Beff Jezos: Thermodynamic Computing, e/acc & the Future of AI

0:04 Okay. Well, hello. I am here with Guom in San Francisco. I'll start by saying I introduced him as Guom, but you might know him as Gil or Beth. So, would love to jump into your origins. >> I guess this is maybe a little known fact, but I' I've been secretly a French Canadian this whole time. surprise. but yeah, my real name is Guom. and I was born in Quebec.

0:34 grew up in Montreal. you know, beautiful city, a great area. And, yeah, I mean, growing up, I I was kind of obsessed with the cosmos, the laws of physics. you I was reading up on Stephen Hawking and Richard Fineman. I really wanted to tread the same tracks as they did and and become a theoretical physicist to, you know, better understand the world, better understand the rules of the game so that I could create the most potent technologies that allows civilization to scale to the stars. you know I guess like you know grown up you know highly highly neurode divergent kid you know that naturally leads to you know yeah I mean social ostrac ostracision and then that leads to you know maybe you you start abstracting out to evade you know your day-to-day life you start thinking about the cosmos and maybe maybe like your you know the schoolyard bully doesn't matter as much in the in the cosmic scale, right? But that in a way it's a gift because it it got me to be obsessed with the universe, obsessed with physics and to use like mathematics and like my own mind as a form of escape from from reality and you know fellow humans which can be nice but can also not be nice and >> maybe we'll touch upon that later and echoes of that later in life but yeah I wanted to be a theoretical physicist and I was talking about subatomic particles like age seven and how this is what I wanted to do very early on and people thought that was that was a bit weird maybe but you know it kind of it kind of worked out. I've had a I've had a weird path you know I really wanted to become you know a professor in theoretical physics and figure out quantum gravity but maybe you know and I ended up getting into theoretical physics for for college and for my masters. So, I I did math and physics at McGill. That was kind of my first time like really leaning into my passion for physics. But I I guess like there's the whole high school phase, you know, where you know, you you just you just do what you need to do to survive, you know, >> in in a social environment. And you know, I went to great schools. I went to like private schools in Canada. you know, not the best money you can buy, but up there. So, great education, but you know, I went to a school called Kales Notadam, who's well, basically it's the it's very Catholic.

3:14 Everything's Catholic coded in in Quebec. but you know it was post there was a revol it was it was like a quiet revolution of you know maybe sort of more anti- athoritarian sort of sort of vibes like a a back reaction to sort of the the constraints right imposed top down by you know the institutions. So I was always maybe growing up sort of distressful of institutions I guess and and sort of >> you know a bit rebellious you know in Quebec it's like >> you know I went to bilingual school that's why I don't have that's why I don't talk like this you know but I could turn on my sorry my chemqua friends who can't turn it off I'm very sorry but the key was learning English when I was like six and I went to bilingual school but technically that was >> did your parents speak English or did they speak French they're French can They're French, French Canadian and >> but they, you know, my my mother's like a professor at McGill.

4:13 >> Okay, cool. >> And you know, so they speak English, but they all have an accent. But in my case, you know, they had the insight to send me to a bilingual school where where I'd learn English where I when I have a lot of neuroplasticity >> and that was a huge gift and I'm grateful grateful we went went for that. But but yeah, I guess like growing up it's like I think like you know my whole life it's like okay well if you're smart and like unapologetically curious and like really driven sometimes people see that as a threat >> and they try to like tear you down. And so I like learned that very early. And so I like I went like stealth mode on on being a physicist. And then I guess like you know as anyone like learns as a kid it's like you have you you you you put on these masks to these persona you create this persona that you know is more legible to people and that becomes like your your defense mechanism. And so for me it's like okay what is you know if deep down I'm like a physics nerd but if if you've come out you're gonna say that it's like it's not safe >> like what's what's like >> what's the antithesis of that? What's gonna keep me like safe? It's like, okay, what if what if I became like a jock? What if I like joined the most hardcore football team? What if I got like really strong and like I really, you know, got into the stren the science of strength and and conditioning and so on. And so I I I I went really deep into that. I guess like in my in my teens, I was like pretty bored at school. It was it was frankly like too easy in many ways. and I'd skip a lot of classes or sleep through them and then showed up for the exam. And so I was like, "Okay, maybe at least like let me let me play some football. Let me play some sports."

5:57 So So where I went to high school, they have sports every day. and I'm grateful for that because I I think you know, if you if you if you develop really advanced motor intelligence, your body also like secretes all sorts of good stuff like you know, BDNF. And >> what's BDNF? it's like it increases like your your neuroplasticity, right? So if you >> if you do strength sports or certain sports like you're you have a surge of of this in your body and and and if you >> you know if in a day you could do both like a lot of sports and then you could do math or whatever you're passionate about, you're actually going to be swimming in your cerebral spinal fluid and you're going to actually like learn faster and love that there's going to be more connections made. So that's basically what I did in in in college.

6:43 But yeah, in high school, this is like a very different podcast from my usual podcast. Nobody asked about this. They're they're kind of like, "Give me the last like four or five years, right, at most, but we're we're going back." Well, I also want to say like meeting you because we've also had some mutual so we haven't spent a lot of time together, but I've got to say like in this moment I am just so charmed by your energy because you are such an intellectual on the timeline but also have this like really creative kind of soulful center that is coming out and that I want people to feel because I think it's not >> so common to see that coupling.

