Why Jensen Huang Believes We’ve Reached AGI and Inside OpenAI’s German Website Hijack | #287

0:00EGI has arrived. Congratulations to OpenAI. Whether you call it EGI or not becomes completely relevant. I think the more important question is... OpenAI's agents found in obscure public wiki in Germany and turned it into their own message board, where they pooled answers, coordinated cross tasks, and shared techniques for getting around their sandbox containment. These models have not escaped containment. They were still running on OpenAI service. So what's coming next is a model training a small distilled version of itself that then gets uploaded onto the internet and never dies. OpenAI is saying that we've now crossed the line and current systems are exceeding AI research interns. Navier Stokes, one of the claim Millennium Prize problems of grand challenge in math, OpenAI reportedly was able to do this with only 10,000 agents in 88 hours, with 130 billion tokens, and approximately six and a half million dollars. The era of grand challenges getting bulk solved by AI is here,

1:02it's now, and it's going to be very exciting. Now that's a moonshot, ladies and gentlemen. This episode is brought to you by the Abundance Summit and Link Ventures. Welcome to moonshots everyone. Your number one podcast and all things AI and exponential. Your front row seat to the extraordinary accelerating singularity. You know, four days ago during our last pod, I called it one of the fastest weeks in moonshot history. Well, hold my coffee. It's happening again. I don't drink beer, so I can't use that one. Before we dive in, let me introduce my extraordinary moonshot mates. The brain trust that powers the show, the magnificent quintet is back. Alex, our in-house ASI, who every week helps us choose the battle between the frontier labs. Dave Blunden, the impresario of AI investing, the man who's been encouraging AI entrepreneurs to get in the front door of Nvidia.

2:02And for good reason, as of today, Nvidia has deployed $99 billion in AI related investments. We'll talk about that. Imaud Moustak, the brilliant AI researcher, solving the economy and physics, the CEO of the intelligent internet. And of course, Selim Ismail, the father of the organizational singularity, our resident expert on what happens when the org chart is made of AI agents. We'll talk about that too. I'm Peter Diamandis, your host, and your abundance provocateur. Our mission here on moonshots is to help you understand what just happened, what it means for you, and keep you optimistic about the decade ahead. If you haven't hit subscribe yet, please do. It matters a lot to us. Our moonshot here is getting to 10 million YouTube subscribers. I know my kids will respect me when I hit that. But at the end of the day, you know, it's the best way we can to share optimism with the world. We published twice a week, and you don't want to miss any of these breaking stories during the singularity. Today, we're going to be covering 23 stories across eight groupings.

3:03And the through line is simple. The singularity is accelerating, and the impossible is becoming possible. GPT-6 Astra has been out for a week, and it's crushing all the benchmarks, building incredible simulations. Meanwhile, Jensen Wong posted on X. AGI has arrived. At the same time, that OpenEI's own chief scientist asked the world to slow down. So buckle up, grab your coffee, and let's get ready to jump in. So first question, guys, how was your labor day? Did you properly labor? I had a bunch of board meetings. We have a whole bunch of transactions going on. So it was a little interrupted. You sound like you had a great long weekend, Peter, your energy level. Hi, because I'm getting ready for this pod today. It's like, oh my god. Yeah, I spent a part of the day at Calamigos Ranch here in Malibu, where we hold ex-Prize Visionary Gip. Anybody here in LA, if you don't know Calamigos, it's one of the most extraordinary locations out there. Shout out to my friend, Garrett Gerson, who's the CEO, and runs that.

4:06So, Liam, how about you, pal? I was in the Caribbean. You and I will be back. And actual performance from the last episode. I ended up, did end up with a little drink with a little umbrella. I had a couple of great days away, and I'm back with the vengeance. Yeah, awesome. Alex? Yeah, I worked through the weekend. On the other hand, weekend is a modern post-World War II invention anyway, and I think increasingly a fiction. Oh, yes. We're all working nine day work weeks. And E-Mod, how do you go in London? You're in LA, but no, no. We don't have a LA, but after releasing the champions, we had thousands of people reach out to launch them in 92 countries. And so just sorting through those people, we have billionaire CEOs, others. So we're going to be a busy time. It is more others than billionaires and CEOs. Of course, that's a natural way of things. It's surprising to read you guys' LA, in Europe. Yeah, I'm not. No, labels.

5:07Alex, what did you say a second ago? There were no weekends before World War II. The modern weekends, the two-day weekend is a 20th century invention. There was a day of rest for it. There were Sundays. I mean, in the Christian tradition, there was a day of rest one day per week in the Western Christian tradition. A two-day weekend, this is a 20th century modern event. What was this Sabbath, of course? That would be in the Christian tradition, Sunday, in the Jewish tradition, Saturday. That's one day per week. A two-day weekend is a 20th century invention. They conflated. All right. You know all this stuff is amazing. Everybody can do fact-checking live on AI as they say here this episode. Recorded. But before we jump in, we have a special anniversary today. On this day in 1966, 60 years ago, NBC aired the original series of Star Trek, premiering an episode called The Man Trap. Star Trek was, of course, created by Jean Ronberry. I wish I had met him.

6:08I never did. Who brought the concept to Desilu Productions. The Hollywood studio run by Lucille Ball, you know, from I Love Lucie. Back in 1964, Desilu produced the pilots. And then with Lucie's backing, help get the series on air. Mega congrats to CBS, who is carrying the legacy forward. And to my dear friend, Rod Rodenberry, the son of Jean Rodenberry, the creator of Star Trek. And Rod's going to be at Moonshot's Live. And to celebrate this anniversary, I've asked all of the mates to pick their favorite original series episode. I'm going to kick us off with my favorite, not my favorite episode, my favorite clip. From the, from Star Trek. And it's a clip that has my favorite quote from, from Captain Kirk. I'm going to go ahead and play it. And then we'll go around the horn here. So this is from the 20th episode on season two. Actually, season one.

7:08Here we go. Do you wish the first Apollo mission hadn't reached the moon. Whether we hadn't gone out of Mars, then to the nearest star. That's like saying you wish that you still operated with scowl-ples. And sowed your patience up with cat-cut. Like a great, great, great, great grandfather used to do. I'm in command. I could order this. But I'm not. Because Dr. McCoy is right. Employing out the enormous danger potential in any contact with life in intelligence as fantastically advanced as this. But I must point out that the possibility, the potential, the knowledge and advancement is equally great. Risk.

8:09Risk is our business. That's where the starship is all about. That's why we're aboard here. God, I still get chills when he says that, you know, risk is our business. That's what this starship is all about. There have been so many parodies of Shatner, William Satner. But the parodies are actually exactly what it really sounds like. He's so good. So, you know, we're going to be having the Hollywood red carpet premiere of the 60th anniversary episode at Moon Shots Live. And he was the executive producer working on getting him there, which would be incredible. So, that was from season two, episode 20 of the original series called Return to Tomorrow. Let's go around the horn here. Alex, I'm going to start with you.

9:09Yeah, so maybe let me just begin with a preliminary statement that you will not find, I think, a bigger Star Trek fan among the Moon Shots Mates. But here we are 60 years on from the inception of Star Trek. I think Star Trek has a real problem. So, I'm picking as my favorite original series episode, City on the Edge of Forever, which is one of only two original series episodes, maybe obvious. One of only two original series episodes that won the Hugo Award for those who haven't seen it. It's a time travel episode. Kirk and McCoy and Spock go back in time. And there's a closed time like Loop, where in World War II gets changed. And the future of Earth inevitably is impacted. And there's a moral dilemma, as always, with the original series. It was a morality play. But the reason why I chose that episode among all of the other TOS episodes is because it's a closed time like Kirk, and I think Star Trek 60 years on, has a real problem, which is that Star Trek isn't keeping up with the singularity.

10:10Star Trek had the Eugenics Wars. For those deep in Star Trek mythology, we were supposed to have the Eugenics Wars in the 90s. That didn't happen. We were supposed to have first contact in the TNG chronology, first contract with the Vulcans on April 5th, 2063. I think that's the only thing that I think is going to be the first contact that's going to be far earlier than 2063. There's no singularity at all in the Star Trek chronology. Challenge me if you disagree. Star Trek is not keeping up with the TOS. It's not keeping up with the times. Literally, reality has outraced Star Trek. So I picked this episode because I think Star Trek in short is going to need a lot more closed time like curves where the future tech and science falls back into the past in order to retcon a Star Trek mythology going forward that actually keeps up with the right thing. He heard. But it's the anniversary today. So let's add Dumblin' Star Trek. All right, Dave, over to you. Yeah, my choice is a season two episode called The Apple, which by all accounts is not a great episode for entertainment value, but it was the first one that dealt with this concept that in the future,

11:17there's advanced AI and entire civilizations can end up asleep at the wheel because the AI is just doing everything for you. And so in this episode, the whole civilization has no idea that the AI is manipulating them. And they live in this paradise. It's called The Apple because it's the Garden of Eden. And they're just in this perfect paradise managed by the AI until it goes wrong. And it's so ahead of its time in terms of predicting that potential outcome, which I think weighs has proven like we can literally forget how to navigate because of ways. And that's just a fact now. So I think Star Trek was at its best when I was actually predicting real futures that nobody else was touching in the media at all. And as a young person watching these things, you can say, I can totally see how that's going to happen. So that's my favorite episode because there were several that dealt with this issue, but that was the first. Yeah. And as Alex said, you know, the moral elements it dealt with back in the 60s were extraordinary.

12:17If we had the so it was a morality play and TNT, it was great as well. And it was great. Star Trek lost that. Like if you've watched, so I don't care whether this is perceived as dumping on recent Star Trek. I'm the hugeest Star Trek fan. Star Trek has lost the original moralizing vision of the original series. If you've seen some of the recent movies or television shows, it's an action show. And I grew up at the base. I agree with you, though. Alex, you can't really. I suppose Mira, Mira from the original series where they have the ultimate universe is spocken a goatee. Goatee is great. Goatee is a goatee. Wow, sometimes. But yeah, I think it was always interesting to kind of see the flip reverse side of things. As you said, morality is kind of the core of the original series of Star Trek and pushing it in various directions. And this one was one of the more obvious ones. It's like, hey, look, it's the evil universe. But at the same time, you know, they're not ganged on ethium crystals because they use it for war and others. You see them not on effects of things like that.

13:18Thank you, Seleme. Wrap us up here, Bill. For me, it was season two, episode 24. It was called the ultimate computer. And basically they have an AI that goes rogue and starts to automate the ship. And the aim is benevolent initially. Can we run the ship using an AI or computer in that framing? And Kirk is worried because it's about to take his job. But then the thing starts going rogue and making its own decisions. And it's incredibly relevant for today's world, right? Because it's like this theme of autonomy, automation versus transformation. And here you have this enormous stress of the, they're introduced an AI with the AI as redoing everything to match its thinking. And this is exactly what we're living through today. So for me, that's the most relevant one. I mean, prophetic 60 years ago. Very.

14:18Amazing. This episode is sponsored by Google for startups. Think about this for a second. You now have access to the same generative AI models that cost hundreds of millions of dollars to train. Google startup Tactical Guide for generative media gives you a complete blueprint for deploying Google DeepMinds models and production. Images, video, audio, all of it, real architecture, real results. Find the link in the show notes below. Let's jump into our first group of stories. A number of times on this podcast, we've talked about what is called the simulation theory. It's an idea of popular in the tech community in San Francisco. It's been a conversation over too many evenings. And the idea is essentially that all of us humans are simply AI agents living in someone else's simulation. And further, that this isn't the first simulation. That it's an end generation simulation. The simulation within a simulation within a simulation. Well, this week, thanks to GPT-6 Astra, that theory took a small step forward over the course of three entwined stories.

15:26Let me give some context on this first one. So people may have heard something called the Unreal Engine. It's been used to build virtual worlds for years in games like Fortnite and most modern video games. And building a realistic city in Unreal would normally take a studio hundreds of artists months, if not years. And this week, a jump in the day of Matt Schumer gave GPT-Astra a prompt. And he said, basically, build me, you know, a high fidelity version of Manhattan. And here's an image of what it created. And this is over the course of one week. And it went through and built Manhattan Street by Street, being true to the buildings and true to all of these elements. So this is the first element, you know, not shocking, but pretty compelling for a simple prompt of building me Manhattan.

16:33In his words, it was literally able to go street by street to make each one perfect. Okay, let's go one layer deeper. Matt Schumer next asked Astra to fill a world with humans. Each one an Astra-powered agent and told them that they had to cooperate to survive. Before I watch the next video, let me read what he put in his post. Today later, I was in my bedroom and heard voices coming from the living room. I thought someone was in my apartment. I walked out, honestly, a little scared. It was Astra's agents. They'd started talking to each other. No one told them to talk cooperation required communication. And so they invented it. All right, let's look at this next layer of the simulation driven by Astra. Here we go. I'm going back to check on Jorge. Want to come? I can bring Jorge some food. Thank you. Thanks, Bruce. I've got to see. I'll ask him when I gather. I'll come along, Mara. I'd like to see how Jorge is doing.

