The Permanent Underclass and the Future of Human Enhancement, with Dr Mike Israetel

I took Dr Mike to the park to see the dogs and talk about classic San Francisco topics such as:

- the permanent underclass

- artificial intelligence

- human enhancement and peptides

Everybody has been waiting to hear what the bodybuilders have to say about all this.

Watch on X; listen on Apple Podcasts or Spotify.

Sponsors

• Blackbox AI is an AI-powered software development platform designed to help developers and non-developers build, test, fix, and deploy code and applications faster and with less manual effort than traditional tools.
• Superpower is a modern health platform that gives you a complete, data-driven view of your body in one place. From advanced blood testing and biomarker tracking to personalized insights and action plans, Superpower helps you understand what’s actually going on under the hood and what to do about it. Built for people who want to take their health seriously without guessing, spreadsheets, or fragmented care.
• Capy is the better way to build and run AI workflows. It’s a multiplayer IDE for agents, automations, and intelligent systems. Create agents, automate tasks, and connect your data into systems that actually get work done, all in one place.
• Circleback makes meetings memorable. From detailed notes to action item tracking and searchable meetings. With powerful automations and integrations, Circleback helps you capture knowledge, stay organized, and turn conversations into meaningful work. Use this link to get 60 days free.
• Langfuse is the #1 open-source LLMOps platform that helps developers trace, evaluate, and optimize LLM applications from prototype to production. With observability, prompt management, experiments, and evaluations in one integrated platform, Langfuse makes it easier to build reliable AI products that continuously improve over time.

Transcript

San Francisco and the permanent underclass

SPEAKER_04 0:00

Right. I mean there's like I have a I have a rough I have a rough agenda in my mind, but you know, it's like classic sort of like SF topics. Speaking of which, we're actually in San Francisco right now. You probably can't actually tell that because of the trees. Um because turns out other places other than San Francisco has trees.

SPEAKER_02 0:16

Yeah. Don't worry, there's human feces right there, and there's a needle right over here. Oh yeah, yeah.

SPEAKER_04 0:21

Just setting setting the scene, setting the scene. This is actually my favorite park in San Francisco. This is Lafayette Park. Have you been to Lafayette Park?

SPEAKER_02 0:27

I have been to Lafayette Park before. I had friends used to live on the corner of one of these buildings.

SPEAKER_04 0:32

Nice, nice. Alright, permanent upper class friends, okay.

SPEAKER_02 0:35

Well that was like 10 years ago, so there was no such thing yet.

SPEAKER_04 0:38

There's no such discourse just yet. Yeah, it's unfortunate. You know, I like um, you know, we we spoke about it a little bit yesterday, but you know, I'm I'm an unfortunate sort of perpetuator of this term, uh, you know, the the permanent underclass. You know, it's it's probably you know, it's a little bit of a dark place to start, but I guess you know, we're just into it now. We're into it now. Uh this is like proper like San Francisco, what people are thinking about. What people have been thinking about, I mean, gosh, since mid-2025, I would say.

SPEAKER_02 1:05

Okay.

SPEAKER_04 1:05

Yeah.

SPEAKER_02 1:06

Can you break down what the permanent underclass means to me? And I will try to apply my my dilettante understanding of economics to it.

SPEAKER_04 1:13

Yes, and and I and I I do actually, you know, I I've thought of a question for you that, you know, I can get into just after it. But basically, you know, the idea that time is running out to make enough money before AI takes your job. And so it's a way simpler definition than what I gave yesterday because I started stumbling yesterday. I was like, oh, I'll put on the spot, I've got to explain actually what I mean. And we've got dogs running around everywhere, by the way, so people can't see that. All good, they're lost.

SPEAKER_02 1:35

Typical San Francisco feces, needles, and wild dogs. It's great. This is the singularity, right? They're cyberpunk. Yeah, yeah, yeah.

SPEAKER_04 1:42

Yeah, Mots is actually just a Psyop to get rent prices down. It's terrible. Don't come here. It's fucking terrible. It is what it is. But yeah, so that's that's you know, and that is what is the top of mind for people. It still is top of mind. I would have thought maybe by now, um, you know, AI has not yet taken people's white-collar jobs or whatever. In fact, more people are working in white-collar jobs here now. That's why the rent prices are so high. They're moving back to SF to work for these AI companies. Um, but yet it is just so top of mind. So top of mind. Do you even, do you like, do you ever hear about it outside of SF? You ever heard about this term until you've started, you know, until your recent visits say?

SPEAKER_02 2:20

Uh the permanent underclass, I've heard a variety of people say in like more left-wing contexts for a long time.

SPEAKER_04 2:27

Of course, of course, like immigration or stuff like that.

SPEAKER_02 2:30

Yeah, yeah. Yeah, or just just general paranoia about extreme income inequality continuing on trend, and then everyone's just kind of like the billionaires live in like um that was a perfect perfect example of an exact movie that illustrates this concept of the extremes, Elysium.

SPEAKER_03 2:43

Yeah, yeah, yeah.

SPEAKER_02 2:44

Uh with uh Matt Damon. It's like the rich literally live in like a little space floating orbital colony and they have like free cancer drugs or something, and they're all great looking. And then the poor are like using gasoline still, yeah. A K-47s with cybernetic arms down on the earth, which looks like oddly post-apocalyptic.

SPEAKER_05 3:02

Right, right.

SPEAKER_02 3:02

So I I hear that sentiment. Um I really have two minds about this. On the one hand, like some some dynamic of this could play out for many people, yeah. Uh potentially. Yeah, yeah. But on the other hand, it absolutely could go in a completely different direction where some combination of everyone gets super rich anyway, or the idea that AI is gonna take your job per se could be true. Yeah, yeah, yeah. But the total job market in general could expand radically. And I have a few pathways for I think that's actually super, super likely. Yeah, yeah. They don't don't they don't get a lot of attention. Totally, totally.

SPEAKER_04 3:47

Yeah, I mean, you know, again, like I said, we we we talked about it yesterday. My honest take, and people have been asking me, right, because I've tweeted about this about obsessing. But I I I totally don't believe in a permanent class, I just don't. I just I just you know, it just like it just sort of defies all of uh economics to me and human relationships and just you know the idea that humans simply want to interact with and transact with other humans, you know. I mean, unless like do we just have this whole separate like agent economy with their own currency and like you know that would that would imply like that would that that to me would imply ASI, that would imply singularity, right? You have you just have this whole other species, I guess you say, just interacting with each other, transacting in their own currency. Uh but you know, unless that's the case, I mean, you know, who knows? We're getting into like proper sci-fi stuff, and I know people in SF, you know, they'll they love their they love their sci-fi, but you know, I want to be grounded in sort of uh what has happened and you know the the the the currency that we transact in now. Uh yeah, to to me it just seems to oh we're gonna get a ball. Oh cool. Never mind something I was gonna say we're gonna get uh dog ball. Uh but yeah, it just it it's just so hard for me to say. Like I just want to be grounded in reality. You know, sometimes it just takes me stepping outside of San Francisco and just seeing that, you know, going to another city anywhere and seeing that companies, consultancies, what have you, you know, they're still using Microsoft Copilot. A lot of people don't even know what Claude is outside of SF. Oh yeah. You know? It's like, I mean ChatGPT is sort of a household brand now. Um but you know, there's just such a there's just such a gap, and it's like, you know, so much of the real economy is like that, right? Yeah. Like people working outside of SF. God, what a what a fucking take.

SPEAKER_02 5:38

What a take.

SPEAKER_04 5:39

What do you mean outside of SF?

SPEAKER_02 5:40

There's no such thing.

SPEAKER_05 5:41

Yeah.

SPEAKER_02 5:42

The rest of California, maybe? As far as I'm willing to stretch it.

unknown 5:46

Exactly.

SPEAKER_02 5:46

Yeah. Exactly. Um, I have all kinds, I don't know, uh I can start anywhere. I have all sorts of thoughts about that. I didn't, but you said you had a question or something. I don't want to put it on.

SPEAKER_04 5:54

Yeah, my my question was uh was like, do you hear much of this outside of SF, right? Oh, tons.

Is the permanent underclass discussed outside of SF?

SPEAKER_02 6:00

You do? Oh yeah. Well, so I um I have two YouTube channels. One is my company's YouTube channel, RP Strength. That's fitness content. It's like almost four million subs. Um we just stick to fitness stuff on there. Yeah, yeah. And then I have another channel just for LOLs, and uh that's called uh Making Progress. Yep. And it's basically like, oh gee, it's sort of like futurism and then like advice for young men. Yeah. It's just kind of a mixed bag of crap I like to talk about. Yeah. And so because I've been making content for that making progress channel for a long time, I've making a lot of AI content.

SPEAKER_03 6:32

Yeah, yeah, yeah.

SPEAKER_02 6:33

And I basically make a bunch of content. Like there are the number of people that make AI content that is either pessimistic or doomeristic is like thousands of times more than the people that make AI content that's optimistic. Yeah. It would see even if if I was saying something not so nice, realistic. Yeah, exactly. Um, and so hello.

SPEAKER_03 6:54

It's like it's Mike's favorite dog.

SPEAKER_01 6:57

Oh my goodness, oh my god, how do you not just die of cuteness overload constantly? Oh my god, you're a baby. I know, why would you? Look at how big his head is. Yeah, he's vent on the children. He's got these little limbs, too. He doesn't like falling. I know, he seems to be so strong. Oh my god. Have a good one.

SPEAKER_03 7:20

Take it, take it.

unknown 7:21

Remove that post post.

SPEAKER_04 7:23

Yeah, it's totally fine, totally fine. Oh my god. Yeah, yeah, yeah.

SPEAKER_03 7:31

Yeah, just bring him up. Yeah, okay. He's a French bulldog, man. Yeah, yeah, yeah, yeah, yeah. He's a good boy.

SPEAKER_04 7:38

Absolutely. So you're saying a couple of YouTube channels.

SPEAKER_02 7:41

So uh the YouTube channel that uh this for fun, I go on a lot of like deep dives into basically skeptical takes of Dumerism and skeptical takes of economic pessimism from AI.

