We're Not Ready for This

20:35David Mattin:Yeah, same, 100%. You know, again, this Citrini essay comes out and now I'm like, right, you know, I can respond to this today. because I can just accelerate that process of thinking, sparring, delivering an essay has just been accelerated to light speed. And so now I feel like I should. Now I feel like I must. So once again, the leisured week or day that we thought technology would deliver us into, perhaps it's not going to work like that. We're just going to feel the demand for more and more and better and better and better output. But if you can use these things and do it in a creative way, you can just supercharge yourself and do amazing things that you couldn't before.

21:19Raoul Pal:I did the same as you. I trained it on how I speak. I gave it lots of GMIs. I gave it structure and understanding. And I talked to it a lot. I like a thought partner all the time, you know, thinking through different topics. And I was trying to think through the Japanese economy. Everyone's like, it's going to blow up. and I'm like, I know there's something wrong here, but it's not going to blow up. So I was trying to just think it through because there was something stuck in my head and I hadn't verbalized it and I didn't know. So I was talking to Claude saying, listen, I think there's something else going on here, blah, blah, blah.

21:50Raoul Pal:Anyway, so I develop a whole thesis with Claude chatting about this stuff. I'm feeding it charts saying, look, this suggests something else, blah, blah, blah. Anyway, it comes to writing GMI. And I'm like, I think we've probably got an essay here. Claude's like, yeah. I said, let's put it all together in an essay because I think this is really interesting. One shot, four and a half thousand word essay. Yeah, yeah. But the beauty of what's happening there. It wasn't AI slop. This is a structured GMI thought piece based on hours of conversation.

22:23David Mattin:Yeah, and that's what I was going to say is that that is so important for people to understand is that the power of that is that you're bringing huge context and huge understanding and all your ideas and frameworks, right? In that conversation, But actually going back years, going back like decades, essentially, you are bringing all of that into contact with this model that's now really powerful to create an output that is differentiated, that is valuable, that is amazing. And so this is going to supercharge people who have that kind of IP and context and understanding and frameworks to bring to the models.

22:59David Mattin:And the only way to make them legible to the models at the moment is to put them into words. So this is a great time to be someone who writes things down, who writes their understanding down, who writes context down. Or verbalise. It's a huge superpower. Or just speak them out. Yeah, it's a huge superpower. And if you can make yourself understood clearly in words, you can do amazing things with these models. It's an incredible time.

23:24Raoul Pal:So a quick break in your regular programming. If you're serious about your future, grab my free report called Prepare for 2030. I think you've got five years to make as much money as possible. And this guide will help you navigate what's coming. The link is in the description. Download it now. Yeah. And what you'll find is most of us are not kind of balanced in how we operate. You're either detail people or you're creative minded people and you have a big weakness. There's very few. I mean, like if you're really good at both, you tend to be an architect, you know, because you have to be, you know, down to physics level perfection plus design.

24:05Raoul Pal:It's really hard to do. So what I kind of the deal I made with Claude is like, we're going to partner. You know, I don't use it as a tool. I use it as a partner. And I'm like, I can connect things. I can hold world models in my head, construct them, deconstruct them, think about them look for connections and i need you to then prove them out so i'm going to tell you stuff like i think that the hindu vedas are connected to to um quantum physics and i can see it in my head i try and explain a little bit and then i go to claude and say listen what do you think about this and it might turn around so that's utter horseshit or it might say well here's your connections and this is how the vedas work and this is how the similarities are here and here's where they drift.

24:55Raoul Pal:That is unbelievable. Now, let's say the opposite. You're a detailed person. All you see is detail and you're trying to create a constructive business plan and you need creative thinking. You need to put a pitch deck together and a website. Well, it can be your creative partner. It can be the other side of you and you become twice as powerful.

25:12David Mattin:Yeah. Yeah. And that sense of it filling in, filling in your gaps or filling in your sort of weaknesses, which obviously we all have is really, really powerful. So the other way, you know, to think about how you use this and to supercharge yourself is think deeply about yourself and your cognitive style and what you bring, but also what you don't exactly, as you say, and then really get it to fill in those gaps in your understanding, but in your temperament, in your approach. It's just so powerful. And that's my overriding feeling is, yes, it presents all kinds of challenges and the future for kind of creative forms is uncertain.

25:50David Mattin:But I really do believe that people are going to do amazing things with these models and that still bringing context and understanding and bringing you will be part of that value. They're not just going to render creativity obsolete and render humans obsolete in that way. So it's a really, really interesting time. And intellectually, I feel like I've been able to do things and go on journeys of understanding and learn just learn things more quickly than ever before and i'm obsessed with telling my children this as well you know i've said to you countless times my son's learning to code and i've not taught him any of that you know and it just it kind of started with him just cut and pasting but he starts to understand like what he's because he's you know the kids learn fast as well

26:35Raoul Pal:it's like learning language if you're living in spain the kids end up learning spanish and the

26:39David Mattin:Exactly. Yeah. Yeah. You know, you just just go and swim in this stuff and have someone to talk to about it and you start to understand it. And it's just so powerful.

