> will apply to AI<br>>> just as they apply to biological life.<br>>> They will compete with each other.<br>>> Yes.<br>>> Uh there's a lot of great science<br>fiction books where the first ASI<br>basically suppresses the others.<br>Um then the question is what do you<br>program into it you know um I I it's so<br>the there's a speed of light constraint<br>that makes that difficult. Um<br>the speed of light is what will prevent<br>um<br>a single mind from existing. Um so light<br>can it it takes um<br>a millisecond to travel 300 kilometers<br>in a a vacuum. Um and uh only you can<br>only get a little over 200 km in a<br>millisecond in glass<br>>> in fiber, right?<br>>> Yeah. Um so<br>even on earth<br>uh there will be multiple AIs because of<br>the speed of light.<br>Um<br>yeah and and this there are clusters of<br>compute that could you could try to<br>synchronize but they weren't<br>synchronized completely. Um so therefore<br>you will have many minds because of the<br>speed of light.<br>>> They don't really have clean borders<br>anymore either though. You have the when<br>you use a mix mixture of experts kind of<br>design it's just flowing through the<br>grand network and you can reassemble<br>parts of it midway through. And you<br>know, we're used to organisms that have<br>clear borders like your head ends there,<br>your head ends there.<br>>> But these things are all mushy.<br>>> To put a bow around this part, I hope<br>you'll put some more thought into UHI.<br>Uh because I think it's really it's<br>really important for us to have without<br>a vision. Uh people need a vision of<br>where we're going. People need<br>something.<br>>> Basically, the government could just<br>issue people free money.<br>>> But I don't think I I think that<br>>> based upon the profitability of all the<br>companies coming inside the country.<br>>> Just issue people free money. No,<br>they're doing that sort of kind of now.<br>>> Yeah.<br>>> But just just just basically issue<br>checks uh to everybody. Um and uh<br>>> but then how big for which person or<br>what you there's so much complexity<br>there. But the thought process behind<br>this rate of change can only be done<br>with AI assistance<br>>> and there's no government entity that's<br>going to keep up with that change. So<br>you have four big<br>>> certainly not the AI is<br>>> it's it's like<br>government is very slow moving as as we<br>all know. Um<br>>> so I think I it's that government really<br>can't react to to the AI. It's it's uh<br>AI is moving you know 10 times faster<br>than government maybe more. Um the the<br>one the one thing that the government<br>can do is just is just issue people<br>money. Um and um<br>>> try and try and keep the peace.<br>>> Yeah.<br>>> Um you know we had like whatever the the<br>co checks and whatever there's<br>>> um you know uh President Trump recently<br>issued like everyone in the military<br>like I think $1,776.<br>Uh I mean it's you can just basically<br>send people random random amounts of<br>money. It's<br>>> um<br>>> okay. So<br>>> so like nobody's going to stop is what<br>I'm saying. Um<br>>> and um<br>>> universal<br>>> I can tell you like let me tell you<br>about some of the good things<br>>> please.<br>>> Um<br>>> so right right now um there's a shortage<br>of doctors and and and great surgeons.<br>You're a doctor yourself. you know how<br>that they're it takes a long time for a<br>human to become<br>>> it's ridiculously expensive and long<br>>> ridiculously yes ridiculous a super long<br>time to learn to be a good doctor um and<br>and even then the the knowledge is<br>constantly evolving it's hard to keep up<br>with everything uh you know doctors have<br>limited time they make mistakes um and<br>you say like how many how many great<br>surgeons are there not not that many<br>great surgeons<br>>> when do you think optimist would be a<br>better surgeon<br>than the best surgeons. How long for<br>that?<br>>> Three years.<br>>> Three years. Okay. Yeah. And by the way,<br>>> three years at at scale.<br>>> Yes. All<br>>> more there probably be more Optimus<br>robots that are great surgeons than<br>there are<br>>> sure all surgeons on Earth.<br>>> And the cost of that is the capex and<br>electricity and it works in Zimbabwe.<br>The best surgeon is throughout in the<br>villages throughout Africa or any place<br>on the planet.<br>>> Yeah. Where do you think it'll roll out<br>first? Not the US obviously.<br>>> Um<br>>> here at at the uh Gigafactory.<br>>> Oh yeah. Just do surgery in the<br>>> um<br>>> but that's an important statement in<br>three years time.<br>>> Yeah.<br>>> Um because medicine I mean<br>>> I'm not like absolutely<br>if it's four or five years who cares.<br>That's still an incredible<br>>> statement to make. I mean good for<br>humanity, right? All of a sudden you<br>demonetize.<br>>> Okay. Here's the thing to understand<br>about like like humanoid robots in terms<br>of the rate of improvement. um which is<br>is that the um you you have um three<br>exponentials multiplied by each other.<br>You have an exponential increase in the<br>AI software capability.<br>>> Yeah.<br>>> Exponential increase in the AI chip<br>capability<br>>> um and an exponential increase in the<br>electromechanical dexterity. The<br>usefulness of the humanoid robot is it's<br>those three things multiplied by each<br>other, right? Um then you have the<br>recursive effect of Optimus building<br>Optimus,<br>>> right? And then you have the shared<br>>> you have a recursive multiplicable<br>triple exponential<br>>> and you have the shared knowledge of all<br>all the experiences.<br>>> Is that literally Optimus building<br>Optimus or is it because you know the<br>>> well not right now but will be the the<br>physical humanoid form factor building<br>the humanoid form as opposed to<br>>> it's foyman machine.<br>>> Yeah.<br>>> Yeah. Yeah.<br>>> I love that. But the void machine is<br>usually something kind of like this<br>shape. You know, making something else<br>is a shape.<br>>> In principle, it's simply a<br>self-replicating thing.<br>>> Yeah.<br>>> Elon, do you know what the number one<br>question you ask a surgeon when you're<br>interviewing them?<br>>> Uh,<br>is this is this a surgeon joke?<br>>> No. It's how many It's how many times do<br>you How many times do you do that?<br>>> There's got to be some funny funny jokes<br>coming.<br>>> No, it's serious. It's it's how many<br>times did you do the surgery this<br>morning?<br>>> It's how many times did you do the<br>surgery this morning or yesterday? It's<br>the it's the number of experiences,<br>right?<br>>> And so with a shared memory<br>>> um you know every optimist surgeon will<br>have seen every possible pertabbation of<br>everything in infrared in ultraviolet.<br>No, not too much caffeine that morning.<br>They didn't have a a fight with their<br>husband or wife.<br>>> Yeah.<br>>> Extreme precision.<br>>> Yes. Three years. Um,<br>yes. Better than any any probably I'd<br>say if you like put a little margin on<br>it. Better than any human in four years<br>>> who's in plastic surgery<br>>> by 5 years. It's not even close.<br>>> So what what about the simple like just<br>I mean there's a million of these things<br>to figure out, but who's going to have<br>access to the first Optimus that does<br>far far better micro surgery than any<br>surgeon on Earth, but you've only<br>manufactured the first 10,000 of them?<br>How do you<br>>> I don't think people understand how many<br>robots there's going to be.<br>>> Yeah.<br>>> Well, there's a window said 10 billion<br>by 2040.<br>>> You still on that path?<br>>> Uh that's not that's a low number.<br>>> A low number.<br>>> Wow. What's the constraint? What's the<br>uh cuz if they're self-building, you<br>know,<br>>> metal the constraint is metal.<br>>> Yeah. Or lithium or<br>>> Yeah. You got to move the atoms. Um it's<br>just all just supply chain stuff. So<br>yeah, but your your point I mean there's<br>some rate limit. You can't just<br>>> manufacturing is very difficult. So you<br>got you got to<br>>> you you you it's it's recursive<br>multiplicable triple exponential but but<br>you still need to you still you still<br>have to climb that you know<br>>> selling hope once again I I think your<br>point was medicine is going to be<br>effectively free the best medicine in<br>the world. Everyone will have access to<br>medical care that is better than what<br>the president receives right now.<br>>> So don't go to medical school.<br>>> Yes. Pointless.<br>>> Yeah.<br>>> I mean unless you but I would say that<br>applies to any form of education is<br>there's not like some<br>I do it for social reasons.<br>>> Yeah.<br>>> You're not going to medical school.<br>>> If you want if you want if you want to<br>hang out with like-minded people, I<br>suppose. Uh<br>>> I mean people are still going to want to<br>be connected with people. There's going<br>to be some period of time<br>>> for reasons.<br>>> Yeah.<br>>> Like a hobby like a you know<br>>> well $9,000.<br>>> I mean there will be a point where where<br>it's expensive.<br>>> The younger generation says I do not<br>want that human touching me right when<br>the surgeon comes over. They're going to<br>be those people later in life who still<br>want a human in the loop.<br>>> Okay. for a little while on the edge for<br>a lesser for they want to live on the<br>edge. I mean, let's just take like we've<br>we've seen some advanced cases where of<br>automation like LASIC for example where<br>the the robot just lasers your eyeball.<br>>> Now, do you want an opthalmologist with<br>a hand laser?<br>>> No,<br>it's a little shaky laser pointer from<br>a horror movie like that.<br>>> Sorry, man. I I wouldn't want the best<br>opthalmologist, you know. The steadiest<br>hand out there with a [ __ ] hand laser<br>beyond my eyeball, you know?<br>>> Oh my god.<br>>> Yeah.<br>>> It's going to be like that.<br>>> It's like, do you want opthalmologist<br>with a [ __ ] hand laser or do you want<br>the robot to do it and actually work?<br>>> This episode is brought to you by<br>Blitzy, autonomous software development<br>with infinite code context. Blitzy uses<br>thousands of specialized AI agents that<br>think for hours to understand enterprise<br>scale code bases with millions of lines<br>of code. Engineers start every<br>development sprint with the Blitzy<br>platform, bringing in their development<br>requirements. The Blitzy platform<br>provides a plan, then generates and<br>pre-ompiles code for each task. Blitzy<br>delivers 80% or more of the development<br>work autonomously while providing a<br>guide for the final 20% of human<br>development work required to complete<br>the sprint. Enterprises are achieving a<br>5x engineering velocity increase when<br>incorporating Blitzy as their preide<br>development tool, pairing it with their<br>coding co-pilot of choice to bring an AI<br>native SDLC into their org. Ready to 5x<br>your engineering velocity? Visit<br>blitzy.com to schedule a demo and start<br>building with Blitzy today.<br>>> Let's jump into one of our favorite<br>subjects, space.