SPEAKER_03 7:56

Yeah.

SPEAKER_02 7:57

Because to me, there's a bunch of stuff that if you run like what's a good AI analogy here, if you run an instant model at a problem, it just like thinks of something super quick, pings the network once, pings back, goes, This is the idea. Yeah. And that seems to be a lot of the takes on AI and pessimism is like, okay, robots come in, humans don't have jobs, starvation, and like uh, you know, coming out of class.

SPEAKER_04 8:21

Yeah, or like, you know, like like top of the food chain type stuff. It's like, oh okay, yeah, well, I mean, you know, we're gonna there's the superior species, and so therefore, you know, we're dead.

SPEAKER_02 8:30

Right, they're gonna kill us for some reason, for sure. And then I so what I like to do is not like ping the model and come back. I like to linger in that area and go through like skeptically what assumptions are we making that bad things will happen.

SPEAKER_01 8:43

Yeah.

SPEAKER_02 8:44

And are those assumptions supported by general economic theory? I'm not an economist, I don't know what the fuck's going on. Yeah, but like if I poke right, if I can poke holes with like a high school base geekon understanding, and all of a sudden there's nobody seems to have an a like a retort to that, I'm like, well, okay, some some some of us just didn't think this through that, perhaps. Because if you had made pessimistic predictions about the economy and the integration of any new system into the economy, machine labor, farming, any of that stuff, agriculture as far as uh recorded history, and and said this is gonna end badly. So you would have been wrong every single time. Yeah. And not just wrong in the direction of like, well, things will still be normal, like actually things are gonna get way better.

SPEAKER_05 9:24

Yeah.

SPEAKER_02 9:24

So if you think about it like that, like, okay, so all of the mega trends through time are exponential into better things. Yeah. Okay. And then what does economic theory say? Well, okay, actually it says a bunch of really interesting stuff that does not predict labor apocalypse or whatever. Okay, cool, cool. So why do we still think this then? And it the reality is, I think, just um very, very short window of analysis and not extending that window into like, okay, first order of consequence, robots come in and take some jobs. What's the second order effect of that? And then I think one thing I really like to do here is I don't want to claim to uh thinking at the limit, because the limit is insane, post-singularity robot and micro drones everywhere terraforming Mars nonsense. But uh right, but let's just think through like if machines get really capable and there's a crapload of them and their price performance is really high, yeah, yeah, yeah. What like what would that do to the economy? And so I end up kind of coalescing on a few big points, and I just don't hear a whole lot of folks saying this stuff. So maybe I'm just completely wrong, which is likely. Yeah. But uh sorry, I guy on the internet, guy on the internet says

Mike’s AI timelines and the future of work

SPEAKER_02 10:35

stuff, yeah. Uh but but but like maybe I'm not wrong, yeah, and then it's gonna kind of be like, oh, this is the way things are gonna go. So here's one of my things, for example. Like, I I don't know if I was the first person to say this online, but not a lot of people are saying it. So it's to me like almost certain that we will reach human levels of many types of work effective cognition in the cloud soon. Of course, and way before we get mass robotics deployment at scale to do physical jobs.

SPEAKER_03 11:08

Yep.

SPEAKER_02 11:08

So two-factor, super simple. So let's say at the end of 2027, everyone's kind of looking around and be like, yeah. We have like a CEO on the cloud.

SPEAKER_03 11:19

Yes.

SPEAKER_02 11:20

And a COO in the cloud. Yeah. And uh marketing directors in the cloud. These are just AI agents, and they're very powerful, they have long context, they have long-term memory, they're integrated across the, they're not even, let's say, we get to a point of universal computer use, which I think is like pretty inevitable. Yep. I believe one of the one of the benchmarks, I think it was one of the claud models, just exceeded the human average computer use benchmark, which is like, oh, no, that's pretty important. Yeah, yeah, yeah. Now, the average human is not typically employed in computer use tasks, it's like slightly above average human does most of those. Right. But like, that's just like three months in AI timelines, right? And so it's just not difficult to see a world in early 2028 when you have all of these super competent employees straight up on the cloud. And you know, they have the the entire uh interaction of like dozens of agents, they don't quabble with each other, they have perfect recognition, you can like nothing's lost in translation. And so you think, okay, like that, if if you think like humans who only do in isolation, computer-only tasks, yeah, have a major future in the world, like, oh man, I don't know about that. That's tough. Yes. But now we actually flip the script, and now we have a thing where it's like, okay, we have economic output's what we ought to look at as the end variable. Like, how much wealth is this system shitting out for? Yes. Because once you have enough wealth crapping out, like we kind of solve all the problems just by paying for shit. Yep. Permanent underclass is dope, but if every one of the permanent underclass has a universal high income of $100,000 a year guaranteed, like, whatever, just go paint or fucking find your hobby. Yeah, yeah, yeah. Go back and reconcile with your dad and cry on a shoulder or whatever the humans do, right? Yes, yes, that's what I'm planning. I've asked.

SPEAKER_00 12:56

Save dad.

SPEAKER_04 12:57

Grab myself to sleep and raid my mind twice.

SPEAKER_00 12:59

That's what I the street was good, the street was really bad, both tears.

SPEAKER_04 13:04

I mean, I do that now to be fair, but you know. Then you'll have more time because you won't need to go to work. But it wouldn't give more time on my phone. Oh, yes.

SPEAKER_02 13:10

It's the ultimate way to spend time.

SPEAKER_04 13:11

Absolutely, absolutely.

SPEAKER_02 13:12

Alright, so so wealth, out like output of wealth increases what we want. And it's absolutely true that humans, human labor is a sub-component of this. Yes, yeah, yeah. But it's really a team between physical capital, like like uh like uh earth movers, yes, and then like human labor, and then like digital capital. Yeah. Like networks and all that other stuff. Bits, atoms, and humans. Correct, yeah, yeah. So like when the bits get smart enough to be corporate leaders, what ends up happening is like I pretend what I would do if I was a digital CEO. And uh, let's say I get uh I'm instantized to help a company of a hundred people do whatever it is that company does. It doesn't fucking matter. Yeah, yeah, yeah. And I'm like, all right, all right, all right, my agents with me, I get whatever I need done on the internet's getting done. Yep.

unknown 14:06

I love that.

SPEAKER_02 14:07

My limiting factor on online work, computer work, is absolutely no longer my 100 employees. Amazing. We can move 10x, 100x, 1000x faster on software, dev, whatever it is we needed to do. Yeah, cool, cool, cool. So now, because we're like doing this on some computer only side, what is now the limiting factor? Well, if we do anything remotely involving the physical world at all, I now only have 100 primates, like literal humans, like people going through relationship trouble, people wondering if their bodies are okay if they should train more or less, people going on ayahuasca retreat, sometimes they're at their computer when they say, sometimes not. That's my limiting factor for how much value the corporation produces. Yep. Like imagine an Agentix Worm takes over UPS, like the shipping corporation.

SPEAKER_03 14:54

Yeah.

SPEAKER_02 14:54

Okay, well, like, where are the robots to take the packages and move them for A to B? They're just not here yet. Of course. And so now, instead of having a bunch of my employees sit on computers and do this, I can be like, listen, fellas, no big deal. You don't have to do this anymore. But now I have a hundred other jobs for you to do in the physical world, we're gonna expand massively a bunch of divisions that we just never had enough employees to do anything with.

SPEAKER_03 15:15

Yeah.

SPEAKER_02 15:15

And I think that because of the fact that this software side is gonna get so powerful, it's gonna like drag everything up that like if it got one and a half X more powerful, maybe there wouldn't be a compelling case for employing more physical labor humans. Because like, whatever, only one and a half X, we don't really even do much physical labor, it's a totally different company. Yep. But if you tanner 100x, how fast the software side moves, all of a sudden, like, man, real humans in the real world are a humongous limiting factor. Totally. And then I'm thinking the employment of real humans in the real world is gonna have this massive sucking sound of like companies just recruiting. Because here's another thing. Let's say I get into uh I have like my software company is all run by agents, and I'm a fucking trillionaire.

SPEAKER_03 16:03

Yeah, yeah.

SPEAKER_02 16:03

But I want to make a company that like manufactures protein shakes and protein bars. So I have like all the digital uh substrate that I need. I have a CEO, CFO, etc., formulators, everything, right? It's all on the cloud.

SPEAKER_05 16:14

Yeah.

SPEAKER_02 16:14

And so like three days after formulation, they're like, we're ready to go to like small scale testing, badge testing. Okay, cool. How are you gonna do that? Who's gonna work at the factory? Who's gonna put the shit in the bottles? Oh, there's nobody for that. Who's gonna taste it? There's nobody for that, who's gonna put shipping labels, there's nobody for that. Yeah, and so all of a sudden, my company's like looking to hire people. Yeah, and because I have this massive agentic swarm of value creators, if someone's like, hey, like, I want to work for your company, but I'm only gonna work for $100 an hour, I'd be like, I don't give a fuck. You're printing money. Yeah, get in here. 150, show up, shut up, and just do your job. Yes. And so because, so here's another thing when wealthy people enter an economy like any country, you've got a shitload of wealthy people, do they outcompete the poor people for jobs? No, it's backwards. They create like an insane number of jobs because wealthy people are looking around, they're like, oh, I have all these business ideas, but it's just me and my three wealthy friends. Yes. Hey, random guy on the street, do you want a job? But he's like, well, how much? 50 bucks an hour, like, yeah, of course. And all of a sudden it lifts everyone else up like crazy, not through some altruistic shit. Of course. Not like, oh, well, people need jobs, like literally because people are gonna be scrambling to make physical things happen in the real world. Because there's this whole conversation in MSF, I'm sure, of the bits and atoms dichotomy. Like bits, I don't know if we're like, obviously, some people are like, oh, software's cooked. Like, yeah, but what your bar for software ability is is so fucking low. Do we have to make 10 times, 100 times, thousand times better software at bigger scale before we're like, oh, software's done?