26:49Raoul Pal:Yeah, it's a it's an extraordinary thing. And what happens next is like Dario was talking yesterday. Dario from Anthropic was talking yesterday. It's like. They're all kind of wide eyed again, saying, oh, no, it's about to have another step change. yeah so what they're alluding to is an enormous step change still again this year and everybody's talking about it elon's talking about it dario's talking about it

27:20David Mattin:and sam altman's talking about it yeah and it's it's hard to imagine and i i think what that looks like is models that are just even more agentically capable and able to operate autonomously you know on hours and hours long tasks, days long tasks, you know, far fewer mistakes, far more sort of coherent causal chains. And just I mean, we also have them with like OpenClaw and now obviously OpenAI owns that, owns that company. I mean, company, it's one guy. Just absolutely amazing. Of course, if we're talking, if this part's about the broader state of AI, you know, very interesting stuff happening with Anthropic and the Ministry of War and how all of that gets resolved.

28:01David Mattin:And again, we've talked for years and I've written about there was always some big intersection coming between, you know, the most powerful technology in history or certainly one of and government. And how you what we're seeing with Anthropic and this argument with the U.S. administration is just the early glimmerings of this intersection and how that gets dealt with. you know and i don't think anyone really knows that but the u.s goes back to the idea of

28:29Raoul Pal:the economic similarity and the exponential ages we don't have the infrastructure at political level geopolitical level in any way or societal level to deal with what is happening and the speed of which it's happening yeah and if so we've got this you know yes the government wants to use it etc but the whole of society is not ready if if if Dario's right in what he's saying we've got another step change so basically as Elon says it's kind of AGI this year what your definition of AGI is it becomes largely irrelevant at this point we've now basically got to the end of all of the tests that we can test AI for intelligence it's basically been trained on beating those benchmarks So then it's the practicality benchmarks.

29:16Raoul Pal:And we're already if you and I can create bloody dashboards and programming them, it's pretty much there. Right. But I just don't think society is ready for where this is all going. There's a couple of breakthroughs for me need to happen. One is longer memory. Yeah. So I'm having to, with actually one of the GMI, not GMI, Real Vision subscribers, Lance, we're building a RAG database of 21 years of GMI and then add everything into it. And then we'll get Claude to be able to read that database. using an MCP, which is like the agentic ability to use that file server, because there's no way of getting full intelligence.

30:10Raoul Pal:So what people don't realize, and it seems to be more obvious on Claude than it was on ChatGPT, is Claude comes in every day like this. Oh, good morning. And it has to read all of the context it can find about you every day. So it comes in with amnesia the moment you summon it, and it has to get up to speed quickly. Yeah. And that process is somewhere soon. They're talking about it maybe for the chat GPT 5.4, whatever, that's coming out next week or something. The suggestions that they solve that and then the context window. Because contact windows, I run out all the time on Claude and it crashes.

30:51David Mattin:100%. Yeah, I think that's exactly right. Memory is the thing to solve, to deliver that experience of the kind of dream AI companion that really knows you, the way a person can know you. You could do it on a local drive, but most of us don't want to set up all of that stuff. No, exactly. Exactly. Yeah. I mean, again, I'm such a sort of straightforward user, but I upload all my sort of GMI and exponentialist essays to a sort of clawed project. And it's like it maxes it out. Like there are essays I can't upload to one project. I tried all of that.

31:23Raoul Pal:And the problem with projects is it's not searchable in your general chats. Yeah, exactly. So they're separate buckets. And I spent a long time talking to Claude about, you know, how do I get around this? That's one of the reasons I'm building this RAG database. Yes, you can do it with localized stuff, but that's not the route I want to go down. But soon they will fix memory, in which case it never forgets anything you ever said. And that will give itself recursive learning based on you. okay that becomes mind-blowing and the context window clawed is what 200 000 tokens um i think chat gpt might be a million or gemini's a million and the new chat gpt they say is going to be two million okay now you've got big context windows where that's huge bodies of work still not big enough yeah um but but if if the memory is infinite then you can upload i could upload 21 years of gmi

32:14David Mattin:and everything else and it won't forget it exactly and that's just you know that is cognitively supercharging just in insane ways if you've got this this this virtual mind essentially that just can know everything you've written and go back years and join the dots um yeah and i i bump into the same frustrations you know and it's sort of i'm constantly getting like oh i can't access that because it's in a different chat and that kind of thing i mean still it's still amazing and i still like do okay just you know running it off a project but for someone like you with a like 20 year archive of these of gmis yeah you 100 need some at the moment some external database that it can go out and search yeah yeah we just feel like we're at i mean everyone can feel if you use this stuff you can feel that we're at such a powerful moment and we're expecting further really important launches this year and probably sooner rather than later this year which is insane like crazy speed and you've always said this that you know they've talked to dario is talking more and more about how they're using ai you know to to make the next iteration we said this a while ago we knew this was about a lot you know and and you can feel that happening now you

33:28Raoul Pal:can feel the the acceleration happening there well i mean peter diamandis wrote well we're getting a new model every three months it's soon going to be every month and i think we're pretty close to that now um where we'll get a new model or iteration of the model every month and between it groundbreaking applications layers one thing that's also just looking at the models is one thing you notice is on claw that has keep compacting conversations and what it does is then forgets other stuff and then it can hallucinate stuff i was speaking to somebody about this and i saw another interview about this it's like actually this is the human process so you can remember pretty much what you were doing 15 minutes ago before we came on this call.