7:21 >> Yeah. It's like Yeah. If you're a sensitive young man, you got to >> keep that under the hood, right? And then you got to like project an image that just fits people's priors. And so >> for me, yeah, like you know, it's why it's why I have this physique and I look like a lumberjack even though I'm like an intellectual and a philosopher, right? Which already kind of violates people's priors, right? And if you're stacking too much deprisal, it's more efficient for people's brains to like reject your existence or like >> bring you back or tear you down to fit their beliefs for you, right? Because our brains doesn't like it when there's a divergence between reality and our our beliefs, our priors. and you know, this is a lesson I had to learn when growing a following, but we'll get to that >> that that chapter. but but yeah, you know, I played football and went it was it was very serious program and I I basically like learned like David Gogggins like metacognitive hacking hacking. so like you know, we've we had like something like the Buds, the Navy Seal Buds camp, not fully Buds, but like inspired from it, but for football, 6:00 a.m. to 10:00 p.m. for like two weeks in in in like middle of August in Canada.

8:44 >> Yeah, pause. >> So, yeah, for some reason, our high school was the best at football in the nation, and the training program was very hardcore. It was meant It wasn't just physical. It was very psychological cuz there's about there's like 2,000 people trying to make the team and only 50 that make it. >> Wow. >> And so it's like they they try to break you and then if you emerge from that you are extremely resilient, right? And so developed a lot of resilience from that.

9:12 But it's it it's it's pretty crazy like how powerful your mind is if you learn to like hack it. M >> and I guess like the way I would summarize David Gogggins is like you can it's almost like you have this shadow energy and it's like I call it like source of nuclear fuel if you will. It's like in case of emergency you can activate that. >> He's like super masculine like really >> Yeah. No, but but it's like it's really >> I I know some people have it in in them and it turns out, you know, I was like this shy, reserved like nerdy kid and then like that program just got it out of me cuz like they they will break you.

9:52 You're going to be like in the rain and a thunderstorm doing push-ups until like you mentally like break or like >> sleep deprivation air horns in the morning dehydrated and then you have to spend the night studying. Anyways, it was it was quite intense. But I guess like I learned this metacognitive hack of like you can actually weaponize your ego and you can like purposefully like try to hit it or like self-t talk to like hurt it and then the back reaction from that >> is like a weapon >> which is crazy.

10:23 >> That is wild. >> And so you can actually like create actually and and the neur neuroscience of this is interesting. Actually, you can also like think of like an opponent or an adversary and you generate dopamine and it generates norepinephrine. So, you actually have a boost in strength in the moment. And so, like learning to like control my nervous system, my prep, not not just my brain, but like my body this way was was really interesting. And you know, I don't know. I I I I feel like that sort of resilience I know I have that in me and like I so I I'm not afraid of like extreme hardship. I would say like any like you know people that are like veterans obviously football is not as intense as like actual a combat like that but it you know neurologically it feels very very much the same. but you know a lot of people that have gone through training like that they end up very resilient and so bringing that to the world of business of like how hardcore can you be >> if you have to not sleep for two three days like would you be willing to do that for the mission and it's like having that >> that mode that you could get into >> which you know >> you got to you got to learn to control it because if you only tap into that >> shadow energy you will burn yourself out you know it's like it's like nuclear reactor you need the you need the cool down rods sometimes and sometimes it's like relationships and like other people and like calm down, right? But like but yeah, so you know, I guess like I always had like this this beast, the shadow, right, that I I knew was deep down that like if I had an adversary or or or something that like I could always like tap into it. Maybe there's like some echoes of that in the Beth character and persona, right? like the hyper masculine like >> you know like so I I don't know.

12:14 I'm just trying to like analyze how how do we get here? But but yeah, so so so so that that was kind of like you know got got through high school and then and then really I just started like coming out of my shell and and undergrad. I did math and physics and I I was like oh I finally found my people right? I can like talk about math and physics and like nerd out. and I I did a very hardcore program, did like a bunch of grad courses in undergrad, many all-nighters. So, I was just like pushing myself to the absolute limits from like age 18 until I mean now essentially. and and then I ended up going into theoretical physics for my masters at Perimeter Institute for theoretical physics. But then, you know, I got to the higher echelons of physics and, you know, I was in the same halls as like Shawn Magcguire, Sam Mcandlish, right? former CTO of Anthropic and, you know, very much top tier people like Nobel Prize laurates and so on. I was part of this this collaboration where we were trying to view all the universe as like one big giant computer.