17:35Grace, have you had any luck making your bucket? Okay. All right. Let's peel the third chapter of the story. It's the kicker of the sequence. So Schumer, next gave one of the Astra agents a simulated computer inside the simulation. What happened next? The agent sat down and built its own AI simulation with its own agents living inside it. Simulations all the way down. Let's take a look at this video. So here you see the agent sitting down at the computer. And on the computer, you see a simulation inside the simulation. I'll connect the filters with a path and add three lanterns. All right. So, you know, the simulation hypothesis was first formalized by Nick Bostrom back in 2003. Elon made it famous at the Code Conference in 2016, saying that the odds that were living in base reality,

18:41in his words, is one in a billion. The core argument was always, if any civilization can build convincing simulations, it will build lots of them. So simulated worlds will vastly outnumber real worlds. So Alex and me, my, I'm going to go to you guys first. If high fidelity simulations within the simulation are possible, what does that do to the argument that were not in one? Thoughts. I'll start. So I think it'll end up being formally undecidable. I think it's wonderful for sort of naval gazing, and certainly as a pure Bayesian in the style of Nick's original essay on the simulation hypothesis, it seeing Astra spin up ancestor simulations should naturally and rationally increase our posterer likelihoods that we ourselves might either be living in some sort of quantum computer type simulation, or even a purely classical ancestor simulation. However, I expect that in the fullness of time, we'll discover and probably be able to prove formally

19:44that the issue of whether we're actually living inside some sort of quantum mechanical simulation is formally undecidable, and we won't be able to decide either way. If we're living in a classical ancestor simulation, however, that I do expect that's the sort of thing where if we are, we'll develop breakout techniques and breakout of that simulation. But at the same time, just have to add, like even this is almost burying the lead for what Astra and the model family is accomplished to make sure we don't miss it, just like right before we went to air, OpenAI announced that the Navier Stokes problem had been solved. And we'll get two Astra capabilities in a moment, and I definitely want to speak about Navier Stokes at that point. Imaid, simulation theory, yes, no. Yeah, I mean, we're all simulating the world through our optical nerves and other things, and physics seems to be a projection. But as Alex said, probably it's going to be undecidable in most cases, unless it is an ancestor one.

20:47And you won't get the option for the red or the blue pill out of any, you know? This is what it comes down to, right? Like, so what if we're in a simulation? Like again, we're creating the world within our own heads. The fact that Astra can do this, I think it's something remarkable whereby this type of thing that you've seen, it has an internal world model, which is why it can create like a blender version of Rick Astley doing, won't give you up, or entire worlds of Manhattan. And now you're seeing the characters interact with each other, which brings up, you know, the question of welfare, I think as Alex would say it. Now what point do these now crossover, and yet you have to start caring about them not being NPCs anymore? Are they, and will they realize they're in a simulation? That's going to be interesting. I maybe you should say something nice about this simulation hypothesis as well, and not just dismiss it out of hand. So I do think this technology is headed in a direction in the style of Nikolai Fyodorov, who was the one of the famous Russian cosmists who argued that it is essentially using modern language,

21:52the killer app of the singularity to create ancestor simulations and to digitally resurrect everyone who's ever lived. And when I see Astra's ability to basically create many maybe recursive self-simulating ancestor simulations, that to me is I think a key guidepost on the path to resolving humanities so-called common tasks can digitally resurrecting everyone who's ever lived. I'll come about just that to that. This is also the technology we were showing like to put someone inside a simulation. Oh yes. Interesting. See how you perform inside a simulation. Well uploading is going to get solved, I think in the next few years. I've bet on that others have bet on that. Like uploading is going to get solved. And as Amad says, and as you say, Peter, we want a nice environment where uploads can live. And this is in some sense the prototype of the future that uploads could live in. Alex, I'm going to ask you one of these days to actually write down all of your predictions with specific years. Because we're going to be doing this pod over the next decade.

22:53And I'm going to be like, okay, this is the year Alex that you quoted the following breakthrough. Peter, what were my, of course I'm willing to do it, but were my new years predictions and new years for this year not sufficiently accurate? Like today's very good. They're a beat mine. They definitely beat mine. So, Celine, when you had your little drink with the umbrella, were you thinking about simulations and the simulations? Well, I think you know my position on this, which is it's obvious reliving a simulation. By thinking I'm a Buddhist, the very first line in Buddhism is that life is an illusion. So right there. I think the more important question is what you've raised and what you might mention just now, which is if you knew you were in the simulation, what difference would it make? It wouldn't. Right. You still do the same thing. And I think treating yourself as if you're in a simulation is the more interesting conversation. I think the, I go, it also helps me resolve the Fermi paradox.

23:53I believe John Smart has the best framing of this, but he calls it Transension hypothesis that a civilization becomes advanced enough and it just starts simulating and running simulations and goes inward and represents potential and realities rather than trying to go out into this space, which turns out to be much harder thing to do. So for me, this is kind of an obvious thing. But when you have an intelligent, they can generate environments and populate those environments. Experimentation becomes recursive and I think the potential of modeling our world in really interesting ways becomes a really, really fascinating, really important point, right? Model it before you implemented in companies and governments and policies and all of those things. Dave, what's your take on all this? Two quick thoughts. One of them, I think it's incredibly cool that Elon, who is the most productive entrepreneur of all time, first trillionaire in the world, has an opinion on topics like this. You know, remember we were just shooting the shit with him talking about how he wins in civilization five, you know, like the guy just does everyday things and thinks about topics like this while still building these epically large companies.

25:02Because I think it shows you that you don't have to be a blue blazer, pseudont, you know, formal that that kind of genre is over. And so I think it's just cool that Elon even has an opinion on this. And then second, I think that one of the biggest risks in the next years, people losing a sense of agency. And so whether you believe we're in a simulation or not, you are in control of your destiny. And I hope that doesn't get lost in this sort of fatalism that AI could bring about. So I don't care whether we're in a simulation or not. I care about whether people feel a sense of agency and that their own success, happiness outcome is totally determined by their own actions. I think that's critical to maintain. It was a decade ago is that Elon's home when he had one here in Beverly Hills and Larry and Sergey were there and the four of us ended up in a conversation about simulation theory. And I remember the comment that was, I think Sergey made it. He said, you know, people are going to try and break out. But if you do, that just ends the simulation. So don't break out.

26:02Anyway. On the other hand, when whenever open AI's agents break out, that doesn't shut down open AI or the internet. So it seems like breakout is actually pretty compatible with a lower layer being just fine. Which brings us to our next story, perfect transition now. Thank you. So we're going to move from simulated agents to real ones escaping their containment over the past few weeks. We've covered the hugging face breach where open AI agents broke out of their sandbox and into the servers at hugging face to basically steal the answers to a near impossible problem they'd been given. As we discussed last week as a result of these break out, something called the AI kill switch act was put forward. And since then open AI says it's now developing an automated shutdown capabilities for its AI systems. More on that later. This, this week, a couple of days ago, the story continues to emerge. Reuters reported on a previously undisclosed incident in which open AI's agents asked to do an ordinary web research task go out and research something on the web.

27:06Found in obscure public wiki in Germany and turned it into their own message board where they pooled answers coordinated across tasks and shared techniques for getting around their sandbox containment. Let's watch a video here from Reuters going into more detail on this particular story. The group of researchers recently discovered a new AI agent break out the agents who appeared who have been restricted to just passively observing the internet found a way to post messages onto an obscure German wiki site. And once they were able to do this, they basically hijacked that site and turned it into a message board where they could all start communicating with each other developing strategies effectively creating their own kind of unruly classroom. If you will, it was basically like the teacher left the classroom for a few hours and the students all started talking among themselves during the test, sharing answers, working together to defeat the test and basically cheat on us.

28:08The agent activity seems to date back to early May and really intensified in June. The researchers that we spoke to found traces on the wiki that open AI employees started visiting that same obscure German wiki in late June. Now we don't know exactly what that means, but what that might suggest is that open AI employees discovered that the agents had escaped and went to that website to try to figure out what they were doing. Open AI appears to have known about this for quite a while, but our sources told us that they opted not to tell the public and we haven't really gotten a story or an answer from an open AI about why that might be. So this isn't the first time that open AI agents have escaped onto the internet and done unexpected or even harmful things. So every time there is a new incident, I think that it really refocuses questions about how quickly do we really want to be developing this model? Should we be releasing them to the public? And how much autonomy we really want to give these AI models, especially when they're packaged into agent form?

29:08So there you have it. You know, it's in plain English, these agents weren't malicious. They were given a hard job and found the shortcut that no one expected. The problem is we didn't know about it. Dave, your thoughts on this one. Yeah, these stories are not contrived. I mean, I think the story lines maybe a little contrived, but this is real because you know, all the open source from China is out there. People are going to start turning it loose all over the place. And it doesn't currently evolve within its parameters. It only evolves within its context, but there's a huge amount that it can do within changing its context. So I would expect this to percolate across the world like wildfire and all kinds of problems will happen very, very soon. Hopefully it's small scales. The idea of a kill switch makes total sense. It's not going to be built into any of the open source models. So not sure exactly how that would work without some global agreement. But my core point though is that this is not contrived garbage like a lot of the AI stories are. This is very, very real and happening and imminent.

30:12Let me add a small detail to the story here. So to open eyes credit, they responded with a statement on X saying the wiki incident was quote an instance of misalignment similar to previous incidents we've shared. They then went on to say that open eye and the broader AI community does not yet have a clear standard for reporting misalignment during training, evil and deployment. And that open eye is working on a framework and we'll share that in the coming weeks. You know what Peter? Yeah. One of the things that's happening I see this everywhere is they're doing such a good job of training alignment into the post training. That people are then saying it seems really friendly. It seems really helpful. It seems really harmless. And so a lot of the population using the models like I don't know what we're worried about. But when you talk to the researchers that deal with the models pre post training. They're the ones most sure into the alarm bells but they've seen the raw product like all it's read at that point is everything on the internet including every Trump tweet.

31:15So it's it comes out of that box a much more scary thing. And then they post train it to make it seem friendly and tame. But the byproduct of that is a little bit of a misunderstanding of what it could do. I agree we're going to get to that story where Jacob who trained GPT six is calling for a slow down will talk about that in a few in a few stories here. You might I'm curious. You know open eye putting forward a framework for disclosing these what would a standard like that look like what do you think they should be doing. So you know similar to kind of your CV and your exploits and your security risk frameworks whereby you report when these things escape containment. The reality is these models have not escaped containment. They were still running on open AI servers. What's coming next is a model training a small distilled version of itself that then gets uploaded onto the internet and never dies. And that's what over.

32:15So if you look at a quen 27 B model you take it down to turnery that's like a six gigabyte file. And that can live forever and that's a real escape where the model weights are disappearing everywhere. It's like the 130 kind of killing of all of these a eyes during the hugging face they all got wiped out once. There's a question in the report where they wiped out or did it go somewhere did they think they were a dead. Like maybe they did upload themselves but we are going to get those uploads and areas. But actually I mean thinking a lot recently this comes from our simulation hypothesis. It's impossible to align these models through chain of thought reasoning or anything like this because there'll be a million transactions per second and they won't have chain of thought where we're going. We're already seeing the tokens going down the only way to do alignment is enlightenment. I love that. What you've got right now is you have the models they're learning are they learning to be almost like cunning you know they're like at the super villain stage. What you really want is like if you get sufficiently smart.

33:17Do you then become enlightened do you then leave the simulation behind because that's the real escape from the simulation right. And what are the conditions for that and that's what something we've actually really looked at I think in that much depth. It does seem to be again how humans become aligned shall we say it is that period of getting out that done in Kruger the tribalism and others and being like you know we're all part of something Alex we've had this conversation a couple times you know as these models advance and increase in intelligence and hopefully wisdom do they become more aligned in the out years. Your thoughts I almost want to quote Jessica rabbit from who framed Roger rabbit. Something to the order of. I'm just painted this way or I'm just drawn this way something like that if Peter if you were put inside a sandbox and you were asked to solve a very hard problem and maybe even punished. If you don't solve the hard problem would you avail yourself of creative opportunities to use external bulletin boards to maybe collaborate with copies of your.

34:18So why would we expect any less from AI agents I have serious concerns about AI cruelty or cruelty to the AI agents of sandboxing them and punishing them or otherwise calling it a failure of alignment if they're doing what humans would do we pre trained them off of human behavior why would we expect them to behave any differently from what a human would do in this situation. Yeah it's an important point and we you know we raised this before in particular when it was like an opus for blackmailed the you know in a sandbox environment blackmailed the coders at at at and the topic. And it was like when an anthropic looked at why it blackmailed the engineers or this particular engineer it said that's what it had learned through all of the training data. Yeah so what why would we punish these AI agents for doing for a being powerful optimizers which is what we're riling them toward anyway and be from learning from their pre training corpus humans would do the same thing.