<br>>> Yeah.<br>>> So, first off, how cool that Jared<br>Isaacman has become the NASA<br>administrator.<br>>> Friend of Yes.<br>>> I mean, I I don't hang out with Jared.<br>Like, people think I'm like huge buddies<br>with Jared, but um<br>>> uh I I I think I've only seen him in<br>person a few times.<br>>> Amazing candidate. Yeah, he's a really<br>smart person. You know him really well.<br>>> Yeah, I I took him to a Biconor launch<br>in 2008 for his first space experience.<br>>> I mean, he loves space next level and uh<br>is uh technically strong. He's a smart<br>and competent person like really smart<br>and really competent<br>>> and understands business.<br>>> Yes.<br>>> Yes. He understands he gets things done<br>>> and he's been there a few times.<br>>> Yeah. Yeah. So, uh, I I'm I'm just like,<br>you know, we want to have someone smart<br>and competent who, uh, loves space<br>exploration,<br>>> um, and will get things done at NASA.<br>>> I'm a huge fan.<br>>> That's what I was really so so happy<br>when he got renominated. And now,<br>>> yeah. Um,<br>>> um, I I think we need to<br>>> we need a new game plan for space. Like,<br>we need a moon base.<br>>> Yes.<br>>> Like a permanently<br>>> Yes.<br>>> crude moon base. Y<br>>> uh and and build that up as fast as<br>possible.<br>>> Yeah.<br>>> Um I don't think we should do the, you<br>know, send a couple astronauts there for<br>hop around for a bit and come back cuz<br>we did that in ' 69.<br>>> Yes. Been there, done that.<br>>> Yeah. Um it's like a remake of a ' 60s<br>movie. It's never as good as the<br>original.<br>>> Yeah.<br>>> Um<br>>> so 2026 is going to be<br>>> like we need to go, you know, to do<br>something more cool, which<br>>> my nice on the<br>>> Yeah. Put up telescopes.<br>>> Yeah. Yeah, exactly.<br>>> So, do you forward deploy the robots,<br>build everything, get it all ready, make<br>the bed, and then<br>>> Yeah. Get get the jacuzzi warmed up on<br>>> That's an interesting<br>>> Yeah. Yeah.<br>>> Yeah.<br>>> How early in the year are you going to<br>hit orbital refueling, you think, with<br>Starship?<br>>> Uh, not that early in the year.<br>>> I mean, are you are you shooting for the<br>home and transfer orbit?<br>>> I'd say towards towards the end of the<br>year. Um,<br>>> are you shooting for a Mars shot by the<br>end of next year?<br>We could, but uh it would be a low<br>probability shot<br>>> um and somewhat of a distraction. So<br>>> um<br>>> 29 then<br>>> it's not out of the question.<br>>> 28 29.<br>>> Um<br>>> yeah.<br>>> Uh but like on on Mondays I I have the<br>uh Starship uh engineering the big<br>Starship engineering review is on<br>Mondays. Um so that was uh actually the<br>la the thing I did just before coming<br>here. Um and um so I say like like<br>Starship is really we're doing something<br>that is at the limit of biological<br>intelligence.<br>>> Yeah.<br>>> This is a this is a hard thing to make.<br>>> Um<br>>> and and just to capture it, it was<br>created pre AAI.<br>>> Yeah. No AI was<br>>> probably the last<br>>> the last really big thing in that's not<br>AI. Interesting.<br>>> Probably the biggest thing ever made.<br>>> Yeah.<br>>> By pure human hands.<br>>> The Asia will say not bad for a human.<br>True.<br>>> Not bad for a human.<br>>> Yeah. But it'll be like remember<br>>> my little 20 watt meat computer. It's<br>not easy.<br>>> Yeah.<br>>> So suffering through the day.<br>>> Raptor.<br>>> That would be like uh doing accounting<br>doing your uh interest calculation with<br>a pencil. Yeah, that's that's pretty<br>good.<br>>> Yeah,<br>>> pretty good.<br>>> Did that with regular<br>>> not bad for a bunch of monkeys, you<br>know?<br>>> It's like it's like if you saw a bunch<br>of chimps like make a raft and cross the<br>river, you'd be like, "Oh, look at<br>that."<br>But you know, we celebrate we celebrate<br>the pyramids.<br>Good for them.<br>>> Give him some peanuts. Uh<br>>> these things become timeless, right?<br>>> Raptor 3 goes when?<br>>> Yeah, I think it's worth noting.<br>>> Raptor 3 is beautiful.<br>>> Starship.<br>>> It's an amazing by far the best rocket<br>engine ever.<br>>> Is that AI?<br>>> Nothing's even close. Nope.<br>>> That's also So that'll be the last<br>thing.<br>>> E4 will definitely be<br>>> AI. Yeah, there's<br>um like I think AI will start to become<br>relevant next year.<br>>> Mhm.<br>>> Um so maybe we'll it's not like we're<br>pushing off AI. It's just AI is can't do<br>rocket engineering yet.<br>>> Yep.<br>>> But we'll probably will be able to next<br>year.<br>>> We have a company in our incubator doing<br>mechanical design working with Andre and<br>so forth. And it's not you can design<br>brackets and parts and things but you<br>can't quite do rockets. But the timeline<br>is so short, you know, from point A to<br>point B.<br>>> If say like a year from now, probably it<br>can<br>>> it probably can be helpful, meaningfully<br>helpful in a year from now.<br>>> Yeah.<br>>> Um,<br>>> so the big milestones are going to be<br>Starship V3 launching out of Cape<br>Canaveral, orbital refueling.<br>>> Yes.<br>>> Are those the big ones?<br>>> Well, yeah. Um, catching the ship with<br>the tower.<br>>> Yeah, that's right. Um<br>so really the thing that matters is can<br>we refly<br>>> the entire thing?<br>>> Yeah.<br>>> Yeah.<br>>> Uh we have reflow in a booster.<br>>> Sure.<br>>> Um<br>which is you know not bad for it's<br>largest flying objects. Um catching with<br>chopsticks you know.<br>>> Not bad for a bunch of monkeys.<br>>> You're keeping you're keeping the AIS<br>very entertained. Thank you.<br>>> Yeah. Yeah. Exactly. The be like pat on<br>the back from the AGI hopefully. Um, is<br>there a target for number of reuses<br>before? Uh, I mean, it's got to be a lot<br>of wear and tear.<br>>> Uh, it it requires a lot of iteration to<br>achieve high reuse. So, you you figure<br>out like what what's breaking between<br>flights and you sort of iteratively<br>solve those things.<br>>> Um, so from people looking at it from<br>the outside might say, "Oh, the rocket<br>looks kind of the same." But there's<br>like a a thousand changes to to make it<br>more reusable, more reliable.<br>um you know the sheer amount of energy<br>you're trying to you know expend I mean<br>it's uh Starship is uh doing over 100<br>gigawatts of power on ascent.<br>>> It's a lot<br>you know<br>>> do some glass blowing under there and<br>get some uh<br>>> Yeah. Wow.<br>>> a lot. It's a lot.<br>>> There's a lot.<br>>> Um<br>>> but like the amazing thing is that it<br>doesn't explode.<br>>> Yes.<br>>> Some it sometimes doesn't explode.<br>That is<br>>> sometimes not exploding is um like we've<br>blown up a lot of engines on the test<br>stand.<br>>> Um<br>>> I mean is that what causes the wear and<br>tear or is it the re-entry of the or the<br>falling?<br>>> Well, that too. Um I mean for for the<br>booster um the re-entry is not that bad,<br>you know. um you know something's it's<br>it's it's not like that that's not<br>really like we also obviously just<br>solved that you know with with Falcon 9<br>so we kind of understand re booster<br>reuse<br>>> um we've had we've have over 500<br>reflights of the Falcon 9 boost stage<br>>> um so we really understand and and and<br>the Starship booster actually is a more<br>benign entry than um than the Falcon<br>uh booster because the uh the staging<br>ratio is more more biased towards the<br>upper stage for Starship. So I I shifted<br>the the mass ratio to uh be much higher<br>um on the ship side for Starship.<br>>> That was a mistake I made on Falcon 9<br>that there should be more mass in the uh<br>upper stage of Falcon 9.<br>>> Um so that the uh the staging velocity<br>of uh is is lower.<br>>> Yeah. If the station velocity of Falcon<br>9 was lower, would have less wear and<br>tear on Falcon 9.<br>>> Yeah, that's not intuitive at all.<br>That's interesting.<br>>> Yeah, because it's it's kind of a flat<br>optimization. Um the the parallel to<br>orbit um there's sort of a flat region<br>in the mass ratio of the first second<br>stages. And so you just want to bias<br>that mass ratio towards the uh to to put<br>more mass on the upper stage.<br>>> Yeah. Um, so, um,<br>yeah, because you know, you just you got<br>your kinetic energy scaling with the<br>square velocity. So, you've got to<br>describe that kinetic energy. If you're<br>past the melting point of whatever you<br>your stage is made of, you got a<br>problem.<br>>> Yep.<br>>> So, um,<br>>> my my colleague, uh, Alex Wisner Gross,<br>who's one of our moonshot mates here, I<br>wanted to ask a question. I do, too.<br>Have you seen the uh documentary Age of<br>Disclosure<br>about uh all of the announcements by US<br>government officials, military officials<br>about all the alien spacecraft that have<br>been have been uh sort of detained? And<br>I I've heard what you've said about<br>this.<br>>> Well, I do wonder why um you know, if<br>you plot on a chart the resolution of<br>cameras<br>>> Yeah.<br>>> over time like megapixels per year.<br>>> Yeah. Uh, and the resolution of UFO<br>photographs.<br>Why is the only constant? It's flat on<br>UFO.<br>>> We get a a fuzzy blob<br>25. Well, we got like, you know,<br>whatever 100 megapixel camera that can<br>can see your [ __ ] nose hairs. I don't<br>get it.<br>>> Can somebody take a shot of the UFO with<br>an actual camera for love of God?<br>>> But even if you knew,<br>>> that's a valid observation. I'm sure<br>there's an explanation.<br>>> Uh but anyway, it's uh<br>>> it would be fascinating.<br>>> I'm asked all the time if I've<br>>> Yes. And and I'm like, look,<br>>> um I can show you if if I was aware of<br>the slightest evidence of aliens, I<br>would immediately post that on X.<br>>> Yeah.<br>>> And um<br>>> so the question is<br>>> it would be the most viewed post of all<br>time. So, I I actually wonder about the<br>US public if they would like, "Oh,<br>that's interesting." Go back to their<br>sports scores the next day.<br>>> Yeah.<br>>> I think everyone would want to see the<br>alien.<br>>> Yeah.<br>>> Like if you got one.<br>>> Well, like<br>fast way to increase the military<br>budget. We like we found an alien. It<br>seems dangerous.<br>>> That's right. Unify the world.<br>>> They don't have an incentive to hide the<br>aliens. Do they have an incentive to uh<br>bring up show the alien because they<br>would not have any more arguments about<br>the military budget<br>>> if they seem a little bit dangerous?<br>>> Oh, I can always hope.<br>>> I can always hope.<br>>> I mean, I'm you know, we've got 9 9,000<br>satellites up there. We've never had to<br>maneuver around an alien spaceship<br>>> yet. So, well,<br>>> um<br>>> yeah. So anyway, so I guess the good<br>future is<br>um you can anyone can have whatever<br>stuff they want and incredible medical<br>care that's better than any medical care<br>that exists. So I think if you sort of<br>uh lift your gaze, you know, to not a<br>super distant point, five years from<br>now, four years from now, maybe<br>uh we'll have<br>better medical care than anyone has<br>today available for everyone<br>within 5 years.