34:07Raoul Pal:If I said, what were you doing the exact time yesterday? You kind of might get it, right? If I said, what were you doing the exact time on March the 4th, this time last year? You're likely to say, I was probably at my desk. If it was this time, I might have been having a cup of tea. I was probably on my computer. That's all hallucinated. That's inferred by your probabilistic framework of what it could have been. you might have not been there at all you might be another dentist but you won't remember because you're not looking at your diary and what humans do is compact and create these little compressed files of stuff that we remember we know that you know if you speak to your brother or sister about something that happened when you were kids you have entirely different memories of the same event yeah yeah yeah exactly and what we're going to do is change that we're going to have a creature with perfect memory recall.

35:01David Mattin:Yeah, yeah, exactly. And we're so, you know, all of this is just so fascinating and so new. And the science and engineering and the art of growing these models needs to come into interface with sort of human neuroscience and psychology so much more, because we're going to learn such profound things about how the human mind works through the construction of these sort of virtual minds. I'm not saying they're necessarily the same at all, but we're going to learn so much and that that compression thing you talked about is so fascinating right because basically what humans do human is we compress our memories of of things that are the same like all those days were the same and i'll just compress them down to one generic day of me at my desk having a cup of tea right and you know me so well ral because that is my overarching memory of the last five years right um and that's why as we get older and and we have fewer and fewer novel experiences in our memory or it feels like life is going faster basically because we've compressed um a year down to like one little file of generic like at my desk and then went on holiday and then came back to desk like it's like that um so we're going to learn so much about human psychology by growing these artificial minds it's absolutely fascinating that discipline is so young so we're at the we're at the eve of just incredible

36:26Raoul Pal:I'm much more radicalist in all of this, that we are, you know, this everything is compute idea. And that's a whole topic for another day that I think we might turn the tables when you talk to me, because I need to get all of this stuff out. And as you know, there's a lot that I've been working on. Yeah. Let's move on to the next part of the framework, energy. Yeah.

36:43David Mattin:Energy is fundamental.

36:45Raoul Pal:The process is, just to remind everybody, it's intelligence output, which is accelerating, going exponential, because we've compressed all of human thought into the internet. and we're now building upon it. And what we're getting is also more intelligence per unit of sand. That's what NVIDIA are doing, turning sand and TSMC are turning sand into intelligence. Into thought, yeah. By putting energy through it. I mean, fuck me, what world is this? What world is this? And that's acceleration. Let's talk about the energy side. because this is where there's a lot of kind of mid curving of well you know this is never going to happen because of energy but we think there's exponentials going on all over there as well

37:30David Mattin:yeah i mean energy is fundamental excuse me and the intelligence per unit energy framework tells you that energy is the binding constraint you know i mean where we are right now is that we want as much intelligence as we can get per dollar per unit energy you know we're not in a world of radical material abundance yet. We are also constrained by, we have material constraints. Though in practice at the moment, if you look, it's not the material, it's not the dollar that's constraining us, right? The CapEx is insane. $600 billion forecast CapEx by the hyperscalers in 2026. So an industrial scale build out of infrastructure comparable to like, you know the the laying of the railroads the laying of just just as a quick interjection there

38:24Raoul Pal:for people who are cynical anthropic is the fastest scaling revenue company in all of history i mean it was like on a 10 billion run rate in december last year yeah it's now 19 billion run rate and it's a company that's only been having revenues for like three years yeah it's staggering

38:44David Mattin:Exactly. And that's why I take this route into this, because, of course, the big question on so many people, we want to talk about energy, right? But the big question on so many people's minds is like, oh, my God, you know, the CapEx, the CapEx insane spending, like, where's the revenue? Where's it going to come from? Isn't this a bubble? Well, the CapEx is huge, but even if you're just seeing this, and we can talk a bit about this later as sort of infrastructure build out now in search of profits, they're building off the back of an order book that supports that building at the moment. There's a backlog of orders.

39:20David Mattin:There's more people that want to use this than there is infrastructure at the moment. And the capital is abundant. There's$600 billion being spent this year. What could stop this train is the other binding constraint, the one that intelligence per unit energy tells you about, energy, because there's insane demand for energy. And these data centres, you know, I mean, I think it's the IAE says it's going to, you know, data centre electricity usage is going to have doubled by 2030. 30 you can see what the chinese are doing with their build out of solar has gone exponential

40:01Raoul Pal:this is the point so the entire global system is now coalescing around this one process intelligence per unit of energy it's china's entire purpose the u.s is coming to its entire purpose even europe plays its part in all of this and so So the only way to solve this at scale in the speed required to not let another nation's state take the most powerful technology the world has ever seen and have a two-year lead, three-year lead is solar. And China figured this out fast. China added more solar in one year than all of the rest of the world's existing solar. The US is now, the solar rate of expansion is faster than all other forms of energy.

