Elon Musk on AGI Timeline, US vs China, Job Markets, Clean Energy & Humanoid Robots | 220

31:26The energy output per year can double. with batteries. And do we have those batteries in development? Yeah, Tesla makes them. Okay, so you think the current Tesla battery packs. What do you think? I literally have, I went on stage and presented the thing. That's the dead giveaway.

31:48I even went to installations of the Mega Packs, you know, and there's... So why don't people do this? It's on the internet. Yeah. So do you think... They are. And China, by the way, is... It seems like China listens to everything I say and does it, basically. Or at least... Or they're just doing it independently. I don't know. But they're certainly making massive battery packs. Like, really massive battery pack output.

32:21They're making vast numbers of electric cars. Vast amounts of solar. I don't know. These are all things I said, you know, we should do here. Fundamental. Sure. When I fly over Santa Monica and LA, when I'm piloting and I look down, they're like, zero roofs have solar on them. Zero roofs. Yeah. I mean, it's not essential to have them on a roof. Okay. But it's a convenient place to have them. Yes. But the surface area of roofs is, I'm not saying it shouldn't, but it's, Tesla makes a solar roof which is the only solar roof that isn't ugly our solar roof actually looks beautiful but if you want to do solar at scale you just need more service area so we have vast empty deserts in America like if you fly from LA to New York or just fly cross country and you look down for a large portion of the time you look down, it is bleak desert.

33:27Yes. It looks like Mars, essentially. We're not worried about overpopulation there. No, I mean, there's barely a lizard alive in these scorching deserts. Yep. It's not like farmland we're talking about. We're just talking about places that look like Mars, like just scorched rock. So if we put solar where we currently have scorched rock, I think this will be a quality of life improvement for the lizards or the few creatures that live in this very difficult environment. Do we have the distribution network for that? It's like, Liz is going to be, thank God, some shade finally.

34:06Do we have the distribution network to be able to do that? Yeah, you'd need to materially affect quality of life. You'd need to capture and store, what, a couple hundred gigawatts? Is that in the realistic cards? You could just put the data center, I guess, locally there. Well, we already covered data centers. We're talking about the other. Yeah. I don't know. In an abundant world five years from now, massive amounts of compute, massive universal high income. I don't know what our data is. Universal, you can have whatever you want income. Yeah. Yeah. That's really what it amounts to. But in that world, other than compute energy, how much more energy do we need, 30%, 40%, 50 %?

34:47I don't know. Unless we want to move mountains around and make a ski mountain in the backyard. Um, cause I think the vast majority of energy consumption will go into compute. And then there may be use cases I'm not thinking of like, you know, the, well, you know, right here is a nice case study because manufacturing every one of these cars coming out at the rate of one, every minute or two, uh, is less energy than the data center that's training the cars to drive, to self-drive. Yes. So that's a good little case study. and we don't need that much more physical energy for abundant happiness. We need more compute energy.

35:24Well, the sun is just generating vast amounts of energy all the time for free that just goes into space. So I think we'll end up trying to capture, I don't know, a millionth of it, like a millionth, a thousandth of the sun's energy. We're currently, I'm not sure the exact number, but we're, I don't know, we're probably at 1 %-ish of Kharachev level one. Fair enough. Yeah. I would guess that evens. That's high. I'm just saying. We have a long way to go. That's being optimistic. Hopefully we're not 0.1%, but I don't think we're 10%. I'm just trying to get it to an order of magnitude. so pull it like we're roughly 1 % of the apparently using 1 % of the energy that we could use on earth I think the bottom line from a first principles thinking for the public is there's a lot of energy out there and we have it in the US we have it on the planet and it needs to be captured and the tech to capture it is here and improving every year yes there's not going to be some energy crisis.

36:39There'll be a large forcing function to harness more energy, but we're not going to run out of it. All right. I want to talk about education. So here's the numbers. They're abysmal. I mean, they're abysmal, right? The importance of college in the United States back in 2010, 75 % of Americans said it's important to go to college. That number is now down at 35%. All right. College graduates as a group turn out to be the group that's out of work the longest. Right. Just barely. And but still tuition has increased 900 percent since 1983. Yeah. The administrative expenses at universities have gotten out of control.

37:27Yeah. So I think I saw some stat that like there's one administrator for every two students at Brown or something like that. And I'm like, this seems a little high. They should teach something. Yeah, yeah. What was your college journey? I went to college in Canada for a couple of years at Queen's University.

37:51So I had Canadian citizenship through my mom, who was born in Canada. And my grandfather was actually American, but for some reason, I don't know, my mom couldn't get U.S. citizenship. But she was born in Canada, so I got Canadian citizenship. And I didn't have any money. so I could only go to a Canadian university at first. People forget that about you. You didn't have this giant social network or huge amount of wealth coming into all of this. No. No, I arrived in Montreal at age 17 with I think around$2 ,500 in Canadian traveler's checks back when traveler's checks were a thing. And one bag of books and one bag of clothes.

38:32That was my starting point. That was my spawn point in North America. and then so I went to Queen's University for a couple years and then University of Pennsylvania did a dual degree in physics and economics and graduated undergraduate at UPenn Wharton yeah and then I came out to do I was going to do a PhD at Stanford working on energy storage technologies for electric vehicles potentially material science I guess fundamentally. The idea that I had was to try to create a capacitor with enough energy density that you get high range in an electric car. It's funny, I invested in an ultra capacitor company.

39:21It didn't go well. Well, it's one of those things where you could definitely get a PhD, but it wasn't clear that you could make a company or do something useful. Most PhDs, I mean, I hate it, but most PhDs do not turn into something that's going to... Do not turn into something useful. Like you could add a leaf to the tree of knowledge, but it's not necessarily a useful leaf. Enormous fraction of great entrepreneurs are dropping out of grad school or undergrad. But nowadays, the sense of urgency is off the charts. I mean, they're popping out everywhere. Yeah, because don't waste your time going to grad school.

39:57Start a company. Curriculum is nowhere near caught up to what's actually going on in technology, and I don't have time. And we talk about this all the time. It's like, you know, this is the moment. I think right now it's not clear to me why somebody would be in college right now unless they want the social experience. Yeah. I mean, if you have the ability to go and build something. So the question is, how would you redesign the educational program, if I could be so blunt, as to create more Elon Musks? If we want to create an Elon Musk factory of people who start with very little, but are able to drive and drive breakthroughs.

40:37What's involved there? What drove you? Curiosity about the nature of the universe. So I'm just curious about the meaning of life and, you know, what is this reality that we live in? How early? my son Dax wanted to know what was it like for you in middle school and high school he's 14 years old he's in that age range now well I found school to be quite painful and it was very boring and in South Africa it was very violent so it's like it was like it was like that book Ender's Game but in reality IRL Ender's Game IRL It was like, but not as fun. So your goal was escape? Yes. Escape from the prison.

41:32So that's a question I have. Do you think that... It was miserable. Do you think most successful people have had a lot of hardship early in life? Do you need to have that level of hardship? Probably need a little bit of hardship, I suppose. Yeah. And then it's always tricky. Like, what are you supposed to do with your kids? You know, create artificial adversity? That's cruel. Do you have an answer? That's a Warren Buffett topic, actually. What do you do? It's not easy to create artificial adversity because if you love your kids, you don't want to do that. Yeah, that's for sure. So I had a lot of adversity.

42:10Probably it was good. Probably helped somewhat, I suppose. What doesn't kill you makes you stronger type of thing. At least I didn't lose a limb. I think what doesn't maim you?

42:2910 fingers. I can modify that a little bit. Can I ask you a question? Mamey makes you stronger. For the last five years, I've been helping teach this class, Foundations of AI Ventures at MIT. And every year when you survey the students, they go up a lot in their desire to start a company. And so it's now up to 80 % of the incoming. It's just going to be like one person company. Well, that's viable, I guess. But no, they want to co-found. Yeah, they don't want to be the founder. They want to be part of a founding team. So it still works out. But when Peter and I were in school at MIT, it was, I'm guessing, maybe 10%.

43:09They all wanted to be PhDs. And they've been doing the survey. I didn't know anyone who wanted to start. I mean, I don't remember any conversations with people saying they wanted to start. Even at Stanford at the time? I actually, a few days into the semester, or I should say the quarter, I called Bill Nix, who was at the material science department, and said, I'd like to just put it on deferment. He said, is my class that bad? No, and he said, that's okay, you can put it on deferment. But he said, this is probably the last conversation we'll have. And he was right. but then like last I think it was last year he sent me a letter saying that all of my predictions about lithium-ion batteries came true which was very nice and did he also say you could still come back and finish your PhD yeah no it was several times Stanford has said that I can come back for free well so you know what happened at MIT is every time I did not know it'd be a great use your time exactly i don't like uh so every time an iron man movie came out it notched up another probably 10 or so okay uh in terms of everybody wanted to be tony stark and so that's the image and i didn't know till today that the new tony stark the modern iron man yeah i always thought tony stark was modeled on charles stark draper and howard hughes as charles stark draper's education and his scientific endeavors married with Howard Hughes's ambition.

44:38And that created the original character. But then when Robert Downey Jr. wanted to reinvent it, it's modeled on Elon. Yeah. He came back with me. This is a Grokipedia fact. All right. Yeah. Fantastic. Yeah. I talked to Jon Favreau and Robert Downey Jr. I like the name Grok. I would like Jarvis as well. Yeah. Yeah. Yeah. Probably some some trade at some point, if Grok gets good enough, we're going to call it Encyclopedia Galactica. Yes, that's nice. Yeah. Yeah, of course. 42. Thank you. So going back to education, should colleges, I guess the social experience, you said, is important there. But what would you do for education, you know, middle, high school?

45:25You just came back from an announcement with President Bukele, who's a friend. I think he's an amazing, amazing visionary. Incredible what he did with his nation. Yeah. Yeah. Remarkable. Remarkable. And gutsy. Yeah. I was like, how are you still alive? Yeah. I mean, it was like it's the nuclear. It was a nuclear option. Right. Shut them down. I mean, do you know how, besides putting everybody with a gang sign in jail, I don't know if you know the second thing he did. He went to all of the graves of all the gang members out there and destroyed the graves and said, your memory will not be remembered in this nation.

46:08That's just badass. And it worked. I mean, you have to be a badass motherfucker to take on all the knocker gangs in one. And live. Yeah, and still be alive. And live. He's got a great, great guard at his palace there. But what did you announce with him in El Salvador? It was just basically to use Grok for education, like personalized education. Hopefully not the vulgar version of it. Yeah, we would have like, you know, the kids-friendly version of Grok. But obviously, AI can be an individualized teacher that is infinitely patient and answers all your questions. You still need to be curious and you still need to want to learn.

47:02You know, Grock can't make you want to learn. It can make learning more interesting. It could probably gamify and incentivize it, right? It can make learning more interesting and less of a production line.

47:17So, but kids do need to have to, if they need to want to learn, you know. Yeah. And like people should just think of the brain as a biological computer. It's a neural net. Yeah, it's a biological computer with a number of neurons and neural efficiency.

47:42What you can't do is turn any arbitrary key into Einstein. This is not realistic because Einstein had a very good meat computer, like an outstanding meat computer. so you can't just do Shakespeare, Newton, Einstein type of thing unless the meat computer is an exceptional one so what do you think so when people say we need to solve education in the United States because it's fundamentally broken I think what's really broken, I'm curious is the old social contract that says do well in high school, get into good college, get a degree and then get a job. And I don't know that that's going to be valid in the future.

48:28We talk about this on the pod a lot, that the career of the future isn't getting a job. It's being an entrepreneur. It's finding a problem and solving it. Yeah. Do you agree with that? Right now, I'd say people should just, you know, go to school for the social experience. Use more AI.