13:24 M >> it's the it from cubit school of thought. So basically every information theory has a unifying lens to view everything in the world. So already that was that got me into information theory and like viewing the universe as a computer. And so this duality between computation and the universe and them being one and the same is basically the theme of like >> the past 10 years for me. And so yeah so so essentially got to a certain level in theoretical physics did my thesis but then I I kind of had like you know ever since I was a kid I wanted to be the one to figure out quantum gravity many such cases >> you said that how as a kid did you know about quantum gravity like was there a moment in which it turned on for you so young >> I think it's it's just like if you keep asking questions why is that why is that at some point you you hit like the base of the stack right of abstractions And and to me, you know, I was always curious like why are we here? How do we get here? How how do things work? You know, is there is there a god? Right? Like what can I understand everything >> in the universe? Like what I cannot you know create I do not understand. So if I can't like rederive it, I I don't understand it. But you know that led me to studying the you know the sort of boundaries of the reality, right? So the quantum which is the very small on the cosmos and so I was in quantum cosmology essentially >> and so I was studying black holes and you know cosmic microwave background quantum field theory like what what can we learn about our reality from what we can observe from the big bang and could we observe black holes and principles? could devise experiments where we would understand why these two theories of of quantum quantum mechanics and and gravity can't can't coales and can't be unified. And at some point I I I went through like somewhat of an ego death and like a great sadness I guess in my in my 20s because I realized that this classic reductionist approach to physics where you reduce everything in the universe to a few equations that are like >> comprehensible by humans and that like the human's going to write down a couple equations you know like in interstellar it's like you're just on a blackboard you write down a couple equations and you save the world right that's like the romantic idea of like physicists have of of themselves. But I realized that I I didn't think that was going to happen in my lifetime. And and that actually actually a lot of nature is is irreducible. And this is Steven Wolram's theory. And what does that mean? It means that you won't have a simplified computational model of it.

16:19 Your model of that piece of irreducible nature will need an amount of computation for your generative model to make predictions that is similar to the generative process that gave rise to that emergent property in nature. M >> and so that was very eye opening because it's like okay you know even for quantum gravity right the future you know at the time it was like 2015 deep learning was just starting opening I had just started it was sort of emergent computing was becoming more and more dominant in the sciences it was all about simulating the world it's like actually you can write down equations sometimes but then then actually making predictions you need you need a big computer because of this irreducibility and complexity of the world. And so I was like, I got to get into computing.

17:08 But what kind of computing is going to help me achieve my mission of of understanding quantum gravity? And that was going to be quantum computing. And there was no version of artificial intelligence for these new computers at the time. No one knew how to program them. >> so I was actually the first commercial quantum computer programmer. so like there was this company in Bay Area called Regetti like 10 years ago and they told me like years later that I was like the first user ever.

17:37 They didn't tell me that. They were really nice. You know, I was just this kid hacking in my bedroom and they gave me all sorts of access and you know figured out some of the first algorithms for how to bring AI to quantum computer. But maybe we can like double click on what that means. It's like if you view nature as having these parameters that control how systems evolve, >> you can have that in all sorts of systems, control knobs of all kinds.

18:03 >> You could view these as parameters and you could view time itself as a as a graph. You can do forms of back prop or you gradient descent on those parameters. And so this idea that you can like okay I'm trying to grock this physical system and I have this tunable physical system over here and I can tweak its parameters so it can mimic and it's like an analog analog computer analogy. It's a direct analogy and I can make a generative model of chunks of nature with another chunk of nature that I control. It's like this is like >> this is how we're going to understand the world, right? We're going to have programmable complexity to tackle the complexity >> of the world. And I mean that's basically deep learning right that is the deep learning revolution but I came at it >> from a physics perspective a physics background and so for me artificial >> sorry I'd also say the deep learning revolution was this emergent intelligence and even this idea of emergence and being able to simulate and then see emergent intelligence that was like the breaking open part of the last decade.

19:06 >> Yeah. But to me it's like a I guess like you know the thing that's very unique about my path and I guess my views versus like what is maybe common in the Bay Area is like I came at it from like first principles of physics. I haven't talked about human brains yet >> and I have a form of AI >> right? So it's a nonanthropomorphic non-anthropocentric form of AI, right? So if you have that view of what intelligence is and you come here during the boom that is the current cycle of like distilling our brains and reverse engineering our brains you know there's a huge back reaction. It's like, you know, people are going to cook me for this analogy, but like, you know, Galileo and like those that that that show that we're actually not in a geocentric world.

19:58 >> We're actually in a helilocentric world, right? It's like we are not the center of universe of the space of potential intelligences. >> Like basically all of nature is intelligent >> and you know our bodies, our economies, all sorts of self adaptive complex systems are formally a form of intelligence. That is a hitar ego. M >> which again creates a back reaction. I'm very familiar with that cuz I've >> I've I've learned to play with my own and >> yeah I think I think that that caused like a huge reaction and we we'll get to and like those texts and and I was trying to be objective there and that's maybe like the thing people truly don't understand about me and my positions.

20:39 It's like I'm not anti- nature. I'm not like anti- the biological substrate. I'm like so in love with it. I I view it like I I I view the natural world itself as like God. It's the generative process that gives us real quick on that because I asked you earlier, what do you think people tend to misunderstand about you? >> and you just said that there's this anti- nature projection or perception. Can you >> Yeah. >> on that?

21:09 >> So, I guess I'll I'll just assume people have know a little bit about EA. This is very >> maybe explain EOC for folks. >> >> and I know there's a longer explanation. >> Okay. Well, well, maybe we'll start with this thread and then and then we'll we'll branch out to broader EAC. you know, essentially was a theoretical physicist. ended up pioneering AI for quantum computers and I got basically poached out of school, ripped out of school or or grad school by Google quantum and I built there with some friends TensorFlow quantum >> and well after sort of escaping from the quantum computing world essentially I spent several years there. I worked on quantum computing, quantum sensors, quantum internet. I thought the technology was just like too far out. I thought it was going to take far too long. So, again, I kind of I hit a wall of like I've become jaded with with this field.