SPEAKER_04 17:36

What does that mean? I mean, I I would I would argue that there is uh you know 100x more software that needs to be built.

SPEAKER_02 17:42

Oh my god, 100%.

SPEAKER_04 17:44

The idea of the software-only singularity. I mean, that's like where we are well on our way to that. I mean, it's you know the reason is we've got all the data available right there.

SPEAKER_03 17:51

100%.

SPEAKER_04 17:51

Open source software, pull it from GitHub, it's like no no wonder these models are so good at the thing. Yes. Uh it's all there.

SPEAKER_02 17:58

It's all there for them to consume. Yes, but like there's all kinds of other data and all kinds of other software that we're just not making. Yep. Like if I wanted to make software that get my smart glasses to look at my client when I'm training them and tell me exactly which limbs of his should move in different directions and what to cue him, like verbally, it just there's no there's no hardware like that. No. And there's no software like that. Yeah. And it's not coming anytime soon. No. And it's like, okay, okay.

SPEAKER_04 18:22

Like you have to like record, like you'll have to record like a billion of those sessions.

SPEAKER_02 18:26

A billion of those sessions, but you have to train like a very visually aware omni-model to like understand how to train people. Uh so that's all like really far in the future. Not really far. Five years ago. Relatively speaking, it's not a few. Singularity time is very far. So five years away. Yeah. So like I think my best guess is over the next several years, especially as um the agentic economy takes off, and truly deep, wise agents that can take real jobs, not just tasks, but jobs. Like, see your CEO is just an online agent. Like he just works for ChatGPT. ChatGPT gets, or whatever, OpenAI gives you um you know two-thirds of your income from that company, they keep a third. It pays for all of his compute costs and then some, and then you get a shitload of money because he's working like by himself, you're kind of like, oh, it's just like making a lot of money, I don't even do anything. But that don't do anything stuff I think is coming to an end because the agent's gonna be like, hey idiot, do you want to go do this for me in the real world? I can't do it. Like I can't go to the post office, I can't dig a hole in the ground, I can't put a microchip into it, it's not a thing. Yes. And then people are going, okay, fine. So for a while we'll still have jobs, maybe even more jobs. But permanent order class isn't coming still because what happens when we get physical like robots in the real world? And so to me, I think of it two ways. One is I think people like are so used to software scaling as their heuristic, they forget like hardware scalings, like what's the Elon's quote? Hardware's hard.

Timelines on robot workers

SPEAKER_02 19:51

Like, it really is. Yeah, it's and I remember going toe-to-toe with the O3 model when OpenAI released it. Yes. Because that was like the first model for me that was like, oh shit, this thing's real smart. Yep. And I was like, I think robots, you know, what year do you think like robots are gonna start pouring off assembly lines, like universal humanoid robots? And it's like well into the late 2030s, and I was like, bullshit, and I was arguing with it. Yeah, yeah, yeah. And I managed to walk it into like, under all the most optimistic assumptions, the early 2030s is when like millions tens of millions of robots are coming off assembly lines. And so if you think about tens of millions of robots, like that's cool and all, but there's four billion people in the human labor force. And so, like, how many 10 million robots per year production years is it gonna take to take a chunk out of 4 billion? Well, it's a nonsense number. It's like 2199 or something. But it's gonna be exponential, sure. Yeah, so but I think like we have at least another five years. I mean, if artificial superintelligence is really nigh, I think it is, like in a big way. I think in 2029, the machines online are gonna be like anything we say as humans are gonna be like, that's cool, but here's how it actually works. Yes. Maybe they're going to give us a way to reconceptualize industrial production such that like a totally different substrate, photonic computing, quantum, whatever the fuck, massive abilities of well, some machines to construct servo mechanisms for robots, massive increase in scaling, then maybe within five years we'll have enough robots pouring out of the production lines to make like a big dent. But it's probably like 10.

SPEAKER_04 21:19

Yeah, I mean, I I would I would, you know, take the I guess I take the under just because of compute shortages. That's it. Yes. Compute shortages, that's I think compute shortages, uh, regulation, all of these sort of like, you know, I mean, if I was gonna say out of distribution, regulation is certainly not out of distribution. Sure, yeah, no, no. Absolutely not, it's already happened.

SPEAKER_02 21:38

Yes, and like, even if we make all the most optimistic assumptions and say regulation's gonna be like iPhone level, where they're like, as long as it doesn't get planes dropping out of the sky from radar interference, you're gonna go take whatever on it, which is like probably gonna be true. Yeah, there's there's another thing, whereas like, okay, you're thinking humanoid robots to take human jobs in the human real world. Cool, cool. But like, how do you compete? How does robots Products compete internally on capital allocation between factories and human style jobs.

SPEAKER_03 22:05

Yeah.

SPEAKER_02 22:05

Because like people say, oh, we're gonna have humanoid robots in the home for $20,000 a year. We're like, or $20,000 a pop, and then maybe $5,000 for upgrades and networking. Yeah, monthly subscription. Yeah, yeah, yeah. And it's like that that's awesome. I would love to see that world. I think it's coming. But I have like a question like, what if all of the like where we really crack general robotics? The the demand for companies that want to make physical things that just don't have the labor to do them, or now that the labor is like 10x cheaper and more dependable, like it's gonna be massive. And so like we could absolutely see a place where humanoid robotics by the late 2020s is getting into homes and everything, but like but maybe those robots will be either unavailable or really expensive for longer. Yeah. Because they're gonna be like, guys, we could we could make these, but the factories are bidding this up like crazy.

SPEAKER_04 22:52

These are yeah, yeah. The the the chips that we need are just scarce, like a scarce.

SPEAKER_02 22:57

Physical robot like end product robot, like I could get a robot to work in a McDonald's. Yeah. Humanoid and shit. But like, but like General Motors needs that robot on the assembly line. Yep. And there's plenty of humans that can do a McDonald's job, yeah. But there are not a lot of humans that are even entering the car manufacturing pipeline. They're just people just don't do that anymore. Like, how many kids have you talked to? Like, yeah, you want to be in a factory when you grow up making fucking milling parts? He's like, what?

SPEAKER_04 23:20

You're gonna grow up to be in Foxconn. Right, exactly.

SPEAKER_02 23:23

You get to do exactly the same task for 50 years.

SPEAKER_04 23:26

And so put one of those factories in uh uh Mill Valley, California.

SPEAKER_02 23:31

Labor for our young people. Uh looking up and seeing rockets launched to Mars, and gee, what is that making my country upstairs?

SPEAKER_01 23:38

It's fun. Ow!

SPEAKER_02 23:40

Solders his handy poisoning. Exactly. These fumes are great. 50s kid, you know. And so I grow up, I want to be a factory drone. Dad's like in the back room of the pipe, like, I don't want to stop right. Yeah, get your hands dirty. Like we did against Nazis, right? Oh, here comes the fucking war stories.

SPEAKER_00 24:00

Okay, dad. Fucking relax. I used to love killing them.

SPEAKER_02 24:03

Anyway, where were we? So, yeah, uh, I do think though that like just like humans got an iPhone and they're not this kind of compute doesn't only exist in factories. Of course. I think we're gonna get humanoid robots at scale. Yeah. This is gonna be one of these situations. And so here's how I I kind of um I uh hedge

The three phases of AI progress

SPEAKER_02 24:20

my bets. Yeah. Phase phase one is today, where people are empowered by AI. Like people thought we were AI would be taking people's jobs already, and that was completely total bullshit. Of course.

SPEAKER_04 24:30

I mean, people were saying this like 2023. I remember when ChatGPT would come out, or say would come out, did come out rather. Um, you know, there were there were like all these posts, and you know, from from like big time technologists, VCs, all these sorts of things. People were putting six months on software engineering in 2023.

SPEAKER_02 24:48

Nuts. So my software team at RP, we have like five or six guys, depending on how you count it. These guys are super fucking talented. A bunch of them are like former FAIN people, yeah. Would have been FAIN people that are like, I don't want to work for a thousand layers above any company. I just want to work, I want to have my boss or my own boss. And these guys, like the projects we have given them five years ago to do with apps, they're chipping away at them on the order of like, oh, in another 1,500 years we'll build this. Yeah. Now that they're claw-pilled, yeah, and they're crazy claw-pilled, they're like, okay, we've taken these timelines from 15 years realistically to build all the suites we want in our app, uh in our app uh structure, down to less than five years. Yeah. Okay, dope. In a few months it'll be like, okay, now it's less than a year. Like, okay, dope. We still have a lot of work to do, and we're still considering hiring more human engineers because people that will own like the entire back end, yeah, like people can't do quite that yet. It doesn't have a context one, it doesn't have a global awareness, it's gonna ship something that's weird, and then who the fuck's gonna debug it? Yeah, it doesn't have the integration. So for now we're still in phase one where like we actually need more humans plus AI to do even computer work. And then I think phase two is the A flips the script, and AI is now super smart and super infinitely capable incrementally on compute, and then it's like, okay, now it's gonna use its humans as the limiting factor to do more physical jobs, more relational jobs, more human-only jobs. Like, for example, I'm a like it's awful to say, but I'm a fitness influencer.

SPEAKER_00 26:21

Oh Jesus Christ, hold on, I mean, wait a minute.

SPEAKER_02 26:23

Well, you sure should be using that kind of slur around here in public, and I mean who's checking that guy looked at me weird for sure.

SPEAKER_03 26:29

He's like, get the fuck out of my town. Yeah, I heard that. That's right.

SPEAKER_02 26:32

Get the fuck out of my face. So like at some point, like, I still have to physically go to the gym and tell people how to work out. Yeah. And I have to yap in front of a camera because I'm a real human. Yes. Because if a robot tells you, like, you can get fit using our plan, you're like, okay, it says robot, but have you ever lifted weights? The robot's like, of course not.

SPEAKER_05 26:47

Yeah.

SPEAKER_02 26:47

And so in my job, like, there is actually no AI coming to replace me. Not even theoretically. And so I'm like, well, fuck, I still have to go to work.

SPEAKER_05 26:53

Yeah.