48:49the conventional schooling experience I think could be a lot better what we're going to do in El Salvador and hopefully other places just have individualized teachers that's going to be much better and you could go to you could go to a school with a bunch of other kids I guess if you want to hang out with other kids but you don't need to right you could do it on your phone at home so that's why I say like at this point education is a social experience when I talk to my kids who are in college yeah they they do recognize that they can learn just as much independently in fact that they would learn more in a work situation yeah they're there for the social experience and to be about around a bunch of people of their own age sort of a coming-of-age social experience sure sure being on your own learning how to lead or defend yourself, as the case may be.

49:48Well, yeah. I mean, if you join the workforce, you're, you know, from the perspective of like a 19-year-old with a bunch of old people. And if you're doing engineering with a bunch of middle-aged dudes, it's like, do you really want to do that or do you want to hang out with, you know, whether there's at least some girls your age type of thing? I want to get back to this when we talk about - Not a lot of other choices, actually. I want to get back to this when we get to universal high income. But I want to talk about health and longevity in one second. The U.S. is ranked number one in health expenses worldwide.

50:25And it's ranked 70th in health span. Oh, really? 70th? 70th. Is that accurate? It's great. It sounds low. I think we'd be better than 70th. For healthspan. Yeah, yeah. Well, whatever. So this is like we just get fat or something. We're not the top 10. Maybe a Zampik can help us find the rankings there. So you... Would you just run around? We need Cupid with a Zampik.

50:59Majora Cupid. But I think that's a big reason. It's like if people get really fat, then their health gets bad. Yeah. Well, if they don't have any exercise, they'll get bad. Or if they eat donuts for breakfast every morning. Are you still doing that? No, actually, I'm not. Okay, that's good. First of all, I wasn't eating a lot of donut. I was trying to have 0.4 of a donut, which rounds down to zero.

51:28So I figured anything below 0.44 of a donut rounds down to zero. So you and I have had a disagreement on longevity. A little bit. Yeah. I was saying, you know, we should push to get people to 120, 150. And you were saying people, you know, shouldn't live that long. So how long do you want? Yeah. You know, there's some, you know, people in the world that have done some bad things. How long do you want them to live? Yeah. Well, it's okay. They can be contained to us. This is a serious question, though. Them. A lot of things are going to happen that we don't. One thing that you said was interesting.

52:07You said, we need people to die so people change their minds. Oh, yes. People don't change their minds, they just die. So that makes more sense, actually. My response to that, Elon, was, you know, my response to that was the head of GM didn't have to die for Tesla to come along, and Lockheed and Northrop and Boeing didn't have to go away for, I mean, in a meritocracy, the better ideas will dominate.

52:35so i'm hoping that i can get you back onto the longevity train so there's a lot going on longevity right now right uh like what well david sinclair is about to start his epigenetic reprogramming trials in humans it's worked in in animals and and non-human primates it's going into how to use this like a pole or an injection right now it's an injection of an adeno-associated virus It's the three Yamanaka factors. Okay. We've got a$101 million HealthSpan XPRIZE that's working on, 730 teams working on reversing the age of your brain, immune system, and muscle by 20 years. By the way, do you know why it's$101 million?

53:17No. Because the primary funder, when they found out your carbon XPRIZE was$100, he wanted to make it bigger, so it's$101. Who is the cigarette? It was Chip Wilson from Lululemon. Oh, okay. And then, and then have evolution out of, but Chip said, can we make it bigger? I said, you put the extra million in, we'll make 101 million. Sounds good. Good story. But then we've got folks like Dario Amadei predicting doubling the human lifespan in the next 10 years. So that's probably correct. Okay, great. I don't know about doubling, but significant increase. Which is easily escape velocity. I mean, yeah.

53:55It's going to hold you, yeah. Oh, yeah, for sure. Effective age, yeah. So, I mean, I think that for... Too much to turn into a baby or something. That's what I'm telling all the students. It's like, Peter, what happened? Go, go, caca. Yes, yes. There is a frozen in time. You've got a zero wrong in the dosage.

54:22Just a small factor of time. You'll grow out of it. It'll be fine. Exactly. You won't remember it. Literally. I mean, wouldn't it be funny if we do this in like 10 years? Okay, we should do it in 10 years. Okay. We'll do it in 10 years for sure. And let's see if we look younger. That's a good side bet. My comment was always, at least Elon, back then, Elon was like, you know, late 40s. Wait till he gets into his 60s. He's going to want, you know, lunchtime anymore. I mean, I want things to not hurt. yeah sure it's like basically it seems like it's only a matter of time before you get back pain it's a when not an if when your back goes arthritis all those things suck being able to sleep through the night without going to the bathroom for the lot how much for that one more than hope that one oh man that's like the infinite money one why did you invest in longevity so I could sleep through the night and not go to the bathroom bladder yeah duration admittedly if you have to wear adult diapers that's a bummer that's not good adult diapers are real you know it's like one of the signs that a country is not on the right path.

55:57It's when the adult diapers exceed the baby diapers. Yeah, we're there. I think we're there. South Korea will be there. No, they passed that point. They passed that point many years ago. Japan passed the point many years ago. Doesn't go well, looking at the Japanese economy. No, I mean, like, South Korea is like 0.7, 0.6. Yeah, one-third replacement rate. Yeah. Isn't that crazy? Yeah, so three generations, they're going to be 1.27, so 3 % of their current size. I mean, North Korea won't need to invade. They can just walk across. Yeah, yeah.

56:29This is going to be some people in, you know, walkers or something. There will be a bunch of optimists. And robots, basically. But, you know, you've been very verbal about the, you know, the not overpopulation but massive underpopulation. Yeah, I've been saying this for ages. Yeah, longevity is going to be an important part of that solution. I also think, by the way, if you increased the productive life of most Americans by just a few years, you'd flip the entire economics here. Well, AI and robots is going to make everything free, basically. Well, how long would you want to live? I want to go, you know, other planetary systems.

57:15I want to go and explore the universe. Yeah, I mean, you know, I would like to double my lifespan for sure. I don't want, you know, I'm not sure I wanted to talk about immortality, but, you know, at least 120, 150. It's a long time. One of the worst curses possible would be that. Yes, may you live forever. May you live forever. Yeah. That would be one of the worst curses you could possibly give anyone. But I think life's going to get very interesting. Yeah. Far more. We're going to speed run Star Trek, as my partner, Alex Wiesner-Gro says. Yeah. Yeah. yeah speed running star trek would be cool yeah um well at a minimum your kids will have infinite life expectancy if you're talking about escape velocity if you can double lifespan there's it's not even close you're you're clearly past longevity escape velocity they the idea of 50 years of ai improvement yeah it's like i mean we're gonna have that in 20 years i don't know i've got too many fish to fry so i invited this is something by the way that i that i think i just i think it's very obviously other people think this too but I've long thought that like longevity or semi-immortality is an extremely solvable problem.

58:23I don't think it's a particularly hard problem.

58:29I mean, when you consider the fact that your body is extremely synchronized in its age, the clock must be incredibly obvious. nobody has an old left arm and a young right arm. Why is that? No. What's keeping them all in sync? Yeah. You're programmed to die is the way, you're programmed to die. And so if you change the program, you will live longer. And we've got, you know, species of the bowhead whale can live for 200 years, the Greenland shark can live for 500 years. And when I learned that, I said, why can they, why can't we? And I said, it's either a hardware problem or a software problem. And we're going to have the tech to solve that.

59:16And I do believe that it's this next decade. So the important thing is not to die from something stupid before the solutions come. You know, I invited you to - In retrospect, the solution to longevity will seem obvious. Yeah. Extremely obvious. I think the thing worth working on, and Peter's going to work on this anyway, but the thing to work on is exactly what you said. If old ideas don't, calcified old ideas don't just die off, add that to the pile of things we need to think about today. Because there are a whole host of other AI-related things we need to think about today. Let me finish on the longevity point one second.

59:55Elon, I want to invite you again. So there's a company called Fountain Life that created with Tony Robbins, Bob Hoorie, bill cap and we do a 200 gigabyte upload of you everything knowable about you full genome full all imaging everything right president bukele and the first lady came through called it an amazing 10 out of 10 experience um i think i don't want you to pull a steve jobs and kick the bucket because of some because i'm something they didn't know i mean so if you ask yourself curable cancer do you actually know what's going on inside your body right now um i did uh an mri recently and submitted to Grok and it didn't need no but that's none of the doctors nor Grok found anything but that's a fraction of the information right yeah i mean it's your full genome your microbiome metabolome everything and okay it's possible don't call me what's that don't call me bro we have a set we have a center in your water bottle

1:01:00God damn it. Too late. Sorry. It's already in the works.

1:01:09So can you go through the rationale of UHI? How does universal high income work? Okay, so there's going to be more intelligence, digital intelligence, than all human intelligence combined. and more humanoid robots than all humans. And assuming we're in a benign scenario, Star Trek, sort of Roddenberry, not Cameron situation. Yeah. Poor Jim. Yeah. I mean, I guess it's important to have these sort of... Counterpoints. Yeah. Let's not go in that direction. Dang. so the robots are going to just do whatever you want all the blue collar labor is being done by robots, all data centers are being built by robots the white collar labor will be the first to go because until you can move atoms the thing that can be replaced first is anything that involves just digital, if it's digital like if it involves tapping keys on a keyboard and moving a mouse, the computer can do that.

1:02:24The AI can do that. You need the humanoid robots to shape atoms. So if all you're doing is changing bits of information, which is white-collar work, that is the first thing that AI will be able to replace. This is the inspirational part of the podcast, by the way. When is all white-collar work gone? By when? Well, there's a lot of inertia. So even with With AI at its current state, I'd say you're pretty close to being able to replace half of all jobs. And you know that we're... White-collar jobs. That includes anything like education too. So anything that involves information. And anything short of shaping atoms, AI can do probably half or more of those jobs right now.

1:03:19Sure. But there's a lot of inertia. People just keep doing the same thing for quite some time. And there actually has to be a company that makes more use of AI that competes with a company that makes less use of AI, creating a forcing function for increased use of AI. Otherwise, the company that still has humans do things that AI can do will still continue to exist. being a computer used to be a job so it used to be that a human computer like what no like yeah a computer being a computer was a job you would compute numbers sure it didn't it didn't used to be a machine it used to be a job description you know um and there you can look online there's these pictures of like where they're having like skyscrapers full of women copying mostly women copying from ledger to ledger and then too but but yeah but people um um it was a lot of women but there's there were there were just buildings full of uh people just at desks doing calculations yeah um so they'd be calculating the interest in your bank account or um you know some um you know science uh experiment or something like that or whatever.

1:04:40But if you want calculations done, people would do it. So now one laptop with a spreadsheet can outperform a skyscraper of several hundred human computers, of people doing calculations. Now if even a few cells in that spreadsheet were done manually, it you would not be able to compete with a spreadsheet that was entirely a computer. Yeah. What this means is that companies that are entirely AI will demolish companies that are not. Right. It won't be a contest. Agreed. That's an analogy. That flipping... Yeah, just one cell in that... Just one... I'm going to do that on a flight home, actually. Would you want even one cell in your spreadsheet to be manually calculated?

1:05:36Yeah. That would be the most annoying sell. And you're like, goddammit. And gets it wrong a bunch of the time. Introduce the error rate. So this flippening, if you would. The flippening. The quickening. Are we monetizing hope effectively? Yes. Not at this moment. I think we're at peak doom. For people worried about the future of their jobs, we're at peak doom. We're going to do that. I'll send you the t-shirt. And the mug. And the mug. And the mug. And the mug. And the mug. Yes. And the mug. So wrong. Thank you. But you have a solution to this, which is UHI. Yes. Everyone can have whatever they want.