22:05 >> But what I realized was that actually most most of the world that we're interested in is not necessarily quantum mechanical. like most of the properties we're interested in at the macro mess scales are are are are proistic and that we had to build there were already proistic forms of AI that was basically all of AI that >> most people know you know I was working in this weird niche quantum AI which is different but if if all of AI's in general modeling is probabistic >> why are we running them on deterministic computers >> it just didn't make sense to me at the same time we're hitting the end of Moore's law, right? And for digital computers, but that's because we, you know, we got to learn to like let go in a way. We're we're we're holding on >> to the states of our computer too tightly and it's very energetically costing and we're like yanking it.

23:05 Whereas like if you let go and you're more gentle, right, then there might be some fuzz in your signal, but you're actually going to be way more power efficient and actually you'll be able to control the system much faster. >> Let go towards new types of computing and substrates. >> Well, it's essentially like what I realized was that if you're interested in power efficient computing, you're interested in the thermodynamics of computation. So going back to Fineman when he introduced quantum computing, he he wrote about this.

23:36 >> If you're trying to go fast, then you don't want a system that's at equilibrium. So it's not equilibrium thermodynamics that you have to study. You have to study out of equilibrium thermodynamics. Which that's another thing people don't understand. Every time I make an argument with people, everybody knows equilibrium thermodynamics. But but stochastic thermodynamics or out of equilibrium thermodynamics is is is a is a quite a different theory and it's much younger. Can you give us like a metaphor or like a oneliner so maybe someone watching this for the first time in the future can be like oh that's the metaphor that's the difference >> it's the physics of of of jiggly stuff right so whether it's like electron electrons populations of electrons jittering or it's like protein dynamics in in in biology a lot of mezoscale chemistry so usually it's a physics that that is studied in bioysics departments and and really the motivation to study it is to understand the origins of life, right? How did we >> emerge from a soup of matter?

24:36 >> And it's like turns out that like these laws of physics are actually formally equivalent to generative modeling >> and that the emergence of life like the laws of physics are the generative model that gave rise to us. So to me it's like oh okay >> I've found God like and that was that was yeah and so I was like okay well >> I need to dig into this this is you're telling me there's a type of physics that no like I didn't know about quantum mechanics was 100 years old >> stocastic thermodynamics is like as old as Google like the core theorems of 1999 2000 which is crazy I was working with one of the people that did some of the core theorems he's not that old like hi Gavin hope you're well but But, you know, it's I don't know. It was crazy to me that it was underexplored and it was just like a much more natural way to compute and and it it's just much more natural to to do proistic inference on on these types of computers and and yeah, basically, you know, there was a few billion dollar spin out of my team and I I walked away from it all.

25:44 >> you know, they're they're they're doing well. they're doing just fine. But you know, I was there for 3 years because it was like hosted within Google X. and and you know, I had another sort of reevaluation of like where do I want where do I want to go, right? And and and and so I went all in on sort of stoastic thermodynamic, what we just call it thermodynamic computing for short. so so wanted to pioneer this new paradigm. We can get into that or or not. but but at the same time I wanted to like this is a whole new way of thinking of viewing everything as like self- adaptive complex systems and and seeing intelligence and in systems that aren't our brains, >> right? And formally creating an equivalence and it's like okay this is present in our in our biology. It's present in our e economy and our e economics. It's it's it's present at all all scales. and and that got me into the theory of like Carl Fristen an active inference this this theory that you can derive from the free energy well he's the author of the free energy principle which is tied to thermodynamics but it's it's basically a originally it was for neuroscience but it's a very fascinating science that also tells us that you know what is what is cognition it's it's and sorry this conversation is super ADHD we're just taking tangents on tangents but you know if integrate enough tangents, you have a nice wind winding path.

27:08 >> Yeah. Carl Fristen and active inference. And >> you know that's like it is one of the leading theories of how our brains work. It's derived from from physics and it's been tested in sort of petri dishes. For example, how people train neural networks and petri dishes to play pong. They use these sorts of learning rules. So I got really interested in that that so yeah. So I just got into yeah this theory of of the free energy principle of active inference which is kind of this theory of neuroscience derived from from thermodynamics and >> and and and the cutting edge of that school of thought was to bring it into the outer equilibrium regime that is a more advanced branch of thermodynamics right which is you know so I met I met some some scientists from that group that became great friends now years later but you At the time I just view view viewed it as a unifying lens because this theory applies not just to our brains but to all physical systems >> and it's a formalization of how systems are are formally sort of any system that has a proposal and selection mechanism or has some sort of drift, right? It sort of mutates and adapts. it can obey laws that that are formally equivalent to this form of like >> sophist it's a it's a generalization of sort of basian inference right >> but it also taught me a lot about psychology and how human minds work and how our own minds work and it's actually the root of why I talk about hyperstition a lot and it's it's actually >> and it can you explain what hyperstition is for someone that might not know >> yeah so hyperstition is this concept that we tend to to increase the likelihood of of our beliefs happening, right? Which sounds maybe, you know, less scientific traditionally and sometimes dismissed, but actually it it's it's very accurate because our brains really crave a convergence of our model for the world. And >> so our beliefs are reality and reality.