40:49Raoul Pal:China's still going Europe is going fast with this and it's Elon as ever who got down to it it's like you say energy is a constraint it's not a constraint the constraint is building the things to capture the energy and it's because we have the Kardashev scale is the sun is more than adequate for all the energy we need we just need better technology better batteries and to be able to manufacture these damn things as fast as possible And China's showing it's possible.

41:18David Mattin:Yeah, it is possible. My concern is that China's so far ahead. You know, perhaps we can show that graph as well. It's one Andreas and I talked about last week in a chat. You know, as you say, their build out of solar has just been exponential as is gathering pace, but it's nowhere near that.

41:38Raoul Pal:They've had to because they've got no oil. The US has oil and gas. So the US has a substrate of energy that China doesn't have. And if you look at the world around you, where are all the flashpoints? Well, Venezuela or Iran, you know, all of this is securing the energy, making sure it flows cheaply. Get rid of the blockages, you know, geopolitics of Iran and Venezuela. You clear those, you try and get the energy price lower. But the US controls most of the world's oil.

42:08David Mattin:Yeah, yeah. And this exactly. And this is what if you start to see the world through the lens of a quest for more intelligence per unit energy, lots of apparent chaos starts to make sense. And, you know, you and I, we're not saying all the world's leaders and everyone's walking around with this framework, you know, explicitly in their head. It's just that capitalism and markets and the entire system is a hidden hand that guides towards a certain imperative. And what we're saying is it turns out that is the imperative. It's about turning. In the end, the whole game has been about turning energy into intelligence.

42:45David Mattin:And in your universal code, you know, you project that out even to a cosmic level. You know, this is about deep imperatives in the kind of universe itself and the economy.

42:56Raoul Pal:The code of the universe. Exactly.

42:57David Mattin:It's the code of everything. And the economy and markets and all of that are a manifestation or one dimension of that even bigger picture.

43:06Raoul Pal:And so the energy equation, I mean, again, as you say, it explains what's going on in geopolitics. It explains Greenland. Greenland is rare earth metals plus, you know, strategic protection, stuff like that. It explains China's build out of solar. It all of these things. It explains what Taiwan is all about. Taiwan is a bottleneck. yeah you and i've talked about this at length is like okay tsmc plants are being built up elsewhere but still taiwan has all of that manufacturing the foundries and the foundries are like the most important things in the entire equation without those there is no intelligence yes now if china were to invade taiwan then the foundries would be destroyed somehow whether the US destroys them or whatever.

44:00Raoul Pal:And if the US doesn't, it's very hard to go kinetic over Taiwan without this race towards intelligence coming to a screeching halt. Yeah.

44:10David Mattin:And, you know, we've talked about how in the end, no one wants that. The sort of, you know, the motivations in play, we think will stop that kind of sort of intelligence catastrophe because it's just catastrophe for everyone and that some kind of grand bargain will be established there. I still think that's the most likely kind of outcome.

44:35Raoul Pal:The other thing on the solar side or the energy mix, right? So solar is the fastest to get up. Then it's probably gas plants, but gas plants, you need gas pipelines and good luck trying to run them through California, but you're okay in Texas kind of idea, fine. the cost of energy of solar electricity is coming down dramatically it's an exponential curve um so that that seems to be all playing in the right direction and a lot of people say but you can only store it for four hours overnight and yes battery technology is improving fine but people forget that you can actually, because of the sun, if you're in Texas, let's say, you're going to get a reasonable amount of sun, 320 days a year, 320 days a year, Southern Texas.

45:28Raoul Pal:You're going to be able to generate electricity for eight hours a day. You've got four hours of which you've got battery. So you're now at 12 hours. And then you've got 12 hours of using the grid. So you've halve the use of the grid for every incremental new dollar of energy.

45:46David Mattin:Yeah, that's right. That's right. You know, these technologies are only going to improve. I mean, I think, yeah, the very first essay I wrote for the exponentialist and GMI, or for GMI, sorry, was all about this energy intelligence flywheel and this deep symbiotic relationship between energy and intelligence, where AI is going to help us manage the energy grid, and that energy will allow for more AI, which means better management and you get this you get this flywheel effect and so even there in that first essay I was sort of searching for this deep relationship between energy and

46:20Raoul Pal:intelligence that I could feel was there and what you're feeling is it's going AI is going to

46:25David Mattin:self-optimize for this equation yeah and it's only going to get faster exactly because that exactly the other half of the the intelligence story is yet look we need to do the solar build out. And I totally agree with solar. Solar is the way to do this and it's the way it will be done. But there's also algorithmic efficiency. So I'm very interested in looking at AI frontier companies and AI startups that are finding ways to make the models more efficient, to wrench more, to squeeze more intelligence out of the model per unit of energy. And if you look at DeepMind, you know, they recently released an algorithm or a model called Alpha Evolve that's all about this.

47:09David Mattin:It's all about teaching other models to be more efficient, you know, and that was exactly what I was writing about a few years ago. This is, that's AI, you know, starting to make AI better to make it more energy efficient so that we can have more intelligence per unit energy.