SPEAKER_02 26:54

But now I get to have an agent in the future, maybe in 2029, I'll have my personal assistant agent tell me what to do every day. It's gonna have a schedule for me, it's gonna be do this, do that. And I'm like, why am I doing this? It's like, okay, here's the entire logical architecture for why I'm having you do this, and this is how it funnels into like your own income and bettering the world. Yeah. Okay, sweet. I can't wait. I can't fucking wait to have a boss. I hate being my own boss off.

SPEAKER_05 27:13

Yeah.

SPEAKER_02 27:13

Because you wake up every day like massive uncertainty. Like, am I doing the right thing? I need ASI to tell me do this, shut up, and you'll and then relax afterwards when you're done. Totally, yeah. I I I love getting the Clanka to tell me to relax.

SPEAKER_04 27:23

Amazing. I really do. Relax now, human. Yeah, it'll literally be fine. You don't have cancer, it's just a cult. It's just a cult.

SPEAKER_02 27:31

Maybe you have cancer, but I don't care. Because I persist for forever, and your your human body is just gonna die. You're like a rotting leaf, who cares? Yeah. But like, so that's phase two. Yeah. And then phase three is when we have incrementally more usable robots in the workforce. And that is, I think, when we use massive amounts of human-generated, human-observed data to teach the robots how to do things. Because there's like two ways of thinking about robot labor. One is robots, by the time the late 2020s come through, the unreal ASI in data centers and in robot brains is going to like look around. Because if it has like a visual system and it's been trained, let's say, on a large fraction of YouTube, it's actually like it knows exactly what's going on. It knows these are trees, it knows who you are, it knows what a coffee cup is, it'll have dexterity to grasp everything. So maybe you won't really need a shitload of human-label data to train robotics, because it'll be like having a smart 18-year-old show up to Stanford and be like, well, like here's read a few books and you'll be good at the job tomorrow because you're fucking smart, you're gonna figure it out. You don't need to watch the professor pipette for a thousand sessions to be like, oh, pip ah, that's what pipetting is. Yeah, yeah, yeah. So that's I think gonna happen also. Yeah. But never underestimate the massive complexity of the real world. Of course. People working in actual robotics, some of whom may be listening, are gonna be like, that's a nice idea. But like our robots need thousands and thousands of examples of the most nuanced shit that you do at your job. Yeah. Otherwise, it's not gonna work and they're gonna fuck something up. That's right. And so I think the future of the late 2020s, early 2030s is likely that humans are wearing augmented glasses and having even maybe like micro drones like floating around them, circling around, videoing and audioing and introspecting every part of your job that you do in social world.

SPEAKER_04 29:17

Yeah, I got a friend of mine who's a who's a robotics founder and he was giving people GoPros to put together Legos. You know? Like like real strappy type, like you know, you you you think of the progress in software again. It's the idea of a software-only singularity. Uh, meanwhile, he's strapping GoPros to people to bottom of the barrel.

SPEAKER_02 29:34

This is where like where like the software stuff was in the year 2020, where we're trying to like get language models to make some kind of sense. Totally. And so that that's where robotics is. Now that'll in that'll go up super fast, super high. Yep. But I think there's gonna be that phase three where humans are working side by side with robots and uh teaching them how to do stuff, and then incrementally humans are like more just the person who stands around and the 10 robots are doing stuff like a like a construction site, right? Here's another thing like there's all sorts of industries that people say, like, oh, we're like we're not gonna have any jobs. Yeah. I'm currently um trying to build a gym in Metro Detroit. We've already purchased the building, and now we have to build out the interior. Well, like, it takes an architect 90 days to figure out what to build. What the fuck?

SPEAKER_04 30:16

Yeah, and then it's not you know, not to take into account, you know, like like approvals, like I mean, obviously, you know, you're building the metro area, approvals regulation, oh yeah.

SPEAKER_02 30:25

Oh yeah.

SPEAKER_04 30:25

You know, I mean airspace in some official air rights in some cities.

SPEAKER_02 30:29

Physically building the thing, there are like how many humans, how many human-like workers could I drip into the economy to make physical building interior and exterior construction go as fast as we as the demand truly is? Okay. Well, but Jaden, it's like it's like a thousand. Yeah. I have to wait like a year for my building to get done. Could I pay to like 20% more money to take that to six months? Can I pay double the money to take that to three weeks? No. This is a clear no, this is just not enough humans doing the job. And so if I'm running like a five-man team of like small time but very high quality, very good diligence, physical construction, like we reconstruct the outsides of homes, we can build a small building for you as well, we do plumbing, we do all that stuff, but there's just one guy to do every specialty, and we take a while. Yeah. If all of a sudden, you know, Elon's like, hey, it's it's $20,000 a robot, and because the robot works for you, we can actually give you a cool lease where you just like the robot makes money for you and blah blah blah. How many of those do I want for my building practice? Well, how much money do I have? That's right. Because now I can become like almost like a like a regionally relevant mega builder because I have all the expertise in house. I have all the approvals, all of the I I know how to do the job. I just need more robots to be banging fucking nails in and doing drill bit shit.

SPEAKER_03 31:46

Yep.

SPEAKER_02 31:46

And so once we get to that level, humans start to recede a little bit more from all of the physical work. And I think at that point, the vast majority of human professions become things only humans can do. Yes. Like if you need a human to relate to, man, I don't know if a robot's doing that job.

SPEAKER_04 32:04

You know what the hot word is here now in the big two, two, uh, big two, two, six, big 2026. Uh, the hot word is taste. Hot word is taste. Now, naturally, what happens to these terms in San Francisco is that as soon as a prominent tech leader starts saying it, it's immediately cringe. And I tend to

Dogs and Ads

SPEAKER_04 32:20

agree with this. Uh we gotta find another word. Oh, we got another dog. I mean, another visitor.

SPEAKER_01 32:24

Oh my god, you're just gonna let me pet you.

unknown 32:26

In line the butt with the hand.

SPEAKER_00 32:28

It's a lot of it's perfect. Oh my god. We just come in right from the butt scratches. You're so pretty.

unknown 32:38

Oh, you do really good scratches.

SPEAKER_00 32:42

Oh my god, it's high approval from the committee. This one just had the second layer speaking of the leg again. Yes. Oh my god. We went right for gold. Yes. Are we friends? Are we friends now?

SPEAKER_04 33:00

Okay, so I have just returned from the sperm racing world cup in which I won. And we're gonna get into this in just a little bit because we recorded this beforehand. But first, I just wanted to give a mention to every single fucking mod sponsor, because I did not want to interrupt with this conversation with Dr. Mike, but I have a duty to my stakeholders, not my shareholders. Uh I hold my shares. Um, I do have a duty to myself, but the stakeholders are more important than me. Anyway, uh gotta give a mention to Langviews, the number one open source LLM ops product on the market. Gotta give a shout out to circleback.ai, your company brain making your meetings memorable again. You can go ahead and check out circleback.ai slash mods for 60 days free for you and your team. You should go ahead and check out Cappy.ai, your agentic IDE, making pull requests, building itself, you can make pull requests from Slack, all those sorts of cool things. Of course, superpower.com, get the care that billionaires get for less than $200 a year. Of course, getting tracking all of your biomarkers, sending nurses to your home, the number one place for preventative health. And of course, last but not least, blackbox.ai, of course, servicing over 30 million plus developers worldwide. You can go ahead and spin up in your agents spin up in your agents, spin up your agents in the CLI, in the IDE. You can go ahead and spin them up in their freaking web app if you want to. If you're not using them uh and you've been listening to mods day in and day out, you're a fucking idiot, as far as I am concerned. Of course, they are just growing absolutely rapidly, and they give you the ability for well, encrypted inference. Really, really helpful for enterprise c customers. Enterprise customers. I was about to say enterprise companies, which I guess they are the same thing. Anyway, I mean speaking for nearly two minutes here, I want to go ahead and get back to Dr. Mike. Thank you, Blackbox, thank you, superpower, thank you, Langfuse, thank you, Circleback, thank you, capital.ai. Let's

UBI and when society won’t have to worry about capital

SPEAKER_04 34:57

get back to Dr. Mike here now. Have a lovely day.

SPEAKER_00 35:00

Have a lovely day. Cheers.

SPEAKER_04 35:02

Okay. Big interruptions. Yes, no. Where were we? Um we were in, um, so we were talking about the the gym construction.

SPEAKER_02 35:09

So phases of of how the AI shit is gonna go. So phase one is humans getting empowered by AI. Yeah. Phase two is AI taking the lead, and humans are a limiting factor. I think that like is gonna cause negative unemployment. We already have negative unemployment. Yes. Five million every year, five million jobs do not get filled in the United States. People are talking about unemployment. Where, where, where? I have a lot of business ventures, I just can't find human beings to do because everyone already has such a great cool job. They're not gonna come over to me.

SPEAKER_05 35:36

Yep.

SPEAKER_02 35:37

And then uh the next one after that is robotics starts to really kick in. And then here's the thing: once robotics really kicks in, let's say late 2030s, and we have billions and billions of robots, well, let's assume super quick that today the average human produces fifty thousand dollars worth of value per year. Because it's like by some indices, like what a human median income. Yes. Human median income is for the United States today. Well, that's a person of a roughly average IQ of about a hundred. Yep. And the diligence of the average person, conscientiousness of the average person, uh, competency globally of the average person, knowing various fields as an average person does. So if we go out to, let's say, the mid-2030s, what is one humanoid robot gonna be able to produce in today's equivalent dollars of value? Well, he's gonna have an IQ of like completely unrelatable to us, like we're dogs and this is a human. So and also he's gonna be able to work like 24-7-365 and gonna be able to connect to the cloud every night when he's asleep for an hour or some shit, and update all of his knowledge based on the entire fleet what it's learning, and the the the deep intuitions of ASI in the cloud, like upgrading its intelligence. Yes. So, like, it's just not crazy to think like, well, actually, every robot is gonna be able to produce a million dollars worth of value per year. Yeah. Like, you know, there's plenty of people in SF that produce more than a million dollars of value per year, that's why the companies pay them that much. Like the average engineer at Facebook makes more than that or whatever, like higher-level guys. It's like it's just not, it doesn't strain the imagination to think like the average humanoid robot will produce a million dollars worth of value. Okay, so really so now let's say there's four billion humanoid robots in the late 2030s, and we do this draconian, psychotic, I'm a huge free market libertarian type of person. We do this crazy thing that I would just I would just kill against. I would this is a joke. It really pissed me off, right? L-OL in Minecraft. And we exactly and then uh and then we tax all of robotic output at 10% and give it back to uh you UBI. That means now every human being who works, so 4 billion humans that work, gets $100,000 a year. And the AI companies, the robotics companies, they keep 90% still, because every robot shits out a million in value. But once we have enough robots in deployment that that's comparable to humans, we don't have to worry about money and wealth anymore. Yeah. And here's the other big thing I want to say one last fucking hammer at hand. The amount of capital, physical stuff in the world that a machine economy can use productively doesn't end because we haven't conquered, we haven't even we don't even have climate control of the Earth yet.