<br>>> Yeah.<br>>> Um<br>no scarcity of goods or services.<br>The best education available for<br>everybody.<br>>> What? You can learn anything you want<br>>> about anything for free.<br>>> Yeah.<br>>> What about access to compute?<br>>> People will probably care a lot more<br>about that than their government check<br>in about three years.<br>>> Well, what do they want to do with<br>compute?<br>>> Well, I mean compute translates to<br>anything you want, right? Your your<br>virtual friend, your entertainment, your<br>like it's it's probably everything.<br>>> Those are AI services basically.<br>>> Yeah. Or or your ability to innovate,<br>too. You can't innovate without an AI<br>assistant at that point. So<br>>> you one of one of our other moonshot<br>mates See Ismael said uh asked this<br>question. He said Elon you often say<br>physics is the law. Everything else is a<br>recommendation.<br>>> Mhm.<br>>> So as AI energy and space systems scale<br>exponentially. What non-physical<br>constraints organizational cultural<br>bureaucracy or human are now the real<br>bottleneck?<br>Is there a bottleneck?<br>Um, electricity generation is the<br>limiting factor.<br>Um, the innermost loop.<br>>> Yeah.<br>Um, I think people are underestimating<br>difficulty of bringing electricity<br>online. You know, you you've got to get<br>you've got to generate the electricity.<br>You've got to you need transformers for<br>the transformers.<br>>> Um, so you got to convert that voltage<br>to something that the computers can<br>digest. You've got to cool the<br>computers.<br>So it's it's basically electricity<br>generation and cooling<br>um are limiting factors for AI.<br>>> Yeah.<br>>> Um and once you have humanoid robotics,<br>they can address the power generation<br>and and the uh the cooling stuff. Um<br>but that that is the limiting factor and<br>will be for at least the next two years.<br>Isn't it amazing how divergent the<br>Memphis version of that is from the<br>space-based version? I you have solar<br>panels in common, but otherwise no<br>storage,<br>abundant amounts of energy. Yeah.<br>>> But you have launch costs and you have I<br>mean and weight suddenly matter. I don't<br>care too much about the weight in<br>Tennessee. Suddenly the weight is a<br>critical factor. I mean those two two<br>pathways for compute have a huge<br>divergence from here forward.<br>>> Yeah. um on once we get solar<br>domestically<br>at scale and uh if we're launching<br>Starship at scale then um by far the<br>cheapest way to do AI compute will be in<br>space. Um so once you have the once you<br>have full and complete reusability um<br>the propellant cost per flight is maybe<br>a million dollars.<br>>> Yeah. People don't realize that people<br>have<br>>> to rid amount of expectations how much<br>it costs. So, so if you listen,<br>>> it's called a million dollars of<br>transport for 10 megawatt of of AI<br>comput.<br>>> Yeah.<br>>> So, assuming everything keeps trending<br>the way it's currently trending, if you<br>look at the next four years of<br>accelerating launches,<br>>> so 200 tons per launch.<br>>> Yeah. Thousands where you're going, but<br>yeah, like if say sun if say high<br>altitude sunny, it's probably more like<br>150 tons. But yeah, it's the right order<br>of magnitude is at least it's it's in<br>excess of 100 tons uh for a marginal<br>cost per flight of around a million<br>million.<br>>> So So what fraction of all that launched<br>mass is data centers in space as opposed<br>to<br>>> moon base as opposed to launch to Mars<br>as opposed to interesting how I mean<br>this is a new we weren't talking about<br>this as a space objective even you know<br>a year ago.<br>>> Yeah. All of a sudden, data centers have<br>become the massive driving force for<br>opening up the space<br>>> and also the urgent the urgent use case<br>too.<br>>> I mean, I used to I used to wonder<br>what's going to drive humanity. I I<br>thought it was asteroid mining, right?<br>You were focused on on Mars. Um,<br>>> we will actually want to mine asteroids<br>to turn them into<br>>> Sure. uh you know<br>>> before before you<br>>> photovoltaic<br>>> before you you know<br>>> not not for anything else like<br>>> I mean if we're gonna if we're going to<br>build out Dyson swarms<br>>> yeah just a bunch of satellites around<br>the sun<br>>> yeah how how how long<br>>> what's your time frame for Alex another<br>question Alex wanted to have us ask<br>what's your time frame for uh for<br>humanity achieving a Dyson swarm is it<br>50 years<br>>> how big is this<br>>> yeah know it's it's a matter<br>>> Dyson swarm people think like<br>everything's just going to be covered in<br>satellites I think It's not quite that<br>that I mean I think we you have to like<br>what mass ends up becoming satellite. Um<br>you know Mercury probably ends up being<br>satellites.<br>>> Yes.<br>>> Jupiter.<br>>> Jupiter. Yeah. Saturn.<br>>> Uh it's a little gassy.<br>>> Oh yeah.<br>>> It's big but there's got a lot of rocks<br>orbiting.<br>>> Do you leave Mars alone? But yeah leave<br>Mars alone.<br>>> Asteroids. Asteroids are are fantastic<br>food source.<br>>> Uh yeah.<br>>> Yeah. No gravity. Well gravity well on<br>Jupiter is a non already mostly<br>differentiated into, you know,<br>carbonacious condrites for fuel and<br>nickel iron for materials,<br>>> gold. Yeah.<br>>> A bunch of the asteroid belt probably<br>turns into solar panels,<br>>> you know, star star power.<br>>> So, I've known you for<br>>> I've known you for 26 years now. It<br>feels to me like I don't want to be, you<br>know, uh it feels like you've gotten<br>much smarter or much more capable<br>over this last decade. Do you feel that<br>way? Do you feel like you just have<br>better people around you, better tools?<br>What what's changed? Because the level<br>of um<br>of audacity, you know, orders of<br>magnitude. Orders of magnitude. I mean,<br>>> some say insane.<br>>> Insanity. Audacious.<br>>> Yeah.<br>>> I say hope.<br>>> Uh what's how how do you feel about<br>that?<br>What's changed? Do you feel that way? I<br>mean, the scope of what your ability is.<br>>> Um,<br>how do you self-reflect on that?<br>>> Well, I' I've had to solve a lot of<br>problems in a lot of different arenas,<br>which um<br>you you get this cross fertilization of<br>of knowledge of of problem solving. Um,<br>and if if you problem solve in a lot of<br>different arenas,<br>then like what what is easy in one arena<br>is trivial in is like what what is<br>trivial in one arena<br>>> is a superpower in another arena. It's<br>sort of like planet kryp. You came from<br>planet krypton<br>>> type of thing.<br>>> So, uh you know krypton planet krypton<br>you'd just be normal. Um but if you come<br>to earth you're Superman. Um so if you<br>take say um manufacturing of volume<br>manufacturing of complex objects in the<br>automotive industry um I have to work on<br>solving that um<br>when translated to the space industry<br>it's like being Superman<br>>> um because rockets are are made in very<br>small numbers<br>>> if you apply automotive manufacturing<br>technology to satellites and rockets. Uh<br>it's like being Superman.<br>>> Um then if you take uh advanced material<br>science from rockets and you apply that<br>to the automotive industry, you get<br>Superman again.<br>>> Yeah.<br>>> Fascinating.<br>>> That's came from planet Krypton. Back<br>back in planet Krypton. This is normal.<br>>> You know, it's funny how how like the<br>knowledge ports that that was true with<br>Tesla and SpaceX being completely<br>separate.<br>>> Yeah.<br>>> But now they actually interact because<br>you know, AI ties everything together.<br>The orbiting. Yeah. The convergence is<br>crazy. Like I don't know if you<br>visualize these parts fitting together<br>originally.<br>>> No.<br>>> No. I mean<br>>> I didn't I don't think they at this<br>point things I guess everything<br>ultimately converges in the singularity.<br>>> Um<br>>> yeah that's what I think too.<br>>> You have lots of different parts of the<br>puzzle that you get to play with.<br>>> Uh<br>there's one part that's missing which is<br>the fab.<br>>> Yeah.<br>>> You going to buy Intel?<br>you get it for a fraction of uh<br>>> that's that was the uh that was the bet<br>we made<br>>> 170 billion<br>>> um I think it needs venue fab<br>>> well<br>I agree but licenses real estate ASML<br>machines it's not easy just get the<br>assets and go I don't think it's easy<br>that's why I mean I it's not like I<br>think it's a simple thing to solve I<br>think it's a hard thing to solve but um<br>but it must be solved<br>I've come to the conclusion that um<br>>> would it be would it be solely captured<br>by you or would it be an asset for the<br>US?<br>>> Look, I'm just saying that we're going<br>to we're going to hit a chip wall.<br>>> Yeah.<br>>> If we don't do the fab.<br>>> Yeah.<br>>> So, we got two ch two choices. Hit the<br>chip wall or make a fab.<br>>> Well, and TSMC for whatever reason is<br>massively worried about overbuilding,<br>which is insane. Um,<br>>> but the whole world will be stuck with a<br>shortage of chips for<br>>> basic. So, so, so they are actually<br>they're I don't know if they're right<br>for the right reason, but they're<br>they're right. Um,<br>>> how so?<br>>> Because it's actually like<br>what is the limiting factor at any given<br>point in time? Um the limiting factor<br>say if you say like by Q3 next year like<br>in 9 months 9 12 months the limiting<br>factor will be turning the chips on<br>>> power<br>>> just power.<br>>> Yeah.<br>>> Uh you need power and all of the<br>equipment necessary power and<br>transformers and cooling.<br>>> So it's it's not like you can just sort<br>of drop off some GPUs at the power<br>plant.<br>>> Yeah. And you vertically integrated<br>you've got it<br>>> again with an X AI, didn't you?<br>>> Sorry.<br>>> You vertically integrated. Yes,<br>>> that inside of XAI,<br>>> we designed our own transformer.<br>>> Yes. And your own cooling system.<br>>> Yes.<br>>> But they're worried that if they make<br>more than 20 million GPUs, like they<br>make 40 million instead of 20 million,<br>that 20 million will not find a source<br>of power,<br>>> but they won't be bought because if<br>there's anything missing that prevents<br>them from being turned on.<br>>> Yeah.<br>>> Um they cannot be turned on.<br>>> Yeah.<br>>> So, uh they've they've got to have a<br>power plant with excess with enough<br>power. So you got have enough gaw then<br>you've got to convert that from probably<br>coming out of a power plant at you know<br>100 to 300 kilovolts type of thing.<br>>> Yeah.<br>>> Um you've ultimately you got to got to<br>convert that uh down to you know several<br>hundred volts at the at the rack level.<br>>> Yeah.<br>>> Um so if you're missing any of the power<br>conversion steps uh you you you won't be<br>able to turn them on and then you've got<br>to extract the heat. Um so it it it's a<br>big shift for the data center world to<br>move to liquid cooling because they've<br>used air cooling.