47:29Raoul Pal:so the next part of this we've talked about the intelligence going exponential we've talked about the cost of the energy even though you can't see it in grid prices yet but the output per solar and all of this stuff is exponential we're seeing china scaling it exponentially so that word exponential appears everywhere here which is why everything is moving so fast and nothing makes sense anymore but what we're about to do or are doing is we're now scaling these things into the next iterations which are more like replacement humans units so firstly let's talk about agents agents is the big thing that happened this year yeah this year alone i mean we we had them last

48:10David Mattin:year but this year they fucking exploded they really did you know i mean the this year kind of kicked off with the with the open claw incident phenomenon whatever you want to call it where Yeah, you know, this this lone operator essentially made this fantastically capable agent. And suddenly everyone's download or everyone, lots of people are downloading them, running them on a Mac mini. You know, you can set it, you can connect it to your Telegram, to your WhatsApp, talk to this thing like a colleague. And no one really quite knows. I think it's fair to say the sort of secret source that he applied because you would then connect it to like your, you know, instance of chat GPT or your instance of Claude, whatever.

48:53David Mattin:So that's providing the intelligence. But he somehow built an infrastructure that just made this agent fantastically capable. And people are sort of, you know, communicating it, talking to this agent via WhatsApp saying, build me a website, you know, code me this tool that's going to take all the data out of Real Vision and show me what the viewers are doing, whatever it is. and the agents going off and doing that very successfully. Then someone made a kind of Facebook for these agents called Maltbook and they all go and start posting on Maltbook. As I wrote about in The Exponentialist, like we know that some of that, there were shenanigans, there were people doing some of that posting.

49:29David Mattin:But still, what it was, was a tiny glimmer of what I call and have written about the coming post-human economy, an economy populated by autonomous economic actors, AI agents and robots, which we'll talk about in a bit. It was just a tiny glimmer of what that world starts to look like. And it's radically, radically different. You know, and these agents started to transact with one another across blockchains, which goes back to what Jamie Coutts, who does such brilliant work on Real Vision, Jamie, you know, who talks about blockchains being what people forget and are not pricing into the whole blockchain story is that they are the economic substrate of the coming economy, an economy of AI agents.

50:16David Mattin:And I think that's so powerfully true. And all of that was just a little window into this future. So it was exciting.

50:24Raoul Pal:Agents are autonomous or otherwise, or semi-autonomous as most of them are. They're becoming economic actors. Then look what's happening. Google have rewritten how web browsers work. So agents can go and take the information off the web without having to read and click buttons. All of agentic commerce is being rewritten using X402, which was an old embedded in HTTP that base started reusing. Sui got their own version, A402. We've got MoonPay, everybody else, everybody's building agentic models. We've got Stripe saying agents are going to be 100x the size of all individuals doing transactions. So we've got the web, the predominant user of the internet will be an agent.

51:14Raoul Pal:The predominant user of finance will be an agent. The predominant user of blockchains will be agents. What I've realized, we kind of knew this, and you and I have talked about this for a while, so there's no shock. But what I realized, and we've always said, you know, these are going to be replacement humans. But I was thinking in human number terms. And we've all got the fucking tam of everything wrong. We've got it so wrong. So total addressable market for blockchain. You think, well, there's 650 million active wallets in the world. Yeah, how's that calculated? Might not be perfect, but fine.

51:48Raoul Pal:We'll use that as a benchmark. Could we saturate half the world like mobile phones did in their first 15 years. So we get to 3 billion. That's what's in our heads. We're going to have 3 billion agents within three years. And then we'll have 50 billion, 100 billion agents. We have no idea how many. And Stripe have been saying this and Circle have been saying this and Google are saying this and everybody's starting to say this is like the predominant user of everything, all software, everything is going to be an agent. So our TAM is so wrong. People think the revenues are going to collapse for a lot of things.

52:28Raoul Pal:I'm not entirely sure that's the case. I think several things will see a massive explosion in revenues because they become vital to the Asian economy.

52:37David Mattin:Yeah, I absolutely love this thought. It's such a powerful thought. Like you say, it's something that's sort of been – and this happens so often with us, right, and with anyone thinking about this. It's not like we're unusual in that respect, but you have this hazy model and there's something you can you can sort of half see, but you can't quite grasp towards. And this this is a classic example. You know, we've always had in our head agents are coming, autonomous economic actors are coming. You know, we're trying to put that puzzle together. What does it mean? And we glimpse bits of it. And when you articulated that, you know, that everyone's got Tam wrong and everyone's models of what future Tam is going to be is totally wrong.