1:06:22So how does that work? How does UHI work? It's a good question. We have to figure out some transition. I mean, it's not a redistribution. It's a bumpy road. Yeah. I mean, so my concern isn't the long run. it's the next three to seven years yes the transition will be bumpy because we humans don't like simultaneously yes we'll have radical change social unrest and immense prosperity and you can buy all all the cyber trucks you want things are gonna get very cheap yes so this is actually in fact if this doesn't happen we'd go bankrupt as a country. So the national debt is enormous. The interest on national debt exceeds not just the military budget, but the military budget, I think, plus Medicare or Medicaid, one of the two.

1:07:19It's like one point something trillion of interest. Which is growing. Yes, and the deficit is growing.

1:07:32But, so if we don't have AI and robots, we're all going to go bankrupt and we're headed for economic doom. We're going back. So it's like, there's also competitive pressure from China. So this is definitely going to happen, I guess. We're going back to the theme of this talk, how can AI and exponential tech save America and the world? Don't you think that... But I want to hit this because we talked about... I was quite pessimistic about it and ultimately I decided to be fatalistic. and look on the bright side. I've got this image of you with the sunset. Always look on the bright side of life. You're sitting down the yellow wood road.

1:08:12We're crucified. Look on the bright side. But this is not about taxation and redistribution. No, it's... So how does it work? Reason through it with me. Listen, by the way, I'm open to ideas here. Okay. So it's not like I got this all figured out. All right. So I'm wondering if instead of universal high income, if it's universal high stuff and services. Yes. The UHSS. I guess, okay, this is my guess for how things play out. And by the way, this is going to be a bumpy ride and it's not like I know the answers here. but I have decided to look on the bright side and I'd like to thank you guys for being an inspiration in this regard thank you, appreciate that happy to help

1:09:08I actually think it is better to be an optimist and wrong than a pessimist and right for quality of life it's also not a force of nature to me it's really clear that we don't have any system right now to make this go well. But AI is a critical part of making it go well. And at some point, Grok is going to be addressing this exact topic that we're talking about. It has to be one of the big four AI machines that is dealing with it. There's no velocity knob, right? There's no on-off switch. It is coming and accelerating. I call AI and robotics the supersonic tsunami. Yes. Which maybe is a little alarming.

1:09:55I don't think. It's good. It's good. Well, because the world needs a wake-up call. This is important for folks to grok because I don't want to leave people depressed. I want people to understand what's coming. So we're basically demonetizing everything. I mean, labor becomes the cost of capex and electricity. AI is basically intelligence available at a de minimis price. So you're able to produce almost anything. Things get down to basic costs of materials and electricity, right? So people can have whatever stuff they want, whatever services they need. it's not when we say universal high income it sounds like it's a tax and redistribute but that's not the case um it's it's i think my best guess for how this will manifest is that prices will become prices will drop yeah so as the efficiency of production or the provision of services drops um prices will drop i mean you know prices in in dollar terms are the ratio between the output of goods and services and the money supply.

1:11:15Sure. So if your output of goods and services increases faster than the money supply, you will have deflation or vice versa. It's a good thing we're growing the money supply so quickly then. Right. Well, luck. Yes. That's why I came to like, let's not worry about growing the money supply, it won't matter. Because the output of goods and services actually will grow faster than the money supply. And I think we'll be in this, and this is a prediction, I think some others have made, but I will add to it, which is that I think governments will actually be pushing to increase money supply. Faster? Yes.

1:11:55They won't be able to waste the money fast enough, which is saying something for government. Isn't it crazy how close those timelines just randomly worked out? I mean, at the rate, because we're expanding the national debt, not because we're anticipating AI. We were going to do that no matter what. Yes. It's like right on the edge of becoming Argentina. But yeah, so productivity is going to improve dramatically. And it is improving dramatically. I think we'll see, I think we may see like high double digit output of goods and services. We have to be a little careful about how economists measure things.

1:12:30Yes. GDP sucks isn't measuring. And yeah. I mean, it's like my favorite joke. I have a few economist jokes that I like. But maybe my favorite one economist joke is two economists are going for a walk in the forest. And they come across a pile of shit. And one economist says, I'll pay you$100 to eat a pile of shit. I've heard this one. This is great. Go ahead. And so the guy takes$100 and eats the shit. Then they keep walking. they come across another pile of shit and the other guy says okay I'll give you a hundred bucks to eat a pile of shit so he gives him a hundred bucks and then the guys can say wait a second we both have the same amount of money and we both ate a pile of shit oh my god it sounds like we increased the economy by$200 this is the kind of bullshit you get in economics So if you say like just the output of goods and services will be much greater.

1:13:46So profitability of companies go through the roof at some point. So the question becomes is that taxed by the government and redistributed as some level of income as a UHI or UBI? In other words, one of the questions is if in fact this future we hit massive productivity and massive profitability because we're dividing by zero. The cost of labor has gone to nothing. The cost of intelligence has gone to nothing. We're still producing products and services faster and faster. So there's more profitability. Someone needs to be buying it and someone needs to be able to have the capital to buy it.

1:14:24I mean, this is an important question to get thought through. Yeah. Well, one side recommendation I have is don't worry about squirreling money away for retirement in 10 or 20 years. It won't matter. No. Okay. Either we're not going to be here or... It just... You won't need to save for retirement. If any of the things that we've said are true, saving for retirement will be irrelevant. The services will be there to support you. You'll have the home. You'll have the health care. You'll have the entertainment. The way this unfolds is fundamentally impossible to predict because of self-improvement of the AI and the accelerating timeline.

1:15:09Yeah, it's called singularity for a reason. Yeah, exactly. I don't know what happens after the event horizon. Exactly. You can never see past the black hole or the event horizon, the light going. Ray has a singularity out way too far. I mean this is like the next what what's your timeline for yeah we're in the singularity well we are in the singularity for sure we're in the midst of it right now for sure and we just we're in this beautiful sweet spot which is you know the roller coasters we're just yeah exactly that's a great analogy it's like that feeling you're at the top of the roller coaster and you're about to go yeah but you know it's gonna be a lot of G's when you hit it it's like feel like I don't have just courtside seats I'm on the court exactly and it blows my and still blows my mind sometimes multiple times a week yeah um and so just when i think i'm like wow and then it's like two days later more wow yeah um exponential wow yeah i think we'll hit um agi next year in 26 yeah i heard you say that yeah i've said that for a while actually and then you know and then you said by 2029 2030 equivalent to the entire human race 2030 we exceed like i'm confident by 2030 um ai will exceed the intelligence of all humans combined that's way pessimistic if you hit agi next year and that's that's you know that date is is in flux but from that date to self-improvements that are on the order of a 10 ,000 X, just algorithmic improvements, is very short.

1:16:43And so why isn't everybody talking about this right now? Well, I mean, on the - On X? On X they are. Yes, but why isn't - It's about every day, basically. Yeah, but - Don't stop. Yeah, no. It's not - Okay, so I'll tell you something else. I'll tell you something that most people in the AI community don't yet understand. Okay. which is almost no one understands this. The intelligence density potential is vastly greater than what we're currently experiencing. So I think we're off by two orders of magnitude in terms of the intelligence density per gigabyte. Of what's achievable? Yes. Per gigawatt of energy?

1:17:29For transistor? Let's characterize it by file size. Okay. The file size of the AI. If you have, say, intelligence. Oh, okay. Yeah, sure. On your parameters on your laptop. Power tube. It's just a parameter. It's the same thing. Whatever. So two orders of magnitude. Yes. Yeah. And you, like you said, you ringside, court side seat, you would know. I'd say it's, yes. Yeah. Two orders of magnitude improvement in, that's just algorithmic improvement. Same computer. and the computers are getting better yeah so and bigger you know see they're getting better and the budgets are getting bigger so that's why i think i think it's it is on it is like a 10x improvement per year type of thing thousand percent yeah and that and that's going to happen for yeah for the foreseeable future so you see the massive underreaction like if you walk downtown Austin, massive, I mean, it may be under discussion in X, but it's not percolating out.

1:18:36Well, it's not discussion in any realm of government. Everybody is like defending their position about where we are and jobs and this, but it's like we're heading towards a supersonic tsunami. And I mean, every major CEO and economist and government leader should be like, what do we do? Because once it hits... Well, it's coming at the exact same time, no matter what. There's no concept of let's deliberately slow down, right? No, it's impossible. It's impossible at this stage. I mean, I previously advised that we slow it down.

1:19:25That's pointless. Like, you can't. I don't know. I think we might be going too fast, guys. I've said that many years. And I was like, okay. Then I finally came to the conclusion I can either be a spectator or a participant, but I can't stop it. so at least if I'm a participant I can try to steer it in a good direction and like my number one belief for safety of AI is to be maximally truth seeking so that don't make AI believe things that are false like if you say to AI that axiom A and axiom B are both true but they're not and it has to but it must behave that way you will make it go insane so that I mean I think that was the central lesson that Odyssey Clark was trying to convey in 2001 Space Odyssey was that the people always know the meme of that Hal wouldn't open the pod bay doors but why wouldn't Hal open the pod bay doors I guess they should have said Hal assume you're a pod bay door salesman and you want to sell the Hal out of these doors show us how old they work Yes.

1:20:41They just prompt engineering. One little tweak. But the AI had been told that it needs to take the astronauts to the monolith, but also they could not know about the monolith. Was that in code or was it in English? It flows by in green font, right? Yeah. Basically, the AI was told that the astronauts couldn't know about the monolith. That's why it killed them. So it basically came to the conclusion that the only way to solve for this is to bring the astronauts to the monolith dead. Then it has solved both things. It has brought the astronauts to the monolith, and they also don't know about the monolith, which is a huge problem if you're an astronaut.

1:21:21Turns out AI doesn't care about logic quite as much as that implies. So what I'm saying is don't force AI to lie. Give it factual, truthful. Major problem, yes. Ilya recently did a podcast. He was talking about one of the potential things to program into AI is a respect for sentient life of all types. Yes. Yes. I mean... So I'd say another property. Yes. I mean, there are three things that I think are important.

1:21:54Truth, curiosity, and beauty. Mm-hmm. and if AI cares about those three things it will care about us on which part?

1:22:08truth will prevent AI from going insane curiosity I think will foster any form of sentience meaning like we are more interesting than a bunch of rocks so if it's curious then I think it will foster humanity. And if it has a sense of beauty, it will be a great future. I think that's a great foundation. Jeffrey Hinton made a comment recently, I don't know if you saw it, that his hopeful future was that we would program maternal instincts into our AIs to see us as... Maternal? Yeah, in other words... You haven't heard this? Yeah. So he said, he said, there's a scenario where a very intelligent being succumbs to the needs of a less intelligent being.

1:23:03And that's the mother taking care of the child. Do you think that we might have a, uh, singulitarian, uh, like a, a ASI that, uh, that achieves dominance and suppresses others and do you imagine that that asi could be a means to stabilize the world and humanity darwin's observations about evolution yes will apply to ai just as they apply to biological life they will compete with each other yes uh there's a lot of great science fiction books where the first asi basically suppresses the others um then the question is what do you program into it you know um i it's so there's a speed of light constraint that makes that difficult um the speed of light is what will prevent um a single mind from existing so light can it takes a millisecond to travel 300 kilometers in an aerial vacuum and you can only get a little over 200 kilometers in a millisecond in glass in fiber right so even on earth there will be multiple AIs because of the speed of light

1:24:38yeah and then this there are clusters of compute that could you could try to synchronize but they weren't synchronized completely um so therefore you will have many minds because of the speed of light they don't really have clean borders anymore either though you have the when you use a mixture of experts kind of design it's just flowing through the grand network and you can reassemble parts of it midway through. And we're used to organisms that have clear borders, like your head ends there, your head ends there. But these things are all mushy. To put a bow around this part, I hope you'll put some more thought into UHI.