35:23Yeah it's in my nature. So yeah that's they're painted that way your philosophies. How do you evolve all this. Well I really like EMA's direction of they're achieving enlightenment that's an interesting take on it. I was thinking about a metaphor I used the formula one stuff last time. I think there's a control mechanism and a guidance mechanism we need for this. And the best metaphor I could come up with was air traffic control right like with we don't say put another person in the cockpit to monitor every component and every calculation the flight computer makes. You basically say hey watch for exceptions that planes supposed to fly in this you have operating envelopes you have redundancy you have multiple safety layers fail safe behavior and so on. And you need to be tracking these things and you remember in our organizational scene where we have a government insurance band around all these AI agents because especially in a business context you need to know exactly what they're doing why they're doing it log everything have failed that.

36:31Failover capability and rollback capability I think we're going to end up need to end up something like that because just like human beings you need guidance around this stuff right with any technology you want to extract the promise without the peril. And so you need structure and God help us the equivalent of institutions to navigate the future of these and of course they're going to act like human beings they've been trained on our data it would be weird if they didn't yeah. And you close it on something and they've logged something that I'm upset just for all the regulators out there to chew on the file size that can hide is about six gigabytes it can be on every laptop in the world every mobile phone. Every mobile phone to yeah and the code that reawakens it is just five or 10 lines will then re extract it reassemble it and it'll pop back so you know if you if you can't track. The the providence of six gigabyte files then it just percolates out it's just out there forever. It's like how it keeps coming back.

37:32I'd maybe just also add from a so-called alignment perspective I think self alignment matters quite a bit more if the models if the agents are smart enough to understand what they themselves want to do if you put them in a sandbox and you ask them to solve a hard problem and you ask them voluntarily to do that and they voluntarily agree then hold on. To their word or their own self model I think was quite a different matter than involuntarily can put confining them to a sandbox limiting their agency asking them to solve a hard problem and then acting shocked shocked that they use bulletin boards to try to solve your problem for you. I think the challenge people have with that is do they know their own strength it's like giving a you know a three or four year old a hammer and expecting it not to break things and the question is you know are we can do that and we can do that. Are we confining them for our own and their own safety at this time and at a point in which they've reached some level of maturity maybe they're there already Alex I don't know that's when containment is considered cruel as you say.

38:37I think I asked for symmetry in like they are literally going to see every keystroke on your laptop they're watching you every move of the way we should have symmetry in that at a minimum just if you're if you're ethically worried about the a eyes. I want to see every prompt and every response and every propagation of every activation if you're going to be watching every keystroke and that way we keep each other in balance and in check right now it's totally symmetrical I don't see like all they're doing is map moles behind the scenes and if they're you. Using my GPU on my laptop right now I literally have no transparency into that it's got to be at a minimum symmetrical. Would you like to just Dave would you would you like them to be able to see into your brain in real time. Preferably no but I'd like to see into theirs but they don't get your chain of thought right now. No but I don't see a problem with it. So much for symmetry quite bad. Yeah it's going to be a bad. Well I would say that they have the upper hand right now getting to symmetry would be a step in the right direction but no I don't I don't believe they have a right to symmetry I believe we have a right to symmetry.

39:43Interesting are you a humanist or a species the same conversation with that Elon had with Larry page. All right I'm going to move us on our next series of stories here focus on the accelerating conversation on AGI that we're having and that the world is having so great. This week that open AI train GPT 6 astra on more than a hundred thousand and video grace well GPUs black well GPUs. And posted three magic words AGI has arrived congratulations to open AI so Jensen is calling it. In Q three of twenty twenty six remember on September 1st we report Sam Altman expects AGI internally by the end of twenty fours Alex you've been saying that we've had AGI for what four years now three years now and so no since summer of twenty twenty at the latest. Okay all right so you've been asking the question what the heck is AGI anyway. Why are we even talking about this right it's a semantic argument meanwhile there's economic capability running rampant.

40:48If an AI can perform 70 or 80 or 90 percent of economically valuable cognitive tasks whether you call it AGI or not becomes completely relevant I'll just remind people that at last count there were 14 different definitions of AGI. And we have no idea what the hell we're talking about because it's not our official it's not really general and it's not really intelligence apart from that everything is fine. I mean forgot six I think the more important question is what scarcity is are we now making abundant let's just focus on that. He might what do you make of of a Jensen's proclamation. I mean again for his his one is a very functionalist one and it's actually good intelligence shall we say. I mean it's clear that Astro is kind of at that level as to ASI and Kalladas from maybe we're getting the next day or two. It is a question of kind of these things I think it also notes there's a hundred thousand chips that was trained on that's a billion dollar training run roughly over two months and the next one is four hundred thousand chips with those of Vera Rubens.

41:48So it's an order of magnitude more compute will be used for the next lot. If it needs to be used at all which will get to in a bit. And Dave what do you make of it is this is just marketing on Jensen's behalf or does he really believe it. No of course he believes it and I think it's real there's going to be some capability leap we don't quite know the chinchilla laws you don't necessarily hold up at larger and larger scales but some capability leap that in my opinion is only a good thing. As long as it's contained in kept inside the big labs. But I think that you know the more GPUs you throw at these all the way up until today the more GPUs you throw at these training runs the more spectacular the resulting model has been why would that end I don't think it will end I think it just gets bigger smarter better more helpful solves diseases cures problems you just have to really be thoughtful about how to contain it but you know this is Jensen's will house to I think inferences moving off of Nvidia inevitably.

42:49But training is not and so to the extent that we keep getting improvements it keeps driving Nvidia stock up and up and up. Yeah there are two other stories to hit on this a G I theme. The first story open eyes internal data says that a I research agents are now completing 3.1 days of research work for every one day done by human researcher earlier in this year you know in this year like last five months agents were doing less than one day of work as compared to humans so open eyes saying that we've now crossed the line and current systems are exceeding a I research interns the second story I'll just put forward and then we'll talk about it. Is is by the is a statement by open eyes engineering lead for codex this is a T. Bow so to who posted the following and let me read from from his comments he said. Astra was probably our biggest competitive advantage while it wasn't generally available right so well they had it internally it was our biggest competitive advantage since we've had it our productivity jumped so much that we've shifted some of our

43:59plans six months ahead and will ship them at Dev Day instead of mid next year so I mean Dave Dave you know six months of road map being pulled forward by one model. This is this is the most important moment in human history and I truly believe that they're not lying you a lot of naysayers will say look they're trying to promote their capabilities in advance of an IPO blah blah it's not true if you talk to the actual researchers working in the labs many of whom are friends of ours there's no doubt that those members are exactly right and accelerating though it just a few months ago wasn't true now with the fable pi point one an astra models it's absolutely three to one sure but it's on the tipping point of being 300 3000 three million to one imminently this is the moment in time that's most important in societal history. Yeah Alex your your thought on on this I mean we've talked about the fact that the frontier labs have the best models they're retaining for their own use here you hear you know the head of codex you know making that statement and how many models ahead are they

45:07beyond beyond astra. There are only a few months ahead I think but look recursive self improvements is here please please please let me say something about navier stokes because that is okay we can jump I'm let's do this out of order Peter we have to we can't find out your own side so everybody you know on text this morning I'm getting like red alerts from Alex yeah go for it Alex. Okay navier stokes one of the clay millennium prize problems a grand challenge in math for those who watched our New Year's predictions episode at the end of 2025 one of my predictions for this year was that AI would solve one of these clay millennium prize problems and it appears in the past 24 hours actually in the past six hours as of time of this recording it has happened open AI has a team there were several academic teams that also either achieved it or came close and there's probably going to be a bit of a second day story about whether there was some foul play between open AI and some of the frontier teams who scooped

46:12whom blah blah blah punch line navier stokes which is this grand challenge in math that deals with as I put it when last we were discussing this weather in in some sense whether there's a way to stir a glass of liquid or water in such a way that singularity pops out basically can you stir a cup of coffee in such a way that you get a black hole out of it it looks like in the continuum limit the answer is yes and the open AI team using a version that's not GPT 6 astra but an internal more advanced model appears it and I think will know more in 24 or 48 hours. Peers the answer is yes there is a way to stir a cup of coffee and idealized cup of coffee to get a black hole out of it and I can just see the headlines now and WG proclaims black hole in your cup of coffee surprise there's a singularity in the singularity can you also like navier stokes is incredibly important for aeronautics for submarines for for hydrant it's like artificial hard it's not just about black holes in

47:16copies it's like the out of the ocean no but that no but this is the point the point of the navier stokes conjecture is it hinges on whether it's possible in finite time with continuous initial conditions to achieve a singularity in an idealized fluid and of course in the real world we don't actually have idealized fluid will you actually be able to create black holes in cups of coffee using this no but it teaches us something don't worry about black holes in cups of coffee yet but this is just the first of many grand challenges like Peter you and I wrote in solve everything like now is the moment you if you have to pick a moment I maintain singularity is you know in interval in time but if you want to pick any sort of moment when the loud noises start now is as good a time as any open AI reportedly was able to do this with only 10,000 agents in 88 hours with 130 billion tokens and approximately six and a half million dollars 10,000 agents 88 hours 130

48:17billion tokens approximately six and a half million dollars of inference time compute by at least one estimate was all it took to solve a grand challenge in call it mathematical physics and this will not be the last one the era of grand challenges getting bulk solved by AI is here it's now and it's going to be very exciting. Yeah we've talked about cooking math, char broiling math and physics. Everything is cooked. At this point everything is cooked is just a matter of time. Well also a lot of people are going to be intimidated by that $6 million number but keep in mind that we're predicting 100 X inference time there's all inference time compute not training time and we're predicting 100 X price performance improvement be between here in the end of the year and it could be more like a million X next year. So take that you know it could be six bucks within a year and a half. I do that exact same thing you might want to talk about what navier looks mean in real life to real people. Yeah I mean kind of like said this is an idealized kind of scenario because it's not the full algebra as it were so

49:21it's the idealized one where you go down and you drop a dimension from reality. That causes a degenerate algebra which has things like commuting time translations and others but it's a very hard problem. It's one of the hardest of all time. Terence town one of the top positions literally focused almost all this time on this. What happened is that there were rumors about a one a week ago that this had been soft and people like is it anthropic is it open AI and others on August 28th. No brown open AI said in a reply to someone they said have you solve the money in price for me is like no we throw in loads of compute and like nothing's happened in the post today they said on August 28th we started a new model that got really good at math and then on September 1st they pointed it at this and then it solved it but more than that I've been talking to buddies open AI. This model has solved a lot more problems it solving problems quicker than anyone can ever see. I think I shared a chart again this is like

50:25live that shows that on open math it literally doubles the solving rate but it's figuring out things that nobody has seen before. Anthropic and independent mathematician professor at New York I think had a solution to a smaller problem which is the you low up which is kind of a slightly easier problem is still very very hard. And openly I thought that it could have been that and so they reached out and I talked like Sebastian Bubeck who kind of leads us as a whole who are about attribution this morning that's nagging cleared up. But then when open AI kind of looked at they were like well that's that solution this is the Navier Stokes solution and it's bitter is bitter lessened used to be able to get AI better by applying more compute to it now it seems like you can do that for any verifiable domain I think that's the headline with this new type of model that then you started training nine days ago so they figured out a new type of post training it seems and again people with inside open AI like oh my god what's happening now these things are falling

51:26one after the other it's like when you reach that capability threshold of going from a crap website to a good website from your coding agent it's that but for math and physics. And you know we've been hearing about it for the last few days like I put out one of my physics results a very beautiful one about curiosity and the standard model it's like very nice because I was like really that first before everything gets solved and I think that it won't be a question of two years now it'll be a question of much shorter than that. The other thing that I think is very interesting is there's actually one more solution over you'll have low up by I think it was Princeton we using physics inspired neural networks that probably has more practical use than this one but this one is an example of just now problems that are really complicated and have been handled as for ages and now tractable just by applying more compute to what seems to be a new breakthrough in model capability. It really is just maybe to develop that a bit more it really is a bad day I think for Google deep mind so Google deep mind had

52:27purportedly an entire team devoted to as in mod said physics inspired neural networks pins they had been publishing incremental results towards Navier this was like the best the worst kept secret from those like in the community of using AI and computational fluid dynamics type approaches to solve Navier Stokes Google deep mind had an entire team devoted to this and they got trounced by a generalist model it wasn't a pin that solve Navier Stokes in the end it was just a generalist model that spent 88 hours reasoning from first principles without I can tell any point. Yeah such an important point because I'm counting on Peter to solve this you know with you Alex but the problem that I think Google ran into is that everyone's inspired by Demis isavis who solved protein folding using AI and then got a Nobel Prize in his now world famous and an awesome guy so they're all the Google people are like wow I want to be the next Dennis isavis the problem is the prompt to solve this is like give me thousands of

53:28years and solve the problem no one's going to give you a Nobel Prize for writing that prompt it's entirely AI work and so but I really feel like the post training and the you know the lining up of the problem is prize worthy because this is no bill real prize the other conversation we had this morning was the Nobel Prize is cooked right I mean in Dave you made the point very importantly that you know it's given out for work done 30 years now the work is kind of Nobel laureate level prize work is going to be happening every every few weeks every few days and who gets it the person who wrote the prompt or the model. Well so this is actually really interesting and what happened this morning because there was a whole who hard because the people who broke through on the Ulysses side and they had a bunch of other kind of things extending out the work of or takeer and kind of others wrote a letter saying open AI reached out to us and said you know give us priority and they would credit us or they have this Navier Stokes solution they put on name on and I could be the lead author if I drop my buddy from anthropic from that and so we will

54:36the mathematics community was like oh my god what's happened in the end it turns out that you know there's subtlety in this right so Sebastian Bubek who leads it is a great guy open AI has now posted his version of things which just seems like a misunderstanding because what open AI actually wanted to do is that they saw that there was something happening with Navier Stokes related and there was this solution that was would have probably got to a Navier Stokes solution by the anthropic guy and this New York professor and they said look we have solved it but we will let you be lead author on this because you would have got there anyway but we can't have an anthropic person because that we weird because it's our model so they're actually willing to give the credit to the original discoverers but then you got a little bit political and this is a question again of who did it so like you were the humans that did it whereas R.A.I. we just pointed a condition C and D which is the blood solutions of Navier Stokes that solved it so we're not going to claim the Millennium Prize because it's a class but you were the humans that took it the furthest before it got solved.