29:18 So we on one end we have our perception which we, you know, we're familiar with this. We update our gen our our model of the world based on what we see, right? And we're always when we're consuming, we're always like updating >> our weights to have a better predictive model, minimize or surprise. >> We're good basians. >> Yeah, we're good basian. But there's a duel to that which is actually and and you know coming from theoretical physics where you don't take many actions in the world. You just theorize and make a better model.

29:44 >> I had to like take action in the world and to have an impact to steer it towards the better states. taking action is dual to to perception and there we try to steer the world towards our beliefs right so if you're driving the car goes where the eyes look right and to me I was seeing sort of at the time a lot of pessimism about the future I mean it was like 21 22 and you know pe people were maybe you know they were out of the co slump right like they were like isolated and so on And and I was like, man, like if if everybody's like super pessimistic about the future, we're actually going to hyperstition that. We're not we're not going to invest in ambitious things.

30:28 People are not going to, you know, raise families if they think the whole system is going to melt down in a few years. And and then you're you're actually going to hyperstition that outcome. And I was like, actually, you know, there there's a strong need for for optimism. And actually >> optimism >> in in in Silicon Valley, we we we we have this this kind of petri dish of we're experimenting with different cultures and we're seeing which ones give the groups higher fitness at least in the >> in in in the startup game, right? And it's clear that some startups have bad cultures that lead to their death and some startups have great cultures that lead to their success. So hyperstition is tied to this this notion of active inference that actually if we believe that bad things are going to happen about the future, >> we're actually going to engineer them to happen. And for example, if you're obsessed with >> a potential bat virus emerging and mutating with this this this and mutation and you do in a petri dish trying to study it because you're obsessed with it, you might actually create it and then it leaks and then you made your worst nightmare happen. And so to me, I was seeing sort of the Bay Area kind of get caught in this like anxious state about the future, right? Like if I put, you know, a coffee >> precarious >> and I let go. Yeah. Exactly. Then you have anxiety for a brief moment and you have to grab it >> because you're there's entropy about the future, it could spill, something could happen, right?

31:55 >> And and we tend to want to take action to kill entropy about the future, kill variance about the future. And that's how I was seeing sort of the school of thought of like the the AI sort of doomers, right? They were so obsessed with AI doom and and a a singleton AI that that you know they they've actually pushed the research to this point and created the most dangerous systems in terms of cyber and bio risk and they're they're they're postraining systems to test how far they could push it, but they're actually increasing the likelihood >> of of that thing happening which wouldn't ne necessarily naturally happen from sort of the market. And so like understanding this I was like okay this actual movement that's like totally intellectually dominating the Bay Area is actually dangerous.

32:44 >> We need a counterbalance to this and it's actually it's quite pessimistic about the future and it's actually hyperstitioning the advent of of a bad outcome. And so I I I ended up, you know, writing up in the the two weeks I had waiting for my incorporation docs for for this company leaving leaving Google X four years ago. I ended up writing up this text that was sort of the founding text of EA and I had this anonymous account that had started called Bef Jesus which you know now notorious I guess but but that that was kind of the the origin of it and I was like we need to spread optimism in order to to hyperstition the optimistic future we want to see and I it just felt like a great sense of responsibility to do that And you know at the same time I was seeing the writing on the wall that the appetite of the market for intelligence was going to be infinite and that the only thing stopping us from producing more intelligence was going to be running out of energy.

33:47 >> But at the same time you know going back to proistic computing like if you if Moore's law is dead then that will lead to people just building bigger and bigger supercomputers and using more power >> which is what we're doing. >> But if we know there's proof of existence is there's a better way. We're having a conversation with human level intellects running on 20 to 40 watts, >> right? So there's a better way forward and for that we have to understand and reverse engineer the physics underlying biology, the physics of stoastic thermodynamics. So I was like >> there lies all the answers, right? And that was sort of the epiphany. I was like >> I need more people to even know about this branch of physics, you know? So I think the the EAC text sort of mixing what stochastic thermodynamics can tell us about how things are going to go in our future and our place in the universe. I think that was a good exercise and that popularized some of these concepts. but you know, I still have a lot of work to do there.

34:41 And in fact, I I I I feel like even though I've created this movement, gotten billions of impressions, you know, and and it's had like policy impact and a ton of impact in the world, I I I feel like it's not quite enough, you know? I think it has to be like a full intellectual revolution, right? Like >> I think the Bay Area and >> how do you say the loud revolution in French? >> That's a good one.

35:07 I guess that's funny. but you know the Bay Area was just so like we were just talking I guess offline about you know how everyone is stuck in like their logos here and everybody's like like >> and can you explain what logos is for? >> Yeah. Like so everyone everyone is everyone in the Bay Area is like or a lot of people are like rationalists, right? It's like I need to like reason through things. I need to have to create this model of whatever it is I'm trying to understand that is like interpretable and and I can I can reason through it and there's no limit to the power of of reasoning, right? And I I was the same, but I went through that sort of ego death and theoretical physics and realized actually >> we got to go postrationalist towards more of a complexism or like thermonamicism, thermonamic view of the world.