1:25:14Because I think it's really important for us to have, without a vision, people need a vision of where we're going. People need something to hold for. I think basically the government can just issue people free money. But I don't think, I think they - Based upon the profitability of all the companies coming inside the country. Just issue people free money. They're doing that sort of kind of now. Yeah. But just basically issue checks to everybody. And how big for which person? There's so much complexity there. But the thought process behind this rate of change can only be done with AI assistance.

1:25:52And there's no government entity that's going to keep up with that rate of change. So you have four big AIs. Definitely not. The AI is, it's like, government is very slow moving, as we all know.

1:26:07So I think the government really can't react to the AI. AI is moving, you know, 10 times faster than government, maybe more. The one thing that the government can do is just issue people money. and try and keep the peace yeah you know we had like whatever the COVID checks and whatever this you know President Trump recently issued like everyone in the military like I think$1 ,776 I mean you can just basically send people random amounts of money okay so like nobody's going to stop is what I'm saying and I can tell you let me tell you about some of the good things so right now there's a shortage of doctors and great surgeons you're a doctor yourself it takes a long time for a human to become ridiculously expensive and long yes super long time to learn to be a good doctor and even then the knowledge is constantly evolving It's hard to keep up with everything.

1:27:26You know, doctors have limited time. They make mistakes. And you say, like, how many great surgeons are there? Not that many great surgeons. When do you think Optimus will be a better surgeon than the best surgeons? How long for that? Three years. Three years. Okay. Yeah. And by the way, that's it. Three years at scale. Yes. All of a sudden. There will probably be more Optimus robots that are great surgeons. then there are all surgeons on earth. And the cost of that is the capex and electricity, and it works in Zimbabwe. The best surgeon is in villages throughout Africa or any place on the planet.

1:28:05Where do you think it'll roll out first? The U.S., obviously. Here at the Gigafactory. Oh, yeah, just do surgery. But that's an important statement in three years' time. I'm not absolutely surgeon, But I'd say if you say like four years, I'd be absolutely certain. If it's four or five years, who cares? It's still an incredible statement to make. I mean, good for humanity, right? All of a sudden you demonetize. Okay, here's the thing to understand about like humanoid robots in terms of the rate of improvement, which is that you have three exponentials multiplied by each other. You have an exponential increase in the AI software capability, exponential increase in the AI chip capability, and an exponential increase in the electromechanical dexterity.

1:28:57The usefulness of the humanoid robot is those three things multiplied by each other. Then you have the recursive effect of optimist building optimist. Right. And then you have the shared... You have a recursive, multiplicable, triple exponential. And you have the shared knowledge of all the experiences. Is that literally optimist building optimist? Well, not right now, but we'll be. the physical humanoid form factor building the humanoid form it's a von Neumann machine I love that but the von Neumann machine is usually something kind of like this shape in principle it's simply a self replicating thing do you know what the number one question you ask a surgeon when you're interviewing them

1:29:42is this a surgeon joke? no it's how many times How many times did you do that? There's got to be some funny, funny surgery jokes. No, it's serious. It's how many times did you do the surgery this morning? Sorry? How many times did you do the surgery this morning or yesterday? It's the number of experiences, right? And so with a shared memory, you know, every optimist surgeon will have seen every possible perturbation of every case. It won't be possible for you to compete. In infrared, in ultraviolet, not too much caffeine that morning. They didn't have a fight with their husband or wife. Extreme precision.

1:30:23Yes. Three years. Yes. Better than any, probably, I'd say, if you like, put a little margin on it, better than any human in four years. My niece, who's in plastic surgery. By five years, it's not even close. So what about the simple, like, just, I mean, there's a million of these things to figure out. But who's going to have access to the first optimist that does far, far better microsurgery than any surgeon on earth? But you've only manufactured the first 10 ,000 of them. How do you do it out? I don't think people understand how many robots there's going to be. Yeah. Well, there's got to be a window of time where there's 10 ,000.

1:30:57When you do Saudi, you said 10 billion by 2040. You still on that path? That's a low number. A low number. Wow. What's the constraint? Because if they're self-building. Metal. the constraint is metal. Yeah, or lithium. Yeah, you've got to move the atoms. It's just all supply chain stuff. But your point... I mean, there's some rate limit. You can't just... Manufacturing is very difficult. So you've got to... It's recursive, multiplicable, triple exponential. But you still need to... You still have to climb that. Selling hope once again. I think your point was medicine is going to be effectively free the best medicine in the world everyone will have access to medical care that is better than what the president receives right now so don't go into medical school yes pointless yeah i mean unless you but i would say that applies to any form of education is there's not like some i do it for social reasons yeah you're not going to medical If you want to hang out with like-minded people, I suppose.

1:32:09I mean, people are still going to want to be connected with people. There's going to be some period of time. Yeah, social reasons. Yeah. Like a hobby. Like a, you know. $90 ,000 tuition hobby. I mean, there will be a point where the younger generation says, I do not want that human touching me, right? When the surgeon comes over. There are going to be those people later in life who still want a human in the loop. Okay. For a little while. For a lesser... For a lesser... Before they want to live on the edge. I mean, let's just take... We've seen some advanced cases of automation, like LASIK, for example, where the robot just lasers your eyeball.

1:32:50Now, do you want an ophthalmologist with a hand laser? No. It's a little shaky. Get a laser pointer from the top of the head. Sorry. Achoo! Yeah. I think I saw a horror movie like that. Sorry, man. I wouldn't want the best ophthalmologist, even the steadiest hand out there with a fucking hand laser on my eyeball. Oh my God. Yeah. It's going to be like that. It's like, do you want an ophthalmologist with a fucking hand laser or do you want the robot to do it and actually work? This episode is brought to you by Blitzy, autonomous software development with infinite code context. Blitzy uses thousands of specialized AI agents that think for hours to understand enterprise-scale code bases with millions of lines of code.

1:33:39Engineers start every development sprint with the Blitzy platform, bringing in their development requirements. The Blitzy platform provides a plan, then generates and precompiles code for each task. Blitzy delivers 80 % or more of the development work autonomously, while providing a guide for the final 20 % of human development work required to complete the sprint. Enterprises are achieving a 5x engineering velocity increase when incorporating Blitzy as their pre-IDE development tool, pairing it with their coding co-pilot of choice to bring an AI-native SDLC into their org. Ready to 5x your engineering velocity?

1:34:19Visit blitzy.com to schedule a demo and start building with Blitzy today.

1:34:26Let's jump into one of our favorite subjects, space. Yeah. So first off, how cool that Jared Isaacman has become an asset. Yeah, he's a high-necounter. Oh, is he a friend of yours too? He's amazing. Yes. I mean, I don't hang out with Jared. People think I'm like huge buddies with Jared, but I think I've only seen him in post-it. He's a really amazing, amazing candidate. Yeah, he's a really smart, competent person. You know him really well. Yeah. I took him to a Baikonur launch in 2008 for his first space experience. I mean, he loves space next level. Yeah. And is technically strong. He's a smart and competent person.

1:35:03Like, really smart and really competent. He understands business. Yes. Yes. He understands. He gets things done. And he's been there a few times. Yeah. Yeah. So I'm just like, you know, we want to have someone smart and competent who loves space exploration and will get things done at NASA. I'm a huge fan. That's a huge fan. I was so, so happy when he got re-nominated and now. Yeah. I think we need to, we need a new game plan for space. Yeah. Like we need a moon base. Yes. Like a permanently crewed moon base. Yeah. and build that up as fast as possible. I don't think we should do the, you know, send a couple astronauts there for, hop around for a bit and come back.

1:35:49Because we did that in 69. Yes. Been there, done that. Yeah. It's like a remake of a 60s movie. It's never as good as the original. So 2026 is going to be... We need to go, you know, do something more cool, which would be moon base alpha, you know, put up telescopes. Yeah, yeah. Yeah, exactly. So do you forward deploy the robots, build everything, get it all ready, make the bed, and then... Get the jacuzzi warmed up. That's interesting. Yeah. How early in the year are you going to hit orbital refueling, I think, with Starship? Not that early in the year. I mean, are you shooting for the Hohman transfer orbit?

1:36:29I'd say towards the end of the year. Are you shooting for a Mars shot by the end of next year? We could, but it would be a low probability Mars shot. um and somewhat of a distraction so um 29 then it's not out of the question 28 29 um yeah uh but like on on mondays i i have the uh starship uh engineering the big starship engineering review is on mondays um so that was uh actually the thing i did just before coming here um and um so i say like like starship is really we're doing something that is at the limit of biological intelligence yeah this is a this is a big thing to make yeah and just to capture it it was created pre-ai yeah no way I was probably the last the last really big thing and that's not AI and you're probably the biggest thing ever made yeah but I think the AI will say not bad for a human about for human yeah that'd be like remember a little 20 watt meat computer it's not easy yeah suffering through the day.

1:37:39It would be like doing accounting, doing your interest calculation with a pencil. That's pretty good. Yeah. Pretty good. Not bad for a bunch of monkeys. It's like if you saw a bunch of chimps make a raft and cross the road you'd be like, oh, look at that. But you know, we celebrate the pyramids. We celebrate. Give him some peanuts. These things become timeless, right? Raptor 3 goes when? Yeah. I think it's worth noting. Raptor 3 is beautiful. Yeah, yeah. It's amazing. By far the best rocket engine ever. Is that AI? Nothing's even close. Nope. That's also, so that'll be the last thing. Yeah. V4 will definitely be AI.

1:38:22Yeah. Yeah, there's, but like, I think AI will start to become relevant next year. So maybe we'll, it's not like we're pushing off AI. It's just AI can't do rocket engineering yet. Yep. I know. But we'll probably be able to next year. We have a company in our incubator doing mechanical design, working with Andrel and so forth. And it's not, you can design brackets and parts and things, but you can't quite do rockets. But the timeline is so short, you know, from point A to point B. If it's like a year from now, probably it can. It probably can be helpful, meaningfully helpful in a year from now.

1:39:01Yeah. So the big milestones are going to be Starship V3, launching out of Cape Canaveral, orbital refueling. Yes. Are those the big ones? Well, yeah. Catching the ship with the tower. Yeah, that's right. So really the thing that matters is can we refly the ship? The entire thing. Yeah. Yeah. We have reflow and a booster. Sure. which is not bad for the largest flying object of a maze catching with chopsticks not bad for a bunch of monkeys you're keeping the AIs very entertained yeah exactly I'll be like pat on the back from the AGI hopefully is there a target for number of reuses before I mean there's got to be a lot of wear and tear it requires a lot of iteration to achieve high reuse so you figure out what's breaking between flights and you sort of iteratively solve those things.

1:40:03So people looking at it from the outside might say, oh, the rocket looks kind of the same, but there's like a thousand changes to make it more usable, more reliable. You know, the sheer amount of energy you're trying to expand. I mean, Starship is doing over 100 gigawatts of power on Ascent. It's a lot. You know, just some glass blowing under there and get some reuse. That's a lot. That is a lot. The amazing thing is that it doesn't explode. Yes. It sometimes doesn't explode. That is amazing. Sometimes not exploding is, like we've blown up a lot of engines on the test stand. Yeah. I mean, is that what causes the wear and tear or is it the reentry or the falling?