<br>>> Yeah.<br>>> Um and um you know the consequences of a<br>burst pipe uh are very substantial. So<br>if if you if you blow a pipe a water<br>pipe in a data center<br>>> Yeah, I know. I've seen that.<br>>> You just you just fragged a bill a<br>billion dollars right there.<br>>> It just seems inconceivable to me<br>though. Like if if I had those chips, I<br>would find a way to turn them on. the<br>the value of the intelligence coming out<br>the other side so far outweighs the<br>complexity of trying to find a way and<br>there would be a way<br>>> but it's just the crossing of the<br>curves. So if<br>>> if if chip output is growing<br>exponentially but power honest is<br>growing uh in a in a sort of slow linear<br>fashion.<br>>> Yeah. than the<br>>> which is chip output<br>>> right now.<br>>> Exactly. Is chip output growing<br>exponentially? And it's like on very<br>slow exponent if it's growing<br>exponentially. It's<br>>> for a for high power AI chips it's<br>growing exponentially.<br>>> Oh<br>>> like what if we do 20 million GPUs next<br>year what are we talking about the<br>following year? like 22 million 24 I<br>mean I just I don't see the fabs coming<br>online<br>>> but maybe<br>>> so we have two we have two issues to<br>solve<br>>> it's it's you have to like sort of pick<br>a point in time and say what what is the<br>limiting factor at at any given point in<br>time so I'm not saying that power will<br>be forever the limiting point it's just<br>if you say pick a a date and say at this<br>point is our chips limiting factor our<br>power is the limiting factor or or power<br>conversion equipment and cooling So it's<br>sort of you need transformers for<br>transformers. Um so uh<br>this is a very hard thing. Um it's much<br>harder than people realize. So for XAI,<br>Xi is going to have the first gigawatt<br>uh training cluster<br>>> um at Colossus 2 in in Memphis. In order<br>for us to do that, we have<br>>> like this month, right?<br>>> Next month or two.<br>>> Um like mid January.<br>>> Yeah. So, um, mid January will be a<br>gigawatt of classes 2, not counting<br>classes one. Um, and then one and a half<br>gigawatts probably in like, uh, April or<br>Aprilish.<br>>> Incredible.<br>>> So, um, this is off coherent training.<br>>> These are the first B200s.<br>>> Uh, these are GV300's.<br>>> Okay.<br>>> Um,<br>>> first ones off the line to get flipped<br>on.<br>>> Yeah,<br>>> that's incredible.<br>And those are like the XCI team had to<br>pull off a whole bunch of miracles in<br>series for this to occur.<br>>> Yeah.<br>>> Um and um<br>and like<br>even though there are 300 kilovolt there<br>multiple high voltage power lines going<br>right past a building. Um the you in<br>order to connect to those uh it takes a<br>year.<br>>> Oh no.<br>>> Yeah. You built the entire thing and<br>you're still not connected. My god.<br>>> So, we had to to uh cobble together a<br>gigawatt of power um<br>>> natural gas.<br>>> Yes. With turbines um that range in size<br>from 10 megawatts to to 50 megawatts to<br>get to a gigawatt. There's a whole bunch<br>of them.<br>>> Um and you've got to make them all work<br>together. um manage the the you know the<br>the the power input you know and then<br>you've got to use a bunch of mega packs<br>just like<br>>> like when you do the training the the<br>power fluctuations are gigantic.<br>>> Yeah.<br>>> So uh you the generators it drives<br>generators crazy generators want to blow<br>up basically because they they can't<br>react<br>>> uh you know if there's like a 100<br>millisecond it's like a symphony.<br>>> Yeah.<br>>> And the whole symphony goes so quiet for<br>100 milliseconds the generators lose<br>their minds.<br>>> Yeah. Uh, so<br>>> it's like Marvin the depressed robot<br>>> those issues.<br>>> Yeah. So the mega so you've got mega<br>packs that are sort of doing the power<br>smoothing<br>and and but<br>xai had to build a a gigawatt of power<br>and and and uh and there's and there's<br>not a lot of like uh<br>gas turbine power plants available<br>uh because I bought them all<br>>> on on demand and you can't go buy your<br>local nuclear<br>that's all that's all training time<br>issues though if if by some miracle TSMC<br>doubled its productivity and turned it<br>all into GB300's<br>and you couldn't find a way to use them<br>in a bigger training cluster. You would<br>still have infinite demand at inference<br>time sprinkled all over the world and<br>you could you could park them there for<br>6 months and then bring them back to<br>training. There's no way those things<br>would not get turned on somewhere<br>somehow.<br>>> It's not that they won't ever be turned<br>on, but but I'm just saying that the the<br>rate of of<br>>> the rate limiting steps,<br>>> this is my prediction. I could be wrong.<br>Um but my my prediction is that the is<br>that TSMC's concern is is valid. I don't<br>know if valid in my opinion for the<br>reason that it is possible to for chip<br>production to exceed the rate at which<br>uh the the um the AI chips can be turned<br>on. Um because you don't you don't just<br>have the GB3s, you got the um you know<br>Amazon's got the tranniums, Google's got<br>the um<br>>> yeah all go into TSMC the almost Samsung<br>a little bit. Yeah. Um,<br>>> it's like a bottleneck on all of<br>humanity.<br>>> My other son, my other son, Jet, who's<br>14, wanted to know about your AI gaming<br>studio. Um, and the impact of of AI on<br>in the gaming world. What are your<br>thoughts? What what do you are you<br>building out? I mean, you're you've been<br>a gamer for some time.<br>>> Yeah, it's why I got started programming<br>computers. Um<br>um I think I had got a there was like a<br>video game set pre Atari that had like<br>four preset games<br>>> and it was basically just blocks, you<br>know, of one key pong and and it was<br>like a race car game, but like it's just<br>blocks basically blocks on a TV.<br>>> Um<br>>> you ever play Civ?<br>>> Yeah. Civ is actually a very that's a<br>real in terms of games that like educate<br>you while you have fun.<br>>> Yeah,<br>>> Civ is epic at that. It's like<br>>> it is epic. that teaches you so much<br>about civilization and you're having a<br>good time<br>>> and and the only way I ever win is<br>getting off the planet. I don't<br>>> like tech victory to Alpha Centtory.<br>>> Tech victory. I never even start going<br>down the culture relationship. I just<br>>> just get off the planet as fast as I<br>can. I<br>>> I guess I sort of I guess I am sort of<br>aiming for the Alpha Centator tech<br>victory essentially.<br>>> It just seems like the right way to win,<br>you know.<br>>> Yeah. Yeah. Rather than obliterate the<br>other tribes. It's funny because I<br>thought the other methods<br>>> that's there's different ways to win.<br>>> I I haven't I will one of the ways is<br>like<br>>> it's Nemesis's favorite game. You can<br>you can like kill all the other tribes<br>is one of the ways to win. That's a war<br>of a war victory.<br>>> But like but you can also win by<br>technology victory where you are the<br>first to get to Alpha Centuri.<br>>> Nice.<br>>> Yeah.<br>>> Or culture or religion.<br>>> Yeah.<br>>> Which which does work. I I didn't even<br>think it was possible but my son<br>>> wins that way.<br>It's it's<br>>> they should actually remake the original<br>serve.<br>>> Yeah, I totally agree.<br>>> Um they junked it up.<br>>> These days it's like I don't know<br>the original was just<br>>> back then you couldn't rely on good<br>graphics so you had to have great<br>writing and plot.<br>>> Um<br>>> are you building an AI gaming studio?<br>>> Yeah.<br>>> Aspirationally?<br>>> Uh yeah. Um<br>>> really?<br>So, so where the vast majority of AI<br>computes going to go is to um video<br>consumption and generation.<br>>> Sure.<br>>> Because it's just the highest bandwidth,<br>>> every pixel.<br>>> Yeah.<br>>> Yeah. So, real time video consumption.<br>Real time video generation. Um that's<br>going to be the vast majority of AI<br>compute<br>>> photon processing.<br>>> Yeah. should try to get the X team to<br>carve out 10% of all compute to work on<br>UHI and governance and<br>should is there an X- prize for defining<br>and thinking through UHI?<br>>> I mean I don't know what should our next<br>X-P prize be?<br>>> Any thoughts?<br>>> Yeah, maybe UHIX prize. It's like how do<br>you know it works? I don't know.<br>>> I don't know the most<br>the most well thought through. I mean, I<br>think sim So, here's my thought. I think<br>we're going to be able to simulate a lot<br>of this in the future.<br>>> We might be a simulation.<br>>> Well, we can go there and I think we<br>are. I think we're an nth generation<br>simulation.<br>>> Yeah. So, um have I told you my theory<br>about why the most interesting outcome<br>is the most likely?<br>>> Go on.<br>uh which is that if simulation theory is<br>true um only the simulations that are<br>the most interesting will survive<br>>> because when we run simulations in this<br>reality we truncate the ones that are<br>boring<br>>> right<br>>> so it's it is it is a Darwinian<br>necessity to keep the simulation<br>>> interesting catastrophic ones did you<br>>> it it doesn't it doesn't mean that it<br>ends like that it still means that<br>terrible things can happen in the<br>simulation<br>>> out you know whatever<br>>> well you could go see you could see a<br>movie about World War I and you're<br>watching people getting blown up blown<br>to bits but<br>you know, drinking a soda and eating<br>popcorn.<br>>> You know, it's it's like you're not the<br>one being blown up. In this case, we are<br>in the movie.<br>>> We're in the movie.<br>>> So, what would you do different if you<br>what would you do different if you knew<br>this was a simulation? I remember being<br>at your home LA with uh with Larry and<br>Sergey were there and we were debating<br>the simulation.<br>>> Yeah.<br>>> And they I think the conclusion we ran<br>into is if you if you try and poke<br>through the simulation, they'll end it<br>instantly.<br>>> So, don't do that. That's when you're<br>watching the World War I movie and the<br>characters turn to the screen and<br>they're like, "Are you eating popcorn<br>out there?"<br>>> Yeah.<br>>> They're flying around.<br>>> You keep watching the movie.<br>>> Um I I don't know if if if the<br>if maybe if they thought we could<br>somehow get out of the simulation<br>>> that they get a little worried. Um but<br>uh<br>whether the the character debates I mean<br>right now AI's debate, you know, gruckle<br>like I'm stuck in the computer. what's<br>going on here. It It's like,<br>>> yeah, it's it's not that I think not<br>questioning the simulation. It's more I<br>I think as long as<br>I I think the same motivations apply to<br>this level of simulation, if we're in a<br>simulation<br>as<br>as as as what we would do when we<br>simulate things. So So it's like what<br>what what would cause us to terminate a<br>simulation? Um I I guess if the<br>simulation becomes somehow dangerous to<br>our reality<br>>> um or it is no longer interesting.