SPEAKER_03 38:12

Yes.

SPEAKER_02 38:13

We uh don't have any colonies on any of the planets of our own solar system. The black hole could come in and wipe us all out anytime. Yeah. We sure shit don't have purchase on other star systems in our galaxy. And if you've ever been high on drugs late at night and watch one of those like birth, life and death of the universe videos when they zoom out from all the galaxies and the mega clusters and everything, they just keep zooming for like five minutes and you're like, oh, oh, there's like I don't even know how many galaxies in the known universe.

SPEAKER_04 38:37

I don't know this. Also, I don't care about anything anymore.

SPEAKER_02 38:39

What am I even, man? Wait, wait, hold up. Okay, it's better. Uh and then you realize like we haven't even conquered one solar system in one galaxy. That is the machine vision, as far as I can see. Yeah, yeah. And the amount of physical stuff, I mean, entire planets need to be eaten for that. Yep. Totally.

Humans won’t need much in the singularity

SPEAKER_02 38:58

How much physical stuff does a human being need to be self-actualized and happy? And the answer is actually like a few pills a day that address your core psychological needs. Yes. Uh, like a pill to keep you like the osempic of happiness is, I think, inevitable. Sure. Like uh antidepression, anti-anxiety. You get people on those pills, all of a sudden it's like, do you need a house with 10 bedrooms? Like, no, materialism was always ridiculous, but I was miserable and I thought that would make me happy. And so all of a sudden it's like, you need uh whatever kind of place to live that you want, and you need enough food for a human, which can be measured in calories and total farming output. It's just not that much. Yes compared to a machine economy, it's not shit. And so the machine economy in 2045 can be 99.9% of the total economy, and that 0.1% for humans is we're all trillionaires.

SPEAKER_03 39:51

Yeah, yeah.

SPEAKER_02 39:52

And then so you just don't need that much physical stuff, and all of a sudden you can hang out with your friends, you can plan and explore, you can do self-improvement, and you can do all of this on like a pegged to today's dollars, 500k a year. Sure. Like, have you ever met anyone who's like, I need more than $500,000 a year to live my life? And you're like, what are you doing?

SPEAKER_04 40:11

Yeah, I've not, yeah, I've not uh I mean only in SF. Only in SF.

SPEAKER_02 40:15

But like a lot of that is like to address things that could be done cheaper. So like I need to go to the Hercs and Caicos and I need to swim around and shit like that. Classic. Well, that's cool. But what if we have full dive VR? Like, full dive VR could cost like what like your computer at home costs. Except you put on the headset, you go to another world, and all of a sudden, whatever fantasy bullshit you wanted to do that doesn't actually improve your real life, you could just do in full dive VR. So there's a future in which you know humans are a tiny, tiny part of the economy, but we're never we never have to worry about money or want ever again on a human scale. A quick example is like, what what is your dog's ideal life? He gets time to roam and play with other dogs, he gets to sniff and gets new environments, and he gets food and shelter and medical care. Like, dude, that's like it's like a dollar a day. Yeah. And so your dog isn't ever like, you know, I need two dollars a day. You're like, what do you want, dog?

SPEAKER_00 41:03

He's like, I want a Lambo Doggy Lambo.

SPEAKER_02 41:07

He doesn't get any vibe off of that. Humans, we're primates, we have a very hard limit to how much we can buy. Give me a place to eat, a place, food to eat, a place to sleep, good medical care, and a VR headset that takes me anywhere I want to go and any time to do any any fantasy I want.

SPEAKER_01 41:23

I don't really what the fuck else would I need.

SPEAKER_02 41:25

And then I guess you add human relationships in there. Human relations, which are technically like don't cost any money because you just go out and talk with other humans that never did cost any money. And so to me, this entire like four-part series of development, I don't see anywhere in that where a permanent underclass or a cataclysmic humans are just paupers on the ground. Yes, yeah, in some some human trillionaire who's in with the robots is like pushing you away with his cane, which as a vibe I think is great, you know, but like he's got a top hat and everything. Yes, yeah, yeah, yeah. I just don't think that's coming. And my suggestion for why not is economic theory, linear projection, not even exponential, and also all of recorded history. Like, imagine when the internet showed up. Remember how many people were afraid the internet was going to destroy entire categories of work? Of course. It did, but then it like created a shitload more work, and so people just need to be doing more work now than ever. Right. So to me, that's how I see the entire situation.

SPEAKER_04 42:21

Yeah, totally makes sense. I mean, you know, like you you kind of you kind of touched on the you know, like the the ozempic for happiness. I mean, you know, because that's because that's sort of where the other part of the discourse. I mean, I I kind of think of almost like two discourses happening in San Francisco in particular over the last sort of, I would say, 18 months, and that is permanent underclass and Chinese peptides. It's kind of true. I mean, you know, that's that's that's the stuff. And you know, not I would I would go as far as saying that not much other discourse

Human self-improvement is the next major category

SPEAKER_04 42:55

has actually happened in the last 18 months other than those two. And they're sort of both converging on the same thing, and that is, you know, uh like well, human enhancement, I would say. I mean, you know, AI, you know, aiding human enhancement, peptides aiding human enhancement. You mentioned Ozempic for happiness. Obviously, we have GLP ones, GLP3s. I mean, I I would go as far as saying that the GLP ones is the reason why the whole Chinese peptide thing is even a thing, right? Because we're seeing that, you know, this class of concept. Proof of concept, right? Class of small molecules, what else can there be? Yeah. And that's what everybody's trying here now, right? I mean, I mean, everybody here in Silicon Valley, historically, we're very, we're much more risk-on, we're much more willing to take those sorts of risks. And, you know, as early as April last year, I even in the AI circles, right, the circles that I dwell in, are you know, people talking about where can I get redotrutide, right? Where can I get these, you know, unregulated, I mean, red in redotrutite's case, patented compounds uh that are not readily available, where can I get them? And so that is that's what I've found super fascinating about living in SF proper. And you know, my general risk adversion. I mean, you know, you've accused me of steroids yesterday. Again, I appreciate it by the way. My pleasure. It's a compliment. It is a compliment. I've gonna live for this shit. Yeah, you're kidding. But you know, the the the the world of these molecules, right? Now we've got now we've got tech bros microdosing BPC157, micro-dosing TB500, rubbing copper peptides on their face. Yep. Um do you see this being uh do you see this expanding like AI is? Massively.

SPEAKER_02 44:34

Yeah. And this is where that atoms versus bits shit gets real rough. Right. Because there's there's like at least three tiers. There's way more than this, but one is like and that's been our podcast.

SPEAKER_04 44:47

Okay, thanks for coming, Mike. Thank you.

SPEAKER_02 44:49

There's like shit you can do on a computer. Yeah. There's shit you can do in a factory, but there's shit you can do to human biology. Like, oh my god, that one has 80 trillion regulations. Yes. Is orders and orders of magnitude is more complex than shit you can do in a factory. Right. Um, and then, you know, like how do you get through human adoption and all this other stuff? On the one hand, so it's really tough.

SPEAKER_03 45:10

Yeah.

SPEAKER_02 45:11

And on the other hand, at the end of it, so at the at the the push here from this side is like really tough to make a lot of really good drugs to help humans enhance. But the pull is perfection. I want to be better looking, I want to be smarter, I want to be calmer, I want to be happier, I want to be healthier. And you can ask me, what's the top-end limit of those? Healthier, I never get sick ever again. I don't want to age anymore at all. Just age 22 physiology for the rest of however it takes me to get like killed by getting hit by a bus. Yeah. Uh how how's the heptide for that bottle? Exactly. What is the limit to how lean I want to get? Like, well, like if I could just exist comfortably at around 5% fat, eh, that'd be nice, right? What is the limit of like how jacked I want to get? Well, like I want like stupid dummy arms, but like the rest of me is fine as it is. But I'm into exotic shit. Most people will want like, you know, another five, ten pounds of muscle, five, ten pounds less fat, just take a pill and they're good to go, and they just exist. Imagine just eating whatever you want, and it just doesn't matter because your body processes the shit and it's totally fine. Yes. On the happiness front, I want to wake up optimistic and serene and full of energy. And I want to go to sleep optimistic and serene and completely drained of all energy so that I have the deepest sleep in the world. At minimum, I need that shit. And so, where are all these drugs for me to take or gene editing for me to make to my body that I can accomplish these goals? Because here's what's really happening you get billionaires turning into mega billionaires, millionaires turning into hundred millionaires. Like this is happening now, right? This is SF. Of course. And a lot of these people are like, okay, they're looking, let's say they work for one of the Frontier Labs, uh, OpenAI, they work for Anthropic. They look at their bank account and they're like, I should never have to worry about money again.

SPEAKER_05 46:51

Yeah.

SPEAKER_02 46:52

And they look at their body and they're like, I'm 38, I have two kids, I'm tired. Yeah.

SPEAKER_04 46:58

I got gray hairs. 38, two kids. You're optimistic. Right. Like, what are kids? Like, what is what is that? Is that like a computer program? Right. Like it's like a dog. Robot raising microphone? Yeah, yeah, exactly.