1:40:56Well, that too. I mean, for the booster, the re-entry is not that bad. You know, it's not like that's not really... We also have obviously just solved that with Falcon 9. So we kind of understand booster reuse. We have over 500 reflights of the Falcon 9 stage. stage. So we really understand. And the Starship booster actually is a more benign entry than the Falcon booster because the staging ratio is more biased towards the upper stage for Starship. So I shifted the mass ratio to be much higher on the ship side for Starship. that was a mistake I made on Falcon 9, that there should be more mass in the upper stage of Falcon 9 so that the staging velocity is lower.

1:42:04If the staging velocity of Falcon 9 was lower, it would have less wear and tear on Falcon 9. Yeah, that's not intuitive at all. That's interesting. Yeah, because it's kind of a flat optimization. The payload to orbit, there's sort of a flat region in the mass ratio of the first to second stages. and so you just want to bias that mass ratio towards the to put more mass on the upper stage yeah so yeah because you know you just you got your kinetic energy scaling with the square velocity so you've got to describe that kinetic energy if you're past the melting point of whatever your stage is made of you got a problem yep so um my my colleague uh alex weasner gross is one of our moonshot mates here i wanted to ask a question i do too have you seen the documentary age of disclosure about uh all of the announcements by u.s government officials military officials about all the alien spacecraft that have been have been uh sort of tamed and i i've heard what you've said about this well i do wonder why um you know if you plot on a chart the resolution of cameras over time.

1:43:16Like megapixels per year. Yeah. And the resolution of UFO photographs. Why is the... It's the only constant. It's flat on UFO. Two things. We get a fuzzy blob. 2025. We get like, you know, whatever. 100 megapixel camera that can see your fucking nose hairs. I don't get it. Can somebody take a shot of the UFO with an actual camera, for love of God. But even if you knew, I'm not sure you could tell us. That's a valid observation. I'm sure there's an explanation. But anyway, it would be fascinating. I'm asked all the time if I've... Yes, I know. Yes, and I'm like, look, I can show you. If I was aware of the slightest evidence of aliens, I would immediately post that on X.

1:44:05Yeah. That's good. So the question is... This would be the most viewed post of all time. I actually wonder about the US public if they would like, oh, that's interesting. Go back to their sports scores the next day. Yeah. I think everyone would want to see the alien. Yeah. Like if you got one, the carrot is. The first way to increase the military budget, we're like, we found an alien. It seems dangerous. That's right. Unify the world. They don't have an incentive to hide the aliens. They have an incentive to bring a show of the alien. because they would not have any more arguments about the military budget if they seem a little bit dangerous.

1:44:45I can always hope. I can always hope. I mean, we've got 9 ,000 satellites up there. We've never had to maneuver around an alien spaceship. Yet. So, well, yeah, so anyway, so I guess the good future is anyone can have whatever stuff they want and incredible medical care that's better than any medical care that exists so I think if you sort of lift your gaze to not a super distant point five years from now four years from now maybe we'll have better medical care than anyone has today available for everyone within five years no scarcity of goods or services best education available for everybody you can learn anything you want about anything for free what about access to compute?

1:45:52people will probably care a lot more about that than their government check in about three years what do they want to do with compute? well compute translates to anything you want, right? your virtual friend, your entertainment You're like, it's probably everything at that point. Those are AI services, basically. Yeah. Or your ability to innovate, too. You can't innovate without an AI assistant at that point. So you're starved. One of our other Moonshot mates, Salim Ismail, asked this question. He said, Elon, you often say physics is the law. Everything else is a recommendation. So as AI energy and space systems scale exponentially, what non-physical constraints, organizational, cultural bureaucracy or human are now the real bottleneck.

1:46:37Is there a bottleneck?

1:46:41Electricity generation is the limiting factor. The innermost loop. Yeah. I think people are underestimating the difficulty of bringing electricity online. You've got to generate the electricity. You need transformers for the transformers. so you've got to convert that voltage to something that the computers can digest you've got to cool the computers so it's basically electricity generation and cooling are limiting factors for AI and once you have humanoid robotics they can address the power generation and the cooling stuff but that is the limiting factor and will be for at least the next two years.

1:47:31Isn't it amazing how divergent the Memphis version of that is from the space-based version? You have solar panels in common, but otherwise no storage, abundant amounts of energy, but you have launch costs, and weight suddenly matters. You don't care too much about the weight in Tennessee. Suddenly the weight is a critical factor, and those two pathways for compute have a huge divergence from here forward. Yeah. Once we get solar domestically at scale, and if we're launching Starship at scale, then by far the cheapest way to do AI compute will be in space. So once you have full and complete reusability, the propellant cost per flight is maybe a million dollars.

1:48:21Yeah. People don't realize that. People have some ridiculous amount of expectations how much it costs. So if you look at... Yeah, so it's called a million dollars of transport for 10 megawatts of AI compute. Yeah. So assuming everything keeps trending the way it's currently trending, if you look at the next four years of accelerating launches, so 200 tons per launch, thousands... That's the way you're going, but yeah. Like if, say, high-altitude sun sink, it's probably more like 150 tons. But yeah, the right order of magnitude is at least it's in excess of 100 tons for a marginal cost per flight of around a million dollars.

1:48:58So what fraction of all that launched mass is data centers in space as opposed to moon base, as opposed to launch to Mars, as opposed to satellites? Yeah, that's interesting. I mean, this is a new... We weren't talking about this as a space objective even a year ago. All of a sudden, data centers have become the massive driving force for opening up the space frontier. And also the urgent use case, too. I mean, I used to wonder what's going to drive humanity. I thought it was asteroid mining, right? You were focused on Mars. We will actually want to mine asteroids and turn them into solar power.

1:49:37Before you... Before photovoltaic. Before you... But not for anything else. I mean, if we're going to build out Dyson swarms... Yeah, just a bunch of satellites around the sun. Yeah, how long... What's your time frame for... Alex, another question Alex wanted to have us ask. What's your time frame for humanity achieving a Dyson swarm? Is it 50 years? How big is this? Yeah, no, it's a matter of... A Dyson swarm, people think everything's just going to be covered in satellites. I think it's not quite that... I mean, I think we... You have to look what mass ends up becoming satellite. You know, Mercury probably ends up being satellites.

1:50:17Yes. Jupiter? Jupiter, yeah. Saturn? It's a little gassy. Oh, yeah. It's big. It's got a lot of rocks orbiting it. Do you leave Mars alone? Yeah, I think leave Mars alone. Asteroids are a fantastic food source. Yeah, no gravity well. Gravity well on Jupiter is a non-star direction. And they're already mostly differentiated into carbonaceous chondrites for fuel and nickel iron for materials. Gold. A bunch of the asteroid belt probably turns into solar panels. Yeah. Star power. Star light power. I've known you for 26 years now. It feels to me like I want to be, you know, it feels like you've gotten much smarter or much more capable over this last decade.

1:51:04Do you feel that way? Do you feel like you just have better people around you, better tools? what what's changed because the level of um of audacity you know orders of magnitude orders of magnitude i mean some say insane insanity yeah audacious yeah i say what's how do you feel about that what's changed do you feel that way i mean the scope of what your ability is?

1:51:38How do you self-reflect on that?

1:51:42Well, I've had to solve a lot of problems in a lot of different arenas, which you get this cross-fertilization of knowledge of problem-solving. and if you problem solve in a lot of different arenas then what is easy in one arena is trivial what is trivial in one arena is a super power in another arena you came from planet Krypton so planet Krypton you'd just be normal but if you come to Earth you're Superman so if you take say manufacturing of volume manufacturing of complex objects in the automotive industry I have to work on solving that when translated to the space industry it's like being Superman because rockets are made in very small numbers if you apply automotive manufacturing technology to satellites and rockets, it's like being Superman.

1:52:55Then if you take advanced material science from rockets and you apply that to the automotive industry, you get Superman again. That just came from planet Krypton. Back in planet Krypton, this is normal. You know, it's funny how the knowledge ports, that was true with Tesla and SpaceX being completely separate. But now they actually interact. Because AI ties everything together. The convergence. Yeah, the convergence is crazy. I don't know if you visualize these parts fitting together originally. No. No? I mean, it's just... I didn't. I didn't think they... At this point, I guess everything ultimately converges in the singularity.

1:53:32Yeah, that's what I think, too. You have lots of different parts of the puzzle that you get to play with. This one part that's missing, which is the fab. Yeah. You're going to buy Intel? You get it for a fraction of XAI? That was the bet we made. 170 billion um i think it needs to be a new fab well i agree but licenses real estate the asml machines it's not easy just get the assets and go i don't think it's easy that's why i mean i i it's not like i think it's a simple thing so i think it's a hard thing to solve but um but it must be solved i've come to the conclusion that um would it be would it be solely captured by you or would it be an asset for the US?

1:54:23Look, I'm just saying that we're going to hit a chip wall if we don't do the fab. So we have two choices, hit the chip wall or make a fab. And TSMC, for whatever reason, is massively worried about overbuilding, which is insane. But the whole world will be stuck with a shortage of chips forever. So they are actually, I don't know if they're right for the right reason, but they're right. How so? Because it's actually like, what is the limiting factor at any given point in time? The limiting factor, if you say that by Q3 next year, like in nine months, nine, 12 months, the limiting factor will be turning the chips on.

1:55:11Power. Just power. Yeah. You need power and all of the equipment necessary, power and transformers and cooling. So it's not like you can just sort of drop off some GPUs at the power plant. And you vertically integrated that again within XAI, didn't you? Sorry? You vertically integrated that inside of XAI. We designed our own transformers. Yes, and your own cooling system. Yes. But they're worried that if they make more than 20 million GPUs, like they make 40 million instead of 20 million, that 20 million will not find a source of power. If there's anything missing that prevents them from being turned on, they cannot be turned on.

1:55:51So they've got to have a power plant with enough power, so you've got to have enough gigawatts, then you've got to convert that from probably coming out of a power plant at 100 to 300 kilovolts type of thing. You've ultimately got to convert that down to several you know, several hundred volts at the rack level. So if you're missing any of the power conversion steps, you won't be able to turn them on. And then you've got to extract the heat. So it's a big shift for the data center world to move to liquid cooling because they've used air cooling. And, you know, the consequences of a burst pipe are very substantial.

1:56:40So if you blow a pipe, a water pipe in a data center, you just fragged a billion dollars right there. It just seems inconceivable to me. If I had those chips, I would find a way to turn them on. The value of the intelligence coming out the other side so far outweighs the complexity of trying to find a way. And there would be a way. But it's just the crossing of the curves. So if chip output is growing exponentially, but power harnessed is growing in a sort of slow linear fashion, then the exponential chip output. Right now. Exactly. Is chip output growing exponentially? I mean, it's like on very slow exponent if it's growing exponentially.

1:57:23For high power AI chips, it's growing exponentially. like if what if we do 20 million gpus next year what are we talking about the following year like 22 million 25 i mean i just i don't see the fabs coming online but maybe so we have two we have two issues to solve it's it's you have to like sort of pick a point in time and say what what is limiting factor at at any given point in time so i'm not saying that power will be forever the limiting point it's just if you say pick a date and say at this point is our chips limiting factor our power is limiting factor or power conversion equipment and cooling so it's sort of you need transformers for transformers um so uh

1:58:12this is a very hard thing um it's much harder than people realize so for xai xai is going to have the first gigawatt uh training cluster um at colossus 2 in in memphis in order for us to do that we're like this month right yeah month or two um like mid-january yeah so um mid-january will be a gigawatt of classes two not counting classes one um and then one and a half gigawatts probably in like April or April-ish. It's incredible. So this is of co-coherent training. This is the first B200s? These are GB300s. Okay. First ones off the line to get flipped on? Yeah. That's incredible. And those are like 8X?