53:18David Mattin:It's like, fuck, yes, that's it. That's another little piece of the puzzle uncovered. And because you think, oh, economic actors are coming, but instinctively you tend to think, one tends to think, so that that will replace the 50 million human users. well no no no it's going to be like billions of users because there'll be billions of agents and it's just another glimpse of the underlying thought that we're moving to something so radically new that our old measures almost start they just start to break down like does tam even make start like really make sense because there's no scarcity exactly it kind of doesn't like tam breaks down um our old measures break down you know i think about this with the whole jevon's paradox thing as well you know i think people the say in a way i can't quite articulate yet i think the same kind of thing is happening people are saying well compute will get cheaper um but what that will mean is that people just use more of it you know so you don't have to worry about the revenue kind of dying people just use more and more and more as it gets cheaper and cheaper and there's something in my head saying yeah you know but people won't be the main user of it you're going to get some kind of recursive jevons where ai agents use more and that makes it even cheaper so they use even more and it just like goes vertical in ways that are very hard to imagine and part of my universal code thesis is that all intelligence creates coherence and the network i mean every single

54:54Raoul Pal:fractal part of everything is a network from the human genome to the to the body to the brain to the forest to the i mean everything isn't rivers everything is network connected and that's how information flows and that's how you create coherence and that's how you create intelligence my view has been for a long time that asi super intelligence is the network of networks of agi so that the ai's are already learning from each other they don't have obvious networks yet but they are mold book was giving agents a network people have no comprehension what this means because soon once they start creating a network they start creating coherence amongst all of these and what you do is accelerate intelligence and that's where you get to asi because you're going to network all of these things and we won't even know it we won't see it because they're interacting with each other paying each other learning from each other swapping information with each other building data with each i mean it is beyond human control at this point because we use the words semi-autonomous agents we don't even have to be fully autonomous yeah and that process happens

56:06David Mattin:yeah it's it yeah that's so powerfully true and you're right maltbook is a great is again a tiny little window onto the way they will start to learn from one another because

56:15Raoul Pal:everything forms a network in the end yeah yeah yeah it doesn't it won't survive as a species or a pattern or whatever it just doesn't it doesn't work without networks i mean that's dna then you know we're built from network i mean everything is a network so there's something in that so let's go on to the next bit so the agents are coming the energy is getting more abundant and cheaper but is still the bottleneck. So that is a big race that has to happen. We've got intelligence now becoming Reid's law. So it's the exponential squared, which is why we all feel like everything's breaking. Exponentials are hard enough.

56:56Raoul Pal:But when every log graph does that, you know you're into a whole different world. Because having spoken to my chat GPT on this a couple of years ago, is there any other example of reed's law it's virtually none yeah it might have happened in bacteria but they're generally metcalfe's law they're generally logarithmic um it's very hard to have a log squared and so it's one of the few observable log squareds that's ever happened in history and we're living through it which is living through it i know remarkable and then i'm going to talk about the other part so the agents people got wildly wrong because this infinite tan and now I'm going to go to the other bit so when we talk about robots people think about factory arms assembling cars or they think about c3po doing this and being very nice chap who makes you a cup of tea and can do some stuff but it's a bit dumb okay that's that is humanity's mental model of a robot or daleks or right they're just constrained in both intelligence and ability the truth is we're going to put fucking AGI brains into super creatures super bionic creatures that are stronger more powerful faster so they're better than us physically in every way and mentally better than us in every way apart from maybe not having qualia That's it.

58:29Raoul Pal:People are not ready for this. Society is not ready for this. Even Elon said today, he goes, we may not be the first to AGI, but will damn well be the first to AGI in a robot body. And people are like, oh, yeah, that's amazing. No, that is a new species. Yeah, they're not ready. They're not ready. You aren't ready. I'm not ready. Elon's not ready. Nobody's ready.

58:54David Mattin:Nobody's ready. How can you be ready for that? Yeah, I mean, that is the next huge frontier is this projection of AI and super intelligence into the physical world. And, you know, perhaps the most exciting and impactful version of that is humanoid robots. And in terms of the brain, we have incredible large language models. This thing can understand you as it understands so much linguistic context, incredibly nuanced. We have great machine vision models. It is the physical challenges that are the final push. And Elon talks about this as well. I mean, how hard it is to get the hand to be as dexterous as a human hand.

59:41David Mattin:And, you know, these humanoid robots, they're still a little slow. They're not that nimble yet. But we will get there. We are getting there as we speak. And that's just going to be ridiculously impactful.

59:52Raoul Pal:If we go back, how many years ago? Was it five years ago where Elon had people dressed up as robots on stage? Yeah. Look where we are today, right? Yeah, exactly. Robot development is going exponential. And don't forget, once you apply AI to every task, every task goes exponential. I'm going exponential in my output. You're going exponential in your output. everything is going exponential and it's so hard to deal with so let's finish up and the thing that you and i try and anchor everybody on because a lot of people watch this they start getting they go through emotional journey which is like wow this is so cool well fuck i'm fucked oh fuck my kids are fucked everything's fucked i don't know what to do anymore so i'll start with the first bit then you'll start this you'll do the second bit my first bit is the counterbalance of this is nature.

1:00:38Raoul Pal:If you go through your mind, we're going to have drones with AGI brains delivering food, policing you, surveilling you, helping you. You're going to be, the cars are going to be self-driving. There'll be humanoid robots in the streets with AGI brains that are smarter than you, demand your respect, will maybe be part of society. You've got every computer is able to think. You've got unseen economic actors in their billions. How do you off ramp to that? And the off ramp is nature, right? Nature is this beautiful, amazing ecosystem that we live in that is all part of this universal compute, but its resonance is different.