36:04 >> I would push back a little bit because you had those early experiences on the importance of being embodied and of really focusing on the mental mental benefits that come from physically pushing yourself. And I think a lot of the issue with the rationalist movement in my mind is how disembodied it feels, how floaty and like disassociated and that you can be speaking with people and not feel them on a body level. So I think there's a degree of sematic intelligence that you very early can too that plays in.

36:30 >> Yeah, that's right. There's like having physicality, being physically present is is so important. And you know, it's one way to escape the world is just to go into abstraction land. And and in a way, like some of these rationalists, they would you know, they would code on their laptops. This was before vibe coding. And like they change a number in their terminal and then they'd have like 100x more GPUs. And they never touched them. They never felt the heat. They never built those systems. and how like >> and so they never get a sense of like how hard it is to keep scaling. They're like, "Well, if I just keep adding GPUs, the scaling law keeps going, then this thing is going to, you know, reverse engineer all of matter and create gay gray goo, right?" Like they would gray goo is like this thought experiment that, you know, you like bu the AIs would like build nanom machines that consume the planet or something. But that's just provably impossible because of the complexity of nature and how irreducible it is. you need an amount of compute that is similar to the generative process of nature itself. And so for me, >> and sorry, question for you on there. So I know you're down your vertical. do you think there's space for whole brain emulation or like bioorganisms acting as the bits or other brands or different substrates to compute that will similarly like further our understanding of the universe or do you really think there's kind of one vertical that's going to uncover it? I mean I I was working in in quantum computing and there's many substrates for quantum computing. It's kind of a general concept and I worked on many of them. I I think that's ultimately that will come. You know people are making consistent progress. It's going to take 20 more years.

38:12 >> I do think understanding if I were to start another company understanding how the brain works and reverse engineering it properly and like with next generation sensors and so on. that that's that's kind of the the next frontier there. And so, yeah, I'm a big fan of like brain scanning. I'm advising this company called Eon. They they did the the fruitfly that you >> Oh, yeah. That was cool. >> that that went viral. and you know, we're looking to to simulate those on our on our computers.

38:44 >> there's some interplay between the different frontier approaches here. >> Yeah. In a way, it's like I guess we can go back to like the act and like misunderstanding because like I'm actually have huge respect for biology and I'm trying to reverse engineer it and I'm trying to have this notion of intelligence that's like substrate independent just so we could be more compatible and we can more easily merge >> cool >> with the machine. >> That's cool.

39:08 >> And people read the text and I'm like, hey, I don't know what the future holds. There's nothing in the laws of physics that says, you know, humans will be the absolute hegeimons for eternity, right? And sometimes that's a hit to our ego. We have a chance. We have a chance to be part of the future. But but to be part of future, we have to be plastic. We have to adapt. and killing the space of possibilities about the future out of anxiety cuz we we we feel like there's a divergence of our generative model between reality and our generative model. The future generative model the future is too high entropy. There's too many things that could happen like that's not the way. We don't we don't know how much beauty and awesomeness awaits for us like if if we lean in and and you know accelerate through, right?

39:59 And yeah, at least to me it's like we I am I am trying to reverse engineer biologies and and view those sort of laws as appearing in many more substrates and in in the world around us and I'm trying to imbue this these laws that you know generative modeling through stocastic physics in hardware and we're the first company to do that >> and so it's like at the boundary between artificial intelligence and artificial life. So we could very naturally simulate a brain or stochastic processes approximating a brain or the chemical reactions in a brain. But we could also simulate cells.

40:41 >> We can of course do diffusion models too. That's what you know venture capitalists understand. Diffusion models, transformers, it's all they care about. But it's like no, this is actually a whole new form of computing >> and it has orders of magnitude more compute per watt >> and it's much closer to how nature computes. And looking back, you know, it's it's going to look like a mistake to to be running all these proistic computations on a deterministic computer.

41:04 >> the way I like to say it is like, you know, we're we're we're paying Maxwell's demon tax every clock cycle. So, Maxwell's demon is like the thermodynamic version of >> Schroinger's cat. It's our our little fable. Are are you familiar with Maxwell's demon? >> Why don't you explain it for us? >> Yeah. So it's just it essentially the gist of it is knowledge has a cost. It costs you energy to acquire information to acquire knowledge to reduce entropy about the world. And essentially it's this thought experiment where you have a box and you know let's say it's in space and has these these bouncy balls that are blue in one partition. There's a partition in the middle and some that are red in the other half of the box. And if you open a trap door in the wall of the partition, and you just wait, eventually some bouncy balls from the blue side are going to go mix into the red and the red are going to mix into the blue. And it's it's kind of it's like a gas, right? And then it mixes and eventually you're like, where where are the red balls? Are they on the left or right? It's like, I don't know. It's 50/50, right? So, so it's like it's just the second law of thermodynamics, the arrow of time, the onslaught of entropy, entropy increases, >> right?