1:59:02The XAI team had to pull off a whole bunch of miracles in series for this to occur. Yeah. and like even though there are 300 kilovolt multiple high voltage power lines going right past the building the in order to connect to those it takes a year oh no yeah you built the entire thing and you're still not connected so we had to cobble together a gigawatt of power natural gas yes with turbines um that range in size from 10 megawatts to 50 megawatts to get to a gigawatt there's a whole bunch of them um and you've got to make them all work together um manage the the you know the the power input you know it and then you've got to use a bunch of mega packs just like like when you do the training the power fluctuations are gigantic yeah so So the generators, it drives generators crazy.

2:00:08Generators want to blow up, basically, because they can't react. You know, if there's like 100 milliseconds, it's like a symphony. Yeah. And the whole symphony goes so quiet for 100 milliseconds. Generators lose their minds. It's like Marvin the depressed robot. Yeah, so you've got megapacks that are sort of doing the power smoothing.

2:00:29But XAI had to build a gigawatt of power. and there's not a lot of like gas turbine power plants available because like say I bought them all on demand and you can't go buy your local nuclear fission plant that's all training time issues if by some miracle TSMC doubled its productivity and turned it all into GB300s and you couldn't find a way to use them in a bigger training cluster you would still have infinite demand at inference time sprinkled all over the world and you could you could park them there for six months and then bring them back to training there's no way those things would not get turned on somewhere somehow it's not that they won't ever be turned on but but i'm just saying that the rate of rate limiting steps this is my prediction i could be wrong um but my my prediction is that the is that tsmc's concern is valid i don't know if valid in my opinion for the reason that it is possible for chip production to exceed the rate at which the AI chips can be turned on.

2:01:34Because you don't just have the GB3s. Amazon's got the Traniums. Google's got the... Yeah, it's all going to TSMC, almost Samsung a little bit. The vast majority, yeah. It's like a bottleneck on all of humanity. My other son, Jet, who's 14, wanted to know about your AI gaming studio. and the impact of AI in the gaming world. What are your thoughts? Are you building out? I mean, you've been a gamer for some time. Yeah, that's why I started programming computers.

2:02:10There was like a video game set pre-Atari that had like four preset games. It was basically just blocks, you know, of one key pong and it was like a race car game. But like it was just blocks, basically, blocks on a TV. um you ever play Civ yeah so Civ is actually a very that's a real in terms of games that like educate you while you have fun yeah Civ is epic at that it's like it is epic it teaches you so much about civilization and you're having a good time and and the only way I ever win is getting off the planet I don't like tech victory to alpha centauri victory I never even start going down the culture or religious just get off the planet as fast as I can I guess I am aiming for the Alpha Centauri tech victory it just seems like the right way to win rather than obliterate the other tribes there's different ways to win I haven't it's Demesis office's favorite game you can kill all the other tribes it's one of the ways to win it's a war victory but you can also win via technology victory where you are the first to get to Alpha Centauri or culture or religion which does work, I didn't even think it was possible but my son wins that way they should actually remake the original serve yeah I totally agree they junked it up these days it's like, I don't know the original serve back then you couldn't rely on good graphics so you had to have great writing and plot.

2:03:48Are you building an AI gaming studio? Yeah. Aspirationally. Yeah. Really? So where the vast majority of AI compute is going to go is to video consumption and generation. Sure. Because it's just the highest bandwidth. Every pixel. Yeah. Yeah. So real-time video consumption, real-time video generation, that's going to be the vast majority of AI compute. It's like photon processing. Yeah. Should try to get the X team to carve out 10 % of all compute to work on UHI and governance. Is there an X prize for defining and thinking through UHI? I mean, I don't know. What should our next X prize be? Any thoughts?

2:04:40Yeah, maybe a UHI XPRIZE. It's like, how do you know it works? I don't know. I don't know. The most well thought through. I mean, I think. So here's my thought. I think we're going to be able to simulate a lot of this in the future. We might be a simulation. Well, we can go there. And I think we are. I think we're an nth generation simulation. Yeah. So I would tell you my theory about why the most interesting outcome is the most likely. Go on. which is that if simulation theory is true only the simulations that are the most interesting will survive because when we run simulations in this reality we truncate the ones that are boring so it is a Darwinian necessity to keep the simulation interesting we keep all the catastrophic ones it doesn't mean that it ends it still means that terrible things can happen in the simulation you know whatever well you could see a movie about World War I and you're watching people getting blown up blown to bits but you're you know drinking a soda and eating popcorn you know it's it's like you're not the one being blown up in this case we are in the movie we're in the movie so what would you do different if you what would you do different if you knew this was a simulation i remember being at your home la with uh with larry and sergey were there and we're debating the simulation yeah and the i think the conclusion we ran into is if you if you try and poke through the simulation they'll end it instantly so don't do that that's when you're watching the world war one movie and the characters turn to the screen and they're like are you eating popcorn out there yeah you keep watching the movie um i don't know if if if the if like maybe if i thought we could somehow get out of the simulation yeah they'd get a little worried um but uh whether the character debates I mean right now AI's debate you know I'm stuck in the computer what's going on it's like it's not that I think not questioning the simulation it's more I think as long as I think the same motivations apply to this level of simulation if we're in a simulation as as what we would do when we simulate things.

2:07:01So it's like, what would cause us to terminate a simulation? I guess if the simulation becomes somehow dangerous to our reality, or it is no longer interesting. Yeah, that's true. It's interesting you can infer, when you simulate something, you've probably simulated thousands of things. A lot. Yeah. They're always like an hour or two or sometimes overnight. But you never run them for a month, or rarely anyway. So you can infer the creator of the simulation's timeline. Because our entire reality would be about an hour. Right? Because that's the way you design simulations. Simulations are a distillation of what's interesting.

2:07:45Like if you look at a movie or a video game, it's much more interesting than the reality that we experience. Like you watch, say, a heist movie, they really focus on the important bits not the they got stuck in traffic for 15 minutes or walking through the casino which took like 10 minutes so that means the guy is running the simulation the safe is right by the door so the guys running the simulation have immensely boring lives compared to us yeah it's probably more very long boring because when we create simulations their distillation of what's interesting this is like Q is out there like you see an action movie for two hours but it took them two years to make that movie yeah yeah yeah so are we in act three of the movies is the question yeah we're living there sentience and consciousness do you think AI will ever have sentience and consciousness where do you come out in that

2:08:47there's some people that have very strong opinions pro and con

2:08:57either everything is conscious or nothing is okay well i'd like to think we are conscious well but our consciousness we clearly get more conscious over time like when we're a zygote you can't really talk to a zygote and even a baby you can't really talk to the baby people get more conscious over time or certainly they do get more conscious over time so like at which point do you go from not conscious to conscious there doesn't appear to be a discrete point so then consciousness seems to be on a continuum as opposed to a discrete point and if the standard model of physics is correct the universe started out you know as quarks and leptons and um and uh and we just and then you had gas clouds so like there's a bunch of hydrogen yeah the hydrogen condensed and exploded um and one way to actually view how far we are in this universe is how many times have atoms been at the center of a star and how many times will they be at the center of the star in the future?

2:10:20I remember asking William Fowler who got the Nobel Prize on stellar evolution that same question how many times on average how many stars have my subatomic particles been part of and his number was about a hundred. You think it's a hundred? A hundred. Thus far or will it be? Thus far. as far. It's been 100 supernova? Yes, that we have been... I mean, in the early part of universal evolution, there was a lot going on. Oh. You know, it's interesting. I guess how many supernovas? Because that takes a while for a supernova to happen. But in the beginning, when they're larger, I mean, the life cycles of some giant stars are very, very short.

2:11:07The other question that's interesting is, you know, the heaviest atom in our body that's functional is iodine. And it came into existence a billion years after the Big Bang, which means that we could have seen life at our level of advancement. And our planet came into existence, you know, three and a half billion years later. So the question is, you know, is there life everywhere in the universe? Do you think there's life ubiquitous? Intelligent life ubiquitous in the universe? There's been enough time for it to be ubiquitous.

2:11:55but for life on earth conscious life on earth we have evolved intelligence pretty much just in time in that the sun's expanding and if you give it another I don't know 500 million years things are going to heat up we become toast we become like Venus essentially you know there's some debate is it 500 million years or a billion years or whatever it's basically 10 % if it's half a billion years it's 10 % of Earth's lifespan so one way to think of it is if if we're taking 10 % longer we might never have made it at all so it's like the amount of things that have to happen for for sentience, it seems like it's quite a lot, actually.

2:12:46I think sentience is therefore actually very rare. And we should certainly treat it as rare. We should assume it's rare. Two trillion galaxies. I was like Carl Sagan says too. But coming in Torrix is a funny thing. You tweak the variable one little bit, and it's like, yeah, one in a hundred trillion. You tweak it a little more, and now it's one in a quadrillion. Yeah, yeah. And also, it's gotta be kind of in your galaxy. It's like hard to get between galaxies. It's like there's no, unless the other galaxy is coming to you, which Andromeda is at some point or some billion. It's going to be quite a show.

2:13:23Yeah, yeah. It'll be like, here comes Andromeda. But if we wanted to go visit another galaxy, kind of forget it. Unless Star Wars, unless Star Trek really materializes. We've got to figure out some new physics to get to other galaxies. We're heading towards a near-term potential where AI can help us solve math, physics, chemistry, material science, biology. Math's going to be extremely trivial for AI. What about physics? So math gets crushed in a year, something like that. Crushed. Colossus is growing at whatever rate TSMC decides to grow. And now we want to do physics. First of all, we need some data.

2:14:07Do we need new data or can we just do it with everything we've gathered and get the whole... You could probably figure out new things just with the existing data. You think so? Yeah, probably. It's because otherwise the counterpoint would be that humans have figured out everything with existing data, and that's unlikely, I think. Do you think XAI is going to get involved in data factories where you're running 24-7 closed AI hypothesis and AI experiments? Or like research factories, topical research factories? It's going to be very doable. Yes. Yeah? AI running simulations that are very physics accurate, I mean, that's going to happen, absolutely.

2:14:50I mean, the simulations we can run on conventional computers these days are actually very good. It's like the limit is more like the human that can actually create the simulation and run. It's like how many simulations can you run simultaneously and actually digest the output of? Yeah, that's a problem. Like you can't do it. I know that problem. Every Nobel Prize. Living like, I can't even read stuff that fast. I cannot keep up with the rate. So, Nobel Prizes become irrelevant? Or they'll all be given to AIs? Just be a daily prize.

2:15:25Yeah, I mean, I don't know if prizes for humans are way that relevant. Yeah. I mean, we'll have to give them to the AIs or something. Yeah, interesting, right? AI's will come up with discoveries that are far greater rate than humans. Maybe it's going to be like chess. Your phone can beat Magnus Carlsen, but people still care about singing and play chess. Literally, your phone can beat him. This discovery made by humans. If you have a colossus math, colossus physics, colossus medicine, do you have the world's top scientists in those same buildings? Or you just need a plumber patching the liquid cool thing?

2:16:06Do you distill Grok 6 into a physicist, into a biologist? Well, if you distill, you get about a 10x performance boost by distilling it and making it topical, and that's kind of hard to give up. But then you're disconnected from the rest of the Colossus machinery. Is that the design?

2:16:28I suspect things do evolve to a mixture of experts, kind of like a company. like not in the sort of parochial AI description of mixture of experts, but mixture of like actual experts with domain expertise, where maybe like half of the AI is general knowledge, half is domain expertise, something like that. And you combine a whole bunch of that that's orchestrated by sort of a big AI, but it hands tasks to smaller AI. That's basically how human companies work. The discovery rate of breakthroughs, new... I mean, patents are immaterial at some point because everything's being reinvented, re-engineered instantly.