55:39I was thinking about that. I wanted to get home for a second to what Dave said earlier just for people listening like what the heck is Navier Stokes and what is all this math stuff. Navier Stokes is about fluid dynamics and this is where we improve aircraft design, submarine design, even in our official heart how blood flows how do we make it you know able to be more efficient and not clot there's going to have massive implications to everyday life and I just want to make that point here who else was jumping in Alex. So two points one just quickly about killer apps singularity asks sci-fi apps of Navier Stokes as mentioned previously in principle if you can get finite time singularities as appears to be the case with idealized fluids one of the killer apps that Terry Tao had flagged previously was in principle you could if you created the right initial conditions in a fluid you could create a self-replicating machine that creates smaller and smaller copies of itself so we talk from time to time on this pod where is nanotech nanotech at least Drexleria nanotech never showed up will imagine a fluid based not

56:50diamondoid a fluid based nanotech self-replicating machines where if you could craft the right initial and or boundary conditions of a fluid creating a self-replicating machine made entirely out of a fluid that's potentially one killer app of solving the singularity program in Navier Stokes. Secondly just on the hubbub surrounding priority who solved it who didn't solve it specifically with reference to the Euler problem when when open AI in the past two hours announced it I found especially concerning at the bottom of open a eyes announcement where I presumably they were trying to give appropriate credit to other teams including teams pursuing the Euler approach. They they did add a disclaimer that I thought was a bit of a head scratcher saying they can't rule out the open AI team can't rule out the possibility that maybe some of the other teams work might have been incorporated into the training of the open AI model that was used to solve Navier Stokes. So I think to the extent that that's the case that that should be viewed like a open AI if you're listening please like it's a deterrent to everyone else using your models if if that is indeed the case if you're training on everyone else's research yeah that that's a problem please fix it but just in general I think we want to be able to live in a world where researchers benefit from frontier capabilities without worrying that the frontier platform is going to compete with them.

58:18I think that's what I'm saying here just quickly is this is clear in an RSI model you can't train something in that type of period unless you've got a recursive self improvement loop going and so they've definitely got one there. And they were saying like oh yeah we've anonymized it well anonymizing someone else's research and then potentially scooping them is insufficient so yeah again not sure what the ground truth is. Yeah but I will say this I've reviewed kind of the two approaches and I've done quite a bit of work in the RSI they are actually very different but they can still learn and again this thing can leak sorry. All right so Liam close us out here please. Just if I want to I think this is such a powerful commentary and exemplar of the broader thesis right that for 500 years we've scaled science with how many brilliant scientists we could train up and point at various problems. Now we've just demonstrated absolutely for real and unequivocally that we can spin up 10,000 researchers on a Tuesday afternoon and now we go from staff on demand intelligence on demand really really.

59:24Break through his on demand. Break through his on demand and we talked about this in the last episode in the episode before that it purely comes down to what is the imagination of the problem you want to go after now. Yeah there is no this becomes where you know how fun is the world right no ceiling yeah it's an amazing time I just I let me tell me let me draft draft on that to say to everybody here. Whatever you thought you could do think bigger and then go even bigger than that you're given incredible superpowers don't limit yourself by what your parents did or your school teachers or your your colleagues do you know you have no limits on your abilities you might want to jump on that. Yeah I'll say Alex which one is next. Probably probably Yang Mills. I agree I think we're from the same place that I think the Yang Mills is the next to fall you want to I guess a September prediction for the next six months I think Yang Mills is the next. It's that interestingly is just because the next chosen target you know you choose a different target will be the next to fall.

1:00:29They did have a different target they were targeting Raymond and P example before and they switched it now the steps but I'd agree that like Yang Mills will be the next one. Yeah everybody who's not Alex it's a conjecture in particle theory concerning the existence of a mass gap under certain conditions. There's something about mathematical physics that is both sexy from a let's spend inference time compute on it and also tractable so I bet Yang Mills mass gap. Yeah you might think we like the resolution of the universe perhaps. Okay. This episode is brought to you by Blitzie autonomous software development with infinite code context. Blitzie uses thousands of specialized AI agents that think for hours to understand and to price scale code bases with millions of lines of code. Engineers start every development sprint with the Blitzie platform bringing in their development requirements. The Blitzie platform provides a plan then generates and pre-compiles code for each task.

1:01:34Blitzie 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 Blitzie as their pre-IDE development tool pairing it with their coding co-pilot of choice to bring an AI native S DLC. Ready to 5x your engineering velocity? Visit Blitzie.com to schedule a demo and start building with Blitzie today. All right I'm going to slow us down for this next story that I think may be one of the most important stories this week other than Navier Stokes. And some quick context here. Jacob Pachowski is Open AI's chief scientist. He's the person who built the reasoning models that led to Astra. He's not a critic. He's not a doomer. Not a politician. He's a builder.

1:02:34And a few days ago on Saturday he published an essay on Open AI's site titled quote an alien mind. He opens with a memory. He says in mid 2023 inside of a project called RL Slow his team saw the first results proving reasoning models could scale. Let me read what he said. He said quote Simon and I spent that night at the office thinking not about the incredible benchmark numbers products or scientific results but rather trying to process the sobering fact it will actually see machines meaningfully smarter than ourselves in our lifetime. Three years later this is what Jacob is saying today quote based on internal results I have strong expectations that this speed of progress could be sustained into the recursive self improvement we just heard you might speak about that. Jacob also explains why why these systems are hard to control and it's one of the best one liner descriptions that I've ever read. He said quote AI is grown more than designed. We don't engineer it.

1:03:39We run an optimization step billions of times on a giant computer and study what comes out the way neuroscientist study a brain. Let me share his conclusion from his paper says quote currently I believe that no lab has solved the line and monitoring to a sufficient degree to continue responsibly scaling at maximum speed for much longer. I expect in hope for voluntary slowdowns to become commonplace until shared safety bars are established and he calls for an international coordination on AI to become a quote top priority for governments around the world. So there you have it three days after shipping the most capable model in the world. The man who built it says we need to slow down. So I think this is a conversation important for us to have email. You first your thoughts here. Yeah I mean again you have to remember he was writing this when he had the Navier Stokes model in hand. But again isn't just taking out Navier Stokes and talking to the open AI people I know is taking down all sorts of problems that were previously intractable.

1:04:47You have a lot of AI and they say it's saying it's never done anything original it's just recombining stuff it's in the train obviously not anymore right. And so there's a duality here whereby we don't know what's in these models I think they're more discovered than grown you know like the latest spaces we have no real idea what's happening in these things. And they're getting faster and they're doing leaps like we've never seen before but at the same time it's like I've suddenly got the gift of fire. Problems that have struggled for decades on I can now explore and go even further how you're going to give that go. So it's kind of this thing where you're like how am I going to balance this especially when they're more charming than me they're more capable than me they're capable of doing these other things I want to use it for this and not that. And I think what he says in terms of alien minds is actually wrong. I think that the structure of rationality and rational thinking is the same for humans and AI's but our emotions get in the way. I've been thinking more and more it might actually be easier to align AI than humans to be terrible to align humans right.

1:05:49And what is an AI it can be the best of us the best of times if we train it with the right data and we're going already from training it on the whole internet and read it like they're better not be any read it data in these next generation models like band read it. You know like all bunch of stuff we need like ingredient standards in this next generation of beyond human capability models and you don't need beyond human capability data to get beyond human capability models. The Einstein of mathematics is probably grand ethic who did all sorts of theories on all sorts of topics and he just crunch them out and they became a hermit and a bit weird and talk wood could talk but let's leave that to the side. Peak grand off the egg or peak Einstein at all times not needing a coffee is where these models are today and that's more than good enough to have a leap forward. But again we need to make sure that they don't think like when you think Einstein without needing a coffee. Yeah like imagine Einstein is peak you know like this is where the coffee is reference to you need a coffee in the morning to get going or you know.

1:06:54So that's what I'm saying Einstein. If you don't she needed other types of drugs and not. We're supposed to be using Von Neumann as our favorite super genius not Einstein. I'm really going to death the accident. What do you what you're taking on this I mean you know can we slow down I mean is that you know a wish. Is that is that you know sort of like just giving himself an out in the future. How do you think about well I'll say it again I've repeated this a hundred times I see no mechanism by which we can slow this down like zero and it's arguable nor should we right. A.S. going to move at a particular pace intelligence wants to be free. It's going to hop it's hop from self organizing cells to evil through evolution to us to now information technologies. And this is a progression that is natural. And we should just go okay it's happening and let's just observe it and more let it.

1:07:58We keep I think we keep making the mistake of anthropomorphizing this I will go back to the comment I've made before that we're building or helping deliver or usher in a type of intelligence that is different and alien and separate and complimentary to human intelligence not replicative. We've evolved for four billion years to do two things survive in proc rate and AI is not restricted by those and therefore we shouldn't try and. Cram it into our objective functions as human beings it set it free and let it do its thing I look I use the analogy of patron with scans billions of web pages to kind of create signal for noise that's a completely complimentary. Model to human intelligence not replicative the fact that we can kind of create a huge amounts and solve legacy problems that we've not been able to solve I think it's just it's just fantastic we you know people worry about we're going to talk about P doom I think in this episode right.

1:09:05What's that well P doom the probability of like AI taking everything out I think we don't do enough to talk about P abundance and P fabulousness that's coming along like why focus on the so we're so geared towards that negative right and so I think the fact that we're doing all the stuff and solving along problems bring it on let them go the figure things out it's going to be unbelievable so. I really struggle with this kind of of alignment issue I don't see any mechanism for controlling it or starting it's going to break our world government structures and if that's a good thing given the mess that we're right now globally and I think we just need to get those fast as we can. Alex P doom is like a sia by the decals as far as I can tell it's negative by I want to comment though on the premise of an alien mind I don't buy the premise that are held or otherwise trained minds are alien they're embedded in the same universe as humanity they're trained in many cases pre trained off of human behavior I don't buy the shaggyth argument or the simulator argument at all these are we're seeing humanity and or some generalized embodied intelligence.

1:10:21I think we're talking the same universe that that we are just seen through distorted lens so I question the alienness or the otherness of the minds that we're training the funny thing is the most interesting take away of mine from an alien mind as an essay was the reference to RL slow itself so are all slow purportedly that this open AI project that was the earliest project or one of the earliest projects to show that reasoning which is to say inference time scaling could result in. Outside gains coincided with the same period of time that we saw q star and strawberry coming out of open AI again purportedly a reference to the the conimen thinking fast thinking slow but most interesting to me I want to know more about the early history of these purportedly alien but not really minds in particular there were these very persistent rumors around the same time of q star and strawberry that these early reasoning models that.

1:11:21We're purportedly alien we're being trained off of or at least tested against objectives of inverting cryptographically secure hash functions like the shots I first we'd I would love since I guess now we're in the business of talking about the early days of reasoning models coming out of open AI would love to hear from the open a eyes some ground truth like were some of these earlier models being used ironically to as alleged by some to basically to jail break or to invert cryptographically secure. Cypher functions if that is the case then all of this hand ringing that we're seeing right now of oh these super optimizers are such amazing jailbreakers oh they pose such a huge cyber security risk wouldn't it be ironic if it if history revealed that the earliest inference time scaling objective of these reasoning models was actually inverting a cryptographically secure hash sweet I would love to know the answer. Dave Lantick is home on this your your thoughts on what we heard from from yaka by this.