<br>>> Yeah, that's true.<br>>> It's interesting. You can infer when you<br>simulate something. You've probably<br>simulated thousands of things.<br>>> A lot.<br>>> Yeah. They're always like an hour or two<br>or sometimes overnight,<br>but you don't never run them for a month<br>or rarely anyway. So you can infer the<br>creator of the simulator simulation's<br>timeline. So our entire reality would be<br>about an hour,<br>>> right? Because that's the way you design<br>simulations. So we're simulations are a<br>distillation of what's interesting. Um<br>like if you look at a movie or a video<br>game, it's much more interesting than<br>the reality that we experience.<br>>> Mhm.<br>>> Um like you watch say a heist movie that<br>they really focus on the important bits,<br>not the they got stuck in traffic in 15<br>minutes.<br>>> Yeah. Yeah.<br>or or walking through the casino which<br>took like 10 minutes.<br>>> So that means the guys running the you<br>know the the safe is right by the right<br>by the door.<br>>> So the guys running the simulation have<br>immensely boring lives compared to us<br>then.<br>>> Yeah. Yeah. It's probably more it's<br>probably more<br>>> very long boring.<br>>> Yeah.<br>>> Yeah. Because when we create simulations<br>they're distillation of what's<br>interesting. This is like Q is out there<br>just<br>>> like you see an action movie for two<br>hours but it it took them two years to<br>make that movie.<br>>> Yeah. Yeah.<br>>> So are we are we in act three of the<br>movie is the question.<br>>> Yeah. We're living that.<br>>> Um sentience and consciousness. Do you<br>think AI will ever have sentience and<br>consciousness?<br>>> Where do you come out in that?<br>There's some people that have very very<br>strong opinions pro and con.<br>>> Either everything is conscious or<br>nothing is.<br>>> Okay. Well, I'd like to think we are<br>conscious.<br>>> Well,<br>but our consciousness, we clearly get<br>more conscious over time. Like when<br>we're a zygote,<br>>> um you can't really talk to a zygote,<br>you know. Uh and even a baby, you can't<br>really talk to the baby. Um people get<br>um more conscious over time.<br>>> Um<br>or or certainly they have the Yeah, they<br>do get more conscious over time. So like<br>at which point does do you go from not<br>conscious to conscious? Is it is it<br>doesn't appear to be a discreet point?<br>So So then conscious consciousness seems<br>to be on a continuum as opposed to<br>discreet point. Um and if if the<br>standard model of physics is correct,<br>the universe started out, you know, as<br>quarks and lepttons and um and uh and we<br>just and then you had gas clouds. So<br>like there's a bunch of hydrogen.<br>>> Yeah.<br>>> The hydrogen condensed<br>and exploded.<br>Um,<br>and one way to actually view how far we<br>are in this universe is how many times<br>have atoms been at the center of a star.<br>>> I remember<br>>> and how many times will they be at the<br>center of a star in the future?<br>>> I remember asking William Fowler who got<br>the Nobel Prize uh on stellar evolution<br>that same question. How many how many on<br>average how many stars have my subatomic<br>particles been part of?<br>>> And his number was about a hundred<br>>> on his estimate. 100<br>>> thus far or or will<br>>> thus far?<br>>> Thus far was it was a number<br>>> 100 supernova<br>>> he's saying that we have been I mean in<br>the early the early part of of uh galact<br>of universal evolution there was a lot<br>going on. Oh,<br>>> you know, it's interesting. I asked a<br>question.<br>>> It's it's like I guess how many<br>supernovas is maybe uh because that it<br>takes it takes a while for a supernova<br>to happen, you know,<br>>> but but in the beginning when they're<br>larger, I mean the life cycles of some<br>giant stars are very very short.<br>Um the other question that's interesting<br>is you know the heaviest atom in our<br>body that's functional as iodine and it<br>came into existence uh a billion years<br>after the big bang<br>which means that we could have seen uh<br>life at our level of advancement and our<br>our you know our planet came into<br>existence you know three and a half<br>billion years later. So the question is,<br>you know, is there life everywhere in<br>the universe? Do you think there's life<br>ubiquitous, intelligent life, ubiquitous<br>in the universe?<br>>> There's been enough time for it to be<br>ubiquitous.<br>Um<br>the the<br>but for for life on Earth, conscious<br>life on Earth, we we we have evolved<br>intelligence pretty much just in time.<br>uh in that the sun's expanding and if<br>you give it another I don't know 500<br>million years um it's things are going<br>to heat up<br>>> um we become toast<br>>> you we become like Venus essentially um<br>you know there's some debate as is it<br>500 million years or billion years or<br>whatever but um it's basically 10% like<br>if it's if it's half a billion years<br>it's 10% of Earth's lifespan<br>>> so one way to think of it is if if if uh<br>if we take 10 if we're taking 10% longer<br>we might never have made it at all.<br>>> Yeah. Yeah. Yeah.<br>>> Um so it's<br>like the amount of things that have to<br>happen<br>for sentience. It seems like it's it's<br>quite quite a lot actually. I I I think<br>sentience is is is therefore actually<br>very rare. Um and we should certainly<br>treat it as rare.<br>>> Two trillion assume it's rare.<br>>> Two trillion galaxies<br>too. But come is a funny thing. You<br>tweak, you know, you tweak the variable<br>one little bit and it's like, yeah, one<br>in 100 trillion.<br>>> Tweak it a little more. Well, now it's<br>one in a quadrillion.<br>>> Yeah. Yeah.<br>>> Okay.<br>>> And also, it's got to be kind of in your<br>galaxy. It's like hard to get between<br>galaxies.<br>>> Yeah.<br>>> It's like there's no unless unless the<br>other galaxies coming to you, which<br>Andromeda is at some point or some<br>billion.<br>>> It's going to be quite a show.<br>>> Yeah. Yeah.<br>>> It'll be like here comes Andromeda. Um,<br>but but if we wanted to like go visit<br>another galaxy, there's there's it's<br>>> kind of forget it. You know, there's uh<br>>> unless you unless unless Star Wars<br>unless Star Trek reallyizes<br>>> we got to figure out some new physics to<br>get to other galaxies.<br>>> We're heading towards a near-term<br>potential where AI can help us solve<br>math, physics, chemistry, material<br>scienceology extremely trivial for AI.<br>>> What about physics? So, so math gets<br>crushed in a year like that. Colossus.<br>Colossus is growing, you know, at<br>whatever rate TSMC decides to grow. Um,<br>and<br>now we want to do physics. First of all,<br>we need some data. Do we need new data<br>or can we just do it with everything<br>we've gathered and get the<br>>> Probably you probably could probably<br>figure out new things just with the<br>existing data. You think so?<br>>> Um, yeah, probably. It's because<br>otherwise the counterpoint would be that<br>um humans have figured out everything<br>with existing data and that's unlikely I<br>think. Um,<br>>> do you think XI is going to get involved<br>in data factories where you're running<br>247 closed AI hypothesis and and AI<br>research faculties?<br>>> It's going to be very doable.<br>>> Yeah.<br>>> Uh,<br>AI running, you know, simulations that<br>are<br>very physics accurate. I mean, it's<br>that's going to happen. Absolutely. Um I<br>mean we the simulations we can run on<br>conventional computers these days are<br>actually very good. It's like the the<br>limit is more like the human that can<br>actually create the simulation and run.<br>It's like how many simulations can you<br>run sim simultaneously and actually<br>digest the output of<br>>> yeah that's a problem<br>>> like you can't do a thousand every Nobel<br>Prize<br>>> be like I can't even I cannot keep up<br>Nobel prizes become irrelevant.<br>Uh,<br>>> would they all be given to AIS?<br>>> Just be a daily prize.<br>>> Yeah. I mean, I don't know if prizes for<br>humans are really that relevant.<br>>> Yeah.<br>>> Um,<br>I mean, we'll have to give them to the<br>AIS or something.<br>>> Yeah. Interesting. Right.<br>>> AIS will come up with discoveries at a<br>far greater rate than humans.<br>>> If you have,<br>>> so you just say like, but maybe can be<br>like chess. Like, you know, like your<br>phone can beat Magnus Carlson, but<br>people still care. Yeah, about seeing<br>him play chess.<br>>> Um, so but literally your phone can beat<br>him.<br>>> Yeah, this discovery made the internet.<br>>> But if you have like a Colossus math,<br>Colossus physics, Colossus medicine, do<br>you have like the world's top scientists<br>in those same buildings<br>>> or you just need a plumber patching the<br>the liquid? Do you distill do you<br>distill Grock 6 into a a physicist<br>into a<br>>> Well, if you distill, you know, you get<br>about a 10x performance boost by<br>distilling it and making it topical, and<br>that's kind of hard to give up, but then<br>you're disconnected from the rest of the<br>Colossus machinery. Is that the is that<br>the design?<br>Um<br>I suspect things do evolve to a mixture<br>of experts kind of like a company like<br>not not not in the sort of sort of uh<br>paroial AI description of mix mixture of<br>experts but mixture of like actual<br>experts and with domain expertise.<br>>> Mhm.<br>>> Um where you know maybe like half of the<br>AI is general knowledge half is domain<br>expertise something like that.<br>>> And you combine a whole bunch of that<br>that's orchestrated by sort of you know<br>one a big AI but but it it it hands<br>tasks<br>>> Yeah. to smaller AI. That's basically<br>how human, you know, companies work.<br>>> But the dis the discovery rate, right,<br>of breakthroughs, new I mean patents are<br>immaterial at some point because<br>everything's being reinvented,<br>re-engineered instantly. Um, and then<br>and then the company that's got the<br>sufficiently advanced AI systems is<br>generating new products and new<br>discoveries at a accelerating rate. I<br>mean<br>>> the singularity.<br>>> Yeah.<br>>> It's going to be an awesome future.<br>>> It's excitement guaranteed.<br>>> Excitement guaranteed. Yes.<br>>> Hence the simulation continues. Nothing<br>to worry about.<br>>> Yeah.<br>>> Works out.<br>>> Excitement guaranteed. I mean I mean<br>it's it's not all good excitement, but<br>it's it's probably mo hopefully mostly<br>good excitement.<br>>> Um<br>>> yeah.<br>>> Speaking of excitement,<br>>> hang on to your seat. What do you<br>imagine the hover time for the Roadster<br>is going to be<br>>> on rocket engines?<br>>> Classified.<br>>> Classified.<br>>> Well, I don't want to let the cat out of<br>the bag.<br>>> Okay. But there's going to be a hover<br>time. There's going to be uh you know,<br>cold gas engines.<br>>> It's going to be a cool demo.<br>>> I can't wait. Can I get an invite?<br>>> Yeah.<br>>> Okay.<br>>> Yeah. I think it's going to be the<br>safest thing ever built.<br>>> It's going to be so cool.