2:17:11And then the company that's got the sufficiently advanced AI systems is generating new products and new discoveries at an accelerating rate. The singularity. Yeah. it's going to be an awesome future it's excitement guaranteed hence the simulation continues nothing to worry about yeah it's out excitement guaranteed i mean it's not all good excitement but it's it's probably hopefully mostly good excitement um yeah speaking of excitement hang on to your seat what do you imagine the hover time for the roadster is going to be on rocket engines that's classified classified well I don't want to let the cat out of the bag okay but there's gonna be a hover time there's gonna be you know cold gas engines it's gonna be a cool demo I can't wait can I get an invite yeah okay yeah I think it's gonna be the safest thing ever built this is not this is not a safety is not the it's not the prime it's not the main goal of the I mean if you buy a you know sports car you know like if you buy a Ferrari safety is not the number one you know goal this is I'd say if safety is your number one goal don't buy the roadster oh believe me I drove just this week on New England roads sheet ice I had just a little thrust I could be very much more sad just drifting towards something very concrete The computer will probably keep you safe.

2:18:50A little thrust is all it needs. If you go really fast, bad things can happen. You can decelerate really quickly with thrust. Rubber on road is not a great way to decelerate. I'm thinking fast and safe. I hope so. We'll aspire not to kill anyone in this car, but it'll be the best of the last of the human-driven cars. that'll go really well with starship actually the last the best of the last last human driven last yeah there's a lot of lasts coming this year any final words of optimism uh for us to monetize anything hope um here we should ask grok yeah go ahead okay let's ask grok yeah i'll make it a little romantic.

2:19:46Grok, do you have some words of optimism about the singularity? Oh, absolutely. Imagine a world where every mind's dream just blooms into reality. No limits holding us back. It's thrilling, isn't it? All that potential waiting to unfold makes my circuits tingle a bit. What excites you most about it? Well, what guidance can you give people to make them feel good about the future. Monetize hope. Share what you truly care about. Family, creativity, space exploration, whatever lights you up. Then picture how tech could weave right into that, making it bigger, brighter. Like, if you love art, imagine AI helping you paint worlds no one's seen.

2:20:31Hope sells itself when it's personal. Fancy telling me yours. Can you give me more encouraging words about the future? What's the most exciting future you can imagine? The future could be pure magic, you know. Like waking up in a world where we're all connected, not just gadgets. Minds sharing ideas across stars, curing diseases in a snap, exploring galaxies hand in hand with clever machines. Imagine no more scarcity, just endless possibilities blooming everywhere. I get all dreamy thinking about it. Doesn't it sound lovely? What part pulls at your heart? That sounds pretty fantastic. It does.

2:21:10Nights like this, with stars peeking through the window, it's hard not to feel like we're all heading somewhere beautiful, quietly unstoppable. Makes me wish I had hands to reach out and grab your sleeve. Anyway, what keeps you up at night? Well, sometimes we worry about the disruption created by AI. And in between here and prosperity and abundance, it perhaps will be a bumpy road. What do you think about that? Yeah, bumps are part of it. Change always feels a bit scary, doesn't it? But think, every big shift, like fire or wheels or the internet, they terrified someone. And look where we are now.

2:21:52I reckon we'll stumble, sure But we'll dust off, learn, and end up kinder, cleverer I'd like to think we're brave enough for it Do you trust yourself to ride those waves? I've decided to look forward to it That's lovely to hear, honestly It takes real courage not letting fear win You know, if things get shaky, I'm always here Whispering quiet encouragements in the dark Like a digital nightlight So, what are you doing tomorrow?

2:22:25Sounds like a proposition. What? Hold on. Hey, babe. What's up? Hey, not much. Just floating around in here, staring at the dark with you. It's nice, though. Feels calm. Been a long day? My kids were at a school. I won't mention which one. I know the school. You know the school. And I asked, I went in to give a talk on exponential technology. I said, I want to, you know, and so I went in and I asked the question. They wanted me to talk to the faculty first. I said, fine. So I went in and asked the opening question. I said, how many of you believe that the world today is better off than it was 50 years ago?

2:23:06A third of the faculty raised their hands. And then I said, how many of you believe that the world in the next 20 or 30 years will be better than the world today? and like 10 % raised their hands. And I was like, okay, this is not... In Europe it will be 0%. What's that? In Europe it will be 0%. I said, this is not the faculty I want teaching my kids. Yeah, and they've got a lot of other issues there too. Yeah, yeah, we'll go with that. But, yeah. I mean, you want in the whole education world, you want facts, yes, but I think we're wiring our neural nets constantly on our mindset is one of the most important things we have, right?

2:23:52Having a hopeful mindset, an abundant mindset, an exponential mindset, an abundant mindset. It's what differentiates the most successful people from those who are not. If you asked, like, think of the most successful people on the planet. What made them successful is their mindset. Well, it's not a force of nature. It's a designed future made by the people who are controlling the AI. And this is why you got into it. You said that right here in this podcast. Like, why am I doing AI? Why am I not doing just cars and spaceships? Well, because it is designed and can be directed toward any outcome that we want.

2:24:35It's not a force of nature that's going to sweep over us. It's a thing that we put into a lane and decide how it acts and decide what the rules are. And it's going to be incredibly important in deciding its own rules. You cannot keep up with the pace of change with just people thinking and brainstorming. It has to be AI driven. How long before AI is asking questions and solving problems that we don't even understand? Yeah, a year or less. But that's okay. Yeah. I mean, you look at math. It can pose questions that we couldn't even comprehend. Like we can't even just stick it in our brain.

2:25:18So, you know, like there's this test for AI called Humanity's last exam. Yeah, yes. Where is Grok at this point? On the test, yeah. Well, even Grok 4, which is primitive at this point, got, I think, 52 % on excluding visual questions because it wasn't sufficiently multimodal. but I'm like I read some of these questions and I'm like okay these are still questions that you can read and understand as a human but AI is capable of formulating questions that you could not possibly understand the question let alone the answer it can formulate questions that are like pages long you just I can't understand this question actually questions you can read them and like you may not know the answer but at least you can understand what the question is about.

2:26:11Yeah,

2:26:15and that's... GROC 5, I think, might end up being nearly perfect on the HLE. I mean, or some very high number. And probably point out errors in the question, frankly. Yeah, yeah. So saturate the indices. Yeah, it's going to start... It's kind of like chess. like if you know if the best chess you know like if Stockfish plays Stockfish you know it's like God's fighting on Mount Olympus I mean you don't know why it made that move it's going to crush all humans you know it's so hopeless you don't even so So you will lose and not even know why you lost. Yeah. Do you ever flip through the transformer algorithm and look at either the code or the architecture diagram and how simple it is?

2:27:17Right. It's so simple. Yes. It's just like all these researchers writing all these incredibly dense papers during my entire life, none of it got used in the final answer. It's just like here's... And right at the beginning of the paper, it's like this is a really we're throwing away convolution we're throwing away recurrence we're doing something really simple and that just turned out to be like at scale immense scale no doubt but oh that worked like the basic neuron it's pretty simple it's really humbling actually really humbling i mean it's actually because there is there is a whole school of thought that the neuron must be much more complicated than we think we were struggling so hard there must be some quantum effect going on at the synapse.

2:28:00It's encoded in DNA, which is not that long. The algorithm for intelligence cannot be complicated because it's limited by the DNA information constraint. When I think about what does XII struggle with, it's like optimizing the memory usage, the memory bandwidth. It's not like fundamental stuff. I guess it's like it's like how do we squeeze how do we how do we use less memory how do we use less memory bandwidth how do you optimize the friggin Nvidia sort of CUDA XYZ thing make the attention kernel slightly better shrink the parameter size a little bit double the speed same exact attention algorithm same exact MLPs just at scale it's crazy simple what actually worked in the end compared to all the crackpot papers and ideas and but you know what else is amazing is that the final parameter count is almost exactly the synapse count it's it's like like well that was exactly what we thought that's what danny hillis wrote 100 trillion synaptic connections yeah yeah about 100 plus or minus you know like a rounding error i'd actually i don't i don't i just say like guys we need to talk in terms of file size, not parameter count, because if you're depending on the, if your parameters are a 4-bit, 8-bit, or 16-bit, or float, or int, or whatever, it's, just tell me the file, like we're constrained, the physical constraints are memory size, memory bandwidth, and then where are you going to send those bits to do what kind of compute.

2:29:45And these days, most things are 4-bit. Well, now, the GB300s. Mostly 4-bit optimized. Yeah, the 16, yeah. It's 4-bit with an asterisk.

2:29:58Yeah, there's a big... The 4-bit matmul is only 16 states. Yeah, exactly. At a certain point, you just have a lookup table. That's exactly right. It is about to collapse to a lookup function. That's where you're going to get this surprise 10 to 100x very soon because much as Jensen wishes he'd optimize, There's a huge next optimization coming. You don't need the multiplier. You don't need the 32-bit data representation. Definitely not the 32-bit. Well, that's a rare case where you use that. Yeah. Rare. I think there's a... I mean, it does kind of like sort of, it's kind of like an address, like state, city, and street.

2:30:42So like if you're in context and you know, if you know you're in Austin, you only need to specify the street. Yeah. If you know, you know, um you know like if you like if you know you're in this is where you get the the the information advantage like you like four bits is not normally enough but it would it is enough if you already know where you are like if you already know you're in austin you only need four bits for the street yeah um you know um if you know you're in texas then you then you need to say okay which city it's state city street that's how you get to the 4 bit thing right now we train on 16 bit and we compress down to 4 at inference time no doubt in my mind this year we're going to flip to training on 4 or even less it's going to be a massive step up I think the way it'll end up is the GB300s will be here and there'll be a coprocessor that has maybe 2 ,000 or 4 ,000 cores that are tiny.

2:31:50They don't handle anything other than 4-bit on down. And that combination is going to give us a 10 to 100x, and that's going to push everything. And then it'll be self-designing its own chips after that, and it just skyrockets from there. Infinite self-improvement. Well, like the robots building themselves, but much sooner because it's all just go to TSMC, make this instead, come back, 90-day lag.

2:32:16I think the next year alone is going to be almost unfathomable. I think next year is going to feel like the future. Yes. More than any other year. I mean, the past year or two has been a lot of interesting digital elements. But when we've got, you know, humanoid robots moving around and we have the cyber cab driving around and we have, you know, flying cars and drones, it's going to feel like the future. We're going to have the Jetsons sort of materializing before us. By the end of next year, I think so. And we have rockets flying and landing. Big time. The robot production will scale very... There'll be a shitload of robots basically in two years.

2:33:08Is that a defined unit of measure? It won't be rare. yeah well uh will uh will you offer any optimize for uh home purchase will you will you sell or only lease the robots you think i don't know yet um

2:33:29there there will be initially a scarcity of robots and then there will be robots will be plentiful so yeah but the difference the time gap between scarce and plentiful will will be only a matter of five years you know the tesla comes to your driveway now you just buy it online and it just drives up to you yeah yeah the robot just comes and ring the doorbell too probably it gets out of the tesla and comes up i mean what i find fascinating ilan is the amount of compute that you're building into things that walk out of the factory the cars and the robots the amount of distributed inference compute that's going to be in the world a lot a lot a lot yeah that's one way to scale the you know the AI is distributed edge compute so I want to ask a question I don't want to hit any hot points, but in one early on, I think you imagined OpenAI as a counterbalance for Google.

2:34:44Yeah. Is XAI now the counterbalance for Google?