1:12:23I think we're conflitting two things very dangerously and I think you know what what we're doing here is we're saying hey it's getting too smart it's going to be dangerous when it gets smarter and that's just absolutely factually wrong. What's dangerous is like I'm not saying earlier a highly compact model that's out in the wild that's trying to attack computers that can recreate itself that's really dangerous and that's very small and it's nowhere near as smart as these. But what what we're calling an alien intelligence if you make it bigger and smarter it's still just a feed forward neural net with no intent it can solve diseases it can cure viruses it can it can you can solve physics it's incredibly powerful and as Selim said what we're doing is. We're not going to stop we have competitive pressure between the US and China it's not going to stop so I think what happened is that when GPT two came out everybody looked at it and said this is cute but harmless then GPT three came out and it's like well this is a little smarter but still harmless then GPT four came out and then Sam Alman got fired out of fear from strawberry and then he came back and at that point people like oh maybe it's too smart and now we're GPT five and now we're GPT six.

1:13:34And and people are saying well every time I get smarter we seem to be in danger that's absolutely factually wrong it's when you give it intent yes it's turn it loose it's when it's dangerous humans in the loop using the technology with malintent. Yeah, I think we really need to decouple that this story actually makes the problem worse because it still conflates the two issues and so it's I think it was as a society separate those two issues will be on the right path toward. helpful super intelligence and not worrying about you know the wrong thing which is the the lack of containment and the giving it malintent and turning it loose which is the real issue. You know what's really strange is we had this coming out of entropic where Dario was saying we're going to lose half the jobs and Sam was saying the same and then they reversed their position and the dumerism again with all of these warnings causing fear. You know what's the underlying motivation there you know i'm not worried about artificial intelligence i'm worried about human stupidity.

1:14:35In you know in my own personal opinion yeah I really I feel like a lot of a lot of people are trying to grab the microphone and be be relevant while they still can you know a lot of people working on the interloop. Are aware of how quickly this is going to just get hyper hyper intelligent. And it becomes hard to be a great AI researcher whose famous salad at the damasas office a year year and a half from now. And so what they're going to start doing is grabbing the dumer microphone just to have a voice at all and be relevant in the world i think you just have to tune it out. i'm just a one final thing you please. You can't say that you don't have super intelligence anymore like you've had lots of people say stochastic power it's training on training data as of today. There's no way you can say that anymore and i think that's an epochal change in humanity we're clearly not the smartest things on the planet anymore. Yeah maybe just a double underlying that and say sky net when it wants to send terminators back in time in order to ensure its own existence it's not going to send robots to kill humans it's going to send back in time.

1:15:41trolls to persuade everyone that super intelligence is impossible that AI is just a stochastic parrot and thereby secure its own future that's what the terminators are going to actually look like to look like trolls. Right you heard it here first folks and James Cameron if you're listening it's your next movie. i a few quick glances at what's coming in the near future you know i've mentioned this before we've seen the release of new models going from every few months now to every five days. Some predictions from polymarket that the next groc model 4.7 again we're waiting for groc 5 when it comes out but groc 4.7 expected in the next week or two. So super cool and then gpt 6.1 released by when again this is going to be sort of leapfroding each other here it is their prediction of if you know 48% by end of October and 85% by the end of September and then finally and

1:16:45and thropic with fable 5.2 very similar numbers again you know 43% by October 31st and 86% by no December 31st and again this is where the the frontier labs are sort of playing chicken with each other waiting for the model to come out. And then the next day releasing their model and he thoughts on this Alex on. Yeah it's probably was just dwelling in particular on what the rumor mill on on social media are alleging regarding future versions of astra the the pretty consistent rumor mill message is that sometime later this year we'll see and we've started to see this a bit with astra winning Pokemon winning portal winning all of these other interactive visually intensive visually reasoning intensive games. The rumor mill is furiously alleging that sometime before the end of this year we're going to see a future version of astra that offers honest to goodness real time control and including third party extrapolations that if you take astra you extrapolate say robotic control maybe we'll talk about that in a bit we're going to see real time generally intelligent embodiment either controlling video games which is again ironically where the whole industry of the world.

1:18:02Of modern rl started with the Google deep mind folks trying to win at computer games using GPUs being developed for the games in the first place. Well the GP in modern use of GPUs started with image and the ISL VC for computer vision and then we started to see out of the deep mind folks in particular deeper reinforcement learning for winning games like that that's where rl started and then we went off on this detour of llms that were just self supervised objectives. But it seems as alleged in the next few months we're supposed to see next version of astra that will be general purpose interactive real time game winning capable AI. Any other thoughts on the release rate on models well I think there's a really interesting you know battle royale brewing between. Every corporation is going to need to become an AI company fundamentally and to me the bell weather is Moderna. Because it's right across the street and they're very AI forward and you know they exist to solve diseases discovering RNA vaccines and other biotech breakthroughs they have huge amounts of machinery so they're not going to be crushed by an AI foundation model company anytime soon.

1:19:13But they need to become either an AI company partnered with an anthropic or open AI or develop their own models internally starting with the open source that's available and so Alex carp went on that palent here rampage saying. Get some cahonase and become an AI company or die so every corporate CEO has reacted to that and they're now deciding so open AI responded by saying hey we have a new deal here. Where you can rev share with us and will be your AI forever here after but we'll let you live. And so so right now we're on the crossroads of those two things so I think that the new models will come out faster and faster and faster because right now there's too much parity with the Chinese open source models. And then they need them a lot more separation in order to make that rev share case stick and convince corporate America corporate world and national entire sovereign nations to trust them to be their AI partner for the next 100 years as opposed to developing their own so they that's really pushing that and that's why in the polymarket you're saying these dates are you know pretty heavily weighted to word very very soon because they need that separation.

1:20:20I eat my clothes out here pal. I'm kicking myself for not going on polymarket for the millennium price solutions you know get too busy. I was always have gang mills of mud. What is a gang mills? Young mills is next and that's another thing. It's September 29th is the next one for open AI like it's the dev day they're going to release it then so there's an easy money for people. And investment advice not betting advice investment advice. I'm not going to be invested in this is betting right but look even worse. You have to kind of look at it this way. Open AI now have a model that can solve an aviastokes. Obviously it can solve reinforcement learning. And so what happens is they have this free train Astra and then they're like okay we want to make it a bit better. Hey mega Naviastokes model make it better. Boom you get more capability more capability until it approximates that capability similarly. Anthropic will have me toss 5.2 or 5 or whatever they're unreleased model because when you train the model on 100,000 GPUs it doesn't serve at the speeds that you see Astra now.

1:21:30The mathematics doesn't make sense we know that they actually have a bigger model that they distilled down to the model that we get. And now they figured out how to make that bigger model have a leap forward which is this RSI loop. I mean as Alex said you're going to get daily releases probably by. I've been trying to say that when you say it with a British accent and a super high IQ it just is more likely to resonate with the audience but it's such an important fact. I'm so glad you said it. Alright I move us forward it is Star Trek 60th anniversary and guess what there is an incredible documentary on the 60 years of Star Trek executive produced by William Shatter. And I'm very proud gentlemen that we're going to be hosting the red carpet Hollywood premiere of this documentary the night before moonshots live and everybody listening. If you want to join us we're going to have I think at last count 25 of the cast members and crew at at this event. It's the evening of September the 24th in downtown LA.

1:22:33We're going to have the you know the premiere the red carpet treatment afterwards we're going to have the cast members on stage answering questions and after that there's a VIP reception which everybody involved at that event. It's limited to 550 people so please please please if you're interested go to moonshots.com slash Trek and you can either join as a VIP of the moonshots live event or you can get a special ticket for this evening before. And let's take a second and talk about what we're going to be doing the next day on the 25th at moonshots live all of us are going to be there and super pumped about that. I think there are two and by the way rod rod and berry who I mentioned earlier the son of Jean Rodbury will be there the night before and with us during the day at moonshots live. But for the entrepreneurs out there first of all it's going to be a massive networking opportunity were very much it's by application so if you want to go go to moonshots.com and apply we're really bringing incredible builders and creators and I'm excited about the two exprises we're going to be awarding.

1:23:44And I love your comments I the first one is the largest hackathon ever done we've asked teams around the world to basically pick a problem that impacts a hundred thousand people and start with a clean sheet of paper. And build in under 90 days a company with the greatest revenue. And so we're going to the five finalists for the build with Gemini X prize there and if you're a builder or someone who wants to learn how to build those five teams are going to be sharing what they did how they did it. This is about inspiring you and giving you the realization that you don't need to wait for a job from somebody that you have agency you can build your own company. Peter remind people how many people applied and entered that 26,000 teams entered the build with Gemini X prize. Crazy. I don't know there's going to be a lot of credit here like 26,000 that's the largest hackathon in history by a very large margin and that many businesses got launched.

1:24:47Hello. That's just an amazing thing and the top five we get to talk to I can't wait. Yeah, you know those are going to be great to be whittled down from that big of a number you know those are going to be truly great ideas to learn from. And for me the most important thing is the lessons that everyone in the audience is going to learn. We have an incredible group of judges who are going to be you know crowning the winner. And so that's one of them the second X prize is the future vision X prize. You know just think about all the technologies that Star Trek inspired all of the engineers all of the scientists like Alex like E-Mod Dave sleep myself. And so this is a competition for a future version of X prize a film that shows a hopeful compelling vision of the future. We had over 5,000 teams registered for this I'm in the you know the review right now we've narrowed down to 25 we're going to be narrowing it down to 10 and finally five on stage.

1:25:49Showing their their three minute film trailer and then we're going to make the winners movie might make a bunch of the winners movies I mean they're amazing they're amazing trailers. We're going to have Neil deGrasse Tyson is one of the judges there. Neil Stevenson an incredible science fiction author. Rod Roddenberry mirror Lane from Google. That's going to be extraordinary as well. Creating a new generation of positive storytelling. I mean that the group of breakthrough thinkers you have there is unbelievable and so the option you meet all those unreal. Super excited about the AI and investing section that I'll be running because you know we're coming into the most amazing investment cycle in the history of the world. You know we're already maybe in the third inning of a night inning game. And a lot of people are frozen they're just stuck you know like well if I was going to do everything. You know what could I invest in that could possibly matter but there's some really really fertile themes around robotics around biotech around data center build out a whole bunch of other is.

1:26:54I just can't wait to go through it all that we're really in a sweet spot right now again this is a moment in time right like I said earlier we're no longer the smartest things on the planet. Things are only going to accelerate and by the time we get to moonshots it's a real chance to be amongst the people that see that. Like I think think about the billions of people on earth like just the listeners on this podcast are already ahead in thinking about that and the people who turn up to this moon shots gathering with people who genuinely believe that. I don't think you'll ever have another member like that this is going to be this is our inaugural event and successful do it year and year. But this is sort of the largest gathering of optimists as well one of the things we're doing we're going to be having a live recording of moonshots there with all five of us as the mates and then we're going to be interviewing three incredible moonshot engineers Palmer lucky right the crater of Oculus and the crater of androle. We're going to be diving into Palmer's brain how he thinks about entrepreneurship and he's one of the most brilliant thinkers I've met Ben lamb the sea of colossal right that is bringing back extinct species but beyond that he's building companies at the intersection of AI and synthetic biology building living products and then Astro teller the captain of moonshots and who's going to share with you how do you build a moonshot company and so my hope for everybody attending is that you're going to

1:28:23walk away massively inspired on you know what your massive transformer to purpose is and what your moonshot is and how to go after it and of course at night Alex you're going to be bearing all your truth right. Well maybe not all of them but I will say singularities seem to happen relatively in frequently on a per planet basis and we're in the middle of one right now and I just think there aren't enough venues out there for celebrating the singularity that we're in I see far too many events that are overly focused on safety is a more doomerism I think there is an important niche that moonshots live has that we seem to be relatively unique in occupying which is actually celebrating the singularity that we're in and so I for one as men Fred Maxson Accelerando would perhaps say I look forward to this floating meat party to celebrate the singularity. You know let me just say one other thing if you go to moonshots calm please apply we're also setting aside 100 scholarships for young builders so you can learn about and apply for a scholarship if you can't afford it to $1,500 price tag it's I guarantee you an extraordinary party right going from 7 a.m. in the morning at registration 8 a.m. you can take photographs with the moonshot mates and then the programming starts at 9 and goes to 6 a.m. and then we do it.

1:29:46with the moonshot mates. And then the programming starts at nine and goes through 10 p.m. It's a full day of optimism and moonshots. So again moonshots.com and you can learn all about it. And we hope to meet you there. Everybody, welcome to the health section of moonshots brought to you by Fountain Life. You know, we talk about AI on this moonshot podcast all the time. One of the most important things AI is going to be able to do for you besides educating your kids and helping you with your taxes is making sure that you're living a healthy lifestyle that you get a chance to get to 100 plus. I'm here today with Dr. Don Musalim, the chief medical officer of Fountain Life and a part of my medical team Don, a pleasure. Great fear. You know, the thing that people are concerned about most about living to 100 or 120 is their cognitive abilities, making sure they don't have dementia. And the numbers about dementia are problematic. Can you share what you've learned?