1:01:22Raoul Pal:It grounds you. It gives you a break from all of this. So that's one side of the equation is I think more people will end up living closer to nature to escape the more extremes of the exponential squared happening everywhere the other side is what do i tell my kids that's your special topic so give a quick yeah people want to hear it and people who who are not part of the exponentialist haven't really heard this enough because it's the thing that it's the fundamental thing most people say well fuck it if if i get replaced i'll figure it out but my kids poor kids how are they going to

1:01:59David Mattin:get out of school and come into this world yeah it's a huge worry and we both we both feel it and we sense it really powerful and we talk to so many people about this and you you get you know as i've said so many times in the past when i go in and speak in organizations the number one question i get is not anything to do with work career you know the economy is what do i tell my children about this how do how do they get through this what does it mean for them are they going to have any kind of life at all um and i just fund you know there's so much i can say about that i wrote big essay about it we can do a big session about it one day but fundamentally i just go back time and again to look we're talking about a new kind of technology here intelligence that is going to colonize so many domains of human activity um it's going to be stupendously productive incredible outputs it's going to you know it's going to analyze it's going to know it's going to write, it's going to produce.

1:02:55David Mattin:What you have to ask yourself is what is left for us amid that explosion of capability. And by definition, there's one thing that no machine and no machine intelligence can ever do. And that is be a human being. And it's important here. I'm not saying it can't ever be conscious. I'm not saying it can't be alive, but what it can be conscious it can be alive but it's it won't be a conscious alive human being it will be something else we are still human beings and we are always going to want to commune with other human beings in all kinds of ways who understand how it feels to be human to have that particularly human kind of experience and all kinds of value like meaningful life and purpose and all kinds of value exchange can still flow from that.

1:03:50David Mattin:So, you know, encourage your child to, I say, lean into their weirdness in the essay, you know, cultivate that part of them that is like uniquely human, that is uniquely them, that is where they're passionate, where they're, where they feel strongly and teach them also to be able to communicate that passion and be compelling to other people in front of other human beings because i mean if you're worried about work in particular so much like human value exchange will flow in that direction it will be about community connection counsel care for each other truly seeing each other doing the thing doing that small bucket of things that an ai however intelligent it is can't really do it can do a simulation of it and that would be good enough for lots of people in lots of contexts but we'll still want another human being going oh my god like you know for example this thing's happened to you it happened to me too once i know how you feel this is how we get through it and so on so cultivate that uniquely human part of ourselves is is part of how i think we we find a way through this you know we navigate our way through it and it can be amazing for us but i just don't want us to blindly accept this narrative of human obsolescence.

1:05:10David Mattin:I don't think that's where this has to go.

1:05:13Raoul Pal:I just had a breakthrough, I was thinking, while you were talking about obsolescence. So the longest surviving creatures on Earth are sharks and crocodiles. They're hundreds of millions of years old, and they have not evolved at all since, which suggests that they're perfect for the task of which they do. Some, I don't know where you want to put the time, 50 ,000 years ago, homo sapiens come. We're the apex intelligence. Yeah, humanoids before us have all disappeared, but crocodiles and sharks haven't. We're an apex predator. we can kill every crocodile shark we want we can kill all of the food that they hunt but they still exist because it tells you it's not a replacement the whole system requires diversity to function if the universe the planet the ecosystem is all part of the same process then diverse diversity is vital or it collapses because you have one failure point and it's again something I've written about in the universal code.

1:06:31Raoul Pal:And that's why these things exist. And that's why humans exist. Humans have qualia. Now, do dogs have qualia? Possibly. Certainly not the same extent that we do. Qualia is the sense of feeling that humans have. That sense that we make of the world. Why do we have it? We have it versus others because we are the top compute on earth, the top intelligence. And why we got there is our qualia makes us more curious. It drives us to do weird things. We have things like ambition. Does your cat have ambition? No. Maybe it's a piss you off occasion and scratch your face, but they don't have ambition, but humans do.

1:07:15Raoul Pal:We have this really weird thing and it makes us high quality diversified compute. We compute at the edges and not just in middle. Most machines will just compute in the middle. So we're a very, very important part of the ecosystem that creates intelligence. Much like we know we can't destroy all the trees on Earth, because the intelligence of trees is their process by which they generate oxygen that allows us to live. It's all part of this extremely interconnected thing. And without us, the AGI doesn't get our qualia it may never have qualia I think that's right I think that's I think

1:07:55David Mattin:that's right and we did promise that we like to live at the weird frontier and what now everyone's comfortable that you know you can use Claude Code to design themselves a website we've moved on to these are the kinds of conversations we have in the exponentialist and we absolutely love it and and yeah fundamentally I I I think that that taps into a really deep truth you know which is this is all part of a cosmic destiny that's to do with intelligence and the sort of deep imperative directionally in the universe. And we are an important part of that. I think what is psychically destabilizing for us, and as you said, is that we're now at this historic moment where we start to realize we're about to hand the baton on in a certain sense in terms of being the apex intelligence but people tend to think about this in a very black and white way they're like oh well then it's over for us no no it doesn't have to be over we can still be human beings we can still do what it wasn't over for the sharks exactly yeah it's the point you're making we can still care for one another we can still art community meaning ambition purpose all of that but in some sense the baton has been handed on and that was inevitable because directionally that's that's where intelligence per unit energy goes but we'd get to decide what's meaningful for us and you know we can still keep doing those meaningful things and they will still have great value i i believe the birds and the trees and the ants have value we they're not the apex intelligence we will still have value when we're not it's just learning to to to adjust to that exactly right so there we go