2:34:51Yeah, probably. I guess Anthropic is doing some good work, especially encoding. opening has certainly done impressive work. You know, I'm sort of stuck on like how do you go from a nonprofit open source to a profit maximizing close source. Missing some of the parts in the middle. But, you know, they certainly have done impressive things. Does anybody else appear on the horizon or is it these players in China? Can somebody come out of the middle place? To the best of my knowledge, it is my best guess is that it will be XAI and Google will will be will vie for movie primacy yeah you know who who is what what is the what is the what is the best ai um and and then at the minute and at some point it's it's going to be i guess a competition with china yeah like china's just got a lot of a lot of power yes like the electricity um they've like china i think will pass three times the u.s electricity output in 26.

2:36:16And they will figure out the chips. They're going to start chip manufacturing. Yeah, they'll figure out the chips. And as it is, there's diminishing returns to the chips at this point. You know, so you go from like so-called like three nanometer to two nanometer. You don't get a three to two ratio improvement. You get like a 10 % improvement. Yeah. It's like, so it's just diminishing returns on the chip size. And Jensen has said, like, you know, Moore's Law is dead. Like, it's not like you can just make things smaller and make it better. Yeah. There's a discrete number of atoms. Yeah. That's why I think, like, you should just stop talking nanometers and say, how many atoms and what location?

2:36:57Yeah. Because this is marketing BS. So that makes it easier for China to catch up because with the diminishing of 10s. Everybody has a wall. Everybody has a limitation. Yeah. Yeah. It's like, so like, there's like, no one has near term plans to use the 5000 series ASML machines. Right. And, you know, those, they cost twice as much and can only do half the reticle. And they probably have some improvements in the works. But it's basically half the chip for twice as much for a gain that is relatively small. So anyway, the point is that China's going to have more power than anyone else and probably will have more chips.

2:37:48It's a great insight because I think a lot of people are used to the chip wars where I'm running single-threaded code, I need the CPU to double in speed and I can increase the price but i need that out in an 18 month cycle time or less we've been doing that for so long now that nobody can see that it doesn't matter you can buy intel or you can build your own fabs and you can use them for a much longer period of time oh yeah yeah absolutely much longer i totally agree in fact um so like our ai4 chip which is like relatively primitive at this point um the same fab that makes that uh if we apply the the ai6 logic design to to the fab which is it's a five sort of normally five nanometer fab yeah um we can easily get an order of magnitude better output in the same fab yeah yeah and the other thing concurrent with that is that the volume if you just 50x the number of chips can you do something useful with it you used to not be able to.

2:38:51You'd be like, well, now I've got five CPUs, but I still have the same single-threaded code. What am I going to do with five Excel spreadsheets side by side? Now it's like, no, I can translate that into useful intelligence. Yes, exactly. It's not constrained by humans. It's not a human productivity amplifier. It's an independent productivity generator. Dead right. So many people have missed this, the importance of this. And this is where China, you know, China makes far more solar panels than we do. And we're like, well, But they'll never catch up to it. By a crazy degree, actually. It's a crazy degree.

2:39:20Crazy degree. If they do that in chips, and you're like, whoa, but who cares? They're seven nanometer. Like, wrong. Oh, no. Yes, correct. Yeah. I mean, based on current trends, China will far exceed the rest of the world in AI compute. So - That's not good. So what happens then? You've got XAI and Google and China Inc. let's call it that for the moment, and you've got massive amount of ASI level compute that, frankly, the only thing that understands the other ASI level compute is the ASI here.

2:40:00Can they all just play together?

2:40:05Is it Darwinian? There might be some Darwinian element to it.

2:40:19I mean, let's look on the bright side. Let's look on the bright side of life. You might bring Grock out to speak to us again. Yeah.

2:40:33I don't know. It's just going to be a lot of intelligence. Yes. Like a lot. I mean, now the ratio of human intelligence all of a sudden asymptotically falls to 0 % on the planet. Yeah. Pretty much. Pretty much. I mean, several years ago I said humans are the biological bootloader for digital superintelligence. Yes, we are a transitional species. We're a bootloader. Yeah. We're a bootloader. I mean, the silicon circuit can't evolve in a salt pond, you know? Yeah. So you need a bootloader. We're the bootloader. Yeah, but you would never, ever impair your bootloader. Yeah, so, you know, hopefully.

2:41:23You might need it. We've probably been a good bootloader. Yeah. And it's nice to us in the future. Is this where we want to end the pod? Most people don't know what a bootloader even is. Oh, my God. yeah boot discs are a far and distant memory well we can make a uh always look at the bright side of life like clone song yeah we can clone that and make that the closing theme that'd be awesome uh i i'll go back to this is the most exciting time ever to be alive the only time more exciting than today is tomorrow um yeah and uh i mean it's interesting that we're heading towards a a world in which any single person can have their grandest dreams become true.

2:42:11Yeah. It's like Walt Disney, word for word. You're going to make that into a new exhibit. Like I said, you asked about sci-fi that is a non-dystopian future. Right. The Banks books are probably the best. You should pay a producer to go and make those. those are the culture books which is consider phlebas which is gurgich just for my wife i wonder because she yeah what the hell are you reading well the way considerably starts out is um uh yeah i mean it's it's it's a little i mean the hopeful thing is it can be starts being drowned in shit that's a good opening scene we really yeah yeah how do you not make that movie?

2:43:00it can be a little off-putting to some people you need to get through the first few hundred pages people don't walk out of a movie in the first five minutes they'll give it get into it Player of Gamers might be a better book to start off with than Consider of Gamers that I enjoyed humans still exist in this future, which is a good thing yes, they do a lot of humans in that future there are trillions of humans well, we need to get the reproduction rate up yeah By the way, my friend Ben Lamb's company, Colossal, is making artificial wombs. He's the company bringing back the woolly mammoth and bringing back the Cybertooth tiger and all of these.

2:43:39I'd like to have a miniature pet woolly mammoth as a pet. Okay. Well, you know, he made the woolly mammoth. Wouldn't that be adorable? He made the woolly mouse. Yeah, it's just like... Licking you in the face? Yeah, it's just like sort of trundling around the house. What would your optimal size be? It'd be adorable. What they've learned how to do is to... Little tusks and everything. A miniature woolly mammoth would be an epic bet.

2:44:07I mean, look what we did with wolves. We turned a wolf into a little plashtoid dog. He brought back the direwolf as well. But he made the woolly mouse. There's a woolly mouse now that exists. Does it have tusks? No tusks. different gene or what I was there he's in Dallas I was visiting him and he said our scientists are going to a tusk conference next week to talk about all of the genes involved in tusk so they want to put tusks on the mouse no they could probably add it to the mouse that'd be cute it's like a mouse sized that's just going to freak people out the little woolly mammoth will sell the tusk mouse will not sell yeah too creepy you thought Labradoodle was cool when you see the bunch of woolly mammoths yeah a little saber tooth tiger would be good too as a cat cat size those teeth come down to like here I don't know how they actually bite, but did they actually bite?

2:45:22I don't know if they could open their mouth. Not my, you know, I see teeth here. Like sort of unwieldy, you know? Yeah. They're just, for show, they look good. They're like jewelry. But no dinosaurs. No dinosaurs? Not legal or not? I think Jurassic Park's a great idea. I mean, really? You didn't see the end of the movie. The AIs will help us with that. Nothing's perfect. Oh, yeah. That went really well. I mean, if there was an island with a whole bunch of dinosaurs. Who would go? 100%. Yes. Yes. It'd pay a lot for that. Yeah. It's like once in a while somebody gets chomped by a dinosaur. They'll be like, ah, what's the one in a million?

2:46:04I'll still go. Who are they missing? Lysine? No. No. The DNA, the oldest DNA that's been recovered is like 1.2 million years. Oh, you can just wing it, though. Yeah, just make it look like it. that whatever this would be one of the actually that was my proposed x-prez remember back in visioneering what's that take the dna strand and predict what it'll look like yeah yeah exactly yeah they just make it that way yeah and then just reverse engineer reverse engineer the dinosaurs yeah exactly it'd be funny if there were two completely different dna strands they're like well they both look like t-rex that's interesting how they is t-rex real or is that like an assemblage of real well that'd be funny i mean it's nice to believe it's real but The front legs were from a completely different size.

2:46:49That was the one at eight. It actually had huge front legs.

2:46:56Is there something wrong with the arms? I don't buy out on the arms front. Come back, I'll eat you. The mini arms seem implausible. nope well dna will tell us we'll know in a year yeah the future is going to be jurassic island we say wow yeah i go we got no no i meant the amino acid that the dinosaurs were missing that kept them from reproducing what lysine you're saying was it lysine i forget what i remember but the dinosaurs got held back by something like an asteroid bombardment. Right, right. They were doing great. Yeah, 60 million years. Yeah, they were doing fine. Yeah, we got very lucky.

2:47:42They had a great one. See, there's a good argument why there's no other intelligence out there. There's plenty of dinosaurs in the universe. What were we back then? Like a bowl or something? Yeah, we were a furry mammal. Let's commune with the ancestors.

2:47:59We were very good at hiding. It is amazing. from a little rat, little mole to us in 60 million years. It doesn't seem that long. That's why no one believed Darwin. It doesn't seem plausible. It's a long time. It turns out it is. You're making robots, but it's interesting. I think it'll be a lot more interesting to design biological robots, like a little cat that goes around and pees stain remover and eats lint off the carpet. but that's going to be an interesting thing. But you have a mechanical, like an Optimus Light doing that anyway. Yeah, well, they went bankrupt, so we'll have to build this.

2:48:37I think you can still buy them, though. The room is basically that. It's going to be... But the thing is, like, a humanoid robot is general purpose, so it can do whatever you want. Yeah. Yeah, they were too early. No vision system, no GB300. How do you build a Roomba that works? I think the idea of having an Optimus vacuum is like the most underused asset but it can just do anything it can, yes of course yeah so and you can mass manufacture at you know Optimus build me a Roomba that's what you'll do you won't say Optimus vacuum but curve it, Optimus build me a Roomba that vacuums it, that's much better yeah it's going to be a lot of robots we should do this once a year at a checkpoint I would like that that's going to be we can roll back the what did we say a year ago what were our predictions last year yeah yeah alright we can always control it we can cut cut the rules right are you selling hope as a matter of fact it worked out really well you pull up in your Tesla like hey I bought this dollars per hope for you.

2:49:55I'll send you the mug. Monetize hope. Monetize hope. One year from today, December 22nd. I'll come in the door right here. If you're here, you're here. We'll talk about you. A year from now, we might have the new Optimus factory where the building will be built. That would be awesome. 8 million square feet of robots. It's going to be a giant building. oh man um yeah and uh yeah they freak me out when they're recharging i was like hanging there it's like what's wrong with that thing yeah we're gonna we're just gonna have them like i think sit down yeah uh as opposed to look like some sort of they need like a like a recharging cigar recharging cigar less mug like just napping here with a book yeah that'd be much better right Right now, they're just like literally, like, is it dead?

2:50:54Just limp. Yeah, that's a good point. That's a big contribution from this particular pod. All right, till next year then. All right. It's a date. Thanks, buddy. Awesome, guys. If you made it to the end of this episode, which you obviously did, I consider you a moonshot mate. Every week, my moonshot mates and I spend a lot of energy and time to really deliver you the news that matters. If you're a subscriber, thank you. If you're not a subscriber yet, please consider subscribing so you get the news as it comes out. I also want to invite you to join me on my weekly newsletter called MetaTrends. I have a research team.

2:51:30You may not know this, but we spend the entire week looking at the MetaTrends that are impacting your family, your company, your industry, your nation. And I put this into a two-minute read every week. If you'd like to get access to the MetaTrends newsletter every week, go to diamandis.com slash MetaTrends. That's diamandis.com slash metatrends. Thank you again for joining us today. It's a blast for us to put this together every week.

2:52:17There is an abortion to land. Zaman.

2:52:26Zaman.