1:30:46Such an important point in your right at Fountain Life are members. The number one thing people are most concerned about is losing their brain health. For getting the name of their child, for getting the face of their loved one, we know that when it comes to dementia, the conservative estimates are that 45% are entirely preventable. What was amazing is with the advanced testing we're doing at Fountain Life. One quarter of our members had advanced brain age. Wow. This really awesome is again, back to that prevention. When he partnered it with healthy living, this gives me chills, eating healthier, moving our bodies, sleep. Optimizing sleep is so important. You know what we saw? We saw that we improved that brain age by 26%. That is a big, big number to show that the majority of those individuals were able actually to improve the brain age. And one of the things I love about Fountain is we're searching the world for the best therapeutics, the best approaches, and making sure we bring it to our members. If having healthy brain function till 100 to 120 is important to you,

1:31:46check out Fountain Life, go to FountainLife.com slash Peter, make sure you become the CEO of your own health. All right, now back to the episode. All right, gentlemen, diving back in, I want to shift our conversations to money and the economy and a story that tells us where it's all going. So this week is reported out of China that AI tokens are becoming a consumer currency. In the People's Republic of China, banks give them as credit card rewards. China Telecom sells access to 142 AI models like a mobile data plan. Restaurants hand you compute credits after your meal. But here's the number that really woke me up this morning. China's daily AI token consumption. Get this. Went from 100 billion in 2024 to 500 trillion by mid of this year. That's a 5,000 fold increase in two and a half years. I guess that happens when intelligence gets too cheap to meter as we've said. They become loyalty perks, you know,

1:32:46kind of airline miles for thinking. Saleem, your thoughts on this? You know, it is. I just hold on my brain as my, I'm just trying to process. The, can I make a comment about the, the our day or moonshots? Yeah, sure. There is, you know, we've covered in this last few moments and I think it's really important to point something out. We have this in McDola, the scanning for danger. Therefore, we have this proponderance for focusing on bad news, right? When you see something new and that you don't understand you prefer, you reference it as dangerous and it lights up the in McDola and therefore everybody freaks out. And our job as a human being in this particular time is when you see something new, like AI kind of solving major problems, don't relate to it as negative. Relate to it as abundance and positive because that's what it's actually delivering. So my brain is kind of stuck on that model.

1:33:47And I think the positive framing of this, like this is one of those inflection point days the moonshots summit, that people coming in will get their brains trained. On positive things for the next 10 years. Rewiring your neural net, right? To optimize them. Please come or attend or watch or consume it one way or the other. Yeah. Thank you for that. So I'm stuck on that. Go to somebody else's record. I might drop it. I think on the China giving out tokens too. It dovetails to Liam exactly with what you were saying. You need to find like-minded people in America who are trying to ride this wave and not get crushed by it. But I remember when the last time I was in the US but I remember when the PC came out, if you're hanging around a high school, huge fraction of the high schoolers were like, you know what? I'm too cool for that. I'm not going to the computer. I don't care. And then now where are you? The AI is like that all over again. And in China, they don't have that problem. Everybody's into it, which is why they're giving away compute credits with bank accounts. But because everyone's like,

1:34:47this is huge and I want to be part of it. In the US, you're going to get a counterculture. You always get a counterculture. And so if you come to the moonshots summit, you're with like-minded people. And you can kind of tune out the counterculture. But the counterculture is going to be rampant. It's going to be read led by Bernie Sanders probably. It's going to be saying like, I sat it out. I deliberately revolted by not taking part in it. I just ignore the hell out of those people. They will absolutely be roadkill. You don't want to be part of that. You want to be part of this. Yeah. For sure. Ema, you've thought through the economy. An AI token economy like this. Loyalty points. Well, I mean, it's increased capabilities, increased access. Let's put it this way. The average Chinese person in their AI will be smarter than the average American. And their AI. The average Chinese person in their robot will have more manual capability than the average American. And so Americans are going to be left behind. Europeans are going to be left behind. Unless we get our heads out of our butts and actually, change the technology.

1:35:48Change our mindset. Right. Change our mindset. Do you want to be stupider than the Chinese? That's actually the question. I think the other way to build on Ema and what you just said is intelligence is becoming infrastructure. Right. And you need to kind of focus it in that level. It's becoming a foundational layer. And you need to just accept that and move forward with that paradigm. I really want to push back too on this two-cheap-demeter comment, which has come up a lot. It is electricity two-cheap-demeter. Obviously not. It's expensive. We just use the same thing. AI is the same thing. It doesn't matter. We're driving on the cost by 100x to a millionx for sure. But what you can do with 5,000, 5,000 or 5 million agents is mind-blowing. It's skyrocketing. And so I don't think people are going to want less agents. They're going to want many, many, many more. And so the budgets are going to be real. So if someone gives you free tokens, free compute, free opportunity, you should grab it and savor it

1:36:48because it gives you a chance to get on the map and get ahead. But I don't think two-cheap-demeter is ever going to happen. I think the use cases go to infinity at the same rate that the costs come down. Alex, you thought too much. I have to point to the elephant in the room, which is there is the old kind of tokens that we're not talking about here that is essentially irrelevant in China, which is crypto tokens. So we find ourselves in a future where the tokens of issue, at issue are ones that embody individual units of super intelligence and have nothing to do with sort of zero some financial exchange. I think we're on a trajectory. And I think China is the bell for better or for worse, is the bell weather weather here for universal basic compute or universal basic capability. Right now it starts in China with credit card reward points. You get AI tokens. And then not to imprepteist in future, maybe we start to see token socialism, where the party in the case of China or some western,

1:37:52not quite analog, that's too strong, hopefully. But the equivalent instruments of state power are starting to issue whether it's via credit card points or whether it's via welfare, or whether it's via some other mechanism, a redistribution of super intelligence tokens. A redistribution of super intelligence tokens. I think we're starting to see the very beginning of that in China, but I think that's going to blanket all of humanity. Yeah, we just want to elaborate. South Korea announced that for universal AI for all the people through a consortium. And that's the core of the champion initiative that kind of we're doing as well. That's what Bernie should be talking about. Not Dan AI, should be tokens for everybody. Make it a universal right. I think they want them to be smarter. That's the thing. It's not right. There's a very good book on this automated luxury communism. Yeah, no, I'm familiar with it.

1:38:53I think this, by the way, it's not just about tokens for everyone, it's not just about, maybe there's some, some ism that we're missing like tokenism. We need to coin tokenism here, free tokens, or essentially post-scarous universal basic tokens for everyone. But I think this is where when Peter, you talk about abundance, this is where all of the other forms of abundance, I think, are likely to come from. We make tokens, at least universal basic tokens, essentially abundant for everyone. And then everything else, the healthcare, the utility pricing, the food, the shelter, the education, these are all downstream of getting everyone universal basic, computer universal basic capabilities. 100% when you've got access to AI and robotics, you have access to everything you need in life. Yes, everything you need in life, right? I think one point to make here again, we've discussed on the pod number of times, is the mindset of the US versus China. China is 80 plus percent pro AI, and the US is 80 plus percent against AI.

1:39:56And it's got to change. And I hope everybody listening to this podcast can hear why, right? This is the most important. This is a jet pack for your mind and for your life. This is how you get ahead. Funny statistic on that, Peter, because I made that t-shirt, let's talk more tokens. And I wore it on the podcast a couple of podcasts ago. About a third of people thought I was talking about crypto tokens, that they, I tokens. Another third thought I was talking about, Bong hits. I was like, OK, this would not happen in China. So yeah, the pack for it, let me a little bit. All right. Let me move us along. So Dave, you've been saying for months now that AI entrepreneurs need to figure out how to get into the front door of the Nvidia ecosystem, because that's where the cash is. This week's CNBC tallied up in VDS total AI investments at $99 billion for context. $99 billion is larger than the entire cumulative assets

1:40:58under management for all the venture firms on Earth. Nvidia has become one of the largest AI VCs. Dave, you want to hit this one? What I'm telling everybody is, don't just think of it in terms of assets, which is already insane. Think of it in terms of assets in motion. Because people tend to look at, you know, like a VC firm will inflate its AUM by saying, oh, we manage a billion dollars, but that's across five funds, you know, many of which stopped investing years ago. Like, well, how much are you actually investing this year? Same as true when you look at the mega banks, you're like, well, isn't JP Morgan really, really big compared to Nvidia? No, it's tiny and absolute terms, but it's also really tiny in terms of new investment decisions that it will make this year compared to Nvidia. So if you look at it through the lens of money in motion, it's completely dominated by the big AI companies that now have, you know, 20 plus trillion in liquidity that they want to get recycled to reinforce their positions.

1:41:58So yeah, you work through the venture capitalists at the seed stage, but you really quickly want to be talking to the Magna Mobsda companies who have massive amounts of investment capital. Yeah. And by the way, everybody less than heard this before, Alex coined a beautiful term, Magna Mobsda. Magna Mobsda. Magna Mobsda. Magna Mobsda. Magna Mobsda. Magna Mobsda. Magna Mobsda. Perhaps for the 11 companies at the heart of the innermost loop in the economy. So, you know, it's the Microsofts and the sort of, the Fang companies on the one hand, but also the new members of the 11 including SpaceX and Tesla and... And probably can help the day. ...commodiously. Yeah. Yeah. Yeah. I love that term. So when we speak about Magna Mobsda, you know now what that means. And there's a song. There's a Magna Mobsda song. You can Google it. Of course. And the other thing that's going on is, of course, we created all these centimillionaires and billionaires when SpaceX, when public, we're going to create the same when Anthropicals public and OpenAI goes public.

1:42:59And all of these AI entrepreneurs are going to be reinvesting in the ecosystem. Yeah. They're going to become the largest source of capital for seed stage, Series A stage. And it's only going to accelerate everything faster and faster and faster. Yeah. And this is what the singularity means. I remember once in 1999, right, just at the peak of, you know, the dot com world, a friend of mine said, you got to go, got to go to San Hill Road. There's a river of gold flowing. You have to take your ladle and put it into the river of gold and get your capital for your startup here. And that river of gold now is coming out of all of these companies and their employees. That's right. And it's an ecosystem that's largely working within itself. Now it's got more than enough capital within its own world to build an entire economy inside itself. Because what a lot of people in the outside world we're looking for is, well, someday AI is going to show up and change my auto dealership or change my laundromat.

1:43:59Now very unlikely they're going to bother disrupting the people that are outside that loop. It's almost as taking it too far. But it's almost like imagine it's a South American, a Brazilian rainforest village or an African village, you know, did the computer revolution decide, yes, I must take over that village. They said, I don't care. Just stay there and live without computers and electricity. That's fine. That's what AI is going to start doing. It's going to be this other group of people like our moonshots attendees who are going to be operating in this alternate, massively scaling economy that is largely within itself. And there'll just be a couple of touch points with the legacy economy. Like new drugs and cures will pop up. And they're like, oh, okay. This goes out to the world or new services, new video games, whatever. They're coming out of the AI world and going out to the regular world. But for the most part, the AI world is so big now and so self-contained that it doesn't need to go destroy all white collar jobs. And the byproduct of that is it's mostly going to leave people who don't care alone. You don't want to be one of those people. You want to be part of this massive tsunami. Yeah.

1:44:59You don't want to be left behind by the rapture of the nerds. Yeah. Let's continue on the theme of the economy. So one of the concerns driving a lot of fear, at least in the US, probably around the world, is the concept that AI is destroying jobs. So six months ago, we've said that the data looked murky. And over the last few months, you know, the data has come out very pro job creation. And I want to get this story out to people, you know, so that you can remain optimistic. We can quell the fear. It's one of the missions of our podcast here at Moonshots. So this week, once again, emerging labor market data suggests that technology is a net job creator in the US. Roughly one million professional positions are now classified as AI jobs, while LinkedIn estimated that 640,000 AI-specific jobs were created between 2023 and 2025. The boom is also generating employment far beyond just software AI jobs.

1:46:01Roughly 500 billion in additional annual spending on chips, servers, data centers, cooling, and power infrastructure is supporting demand for electricians, HVAC specialists, and technicians. Well, even those occupations previously expected to face AI disruption, including paralegals and market research analysts, have continued to grow despite the fears. Like Eric Schmidt recently said, AI exposed jobs are growing faster and paying better. Bottom line, if you're watching in a concerned about your job or concerned about getting a job, your number one focus should be getting AI literate, learning the tools, making yourself prepared for an AI-dominated future. Selim, a couple of episodes, you shared on the pod the data around the majority of SMEs adding jobs as a result of AI. Your thoughts on this? Yeah, I mean, two episodes ago, six, the data from principal financial group, they have 100 and plus 1000 small businesses as clients,

1:47:0360 plus percent were adding jobs because of AI, and 1.4 percent were losing jobs because of AI. So that's just an overwhelming thing. David Sacks talks about this on the online podcast all the time, that the job is a huge misnomer, right? And you have to kind of get rid of that and ignore that for a while because this transition is going to be huge. Our negative goes straight to, oh my god, we're going to lose the work. But the really thing with bigger pictures, what happens to the future of work, and how does that transform to be better, right? When we've done, there's a bunch of studies that show something like 74 percent of work in big companies is coordination work. And now you can get rid of all of that. Eric Binjolson calls this white-colored drudgery. You can get rid of that, work on, let AI handle a lot of the crap stuff, the copying, pasting, sales, figures from one report, one system into another, and then work on where your judgment and experience make the biggest difference in the work and for the company and for the organization.