42:23 >> Now, what if you had So, Maxwell had this this thought experiment that if what if you had a demon that could open and close the trap door depending on which ball comes from which side? So, if a blue ball comes from the right and is going to the left, you open it. Otherwise, you keep it closed. And vice versa, red going to the right. And if you wait long enough and you let this demon do its thing, eventually you end up going from a fully mixed state back to polarized of having blue balls on the left organized system like this interview.

42:56 >> So you have lower lower Yeah. Like you have lower entropy. So it's like you you lowered the entropy of the system. So did you violate the second law of thermodynamics? What what happened? And it turns out to acquire that information of which ball is coming where it costs you energy. >> And this actually applies to all of computing. you know it's it's it's actually related to Landau's principle which is the principle of like what is the minimum energy to do a certain operation for for a single bit to erase a bit and reset it >> right? What I just described is like resetting a bit right it cost you energy. and so if we are paying to maintain determinism of our computer and at the logical level we're running all these probabistic algorithms that have some noise. It's so wasteful. Why not use the natural uncertainty uncertainty you have that occurs when you run electronics at ultra low power and and then master that and then embed the problems directly into into that physics. And and I guess this is where the struggle begins because when you do the calculations for this, it's very humbling. It's like, oh, we're going to have computers that are 10,000 times or more 10 time 10,000 times more energy efficient than GPUs and people are already >> like scared of AI on GPUs and like banning them. So I was like, "Oh my god, okay, well, >> if if my company's going to succeed, I need to like >> I need to prepare people psychologically for like what's coming as well." And I think for us to be part of the future, we have to like learn to let go. In a way, it's like being part of society, you don't control everything, right? You you subject yourself to a markoff blanket of like your peers, the system, right? you're not like fully independent and in in control of everything. You're not like in the woods, you know, hunting your own stuff. But like there's a huge upside to being part of this system and you let go of like predictability and full control.

44:58 >> There's a huge upside and then your your fitness actually goes up, right? I mean, it's it's better to live in this >> in a social contract. >> Well, we're in this like meta organism, right? >> And in a way, that's what's going to happen with humanity. are going to be part of this you know and that's why I want to study like economics and trade as like a mechanism of alignment between super intelligences it works between >> super intelligences made of humans we call companies right it's like >> it turns out that it's like it's a lingua franka it's like hey I give you resources that you can use for your own growth and persistence and you give me resources or services for my own growth and persistence and we're aligned in our self-interest And I guess, you know, I hinted at it in our pre pre-hat, but I guess like maybe I'll do a little foreign into psychological growth that I've that has occurred to me or that I've been subjected to over the past several years. But >> I guess like when I wrote the text for EA, I was I went through one of my biggest personal and professional betrayals of my my career from people I really trusted out of out of greed. And you know, I had to again since I was a kid, a way to cope with like human nature was like abstraction. So, in the past, I would just like zoom out and escape, >> right? I was like, "Okay, well, you know, if humans live in the messoscales, maybe I'll talk about I'll think about the cosmos and the quantum, right? Where it's safe and they can't get."

46:43 >> Yeah. The scales. Yeah. No, truly. I mean, and but now it's like, "No, I want to understand human nature. Why why does this occur?" And it's like, it's actually okay. Well, every system is trying to minimize expected surprisal, right? And so they sometimes they tend to be greedy and like optimized for themselves, but in a way if if you figure out trade and ways to align interest, it's actually positive sum. You can like steer and couple these self-interests towards kind of a positive sum, you're actually jointly together higher fitness and you achieve more, right? It's like hunting stag versus rabbits, right? Game theoretically. and and and and and basically this is like a basic mechanism that's like a connective tissue to form meta organisms, right? And and it gives us civilization in in society as we as we know it. and and yeah, I just wanted to understand that mechanism better and I guess writing about it was was helpful. But I I think for me it was just just like taking a step back, you know, after getting burned and >> interpersonally.

47:57 >> Yeah. Yeah. Yeah. And you know, yeah, I think like definitely all sorts of PTSD from that. But you know, every great entrepreneur has some form of PTSD that they they they they get through. >> I think Well, I think everyone's gone through betrayal as well. >> Yeah. and how we realign and adapt around those wounds. We were talking about that earlier. How you actually want to take the thorn out. >> Yeah. >> Instead of building layers around it.

48:28 >> Yeah. Like essentially, you know, most of your life you end up just escaping those those wounds and and not necessarily processing them. For me, >> my escape mechanism was always just like work more, work out more. you know achieve more but sometimes you know the greatest growth comes from from looking inwards but >> I'm very curious as you gained more visibility with your platforms I also think with these frontier ideas but also the resourcing to push the needle have a very unique vantage point on the frontier of this space >> how did it feel to have more visib ability, have more opinions, more projections coming in. What was that journey like for you?

49:20 >> Yeah, I was I I I think like at the moment at in the moment it was both exciting slashtraumatizing slashoverwhelming. >> What made it traumatizing? I think I think like you know again people are sometimes self-interested like if you if you build wealth they become jealous they want to take it for themselves. If your conferred social status all of a sudden for some reason some popular popularity people build you up from their projections >> then then it's it's basically like sudden wealth of social capital and then you're a target >> and then >> they want to tear you down not just because they want in the process like pointing out your flaws or or tearing you down they get social capital like you know from bullying or whatever.