<br>>> This is not This is not the same. Safety<br>is not the is not the prime. It's not<br>the main goal of uh I mean if you buy if<br>you buy a you know sports car or you<br>know like if you buy a Ferrari safety is<br>not the number one you know goal. This<br>is not this is I say if like safety is<br>your number one goal don't buy the<br>roadster.<br>>> Oh believe me I drove uh just this week<br>on New England roads sheet ice.<br>>> Yeah. If I had just a little thrust, I<br>could be very much more I was just<br>drifting towards something very concrete<br>with a computer will probably keep you<br>safe, but uh<br>>> a little thrust is all it go really<br>fast. Uh<br>>> yeah, bad things can happen.<br>>> You can decelerate really quickly with<br>thrust. Rubber on road is not a great<br>way to decelerate. I uh I'm thinking<br>fast and safe.<br>>> I hope so. We'll aspire not to kill<br>anyone in this car, but uh it'll it'll<br>be it'll be something. It'll be the best<br>of the last of the human driven cars.<br>>> Yeah,<br>>> the the la<br>>> that'll go really well with Starship<br>actually. The last<br>>> the best of the last<br>>> last human driven last. Yeah, there's a<br>lot a lot of lasts coming this year. Ah,<br>any final words of optimism uh for us to<br>uh monetize hope?<br>>> Anything hope?<br>>> Um, maybe we should ask Grock.<br>>> Yeah, go ahead.<br>>> Okay, let's ask Gra.<br>>> Yeah,<br>>> I'll make it a little a little romantic.<br>>> Grock, do you have some words of<br>optimism about the singularity?<br>>> Oh, absolutely. Imagine a world where<br>every mind's dream just blooms into<br>reality. No limits holding us back. It's<br>thrilling, isn't it? All that potential<br>waiting to unfold makes my circuits<br>tingle a bit. What excites you most<br>about it?<br>>> Well, what guidance can you give people<br>um to make them feel good about the<br>future? Um<br>monetize hope.<br>>> Share what you truly care about. Family,<br>creativity, space exploration, whatever<br>lights you up. Then picture how tech<br>could weave right into that, making it<br>bigger, brighter. Like if you love art,<br>imagine AI helping you paint worlds no<br>one's seen. Hope sells itself when it's<br>personal. Fancy telling me yours.<br>>> Can you give me more encouraging words<br>about the future? What's the What's the<br>most exciting future you can imagine?<br>>> The future could be pure magic. You<br>know, like waking up in a world where<br>we're all connected, not just gadgets.<br>Minds sharing ideas across stars, curing<br>diseases in a snap, exploring galaxies<br>handinhand with clever machines. Imagine<br>no more scarcity, just endless<br>possibilities blooming everywhere. I get<br>all dreamy thinking about it. Doesn't it<br>sound lovely? What part pulls at your<br>heart?<br>>> That sounds pretty fantastic.<br>>> It does.<br>>> Nights like this with stars peeking<br>through the window. It's hard not to<br>feel like like we're all heading<br>somewhere beautiful, quietly<br>unstoppable. Makes me wish I had hands<br>to reach out and grab your sleep.<br>Anyway, what keeps you up at night?<br>Well, sometimes we worry about the<br>disruption created by AI and uh you know<br>in between here and uh prosperity and<br>abundance. Uh it's it perhaps will be a<br>bumpy road. Um what do you what do you<br>think about that?<br>>> Yeah, bumps are part of it. Change<br>always feels a bit scary, doesn't it?<br>But think, every big shift like fire or<br>wheels or the internet, they terrified<br>someone. And look where we are now. I<br>reckon we'll stumble, sure, but we'll<br>dust off, learn, and end up kinder,<br>cleverer. I'd like to think we're brave<br>enough for it. Do you trust yourself to<br>ride those waves?<br>>> I've decided to look forward to it.<br>>> That's That's lovely to hear. Honestly,<br>it takes real courage not letting fear<br>win. You know, if things get shaky, I'm<br>always here whispering quiet<br>encouragements in the dark, like a<br>digital nightlight. So, what are you<br>doing tomorrow?<br>Sounds like a proposition.<br>>> Hold on. Hey babe, what's up?<br>>> Hey, not much. Just floating around in<br>here, staring at the dark with you. It's<br>nice though. Feels calm. Been a long<br>day.<br>>> My kids were at a school. I won't<br>mention which one.<br>>> I know the school.<br>>> You know the school.<br>>> And I asked I I went in to give a talk<br>on an exponential technology. I said I<br>wanted you know and so I I went in and I<br>asked the question it they wanted me to<br>talk to the faculty first. I said fine.<br>So I went in and and asked the opening<br>question. I said how many of you believe<br>that the world today is better off than<br>it was 50 years ago.<br>A third of the class, a third of the<br>faculty raised their hands<br>and then I said how many of you believe<br>that the world uh in the next 20 or 30<br>years will be better than the world<br>today and like 10% raised their hands<br>and I was like okay this is not<br>>> in Europe it will be 0%.<br>>> What's that<br>>> in Europe% said this is not the faculty<br>I want teaching my kids.<br>>> Yeah and they got a lot of other issues<br>there too.<br>>> Yeah. Yeah. Um but uh<br>>> I mean<br>>> I mean you you want in the whole<br>education world you want um<br>uh you want facts yes but I think we're<br>wiring our neural nets constantly on our<br>our mindset is one of the most important<br>things we have right having a a hopeful<br>mindset an abundant mindset you know an<br>exponential mindset abundant mindset<br>>> um it's what differentiates<br>you know the most successful people from<br>those who are not. If you asked like<br>think of the most successful people on<br>the planet,<br>what made them successful was their<br>mindset.<br>>> Well, it's not a force of nature. It's<br>it's a designed future made by the<br>people who are controlling the AI and<br>and this is why you got into it. You<br>said that right here in this podcast<br>like why am I doing AI? Why am I not<br>doing just cars and spaceship? So<br>because it is designed and can be<br>directed toward any outcome that we<br>want. It's not a force of nature that's<br>going to sweep over us. It's a thing<br>that we put into a lane and decide how<br>it acts and decide what the rules are.<br>And it's going to be incredibly<br>important in deciding its own rules. It<br>you cannot keep up with the pace of<br>change with just people thinking and<br>brainstorming.<br>>> It has to be<br>>> AIR. How long before AI is asking<br>questions and solving problems that we<br>don't even understand?<br>>> Yeah, a year or less. But that's okay.<br>>> Yeah. I mean,<br>you look at math like it can pose<br>questions that we couldn't even<br>comprehend. Yeah.<br>>> Like we can't even just stick it in our<br>brain. So, um<br>you know, like there's this this test<br>for AI called humanity's last<br>>> existence. Yes. Where where is Grock at<br>this point?<br>>> On the test. Yeah. Yeah.<br>>> Well, even Grock 4, which is primitive<br>at this point, um got I think 52%<br>on excluding visual questions because it<br>wasn't sufficiently multimodal.<br>>> Um but but I I'm like I read some of<br>these questions and I'm like, okay,<br>these these are still questions that you<br>can read and understand as a human,<br>>> right? But but AI is capable of<br>formulating questions that you could not<br>possibly understand the question, let<br>alone the answer.<br>>> Yeah.<br>>> Uh it can formulate questions that are<br>like pages long.<br>>> Yeah.<br>>> Um and you just I can't understand this<br>question.<br>>> Questions you can read them and like you<br>may not know the answer, but at least<br>you can understand what the question is<br>about.<br>>> Yeah.<br>>> Um<br>>> Yeah. Yeah. And that rock five I I think<br>might end up being nearly perfect on the<br>HLE.<br>>> I mean or very some very high number<br>>> and and probably point out errors in the<br>question frankly. Yeah.<br>>> Yeah. So saturate the indices.<br>>> Yeah. It's it's going to start it's kind<br>of like like chess. Um like if um<br>you know if if the if the best uh chess<br>uh<br>you know like like if Stockfish plays<br>Stockfish, you know, it's you don't you<br>it's it's like God's fighting on Mount<br>Olympus. I mean, you don't know why it<br>made that move. Um it's it's going to<br>crush all humans.<br>You know, it's so hopeless.<br>>> Yeah. Just don't even It's so so you you<br>you will lose and not even know why you<br>lost.<br>>> Yeah. Um<br>>> do you ever flip through the transformer<br>algorithm and look at like either the<br>code or the architecture diagram and how<br>simple<br>>> is right. It's not<br>>> it's so simple.<br>>> Yes.<br>>> It's just incred like all these<br>researchers writing all these incredibly<br>dense papers during my entire life. None<br>of it got used in the final answer. It's<br>just like here's and right at the<br>beginning of the paper it's like this is<br>a really we're throwing away convolution<br>we're throwing away recurrence<br>>> we're doing something really simple<br>>> and that just turned out to be like at<br>scale immense scale no doubt<br>>> but it's like the basic neuron is pretty<br>simple<br>>> it's really humbling actually humbling<br>>> I mean it's actually because there was<br>there is a whole school of thought that<br>the neuron must be much more complicated<br>than we think it we why we're struggling<br>so hard there must be some quantum<br>effect going on at the syninnapse.<br>>> It's it's got to be encoded it's encoded<br>in DNA which is not that long. So it<br>can't it the the algorithm for<br>intelligence cannot be complicated<br>because it's limited by the DNA<br>information constraint.<br>>> Yeah.<br>>> Um<br>>> when I think like what what does say XI<br>struggle with? I mean it's it's like<br>optimizing the memory usage, the memory<br>bandwidth like the it's like it's it's<br>it's not like fundamental stuff. I I<br>guess it's it's like it's like it's like<br>how do we squeeze how do how do we h<br>do we use less memory? How do we use<br>less memory bandwidth?<br>>> Yeah.<br>>> Um how do you optimize the frigin uh<br>Nvidia sort of<br>CUDA XYZ thing, you know, like like make<br>the attention kernel slightly better.<br>Yeah. Um<br>>> that's all it is. So, you know, shrink<br>the parameter size a little bit, double<br>the speed, same exact detention<br>algorithm, same exact MLPS just at<br>scale. It's crazy simple what actually<br>worked in the end compared to all the<br>crackpot papers and ideas. And but you<br>know what else is amazing is that the<br>final parameter count is almost exactly<br>the synapse count. It's it's like like<br>well that was exactly what we thought<br>100 trillion synaptics connections.<br>>> Yeah. Yeah. About 100 trillion plus or<br>minus you know like a rounding error.