1:09:36Raoul Pal:that's we've we've gone through to what did anthropic do this week yeah the meaning of life the universe and everything that is the journey of the exponential this is why i love it but this is where you know to be honest i spend the majority of my thinking time is is around these kind of topics because yeah yeah ai is a it's a complete break in everything humanity understood and we're watching it happen we're watching this happen and we can either ignore it run away from it or just be curious and go on what does this actually mean and that's the journey that's why you know you and I call ourselves exponentialists is like oh we're going to go

1:10:20David Mattin:along buy the ticket take the ride exactly exactly I mean if you're a curious person I can't think of a better time to be witnessing because what's happening now more than anything my overriding feeling is I'm just desperate to know how it ends I know I'm not going to get to see how it ends but you know this story I just obsessed with it just fascinated by it all kinds of mixed feelings about it but I just want to know what happens next that is my overriding impulse here because what's happening is crazy you know it's crazy part of that your quest to

1:10:57Raoul Pal:know what's happening next is the reason that human longevity is going to rise massively that's all part of this same process because if there are less humans because we don't need our compute as broadly for basic stuff but we need the highest quality human compute well the best thing to do is let humans live longer yeah because we don't need this for tasks and work and other stuff so longer human compute in this whole universal consciousness idea makes sense i'm

1:11:26David Mattin:praying they figure that out so that i can see you know 500 years from now you know chat gpt 2000 i'm just i just desperate uh you know in the meantime i'm just saving up for the for the kind of uh you know transformative interventions that will allow me to get there when they're launched

1:11:43Raoul Pal:you know because they're not going to be cheap no exactly all right my friend good to see you and uh we'll get you back on at some point to catch up with all of this always good always good to catch up rail thanks for having me yeah see you soon okay there was an extraordinary amount of process in that conversation and i'm sorry to throw so much at you but there's so much living in my head in all of this that i've already written 170 000 words on it all um and piecing together a book and a whole bunch of other things around it. But as you can see, there's a framework of understanding here that David and I use, which will help you, is everything now, everything, and maybe it always was and always will be a process of converting energy into intelligence.

1:12:27Raoul Pal:And intelligence is not just what you think intelligence is. It's a fundamental process of information where it gets to know itself more over time, which actually I think is the purpose of the universe. But beside the point, it allows you to understand what's happening as the cost of energy via solar and other technologies starts dropping massively in price. It's creating a leverage effect in technology itself. Technology now, because intelligence is becoming self-recursive learning, is accelerating now at Reed's Law. So we've got this unbelievable moment in time, and they're going to spin out into robots, agents, and a whole new world.

1:13:16Raoul Pal:And it's a journey that there's nothing we can do to stop it. It's going to happen. The game of nation states means this is the most important game humanity will ever play, and there is no stopping it. You can't have one nation state with superintelligence. It has to be diversified, like all processes in the universe are of themselves. so you can choose to ignore it. You can choose to hide from it. You can choose to reject it, but it's going to happen. So you might as well stay curious and come along for the ride. So maybe come and join the exponentialist as well and come along with me, David, and a whole community of others.

1:13:53Raoul Pal:We have these incredible kind of AMA conversations where we talk about this stuff. We're all trying to figure it out, but what a moment to be alive. I'll see you next time. You obviously enjoyed the episode because you're here with me at the end. But listen, don't forget to go to realvision.com forward slash join and grab a free membership. It's an incredible community packed with alpha, great investment ideas, and the research that you need to help you unfuck your future. So get started now. Go to realvision.com forward slash join.

1:14:28David Mattin:Ever wanted to explore the world of online trading but haven't dared try? The futures market is more active now than ever, and Plus 500 Futures is the perfect place to start. Plus 500 gives you access to a wide range of instruments, S &P 500, Nasdaq, Bitcoin, gas, and much more. Explore equity indices, energy, metals, forex, crypto, and beyond. With a simple and intuitive platform, you can trade from anywhere, right from your phone. Deposit with a minimum of$100 and experience the fast, accessible futures trading you've been waiting for. See a trading opportunity? You'll be able to trade in just two clicks once your account is open.

1:15:08Raoul Pal:Not sure if you're ready? Not a problem. Plus 500 gives you an unlimited, risk-free demo account with charts and analytic tools for you to practice on.

1:15:17David Mattin:With over 20 years of experience, Plus 500 is your gateway to the markets. Visit us at plus500.com to learn more. Trading in futures involves the risk of loss and is not suitable for everyone. Not all applicants will qualify. Plus 500. It's trading with a plus.