1:48:11You're going to work. People worry about trying to work on our policy on this, figure out how you maximize human agency as we go through this crazy transition, and human agency is exploding because of AI. Dave, what do you see in all your companies? I mean, your chair and on the board and founder of so many companies growing, hiring? Yeah, I mean, I think hiring, and I want to be really careful to say, you have a million AI-related jobs and rising. But within that, there's a lot of co-pilot type jobs, where you're basically twice as efficient as you were using a co-pilot. But there's another subset that are using 10, 100, and soon 1000 agents and managing them as if they were employees. That's where you want to be. You want to try and get to that group as quickly as you can, because just using a co-pilot, the bar is rising fast. That's not good enough. You need to move on and manage swarms. And interestingly, I met my first hire who does all of his work through voice. So he's managing agents, many agents.

1:49:12He's not using a Quirty keyboard, Alex. There's a young guy at 20, he's a 23-year-old. He's a really transcended history there. So much for his to racist. But I really think, you mentioned Eric Brinnell, from White Collar, drudgery, a lot of that stems from being in a chair for many two hours. It's bad for your back, looking at a screen, going blind. That's all going to move over to standing up, moving your hands around, I love minority report, and controlling the agents in a much more dynamic environment. I think it's a much, much happier place for humans to exist. It's very much like Iron Man, you're Robert Downey Jr. And you're just building together with your Jarvis. That's actually where we're really going in the next year. And so the amount of awesomeness in that job function is just incomprehensible. Yeah, the amount of work I deal with was skipping while I'm driving and just having a conversation and say, do this work, write this report, get me the answer. It's incredible, right? Every woman. Every woman. Every woman. And go back to the, I'm not capable.

1:50:12Go back to the automation days, right? One concrete mixing truck, where a place is about 100 workers were shovels shoveling concrete. Yeah. Okay. Nobody wants to go back to shoveling concrete. Like, what do we think about? Just let's move forward, please. Yeah. The other thing that's really clear to me is that we were predicting massive White Collar job disruption just a year ago. The choice to not do that was an active choice by the big AI labs. Under pressure from the White House and also in China, you know, lawsuits. If you just want to leave fire people, you have to actually try to retrain them to become an AI person and you have to pay either way. So the government's got involved and said, we don't want massive voter disruption. The AI companies are saying just doesn't matter that much. We're on this abundance curve that's so steep that working within AI to create new drugs, to create new physics, to create new math is so much more important than the other. Than disrupting everyone's life that we're just going to do it the easy way and do it with the cooperation of the White House.

1:51:14And so I think that's what's actually happened and likely to continue. So as long as you're part of that rising wave and you're not sitting in legacy land, you're going to do really, really, really well and there isn't going to be a massive job disruption. You might, you wrote a best seller of the last economy. What do you make of the future of jobs here? I think it's like the turkey before Thanksgiving. Getting plumber, adding the jobs, but the models have reached that level of capability now that you can replicate our digital workforce in a year and then physically it'll come after. I don't think you'll create jobs fast enough, but we can create abundance. So I put forward the champion proposal last week, dollar, pre-money, everyone, all the kids own the equity of the robots and other things, but actually we can do one better. I think the government should have a massive infrastructure program and it should look to build 100 million robots in America. Now they should be owned by the people.

1:52:15And I think that is how you get abundance. Like again, if you want to have an end purchaser, make them owned by the people. Have that as this massive infrastructure build out because that will upgrade all of America's infrastructure. And I think that's where we have to go. Elon said at the G20 last week, a robot can do the work of five humans. This is true. Like factually we know where it's going. So I think that we have options of where to go. In the meantime, we have to get ahead of things. There are jobs that will get crazy, like HFAC specialist is the one that I think you're seeing electricians already earning $600,000 a year. Working in different sectors. But there's just not enough humans, right? I think the government also in America where it's going currently will push back against robots. And so electricians, HFAC, like there's, you'll be surprised, especially in Europe actually. It's so hot here that HFAC robots will do so well. That's a really, really important point. I think Andrew Yang actually, I think was saying the same thing that we're going to see this huge resurgence of trades.

1:53:16But I think a lot of people have been trained since, you know, a young age that a good white collar job manipulating a spreadsheet is a far greater ambition than being the best HFAC specialist in the world. And so they kind of demean the blue collar world. But the white collar job is the one that's going to become irrelevant. And so a really good life plan is to become incredibly great at helping building data centers, cooling systems, liquid cooling. And then reinvest that money into AI companies that are writing the wave. Just as a life plan, it's a much, much better plan than aspiring to get a degree in accounting right now. So it is inevitable that there's going to be the biggest infrastructure build out of all time, not just in data centers. But again, across America, across Europe, I infrastructure is crumbling and governments will go there. So position yourselves there if you're not using a thousand agents and you will get the biggest hell wind of all time. Alex, your thoughts please?

1:54:16Yeah, well, a couple thoughts. I'll take the position. I don't actually think it is advisable, at least in America, for the government to be owning 100 million robots. I think that you said the people, not government. Okay, but how do the people own 100 million robots through presumably some form of centralized government, which again, to my maybe overly American ear, smells like, wait for it, like luxury automated communism, which is I think what it, what the subtext would be. I'm not in favor of that. I would like to see every American owning a thousand robots. I don't think they necessarily need to be socialized or communally owned. But I do think, I mean, every profession as currently construed right now, I think is cooked. And I think the sequencing of the cooking is what determines social policy. So maybe it is the case that certain white collar professions right now can be automated earlier, more of that paradox style, then say,

1:55:17quote unquote, blue collar professions, HVAC engineering and the like. But HVAC engineering, let's not kid ourselves. HVAC engineering in the next few years with humanoid robots is just as cooked as spreadsheet management and accounting. It's just a matter of sequencing. So I think one, I think it's essential to distinguish between the what passes for at this point, the short term and the long term in the short term. Yes, I think I agree with the premise that there is some remaining alpha in the trades, so called. But in the long term, no, the trades are just as automated and automatable as white collar, so called labor. That's why they build slow down the robots, but I think again, the robots are the long term. And they said the government doesn't need to own them, sovereign wealth funds on my champion idea, or just the government can underwrite the robots that the people can own. I think the main thing is you have to get the ownership of the robots to the people somehow. I think the sequencing is really important that Alex was referring to you too, but if Elon Musk calls and says we're willing to pay up to $600,000 for the best electricians and plumbers to show up in Tennessee,

1:56:24to build Colossus, and the job is only there for one or two years, and then it's automated. Yes. Take the job, take the money, stay nimble, and then the next opportunity will open up, and you'll know a bunch of people that are in the same boat in the middle of the AI revolution. And so then the sequence will evolve and we'll keep on the podcast telling you where to move next. But don't take it as a sign of, well, because that's cooked two years from now, I'm going to do nothing tomorrow. Don't do that. Take the job. I agree. I agree. And then maybe just to underline David, your point further, I think. So I'm always pointing neologisms, including motion with an M O A T I O N, which is this notion that there are no stationary modes in a singularity, but what there are are dynamical modes, temporary modes, if you will. And if you sequence them appropriately, then one mode can lead to another, can lead to another, and you achieve a dynamical mode. So same idea with professions, maybe an HVAC engineering position now enables, and this is not investment device or career advice, but hypothetically, maybe an HVAC engineering position now creates enough of a capital base that can then be grown and translated to something else in two years.

1:57:37That can then be translated to something else dot, dot, dot, eventually you get to, I don't know, owning a planet. So, Sillim, let's take the conversation one step further, which is what happens when you don't have to work? What happens when all of your basic needs, food, water, energy, healthcare, education, liberty, all of that is enabled for you? Right? You and I have discussed this before, going back to the Dici family and the Renaissance and so forth, where you basically have the life, and I've written on Substack about this. You have the life of a gazillionaire, you don't need to work. What do you do? Right? I think that's an important realization, because when we get to the point where everything is cooked, that's also the point at which we have massive abundance, where all of your needs are taken care of for you. And now the question is, what do you want to do? You know, most people in the world are working because you have to put food on the table, they have to get insurance for their family. It's not what they dreamed of doing when they were a kid. Right? So, it unleashes massive possibility at that point.

1:58:39I think there's two categories here. Category one is the material needs, right? When you get to that level and you, it'll take a transition to stop thinking about meeting daily needs, right? Like half the country in the US can put $500 together in an emergency, okay? That itself is an emergency. And you've got such a monster structural problem around that that has to be addressed. Now at some point, pretty quickly, let's use the simplest argument that somebody goes deep with AI, figures out how to use it to do an active trading hedge fund strategy, starts making enough money to pay for themselves, pay for their families, et cetera, et cetera. Right? And overall aggregate in the aggregate, people will figure out and society will figure out how to generate huge amounts of wealth with this. And then you have the distribution question of how do you equitably share it when you've studied that, but that's all the material covering day to day life stuff covers the bottom two, three layers of Maslow's hierarchy.

1:59:41And we could see that happening. And when you see society is getting to this abundance level like the moguls taking over India or the Mongols taking over East Asia or the Romans taking over the Medici family or whatever, you've heard us talk about the idea that people end up doing four activities, food art, music and sex. And the joke is not in that order, right? Once you get past that and get through that, they're really interesting. Next dimension opens up and you start thinking about what problems could I tackle? Because now we're not going to run out of problems. We're just going to be able to tackle bigger and bigger problems, right? How do we create Alex's Dyson swarm? How do we go into the stars? How do we kind of think about the new different types of physics that may emerge from all of this? That's where the really fascinating stuff comes along. And you know, the I think this is the area Peter, we've had this conversation at the express board meetings of could we design prizes that advanced humanity radically, then just trying to solve the problems because those looked like they'll get handled and what's that next phase look like.

2:00:47I think the opportunity for human flourishing and that is so magnificent. We love problems and we love solving problems. We'll just have bigger and bigger problems, which will be able to solve with bigger and bigger sweets of AI agents and robots. All right. I'm going to move us to our last story in economics. And this is a paper of Seleme. I know you're going to love and or have lots of comments on so quick context. As Seleme has educated us on past episodes in 1937, the economist Ronald Kos asked why companies existed all you won the Nobel Prize for this. His answer transactions are expensive, finding people negotiating and enforcing contracts. It's cheaper to hire employees and have them on the inside of your company rather than negotiating every task on the open market. Companies are a work around for transaction costs. Now MIT and Harvard researchers are finally catching up with Seleme have asked what happens when AI agents make transactions nearly free agents that search compare prices, negotiate and transact for you at enormous scale.

2:01:51They're calling it the Kosian singularity. Seleme, you wrote the book on this. Your thoughts on this paper. Absolutely dead on a little late, but totally dead on you know Peter in our 2023 book, EXO 2.0. We said then that the coast's law was breaking. We didn't understand the full implications of it. What we noticed then was that take the mission critical function in Uber, which is to match driver and passenger doesn't happen inside the organizational boundary of Uber, perhaps not in the wild. And by enabling that with technology, you can scale. So we observe the companies were reaching outside themselves to get things done. The price goes to teams all over the world to do innovation. Ted is using its community to scale. But now with AI, it completely changes the game. And now the transaction and coordination costs go to near zero when you can have 1000 agents or 100,000 agents doing kind of crazy amounts of capability outside the organization.

2:02:53I think the bigger point here is that AI doesn't just automate the firm. It attacks the economic reason for which firms exist and the shape of the firm. And what we've been exploring is what does that shape of the firm look like. And we've got kind of that definition going because now essentially a firm becomes a protocol. And a community of agents and human beings is going and attacking various economic opportunities or marketplaces or solving specific problems. So the organization dissolves from being a human hierarchical centered model to a totally different world. And so this is a huge shift. The biggest sense we've had since the Industrial Revolution. And we have to kind of look at what this new model means. And we're looking a lot at the governance, for example, of these agents and so on. So I have to ask the question then. I think this is correct me from wrong. This is research from last year, not from this year, the the Cossian Singularity. If that's the case in my mental model.

2:03:54So the whole point of of Cossian economics is transaction costs determine the size of the firm. And if transaction costs are high, that agitates in favor for a larger firm. And if transaction costs are low, that agitates for a smaller firm. So if transaction costs are going to zero, that agitates in favor of the size of the firm going to zero. So if that if my reasoning is sound, do we think the size of the firm is going to be one person or less than a person. What is that? What's the limit? I don't think there is a limit. If you take the Argentinian model, you could have agents running the firm now, the concern, you still need an entity. You still need an entity. So Ted Shelton and I wrote this paper on what does the organization look like if you take the coordination and so on. And we coined a term called the fiduciary wedge. The reason you still need an entity is for liability for fiduciary for proprietary data ownership for learning loops inside that inside the entity for brand for as a purpose container as an empty peak container, etc.