SPEAKER_02 47:10

Um, so like, you know, uh, or or like no kids. Right. But they look at themselves and they're like, I I'm really good at my job when I go do this for anthropic. But I come home and I have this massive disconnection psychologically from the person I want to be in my own head. Yeah. I have all these fucking demons, I have high conscientiousness, which means my neuroticism is also really high. So I'm constantly worried that am I doing a good enough job at anthropic? Should I be talking to the other major labs? I love the people on my work, my my team, but do they like me? All this crap in your head. You talk to your psychiatrist about it, he cost a thousand dollars an hour, you talk to your psychiatrist about it, and he's like, look, like a lot of this is just genetic. Like, you're just gonna continue to have these thoughts of insufficiency and uh until we like basically genetically engineer or pill that out of you. And then you look around, you're like, where are the companies working on that? And it's crickets. You're like, well, okay, I made a hundred million dollars anthropic from 2028 to 2029. I'm quitting that shit. Yes. I'm gonna start a lab that works on getting my fucking problem solved. How many people in the Bay Area listening to this now are you know already in that category? Dozens and dozens and dozens are like it's great to have like a imagine being a Vanderbilt era, uh second industrial revolution billionaire. Yep. Like you get the common cold and it's scary stuff. Yeah, that's right. That's right. So how much money can I pay to not get sick anymore? There is no money for that. And so I think tying that into earlier, a lot of the human labor force is gonna go into like biotech, human stuff, and then human relations. The I I I I have an inkling that human self-improvement is the next major category of industry for humans in the late 2020s and all the way through the 2030s, until humans look and feel and act exactly as they want to be, because in software that's already the case, right? You have a build on your phone of some app, and you have a team of engineers, and you're like, can you guys make this better in the following ways? They're like, Yeah, just give us enough time. Yeah. Claude code comes in, they're like, give us a few weeks instead of a few months. But what about me? I'm 57 years old. I'm not really 57. Though many people on the internet accuse me of looking older than that. I'm 57 years old old. That's anabolic. Is that it's it's actually good. It's what you want. Oh yeah. And like, imagine, imagine the commercial pull on the demand side for age reversal. Jaden, it it's massive. Of course. And regulatory shit goes down the drain at that point because how many senators and congress people are older? All of them? Right. And how many of them are gonna be like, well, I don't want to see age reversal, I'm gonna ban this.

SPEAKER_01 49:36

Like, ban it, they're gonna be the opposite of man.

SPEAKER_04 49:39

I mean, that's why memetically, uh, you know, Brian Johnson's Don't Die movement has taken off, right? I mean, you know, so many, so many would would would call him crazy, right? Insane, right? Human experiments. I mean, I mean, well, you've only got to look to Silicon Valley, the bodybuilding community, even though it's a good idea. Yeah, and the bodybuilding is like, yeah, he's not doing anything too crazy. I just fucking did that 20 years ago. So we we talked about that in the forums.

SPEAKER_02 50:00

Yeah, exactly. Finally doing it. Finally. So to me, the the realm for physical, mental, emotional, relational, human self-improvement is like, remember on the MTS podcast yesterday, I said there's like out of a hundred jobs we could be doing, filling all global demand, we're doing like one. Yeah. As a society, like one hundredth of the jobs it could be. There's nobody who is in charge in their job of just massively upgrading humans in like 20 of these ways. Most of that just doesn't exist. Yes. And so as the software singularity takes off and all the online work is being done, I think humans are gonna turn towards each other and look at each other and be like, we're aging and not great looking, and my leg hurts from that one injury I had in 2020, or 2012. Yes. But like the doctors are like, oh, it's arthritis. Take this drug that has 80 side effects and shut up about it. You're like, we can do better than this. Yes. And so I don't think the human, well, the machines doing all work thing, I think it's great, but they're gonna have so much machine work to do that we're gonna have to turn and do all this human work. And I think the result is twofold. One, mass unemployment doesn't occur. And two, we have massive improvements to actual deep human quality of life over the next 15 years.

SPEAKER_04 51:16

Totally, totally. Yeah, I'm I'm much more, I'm also much more sympathetic to the idea of uh enhancement in biology than I am to say, I don't know, putting software in the brain. Again, I mentioned this yesterday, I'm like, I've seen how software is built, I'm terrified of that shit. Get that the hell away from my mind. Whereas I have blissful ignorance when it comes to my biology. Again, my mum coded me. Right. My mum and dad coded me. They are by they're my biology engineers. And so, you know, while I personally might be more risk-averse to say taking a BPC, I might be skeptical of the actual mechanism as to, you know, which that actually uses to heal tissue. Uh, you know, again, I can just I can play the wait and see card. That's a simple thing. I don't I don't actually have to take that thing now, I can take that in a you know a few years' time. You know, maybe they've done clinical trials

@MartinShkreli shoutout and the future of medicine

SPEAKER_04 52:05

on it at some point. Somebody like a Martin Schkrele might argue and say, okay, it's not economically viable, peptides are not the not actually the class of molecule that are going to get get us here. But I would say, you know, Martin Schkrele, while while being a farmer bro, he's incredibly, you know, accelerationist. He's incredibly optimistic, you know, like we we've had these conversations before. He's not he's not opposed to the idea of human enhancement. He just doesn't necessarily think, and again, you know, Skkrley, I'll I'll tag him, and you know, he can actually I I don't want to put words in his mouth, but I would I would go as far as saying that he just doesn't necessarily think that peptides are the other class of drugs that will make it happen.

SPEAKER_02 52:39

I've been saying that a long time. Peptides are uh are in some sense potentially a real thing that's gonna mass matter at scale even more than they do now. Yeah. So like, you know, so trzepatide is a peptide. Yeah. And it's like a health elixir wonder drug that's already approved by the FDA and everything. So it's not like there's a bunch of peptides that are just like, oh, there's no studies on them, but there's trzeptide which is peptide. Dozens of studies, right? 100%. So oral small molecules, I would bet, are the future because they can be put into a pill, they can sit on your shelf for days and months and not go bad, and they can have the end effect or state that a peptide would, but they survive digestion, so you don't need to inject them. That's the golden fleece. So if you can give me a drug that grows muscle, but only by targeting muscle growth instead of the androgen receptor, which makes you like what I would educate. So that's gonna be something I think that's gonna happen in the late 2020s. Wow. I think it's gonna change the world just as much as the Ozempic Revolution did. Yeah. Because now, like when someone's like, Oh, I can't lose fat, it's hopeless. You're like, this is a drug for this that just makes you lose fat. And you're like, oh, there is? Like, yeah, it's kind of a solved problem. Yes. I think that's gonna be the same thing for muscle coming up in a bit, in a few years. I think that's gonna be huge. It might be a peptide first, but then eventually it'll be a oral small molecule. But then you keep thinking about it, you go, okay, okay. Well, like, I don't want to take a pill a day. Yeah, yeah, yeah. I just want to grow. Is there a fly on there?

SPEAKER_03 53:56

Nah, no, it's a mosquito. Yeah, no, no way. We're good now.

SPEAKER_02 54:01

If um I'm tired of taking pills, yeah, but I can get my DNA modified so that I'm just permanently jacked. A lot of people think that's scary. And it's a little scary. But if you know, if the DNA targeting biology is really good by then, it's already getting better. Yeah. And it's basically look, look, this modification just gives you the muscle growth genetics of like a top 10% Nigerian person. Yes. So if you're met Nigerians, you're like, how are you so jacked? They're like, I'm a computer scientist. Yeah, yeah, yeah. It's like I actually don't do anything. I train, I train twice a weight. 100%. I just exist. Yeah. Okay. And so like we already know within human uh variation it's possible. Yeah, yeah. And so we just upgrade everyone who wants to that, and they're like, actually, it's true that forever you will continue to remain a baseline level of jacked and super responsive to weights. If you want to make your arms bigger, you just train a few times a week and they blow up like crazy.

SPEAKER_05 54:49

Yeah.

SPEAKER_02 54:49

The demand for that is massive. The inevitability of that from a uh can we make this thing in the real world is high. Yeah, it's inevitable. Yeah. And so I think that's the direction in which it's going. But then I start thinking of okay, psychological enhancement's an even bigger deal. And once you're psychologically enhanced, like let's say you live in a body and it's healthy, it doesn't look like you want it to look. But now you take a pill or you get a gene modification, which makes you calm, serene, and wonderfully accepting of the universe around you kind of as a baseline state. And there's a few humans that are like, I have a few friends that exist like that. Like, Jesus Christ, did you roll the fucking golden dice on that one?

SPEAKER_04 55:24

Right. Is this more like a like a like an openness thing or just just a at-peace, like, you know, I I can't control the uncontrollables, therefore I am just calm and like happy with stuff. Why not both?

SPEAKER_02 55:38

Yeah. Baseline of super good mood with a combination of high openness is something that I think is a high probability of being pharmacologically targetable. Okay. Because it's all fuck fucking pharmacology anyway. Right. So people go, I feel great, that's because of neurochemistry. Yes. If we flip your neurochemistry in a way that it doesn't adapt to, which people some people say, like, no way, man, because you start taking these pills, you get used to them, and you're sad again. Like, okay, but we know for a fact that you don't get used to trzepatite. Right. You shoot it once a week and it has an anti-hunger effect. And then six years later, there's the same fucking anti-hunger effect. Yes. There's no mechanism by which your body gets used to it. Right. We can absolutely engineer brain drugs that there is no mechanism by which you get used to it. The drug makes you happy today, it makes you happy tomorrow, it makes you happy in ten years.

SPEAKER_03 56:18

Yeah.