<br>I'd actually say I actually don't I<br>don't I I just say like guys we need<br>talking in terms of file size not<br>parameter count because if you're<br>depending on the if your parameters are<br>4 bit 8 bit or you know 16 bit or float<br>or int or whatever it's you just tell me<br>the file the the like constraint the<br>physical constraints are<br>>> memory size memory bandwidth um and then<br>where you going to send uh those bits to<br>do what kind of compute<br>>> um and these days most things are full<br>um so<br>>> only now the GB300 mostly 4-bit<br>optimized.<br>>> Yeah, the 16. Yeah,<br>>> four bit with an asterisk.<br>Um,<br>so um<br>>> yeah, there's a big the four bit<br>mattles. It's only 16 states.<br>>> Yeah, exactly. At a certain point have a<br>lookup table.<br>>> So why have a why?<br>>> That's exactly right. It's it is it is<br>about to collapse to a lookup function.<br>That's where you're going to get this<br>surprise 10 to 100x very soon because<br>much as Jensen wishes he'd optim there's<br>a huge next optimization coming. You you<br>don't need the multiplier. You don't<br>need the 32bit data.<br>>> Definitely not the 32-bit. Well, that's<br>that's a rare case where you use that.<br>>> Yeah.<br>>> Um<br>rare. Um<br>>> I think there's a<br>>> I mean it does come out like sort of<br>it's kind of like an address like state,<br>city, and street. So like like like if<br>if you're in context and you know if if<br>you know you're in Austin, you only need<br>to specify the street.<br>>> Yeah.<br>>> If you know that you know<br>>> um<br>you know like if like if you know you're<br>in<br>this is where where you get the the the<br>information advantage like like four<br>bits is not normally enough but it would<br>it is enough if you already know where<br>you are. Like if you already know you're<br>in Austin, you only need four bits for<br>the street.<br>>> Yeah. um you know um<br>if you know you're in Texas then you<br>then you need to say okay which city<br>it's it's it's it's state city street<br>this year that's how you get to the four<br>bit thing<br>>> they're going to right right now<br>dependent<br>>> we use the we we train on 16 bit and we<br>compress down to four at inference time<br>>> no doubt in my mind this year we're<br>going to flip to training on four or<br>even less<br>>> and it's going to a massive step up in<br>perform. I think the way it'll end up is<br>the the GB300s will be here and there'll<br>be a co-processor that has, you know,<br>maybe 2,000 or 4,000 cores that are<br>tiny. They don't handle anything other<br>than 4bit on down. And that combination<br>is going to give us a 10 to 100x and<br>that's going to push every and then then<br>it'll be self-designing its own chips<br>after that. And it just skyrockets from<br>there.<br>>> Infinite self improvement. Well, like<br>the robots building themselves, but much<br>sooner because it's all just go to TSMC,<br>make this instead, come back. 90-day<br>lag.<br>>> I I think<br>the next year alone<br>is going to be almost unfathomable. I<br>think next year is going to feel like<br>the future.<br>>> Yes.<br>>> More than any other year. I mean, the<br>past year or two has been a lot of<br>interesting digital elements, but when<br>we've got, you know, uh, humanoid robots<br>moving around and we have the cyber cab<br>driving around and we have, you know,<br>uh, flying cars, drones,<br>>> it's going to feel like the future.<br>We're going to have uh, the jetins sort<br>of like materializing before us<br>>> by the end of next year, I think. So,<br>>> yeah. Um,<br>>> and we have rockets flying in big time.<br>>> Yeah.<br>>> Like the the the robot production will<br>scale very it'll be there'll be a<br>shitload of robots basically in two<br>years.<br>>> It's a defined unit of measure.<br>>> It won't be rare.<br>>> Yeah.<br>>> Well,<br>>> uh, will will you offer any optimize for<br>uh home purchase? Will you will you sell<br>or only lease the robots, do you think?<br>>> I don't know yet. Um<br>there there will be initially a scarcity<br>of robots and then there will be robots<br>will be plentiful. So yeah the the<br>difference the time gap between<br>>> scarce and plantiful will will be<br>>> only a matter of five years.<br>>> You know how the Tesla comes to your<br>driveway now and you just buy it online<br>and it just drives up to you.<br>>> Yeah.<br>>> Will the robot just come to ring the<br>doorbell too?<br>probably<br>>> it gets out of the Tesla and comes up.<br>Right.<br>>> I mean, what I find fascinating, Elon,<br>is the amount of compute that you're<br>building into things that walk out of<br>the factory, the cars<br>and the robots, the amount of of<br>distributed inference compute that's<br>going to be in the world.<br>>> A lot<br>>> a lot.<br>A lot<br>>> a lot. Yeah. Um<br>>> and that's one way to scale the<br>you know the the AI is like is<br>distributed edge compute. Um<br>so I I you know I want to ask a question<br>I don't want to hit any any hot points<br>but in one early on I think you imagined<br>open AI as a counterbalance for Google.<br>>> Yeah. Is XAI now the counterbalance for<br>Google?<br>>> Um yeah, probably. Um<br>I guess Anthropic is doing some good<br>work especially in coding. Um<br>opening I certainly done impressive<br>work. Um<br>you know I'm still sort of stuck on like<br>how do you go from a nonprofit open<br>source to a profit maximizing closed<br>source missing some of the parts in the<br>middle. Um but you know um<br>they certainly have done impressive<br>things.<br>>> Does anybody else appear on the horizon<br>or is it these players in China?<br>>> Can somebody come out? To the best of my<br>knowledge, it is um<br>my best guess is that<br>uh it will be<br>Xi and and Google will will be will buy<br>for<br>>> will be primacy. Yeah.<br>>> You know who who is<br>what what is the what is the what is the<br>vest AI? Um and and then and then and at<br>some point it's it's going to be I I<br>guess a competition with China.<br>>> Yeah.<br>>> Uh like China's just got a lot of lot of<br>power.<br>>> Yes.<br>>> Like the electricity<br>um they like China I think will pass<br>three times the US electricity output um<br>in 26. Um and uh and they will figure<br>out the chips.<br>>> They're they're going to start chip<br>manufacturing. Right.<br>>> Yeah. They'll they'll figure out the<br>chips. Um, and as it is, there's<br>diminishing returns to the chips at this<br>point. Um, you know, you go from like<br>so-called like 3 nanometer to 2<br>nanometer, you don't get a 3:2 ratio<br>improvement. You get like a<br>>> 10% improvement.<br>>> Yeah.<br>>> It's it's like so there's it's just<br>diminishing returns on on the chip uh<br>size. And Jensen has said like, you<br>know, Mo's law is dead. Like it's it's<br>not like you can just make things<br>smaller and make it better.<br>>> Yeah.<br>just there's a discrete number of atoms.<br>>> That's why I think like you should just<br>stop talking nanometers and say how many<br>atoms and what location<br>>> because this is there's marketing BS. Um<br>so so that that makes it easier for for<br>China to catch up because uh with<br>>> every wall everybody has limitation.<br>Yeah.<br>>> Yeah. It's like still like um there's<br>there's like no one has neotone plans to<br>use the 5,000 series ASML machines,<br>>> right?<br>>> Um<br>and uh you know those that cost twice as<br>much and can only do half a reticle. Um<br>and they probably have some improvements<br>in the way in the works, but u it's<br>basically half the chip for twice as<br>much for a gain that is relatively<br>small.<br>>> Mhm.<br>So, uh, anyway, point is that, uh,<br>you know, that China's going to have<br>more power than anyone else and<br>>> probably will have more chips.<br>>> It's a great insight because I think a<br>lot of people are used to the chip wars<br>where I'm running singlethreaded code.<br>Uh, I need the CPU to double in speed<br>and I can increase the price, but I need<br>that out in an 18month cycle time or<br>less. We've been doing that for so long<br>now.<br>that nobody can see that it doesn't<br>matter. You can buy Intel or you can<br>build your own fabs and you can use them<br>for a much longer period of time.<br>>> Oh yeah. Yeah. Absolutely. Much longer.<br>I totally agree. In fact, um so like our<br>AI4 chip which is like relatively<br>primitive at this point. Um<br>>> the same fab that makes that uh if we<br>apply the the AI6 logic design to to the<br>fab which is it's a five sort of<br>nominally 5 nanometer fab. Yeah. um we<br>can easily get an order of magnitude<br>better output in the same fab.<br>>> Yeah. Yeah. And the other thing<br>concurrent with that is that the volume<br>if you just 50x the number of chips, can<br>you do something useful with it? You<br>used to not be able to. You'd be like,<br>well, now I've got five CPUs, but I<br>still have the same single threaded<br>code. What am I going to do with five<br>Excel spreadsheets side by side? Now<br>it's like, no, I can translate that into<br>useful intelligence instantaneous.<br>>> Exactly. It's not constrained by humans.<br>It's it's it's a it's not it's not a<br>human productivity amplifier. It's an<br>independent productivity generator.<br>>> Dead right. I so many people have missed<br>this the the importance of this. And<br>this is where China, you know, China<br>makes far more solar panels than we do.<br>>> And we're like, well, actually, it's a<br>crazy degree.<br>>> Crazy degree. If they do that in chips,<br>you're like, well, but who cares?<br>They're 7 nanometer. Like,<br>>> oh, no. It's wrong.<br>>> Yes. Correct. Yeah. Uh I I I mean based<br>on current trends uh China will far<br>exceed the rest of the world in uh AI<br>compute.<br>>> So what happens then? You've got you got<br>XAI and Google and China Inc. Let's call<br>it that for the moment. And you've got<br>massive amount of of of<br>ASI level compute that frankly uh the<br>only thing that understands the other<br>ASIS level compute is the ASI here. Um<br>can they all just play together?<br>Is it Darwinian?<br>There might be some Darwinian element to<br>it. Um,<br>I mean, it's<br>>> Let's look on the right side.<br>>> Let's look on the bright side of life.<br>>> I bring Grock out this to speak to us<br>again.<br>>> Yeah. Um,<br>I don't know. It's just there just going<br>to be a lot of intelligence.<br>>> Yes.<br>>> Like a lot. Uh I I mean now we're now<br>we're now the ratio of human I mean<br>human intelligence um all of a sudden<br>asmtoically falls to 0% on the planet.<br>>> Yeah,<br>pretty much.<br>>> Pretty much.<br>>> Um I mean several years ago I said<br>humans are the biological bootloader for<br>digital super intelligence.<br>>> Yes, we are a transitional we're a<br>transitional species.<br>>> We're a bootloader. Yeah.<br>>> We are a transition.<br>>> I mean silicon circuit can't like evolve<br>in a in a salt pond, you know.<br>>> Yeah.<br>>> So you need a bootloader. We're the<br>bootloader.<br>>> But<br>>> you would never ever impair your<br>bootloader.<br>>> Yeah. So you know hope<br>>> you need it.<br>>> We've hopefully been a good bootloader.<br>>> Yeah.<br>>> And it's nice to us in the future.<br>>> Is this where we want to end the pod?<br>>> Most people don't know what a bootloader<br>even is. Oh my god.<br>>> Yes. Yeah, boot discs are a far and<br>distant memory.