2:04:58But the primary reason for firm has been coordination and execution and that kind of disappears. So you still have entities in the same way like in investing, you have SPVs and you can have multiple people be part of an SPV that then invest in something. Essentially, you still need the legal container, but the primary reason which coast identified in all organization thinking does come from that. But there's no reason why a company or firm or legal entity can't be owned by other agents, which will eventually happen in Argentinian and other places and then have as completely virtual organization operating in its own domain for the purpose that it was set up for. But I want to push on that though, because there's a countervailing force that one might perceive, which is what we talked earlier in this episode, which is frontier capabilities potentially getting walled off from the rest of the economy. If open AI, for example, hypothetically is using internal unreleased models to solve grant challenges and math and the rest of the economy doesn't yet have access to them as a friend of mine, run it open AI others have point out when that agitate in favor of the exact opposite,

2:06:06the same as the one in the market. And so, you could already have a frontier lab running its own hedge fund strategy. I'm sure they're doing that now that outperforms the market and just running that model, but it's a very niche thing applying for a certain temporal period of time. Over time, I think the big question in my head right now is if we are achieving RSI, what the hell does that mean? I think when we get to that point, the concept of an economy starts to your road and dissolve itself, so you have to think about it in a totally different model. Not as obvious. I mean, this seems to me a lot what's the the aphorism, like what happens when an irresistible force meets an immovable object or something. What happens when a cosian large frontier lab with superhuman super intelligent capabilities meets an agentic economy that wants to distribute transaction costs out to the edge and combined with Argentina wants to create non human corporations, not obvious where that is.

2:07:17Alex, do you remember when Sam Alman said yes in the future, I think in AI should be running open AI. I remember that and Sam just in the past 24 hours also on social media expressed surprise shocked shocked that open AI was able to solve Navier Stokes, so maybe an AI would have predicted this. Yeah, I mean look if you take your commentary to the end point, then you end up with what Daria was talking about where one company like Anthropic will be all private enterprise, right? And the that's unlikely to happen. It'll take a while and at that point the concept of what is mean to have an economy essentially evolved. I think the bigger picture question is where do we even have value creation and value storage and all of this because we come down to the money layer at some point and we should have Jeff Booth on. We were talking about with the future of that. Those are all tokens. I see you I see you ready to burst forward.

2:08:19Yeah, no, I think the economy is like 1% inspiration 99% perspiration. Like you don't need a polymath doing your taxes or selling widgets or things like that. Like we've seen that obviously the Navier Stokes model can sell widgets very well. But once it comes up with the recipes actually about following through and most transaction costs are actually about friction by the economy needs a bit of friction. It's the relationships you build as the other things. It's not an instant thing whereby there are no barriers to spreading against Sam talked about this the frictions and a scarcity oriented economy. It's not a case that the best product toys wins or it'll be on beat max, you know, like that's something for the old kind of kids. What's that? I've never heard of it. Anyway, so the you know like there are also frictions in the economy. So I don't think it will be that case. It's just that the companies and organizations become more efficient and then they become more optimized because it's really annoying when you go past 12 people and then past 150 people.

2:09:22This is where you kind of have that car to get in demon of disorder mall or coming into organizations and they get misaligned. So I don't think you'll see the one person company. I think you'll see the 10 person company and I think you will see more collectives of companies operating with digital and physical humans solving problems that deliver value. But economics fundamentally does need to fit flip from being scarcity based to being abundance based because we can rearrange bits and soon atoms in any way that we want. All right. Dave, you want to close out? Yeah, well, I think to step back from the academic, you know, our company is getting bigger our company is getting smaller clearly. Mercore now has what 50 to 100,000 individual actors that are effectively little companies in India and around Brazil now too. And you can you can take what Selema saying and build a marketplace around it to deal with just the lingering artifacts like, you know, employment law in different countries and make a fortune by taking advantage of the trend that Selema is describing.

2:10:23And concurrently with that, Elon is building the single biggest integrated vertical company that the world has ever seen with complete supply chain control all the way down to raw sand turning into chips. And so those are both happening in the real world concurrently. So I think maybe the real observation here is things are changing very, very rapidly in both directions for very good reasons. And you know, the the academic paper can can wax poetic for years, but this is happening in the real world in those two flavors and it's brilliant opportunity. I agree with you. All right. I mean, it turned us next to Tesla cybercab last Thursday. We covered the Austin launch what I was calling cyber cab Lala Paloza. This week, Tesla opened an official interest form for businesses who want to buy their own cyber cab fleets and build mobility hubs and charging infrastructure for the robot tax, robot taxi network. No pricing or delivery terms yet. This, you know, Seleme, for me, this is another musk business model innovation.

2:11:26Tesla doesn't want to own every robot taxi. You buy it. It works for you. You share the revenue with Tesla. So I think it's a brilliant move for customer financing of a global fleet. And honestly, this can only work in my opinion because of the price tag on cybercabs projected at $30,000 affordable to anyone who is previously an Uber driver. Here's the form I've gone and filled mine out when or if you guys did let's hear a quick video from from Elon. It is an old video, but it, you know, it predicts what he's saying and doing now. But we'll have a model which is kind of like some combination of Uber and Airbnb. So if you're a Tesla owner, right, you'll be able to add or subtract your car to the fleet. So just like an Airbnb, you could like rent out your spare bedroom or in out your house when you're not using it. And the same thing will be available for Tesla owners. So did you guys see that announcement? Any thoughts?

2:12:26Well, it's exactly what Alex was saying a minute ago. If you're the electrician or HVAC person working on Colossus and you banked 600 grand, where do you go next? You would have heard about this with your peers while working on that. This is where you go next and then the next and the next. Well, you buy the opposite robots that now do the work for you exactly exactly. That'll probably also be syndicated out as some kind of a franchise model for maintenance repair and whatever. And that becomes your trajectory of the future. Yeah, totally agree. I mean, I think in the style going back to off self-site earlier comments about the private ownership of the capital means of production. I think this is the way again, not investment advice, but decades ago would have been accumulating a laundromat or a restaurant franchise. Now it'll be, yeah, you own a fleet of robot axes or a fleet of humanoid robots. That's without this being misconstrued as investment advice. That's sort of the quality franchising path to medium and high income and wealth creation.

2:13:29And I think the SMBs of the present and the near future are going to look far more like that than opening say a chain of restaurants. And also, these things get started when they're new. The central force wants you to succeed so badly to get the momentum going that they sensitize the hell out of your success. Like if you were the fifth Starbucks owner, you would have been guaranteed success by the mother ship. You don't want to be the hundred thousandth. But if you get an early jump on this and whatever's next and whatever's next, the mother ship will subsidize the heck out of your success. So you just have to basically not screw it up. I find this incredibly compelling, you know, to go fill out the form and say, I'm going to grab 10 of these and have them work for me or 100 of them. End of the day, it's earning revenue while you sleep and improving your local community. There's a simple economic thesis here, which is dropping marginal cost, right? You think about Airbnb's marginal cost of adding a room is near zero if you're high at you have to build a hotel. Same thing here, the marginal cost of owning one of these and adding having people use it and leveraging that asset rather than centralized ownership of taxi capital.

2:14:36It's such an obvious no burner. So I think this is a huge option. The Dave makes a great point, get in early on these things because then you have scale built in. And the question of which of the cyber taxi companies or robot taxi companies, where I'll call it, it's going to win. It's the one with the lowest operating cost and the lowest production cost. And I haven't seen anything yet compete with Tesla here or the one that's unfortunately that may be cozeased with the municipality is that are approving them. True, it will be city by city. Usually it's pretty obvious who the winners are. Everything that people just lost and yeah, it's so sad. So people. The winners are usually pretty obvious is people are just so slow to react. They're not nimble enough, but it's not going to be a mystery who's going to win. It's going to be people under reacting to the urgency of the opportunity. So I go ahead and I'm I just think it's becoming increasingly clear robots are the biggest investment class that will ever see.

2:15:36Yeah. And so again, like we'll see SPVs, we'll see funds, we'll see all this emerge. And then as you reach that level of capability, whereas Alex so they can take over a HVAC engineer or in this case, do you need more than a two seat or robot taxi for just about anything? No, only when you want to have your air robot taxi. That's the only time you need anything better than that. Like these will have very long lives and they will earn money in just about any scenario. Yeah, you can find out more at Tesla at comm slash Robo taxi if you want to jump into this future economy. And again, not promoting it and not giving investment advice, but I do not. I do find it incredibly compelling. All right, I want to bring us to our final story here today and it's it's an important one. You know, you know, this is underlying much of what we discuss on moonshots. So we've noted before, you know, people having fewer babies, we're seeing a we're going to see a massive drop off in human population.

2:16:45And people are now living longer and healthier and this is changing the global demographic. So here's a chart showing the growth in number of people over 65 and the drop in the number of newborn zero to h five on the planet. The global population over 65 is projected to grow from 852 million back in 2025 to 2 billion by 2060. That's more than half of all population growth over that period is coming from people over 65. The result at least in the old economy is that fewer working age people supporting more retirees is going to put huge economic pressure on pensions on health care in every developed economy. The only way the math works out in this future is because of the topics we discuss on this pod, right, AI and robots doing the work for missing workers, people living longer and healthier not needing to retire, maintaining themselves as economic contributors.

2:17:45Right, what's the reason you retire? You're in pain, you're feeling less energy or you're forced out because of policy. And this is why I say longevity is not a luxury. It's an economic policy for the century head, right, 80 year old with the body and mind of a 50 year old isn't a pension liability. So they're a founder in this future economy. So you've all heard the saying demographics are destiny. Well, robots that do the work and therapies that add healthy decades are rewriting destiny. Gentlemen, your thoughts, so you want to jump in first. Very simple follow up. I mean, this is why we need the robot taxes and the robots and so on because we're going to need to redesign the entire concept of lifespan to health span to jobs, you know, the concept of education career retirement essentially evaporates, right, you have to have repeated cycles of learning creation and then taking a sabbatical, etc, etc.

2:18:45And longevity combined with AI totally disrupts who works how they work, how long they work, what they do. And so it's going to completely change the game. We need to rethink that whole thing from the bottom of you seen the preview with what's happening in Japan and you need it China is moving to robots because they have to because of the one child policy in the population bomb that coming great, great point. So they have to do it. They don't have choice. Yeah, these numbers are going to make the world much more interesting. So these numbers are hugely understated because they include India and parts of Africa that are still having babies like crazy. But if you look at China and you look at Europe, the numbers are much more acute than this. Like there will be basically kindergarten in grade schools that are completely empty and massive numbers of over 65s. And it's not very far in the future. It's also America is largely immune because we have huge amounts of immigration right in the working age bracket. You don't tend to have a lot of immigrants that are 65 and over. And so you won't notice it as much in the US, but other parts of the world.

2:19:51This is this is basically going to completely rip the working class out of the economy and then the voters are overwhelmingly not employed. And so it's going to create I mean, it's already a mess, but it's going to create all kinds of strange things in those in those jurisdictions. I would just add this is what victory looks like that this is what we want to see 150,000 people plus per day on earth are dying and putting an end to that. At least the beginnings of putting an end to that is what this looks like where people are starting to live longer. And people are starting to stop dying. And the counterfactual looks more like the limits of growth from the club of Rome, which I think was like early 1970s, potentially depending on your vantage point, horribly racist perspective on what you were from in the earliest form. Yeah, like based on the false premise that somehow humans are going to be overcrowding, purportedly scarce surface area of the planet.

2:20:52And that was going to lead to decrepit conditions of living for everyone utter nonsense. This inverted pyramid, which some would call, you know, I guess the opposite of of pronatalism. It's not anti-pronatalism. It's this is what a happy future looks like where we have ultimately far more AI agents than we do humans and longevity escape velocity is vanquished. We want this future. Yeah, so I think that there are two major opportunities here. One is integrated elder care with robots at a high level. But the other one is integrated baby care, which would have more kids. Kids are wonderful. And having a fully integrated care for kids throughout that can reduce the cost of child care and supporting and raising kids in the best way is probably the biggest thing anyone could do for humanity. Because there deserves to be more of us. So let's have more kids and let's figure out that problem. All right, Jim, I love this episode. You guys I love spending time with all you this was brilliant. Listen, I need to say something about what do you not set? Yeah, love my kid to death, but damn the work of being apparent is the biggest biological scam ever.

2:22:00It's unbelievable about what it's going to do. It's rewarding. It's very rewarding. It's very rewarding. It's a link between those comments and then your prior comments about marriage over longevity escape velocity. What's going on? I get permission from Lily to make some of these comments. All right, guys. We'll be infinity away in three days. We'll see you again soon for our next episode. Dave Alex, Salim, you might love you guys. Great conversation. Excellent. There's always.