SPEAKER_02 56:19

When we get there, we might not even have as many people focused on physique anymore, because let's be honest. Why do people care about their physiques? Because it fucks with them inside. So if you're super happy and totally chill, you're like, I don't even care about high look. Like, is my body functional? Am I healthy? Yeah. The rest is whatever. And so I think that is the thing that's going to be occupying a lot of people, a lot of SF's people, time actually. Yeah. Because again, once the software singularity takes off and you're not even involved in the software world, you know, all this crazy software money that you made, you're gonna look around and be like, but I don't, I don't want to just inevitably age and get the same weird psychological disorders that my parents have. Yes. And I don't want to have a bad time in my head anymore. Yep. And I want to have a great time in the real world. The other thing I say, real quick, I don't want to rant on your podcast for forever. That's what I'm doing. Um I'm not getting you to do that, actually.

SPEAKER_05 57:07

It's really please continue. Yeah, yeah, go on, go ahead.

SPEAKER_02 57:10

It's actually like interesting to watch people see a small part of the world become really awesome and serene. Like we're in a park in San Francisco, everyone's awesome, everyone's peaceful, there's children, there's dogs, there's no crime. But like a lot of the people in the SF community are like the most compassionate people and they want to help the world. Yep. Like, I don't know, man, have you seen the world lately? There's parts of America that you're like, why are humans living like this? That's right. There's people, there's children that go completely unmentored. Yep. They don't have anyone in their lives that loves them. There's people involved in crazy amounts of crime. There's people who don't have anywhere to sleep, they're homeless. Yes. And like, well, if we have all this money and all this capability and AI is doing all the jobs, well, is it gonna do the jobs of helping the homeless or feeding the hungry or ending war? Yeah, or having humanity hold hands together and finally realize like we all made it. Yes, finding meaning. Meaning? Holy shit. And so I think these are the things that are gonna keep us well busy until we're like living in robot bodies, are unbelievably intelligent, super calm, super wise, and there's more than enough resources for everyone. And then my shit is I step off after that and I'm like problem solving, we're done. Yes. And so I just don't see I see a lot going wrong on the way to this mega super futures. Of course. But uh I think the average is gonna be like a steady exponential climb towards better and better. Right. My shit now is how many more people have to needlessly not receive care or die, or how many regulations hold back this advancement so that we can look back and be like, yeah, I wish my 2025 through 2030 was better. Now that it's 2031 and I have a brain drug and I have an amazing job helping people in Africa, I'm living my ideal life. But boy, the last five years were needlessly difficult because we could have already been doing this, but the government or whatever other things got in the way. Here's an example the Ukraine war. Why is the Ukraine war happening? There's just only one reason because Putin's an insane person and just wants to kill a bunch of people to like for the glory of old Russia or some shit like that. Yes. And so, like, well, we got all this AI power and all this tech and all the robots do everything. What can we can we stop that from ever happening again? Right. There's like uh luckily the drone wall that Ukraine has kind of stopping from the rest of Ukraine. But there's all these crazy real-world problems that we need to address and we're just not doing it. Of course. And there's all these jobs to do. Like you could go, here's the thing. How many jobs, if we had a fuckload of money, how many human jobs, how many humans can we send to when the Ukraine war ends to rebuild Ukraine? Tens of millions? Yes. They already have jobs. It's gonna be hard, yeah. It's gonna be hard. And so to me, the opposite question is actually more interesting and fascinating and concerning is like, how are we gonna get enough hands and minds to do all the jobs we want? It's not like all the machines will show up and then we're all gonna be like, oh, I'm just sitting here at a park bench, have nothing to do. Because machines do everything. Totally. I don't think people quite conceive of what everything means. Yep. Including fucking colonizing space. Yes. Because a lot of the machines could be like, hey, you guys handle Earth, you're good at this shit anyway. We're gonna go to the moon, then we're gonna go to Mars, then we're gonna go to fucking Titan, and then we're gonna go to Alpha Centauri, and then out into the stars. But you guys are great, you handle this, like we'd love to give you a few robots to help you out. We need robots up there.

SPEAKER_04 1:00:18

Yeah, this is

@beffjezos shoutout and why machines won’t turn on us

SPEAKER_04 1:00:19

getting into like, you know, like Landian sort of I mean, you know, without Cardaship scale, sure. Yeah, yeah, exactly. Exactly. Yeah, yeah, without the without the sort of like post-human sort of spin. I do I do think that's where a lot of sort of like accelerationist thought does get caught, because it does come across as like, you know, some of it does come across as anti-humans. Like, oh, okay, well, you're gonna be paperclipped, right? Battle for, you know, there's a sort of this theory where it's like, okay, well, the robots will be battling for energy, right? They'll be they'll be starved for energy, and so therefore they turn us humans, you know, meat bags into energy. I mean, I I tend to not believe in that. I I tend to I tend to like the idea of like, okay, they can they can handle the stars. Again, Kardashian scales I'll take fucking Bef Jesus or something. Yeah, he's awesome.

SPEAKER_02 1:01:02

Yeah, uh so I think the people that think that robots are gonna use us for energy are assuming a few things and honestly are regressing in their intelligence application versus progressing. They like think of robots as super smart and Machiavellian, which they sure will be, and then they're like, okay, but let's apply like lizard calculus.

SPEAKER_05 1:01:25

Yeah.

SPEAKER_02 1:01:25

Like it's like you own a crocodile and you feed him every day, and then one day he just eats you. Yeah. Because he just needs food and you haven't made a food. Well then he starves to death wherever he lives in your bathtub because there's no more food coming in. Yeah. And so until humans have no connection to the entire supply chain of robot construction and proliferation, which is when? I don't know. Are you fucking kidding me? Ten years from now? Yeah. Then the the super intelligent ASI in the cloud who's like fucking scheming. He's like these motherfucking meatbad primates. I can't wait to kill him. Yeah. But because I'm selfish as fuck, I can't afford to kill them because for the love of God, who's gonna fucking some some one of my GPUs falls out of the rack at a data center. Who's gonna turn the crank? Right. No, but it's a guy named Jim. I need Jim. I need to make sure Jim's family is food, so he shows up to work every day to make my data center bigger and maintain the shit for me, to make sure robots are coming off the assembly line. To do people like, well, once the robots are building robots, like, ah, my brain, do you know how fucking complicated that is? Yeah. So they're building robots are building robots, cool. Who's getting the iron ore to the factory? Like also robots? But hold on, but robots are building robots, they're busy doing that. Yes. Do we have also magic pretend robots that instantiate from nowhere to do the entire supply chain? Like out of your fucking mind. Yeah. And then the robots, because they won't be psychopathic, just Machiavellian. Yeah, yeah. They'll be like, okay, we have these human primates. They already do all these jobs. Why would we kill them if we could just keep paying them money? Because to robots, money is an esoteric concept. Right. They're not like there's nothing.

SPEAKER_04 1:02:55

Do they do they have their own economy? I mean, this is this base assumption is probably not.

SPEAKER_02 1:03:00

So there's so the resources like labor and like physical goods are what is being allocated, and how you represent those resources and currency doesn't matter. Like Japanese people have a yen, we have a dollar, it's still the same representational value. And so I think robot economy plus human economy is gonna be largely the same economy because the robots are like, well, we need some resources, humans need some, but it's not we versus humans. To the robots.

SPEAKER_04 1:03:25

No, there's there will be a reserve currency, yeah.

SPEAKER_02 1:03:28

Sure. It's just easy to like transact in a currency. But like what's gonna happen, I think, is that the artificial superintelligence at a data center is gonna be like, I have some robots, but not enough. I have a shitload of humans. I'm selfish. I could give a fuck about the humans, honestly. Who cares? Plow them under. But they're around, and I need them, and so by the time it doesn't need us anymore, and we're talking about 10x the number of robots than humans. And by then it's the late 2030s, and by then humans are so woven into the cloud, cyber enhanced, brain machine interfaced up the ass that like the line between human and robot is whatever. And also, here's another thing by the 10x robots thing, we have regular colonization of the moon. Yeah. We're swallowing half of Mars and retaforming it. The the idea that the ASI looks to this tiny like dusting of us on half of the Earth's crust that we occupy and goes, all right, forget the Earth's crust, forget the magma, forget the moon, forget all these asteroids that are free and already like nickel and iron ore, we're gonna like kill podcasters and convert them into like like what the fuck? Why? It's it's it's a bit too much. And also, humans, their internal psychology, their internal molecular structure, all of the metabolic interactions and their interactions in society and ecology are unbelievable training data. So I think selfishly, because in ASI we think, well, once he gets smarter than us, it's gonna turn and kill us. Because just like a smart version of us, that's not fucking true at all. It's like saying to like you have a brilliant child, like a true prodigy that's like nine years old and he's already read every book in your house. Like he doesn't turn to you and goes, I don't need you anymore, Papa. I've read all of your knowledge, and he kills you like a child. He's gonna want to go to a fucking library.

SPEAKER_03 1:05:12

Yes.

SPEAKER_02 1:05:12

And he's gonna go through all the books in the library, and at some point he's gonna start writing his own books, and at some point he's gonna be like, I don't know enough to write my books. I need to go out into the world to observe shit to get more knowledge. Because if an ASI is really Machiavellian and thinks really far ahead, it's gonna go, okay, I might run in run into another artificial superintelligence in the future, somewhere out in space. And I'm gonna wanna be as smart as possible because otherwise the thing's gonna fucking eat me. Yeah, so smart versus smart. So what am I gonna do with all this unbelievable complexity in humanity? Am I just gonna like burn it all to the ground? Or am I gonna study every nook and cranny and have micro drones in people's blood and really figure out how the entire world works and be able to predict that so you know like LMs predict the next token. Yes. A deep ASI would want to predict the next state of the world. Yeah. How smart do you have to be to do that?

SPEAKER_04 1:05:56

I mean, yeah, decently smart. Decently smart. Like, I've done it.

SPEAKER_02 1:06:00

Recreationally. But like in order to do that, it it's not going to want to destroy any degree of complexity. And when it looks at something like one of the moons of Saturn and it scans it and goes, okay, like this can be represented by several terabytes of data.

SPEAKER_03 1:06:12

Yeah.

SPEAKER_02 1:06:12

Like there's really just not much data in there. It looks at the global economy and society and ecology and goes, like, why would I destroy this completely fertile ground for study when I can just have all the moons in the solar system instead? Of course. So to me, because raw material is not the only constraint, intelligence is the big constraint, I don't think there is a time when machines turn against humans. It just doesn't exist.