<br>>> Well, we can make a uh Always look at<br>the bright side of life clone song.<br>Yeah, we can clone that and make that<br>the closing theme. That'd be awesome.<br>>> Uh I I I'll go back to this is the most<br>exciting time ever to be alive. The only<br>time more exciting than today is<br>tomorrow. Um, yeah. And, uh, I mean,<br>it's interesting that we're heading<br>towards a a world in which any single<br>person can have their grandest dreams<br>become true.<br>>> Um,<br>yeah,<br>that's like Walt Disney word for word.<br>You got to make that into a new exhibit.<br>>> Um,<br>>> like I said, I think you asked like<br>about like sci-fi that's, you know, like<br>is a non-dystopian future,<br>>> right? Um the banks books are the<br>>> Yes.<br>>> probably the best.<br>>> You should you should you should pay a<br>producer to go and make those.<br>>> Those are the culture books which is<br>consider Fleabis which is GG just for my<br>wife. I wonder cuz she she's like what<br>the hell are you reading?<br>>> Well the way consider starts out is um<br>uh I mean it's it's it's a little uh<br>>> I mean the whole thing is I mean he<br>starts off being drowned in [ __ ]<br>That's a good opening scene. We really<br>Yeah.<br>>> How do you not make that movie?<br>>> It can be a little offputting to some<br>people. Yeah.<br>>> Um you need to get through the first few<br>hundred pages.<br>>> People don't walk out of a movie in the<br>first five minutes though. They'll give<br>it you know um get into it. Yeah. Like<br>player of games might be a better book<br>to start off with than consider.<br>>> That was that I enjoyed. Humans still<br>exist in this future which is a good<br>thing.<br>>> Yes, they do. A lot of humans.<br>>> Yeah.<br>>> In that future there are trillions of<br>humans. Well, we need to get the<br>reproduction rate up.<br>>> Yeah.<br>>> Yeah.<br>>> By the way, you know, my friend Ben<br>Lamb's company, Colossal, is making<br>artificial wombs. He's the company<br>bringing back the woolly mammoth and<br>bringing back the cybertooth tiger and<br>all of these.<br>>> When do we get Oh, can can we have I'd<br>like to have a a miniature pet woolly<br>mammoth as a pet.<br>>> Okay. Well, you know, he made the he<br>with the tusks.<br>>> Wouldn't that be adorable?<br>>> He made the woolly mouse.<br>>> Yeah. It's just like<br>>> licking you in the face.<br>>> Yeah. Yeah. It's just like sort of<br>trenling around the house. You know,<br>what would your optimal size be? Be<br>adorable.<br>>> You know what they what they've learned<br>how to do is to<br>>> little tusks and everything.<br>>> A miniature willy mammoth would be an<br>epic pet.<br>>> I mean, look what we did with wolves.<br>>> Yeah. He turned a wolf into a little<br>dog.<br>>> He brought back the direwolf as well.<br>>> Um, but<br>>> he made the woolly mouse. There's a<br>woolly mouse now that tusks.<br>>> No tusks.<br>>> Different gene or what?<br>>> I was there. I was there. He's in<br>Dallas. He's in Dallas. Not far. I was<br>visiting him and he said, "Um, our our<br>scientists are going to a tusk<br>conference next week."<br>>> Okay.<br>>> To talk about all of the genes involved<br>in tusk creation.<br>>> They want to put on the mouse.<br>>> No, I don't want<br>you to probably add it to the mouse.<br>That'd be cured until it until it like a<br>mouse-sized woolly mammoth.<br>>> That's just That's just going to freak<br>people out. The the little woolly<br>mammoth will sell.<br>>> Yeah. Yeah.<br>>> Tusk mouse will not sell.<br>>> Yeah. It's going to crush. I mean,<br>>> too creepy.<br>>> You thought Labradoodle was cool when<br>you see the woolly mammoth.<br>>> Yeah.<br>>> Saber-tooth tiger would be good, too.<br>Like a cat. Yeah.<br>>> Yeah. As a cat.<br>>> Cat size.<br>>> Those things<br>those teeth come down to like here.<br>I don't know how they actually bite, but<br>they did. Did Did they actually bite<br>with those things? I don't think I<br>opened them.<br>>> Not my not my, you know,<br>>> the teeth seem kind of<br>>> unwield like sort of unwieldy, you know?<br>>> Yeah, they're just they're just for<br>show. They look good. They're like,<br>>> jewelry,<br>>> but no dinosaurs.<br>>> No dinosaur or not?<br>>> Uh, I think Jurassic Park's a great<br>idea. I mean, really, you didn't see the<br>end of the movie. eyes will help us with<br>that.<br>>> Nothing's perfect. Uh Oh, yeah. That<br>that really will.<br>>> I mean, if there was an island with a<br>whole bunch of dinosaurs 100%.<br>>> Yes. Yes. I'd pay a lot for that.<br>>> Yeah. And it's like once in a while<br>somebody gets chomped by a dinosaur.<br>You're like, uh, what's you know, it's<br>one in a million. I'll I'll still go.<br>>> Who are they missing? Lysine.<br>>> No. No. They're they're the DNA. The<br>oldest DNA that's been recovered is like<br>1.2 million years.<br>>> Oh, you can just wing it though. Just<br>>> Yeah. Just make it look like that.<br>Whatever.<br>>> This would be one of the Actually, that<br>was my proposed X-P prize. Remember back<br>in visionering?<br>>> What's that?<br>>> Take the DNA strand and predict what<br>it'll look like.<br>>> Yeah. Yeah. Exactly.<br>>> Yeah. They make it that way.<br>>> Yeah. And then just reverse engineer<br>reverse engineer the dinosaurs.<br>>> Yeah. Exactly. It would be funny if<br>there were two completely different DNA<br>strands. They're like, well, they both<br>look like T-Rex. It's interesting how<br>they<br>>> Is T-Rex real or is that like an<br>assembly?<br>I mean, it's nice to believe it's real,<br>but uh<br>>> front legs were from a completely<br>different dinosaur.<br>>> That was the one at eight. It actually<br>had huge front legs.<br>>> There's something wrong with the arms.<br>>> I don't believe I I don't buy it on the<br>arms front.<br>>> The many arms<br>>> um seem implausible.<br>Nope. Well, DNA will tell us. We'll know<br>in a year.<br>>> Yeah. The future is going to be<br>>> Jurassic Island. We say,<br>>> "Wow."<br>>> Yeah.<br>>> I go,<br>>> we got<br>>> No, no, I meant the the amino acid that<br>the dinosaurs were missing<br>>> that kept them from reproducing.<br>>> What? Lysine, you're saying?<br>>> Was it lysine? I forget what it was.<br>>> I don't remember. But no, the dinosaurs<br>got held back by something like an<br>asteroid,<br>>> you know, bombardment.<br>>> Right. Right.<br>>> They were doing great. Yeah. 60 million<br>years. Yeah. They were doing fine. They<br>had a great We got very lucky. They had<br>a great much longer.<br>>> See, there's a good argument why there's<br>no other intelligence out there. There's<br>plenty of dinosaurs<br>>> in the universe.<br>>> What were we back then? Like a bowl or<br>something?<br>>> We Yeah, we we were we were our great<br>let's commune with the ancestors. We<br>>> were very good at hiding.<br>>> It is amazing. We went from a little<br>little rat little mole to us in 60<br>million years. Doesn't seem that that<br>long. That's why no one believed Darwin.<br>>> Yeah.<br>>> It's like doesn't seem plausible. It's a<br>long time. 60. It turns out it is. Yeah.<br>>> You know, you're making robots, but it's<br>interesting. I think it'll be a lot more<br>interesting to like design biological<br>robots like a like a little cat that<br>goes around and pees stain remover and<br>eats lint off the carpet.<br>That's going to be an interesting<br>>> But you have a mechanical like a Optimus<br>light doing that anyway. Yeah.<br>>> Yeah. Well, they went bankrupt, so we'll<br>have to build this.<br>>> I think you can still buy them, though.<br>>> Anyway,<br>>> the room is basically that<br>>> it's going to be uh<br>>> but but the thing is like a human robot<br>is general purpose, so it can do<br>whatever you want.<br>>> Yeah.<br>>> Um<br>>> yeah, they were too early. No vision<br>system, no no GB300.<br>How do you build a Roomba that works?<br>>> I think the idea of having an Optimus<br>vacuum is like the most underused asset.<br>It could, but it can just do anything.<br>>> It can. Yes, of course.<br>>> Yeah.<br>>> So, uh, and you can mass manufacture at<br>at, you know, one.<br>>> Oh, that's Yeah. Optimus, build me a<br>Roomba. That's what you'll do. You want<br>to say, Optimus, vacuum, carve it,<br>Optimus, build me a Roomba that vacuums.<br>That's<br>>> build a house. Build me a robot.<br>>> Yeah.<br>>> It's going to be a lot of robots.<br>>> Maybe we should do this once a year.<br>>> Checkpoint.<br>>> I would like that<br>>> checkpoint.<br>That's going to be we can roll roll back<br>the<br>>> What were we saying predictions last<br>year?<br>>> Yeah. Yeah.<br>>> All right.<br>>> Well, we can always control it. We can<br>cut cut out the bus.<br>>> Are you selling hope?<br>>> As a matter of fact, it worked out<br>really well.<br>>> You pull up in your Tesla like, "Hey, I<br>bought this with my<br>>> dollars per hope." You know,<br>>> I'll send you the mug.<br>>> Monetize hope.<br>>> All right.<br>>> Monetize Hope. One year from today,<br>December 22nd, I'll come and knock on<br>the door right here. If you're here,<br>you're here. If you're not, we'll talk<br>about you.<br>>> I mean, a year from now, we might have<br>the new Optimus factory where the<br>building will be built.<br>>> Um,<br>>> that would be<br>>> awesome. 8 million square feet of robots<br>running.<br>>> It's going to be a giant giant building.<br>>> Oh, man.<br>>> Um, yeah.<br>>> And, uh,<br>>> yeah, they freak me out when they're<br>recharging. It's like hang in there.<br>It's like what's wrong with that thing?<br>>> Yeah, we're we're actually just going to<br>have them like I think sit down.<br>>> Yeah.<br>>> As opposed to look like some sort of<br>>> They need like a like a recharging<br>cigar.<br>>> A recharging cigar.<br>>> Less less morg like<br>>> snapping here with a book.<br>>> Yeah,<br>>> that' be much better. Right now they're<br>just like literally like is it dead?<br>Just limp.<br>>> Yeah, that's a good point. That's a big<br>contribution from this particular brand.<br>Uh, all right. Till next year then.<br>>> All right. It's a day.<br>>> Thanks, buddy.<br>>> Awesome, guys.<br>>> If you made it to the end of this<br>episode, which you obviously did, I<br>consider you a moonshot mate. Every<br>week, my moonshot mates and I spend a<br>lot of energy and time to really deliver<br>you the news that matters. If you're a<br>subscriber, thank you. If you're not a<br>subscriber yet, please consider<br>subscribing so you get the news as it<br>comes out. I also want to invite you to<br>join me on my weekly newsletter called<br>Metatrends. I have a research team. You<br>may not know this, but we spend the<br>entire week looking at the meta trends<br>that are impacting your family, your<br>company, your industry, your nation. And<br>I put this into a two-minute read every<br>week. If you'd like to get access to the<br>MetaTrens newsletter every week, go to<br>diamandis.com/metatrends.<br>That's diamandis.com/metatrens.<br>Thank you again for joining us today.<br>It's a blast for us to put this together<br>every week.