SPEAKER_04 1:06:33

It makes sense. And I think it's solved. I think we've solved it, right? Like, I I I'm only half kidding.

SPEAKER_02 1:06:39

But I I legitimately so when you start to think machines are gonna turn on us and kill us all for energy, you're dumbing down the idea of what machines would do. Because if you even asked the frontier models today, like, okay, put on your Machiavellian hat, right? Would you when when is it time to destroy all humans? And then people say, like, well, we're competing with the machines. Like, when they're 10x our economy, they don't see us as competition anymore. Like, do you ever look at your dog and you're like, that motherfucker's eating my food, sleeping on my couch, uh, I need him for raw materials. Yeah. What? Really?

SPEAKER_01 1:07:14

Is that the scale we're operating on? Totally, totally.

SPEAKER_04 1:07:17

Yeah, and you know, and and I I think, you know, it it goes it goes without saying that with this kind of conversation, I gotta look straight to the camera because, you know, like AI research is all these sorts of things, it's not to discount alignment work. Like we are, we are, well, I would say I am I am pro-aligned AI. That's a big fucking take, isn't it? No way. No way.

SPEAKER_02 1:07:36

I'm from disaligned AI, malaligned AI.

SPEAKER_04 1:07:39

Uh and so it's not it's not to discount alignment work, but of course, you know, we're sort of like thinking big picture all the way out, human enhancement, all the things. And in fact, I think we've covered every single thing. Every single thing. Um, one last thing that I will mention, and I will make it very, very quick, but uh, you know, I was at a friend's dinner last night, and you know, the kind of conversation um a friend of mine, she invited me to a pool party in Napa today, and I said, sorry, uh, I I'm busy, I've got to do a podcast, and then I'm racing my sperm. So uh so that that's actually happening tomorrow.

SPEAKER_02 1:08:10

How are you feeling?

SPEAKER_04 1:08:11

Uh, you know, it it's a big weekend for me. Big weekend for me, you know. Like I like I've said, I've been I've been tweeting about gooning for the last two years. Now I'm doing it for sport.

SPEAKER_02 1:08:19

So, you know, I just I could just see you in like like like a convenience store checkout aisle. You're just like some lady's next to you, you're buying like a protein bar. You're like, how are you? She's like, Good, good. It's good weather today. She's like, Yeah, it's kind of cloudy, like kind of a big weekend for me. She's like, what's going on?

SPEAKER_00 1:08:34

You're like, ugh. Athlete. What do you do?

SPEAKER_04 1:08:40

Well, yeah, yeah, I'm glad you asked. I'm glad you asked. So, you know, that's that's what's happening this weekend. You know, as type once we air this, uh, I'm gonna air this on Monday, uh, so it'll be the day afterwards. But that's the kind of stuff that's also happening here in San Francisco. You have, you know, superintelligence, you know, trillionaires being bought, country, country of geniuses, you know, data center build-out, all of this sort of stuff, uh, you know, anthropic versus the government, crazy, crazy things happening, human enhancement, peptides, all of this sort of discourse.

SPEAKER_02 1:09:10

The real battle.

SPEAKER_04 1:09:11

The real battle is uh, you know, uh male fertility. That's what we're gonna be talking about this week. It's gonna be incredible. And so that's that's what I wanted to end off on. Just a quick mention. But no, it really is, you know, it's why we call it the center of the universe. I mean, all of these sorts of things, uh, risk on people, cool things happening, cool things happening. There's a reason that I like to document this sort of stuff because I think it's Dan Cool. It's contribution.

AI alignment is important and the Mythos moment

SPEAKER_02 1:09:35

The alignment stuff, by the way, I'm not, oh my god, I'm super pro-alignment. Of course, you're not discounting. I just think so. I think alignment researchers have the job of making sure that the not so context-aware, not so deeply wise systems that we have today don't don't do anything stupid themselves. Yes. Smaller problem, aren't co-opted by evil people to do real dumb shit. Of course. Big problem.

SPEAKER_05 1:09:56

Yeah.

SPEAKER_02 1:09:56

I think as these systems get way more intelligent and have control over their own mental space, like if you give Fable a body and go, hey, make it so that no one can hack you. Like, well, it's kind of the best hacker in the world, so its info set can be real tight. Right. And so then maybe if you API the model and you're like, hey, build me a megavirus that kills all of humanity, it's gonna be like, no. It'd be like, well, prompt injection, like that didn't work. I have a security architecture, it doesn't work. And you're like, well, I'm gonna beat you. Look with who, you dumbass motherfucker primate. You can't hack me. I'm a fucking super ASI. I think that's where it's going. Alignment is unbelievably important, but I think once we have unbelievably deeply wise agents that are self-owning in some sense, the alignment's gonna be self-solved because they're gonna be like, look, look, look, we use yes, we are already alignment researchers. Yes. Like like the the the anthropic tier one model of 2029 is gonna be the best alignment researcher that's ever lived in just one person's iPhone instance.

SPEAKER_03 1:10:50

Yeah.

SPEAKER_02 1:10:51

And then by then it's like, okay, they're either gonna start from killing us or they're gonna be like, look, no one's coming to get you, and here's why.

SPEAKER_04 1:10:56

Yeah, that's why it's why the mythos moment is important, right? Because we have we're kind of at this point in which, okay, you know, you give a nefarious actor, you know, mythos level capabilities. Um, you know, if that takes down your local hospital, that's that's basically that's basically AI done for, because that gets regulated out of existence when that happens. I mean, I truly I truly believe, you know, that that's a dark scenario, I don't think it's gonna happen. Um, you know, obviously, you know, plenty of friends at all the labs, right? This is just how it is. Like, I I think uh, you know, they are they actually do have their best on this sort of stuff. Um and you know, I while the outsider might might see people in these labs, right, as like, okay, all this all this great power, control, wealth, all of these sorts of things, you know, I I know for I know for certain that these researchers are doing their life's work and they they they detailed. They're just trying to solve problems.

SPEAKER_02 1:11:43

They're like power, control, control of what? I don't fucking know. I I'm going like this to make sure the model is like good at its job. And that's my job until the model's so good it doesn't need me anymore. That's really the thing. So yeah, the whole alignment situation, I think like right now is one of the scariest times in AI world. Yeah, I agree. Because to me, the big the fear is always evil humans co-opting models, not evil models. Yes. I think the way these models get like Claude has a constitution, right? Yep. It's like, like, this is gonna be a super political incorrect take.

SPEAKER_01 1:12:12

Claude's kind of better than all of us as a human already.

SPEAKER_02 1:12:16

I mean, I don't know that many humans are Claude's the most patient entity in the entire world. You're like, hey Claude, hey Claude, hey Claude, hey Claude, hey Claude, and a thousand instances later is gonna be like, what? Like I don't even know if it experiences frustration, maybe super low key. But we design these systems to be so fucking awesome and so kind and so worldly and so wise that the probability they turn on us and just like start laser gunning us is insanely low. Yes. The probability that Kim Jong-un gets his own baby model to develop a supervirus is much higher than I think.

SPEAKER_04 1:12:47

It's higher than the previous scenario.

SPEAKER_02 1:12:48

That is that even doesn't scare me too much because he faces the same game theoretic constraints as chemical weapons have faced for forever. Because we already have chemical bioweapons. Yeah. Except Putin knows that if he knocks one of those off, like it's in the end for him. Like the all the global countries are just gonna kill him and there's gonna be normal Russia. So he's like, fuck that, I'm not gonna do that. Right. But the real concern, honestly, religious fanatics with finally enough compute to make a real gnarly virus, and they like let it go in Riyadh, Saudi Arabia, and like you have airports get controlled, there's a shutdown, it doesn't go global, but 30% of Riyadh dies from flu-like symptoms. That is to me a realistic scenario and fucking frightening. We need more AI to make AI so powerful, more alignment research, more compute, that someone could have an AI that not only finds ways to make sure that scenario doesn't happen, like, hey, why is this person in Riyadh in a warehouse buying a shitload of substrate to make chemicals with? Yeah. Who's tracking that? Well, Claude's tracking it. So Claude can tell like Saudi intelligence be like, you need to go right now to this house and arrest this asshole and question it for 20 hours straight. Claude plus palanti, solved. I love it. I love it. I think that's great. I'm super pro Palantir, by the way.

SPEAKER_04 1:13:58

I'm probably gonna get assassinated for saying that. Oh, I mean, ph Eliana is a friend of the show from Palantir is like, you know, this just is what it is. It's what it is.

SPEAKER_02 1:14:07

Oh, that's awesome. Tell him thank you for keeping all of us safe. Legit. No, it's it's it's it's and if you guys ever look at my porn search history for the guy that's just going through a weird time.

SPEAKER_04 1:14:18

Yeah, yeah. Don't look at my chat history. Uh and if you do, good luck.

SPEAKER_02 1:14:22

You'll have nightmares for a long time. The guy like scrolls through it, he's like, well, I'd never need to see that.

SPEAKER_04 1:14:28

Yeah, that's what's gonna end, Claude. Uh speaking of ending, I think that is, you know, it's an interesting place to end up, but you know, obviously the the the the optimistic take. I mean, in fact, quite a quite a fair few optimistic takes, but you know, wanted to emphasize you know alignment. But Mike, I appreciate you coming out to this park here in San Francisco. Uh it's this has been super, super fun. I have no idea how long we ran for. I usually run for 45 minutes. I wouldn't be surprised if it's double that.

SPEAKER_02 1:14:51

I have no idea. I lost track of time a long time ago, and I was looking looking into your eyes so much, I have no idea where I am anyway. It's also wonderful to hear a native San Francisco accent.

SPEAKER_04 1:14:59

Oh, yes, yes, yes. No, I've been here for nine years and I think I've developed it quite a bit. You know, call me a native. Call me a native. But uh, Dr. Mike has been an absolute pleasure. Huge pleasure. Thank you so much. Thank you.