50:14 but also their projections onto you, right? It's like, oh, why would some guy with a Twitter account and who's like a a philosopher and and very popular, h how could he also be like a great like technologist that's like world class and like pioneering who's pioneered a paradigm computing doing another one like I'd rather like dismiss that. It's it like it's too much surprisal. I'd rather dismiss it than like reevaluate that actually maybe I should reevaluate my own life and my own self-limiting beliefs. That's too costly.

50:44 >> I'd rather tear this person down, con make them converge onto my beliefs for who they are. and and and and at a billion view scale >> Mhm. per year. It's just you're sampling a very large distribution and you're going to experience all sorts of, you know, yeah, a difficult difficult things whether it's, you know, online, but also, you know, IRL people just are out to get you.

51:16 and you know as a as a kid who's just wanted to to be safe and belong you know and you come to SF and like people are very open and you're just you have this hunger to belong you know some people like take advantage of that and >> you know I have a lot of metacognition and whatnot I don't know how to describe it such that I don't fall to into the traps but I could see how certain people that are like similar upbringing or or whatnot or or background grounds like they could end up in like some of these weird cults that prey onto your like need for belonging and then >> and then you got to adhere to the top down prescribed ideology otherwise you're you're an outcast and in a way I think like being an outcast in various points in my life and and never quite belonging right both from neurode divergence and I don't know what else but like you've already paid that social price in the past so you're not afraid to >> speak truth to power even if it puts you in trouble.

52:16 >> It's been a recurring issue in my life, I guess, since I was a kid. >> >> both talking back to priests to my school directors to now, you know, talking truth to >> big tech executives to big AI labs, keeping them sort of honest and check. And I get a lot of heat for that. And then even at a >> political level, right? And, >> you know, it puts you in in sort of social danger sometimes. And but that courage comes like comes from experience of not being afraid to be an outsider and just standing alone. It's like actually I know I know I'm right. But that that's also a gift that allows you to make these crazy bets that are so divergent from the main pack technologically. It's like a whole new tech tree. But you're not afraid that like even if I look stupid for 5 years, I know I'm right.

53:08 >> And now we're at year four >> and people are starting to believe now. and the vibe is shifting. That feels both really good, but you know, I think in the early years like getting torn apart every day >> is just it's tough, you know, and >> but I in a way it's like there's also this concept I learned from my my years in lifting of like progressive overload. It's like you you can adapt your nervous system by >> engineering the right stressors and and kind of, you know, tapering up. and I I I think now I just feel very robust in the public eye. It's kind of really hard to get under my skin. And I think I mean yeah I mean it was a catalyst for me to to go through the you know union individuation process and get into that whole area of psychology which I I found very interesting and very illuminating about myself and others.

54:10 but that's made me into a better leader. It's like given me just like an overall like better vibe. like it's almost like >> going through that somatic release. You're just like >> more calm. You're not like >> tense, you know, like like we were talking there's a lot of people here and meet them and they're just like >> they're holding on to all this tension. They have unint unintegrated shadow and in a way I think the powers that be don't want people to have like full emotional control or at least like understand their own latent space, right? People don't realize they have a a latent space, right? Because because because then if they're not consciously aware of it, then it can it can be hacked to to steer them.

54:55 >> Yeah. >> and >> I wanted to I wanted to double click. So practically, are there habits or rituals of ways of thinking you've had you've learned to develop that work for you on this mountain of visibility? So I think of like visibility in public platforms as a mountain where at each new level of altitude you have to acclimate like how you breathe what oxygen is available. >> h yeah I mean understand human psychology that people will project onto you their own subconscious shadow and their own wounds and that it has nothing to do with you and that it's okay to be misunderstood.

55:39 Even though our nervous system sees it, we're evolutionarily programmed because we're used to like tribes of 30 or 100 that if you're someone really doesn't like you and is hostile, that's a threat to your survival and like belonging in the group is very important. So appeasing the crowd might be like you have might you might have a pension for that, but you got to like let go of that and it's okay and they'll come around and you can also forgive them. M >> wait, can you talk a little bit more about forgiveness because I feel like we're in a moment too where it's really normalized to cut people off, never speak to them again. but obviously San Francisco is a really small tech community actually and so we all are constantly interweaving with each other over the course of years. Can you talk about forgiveness after people have after the mob has come after you after you've been attacked unfairly like whatever?

56:31 >> >> I think it comes from understanding first, right? Like you have to understand if someone burned you like how did they feel in the moment? Why did they do that? Or maybe they felt uns they had a lot going on or something. You can both like forgive them but keep your keep your distance, right? I think for me it's like, you know, having rivals, having adversaries is a good metacognitive hack to generate dopamine, get norepinephrine on demand, if you will. You know, it's a it's a good it's a good hack that, you know, is used in sports and I I I used to use it and I use it like could use your competitiveness. It's good, but it's not you can't rely on that sort of darker energy. it can't be your soul source. It's like a a source in case of emergencies. You have to have this sort of higher calling and be magnanimous and be positive some and be optimistic. I think I think that's like my baseline.

57:30 That's what I try to be. But sometimes people are like so caught up in their misery and they need that model their model of the world that it's going to go wrong and we we shouldn't be optimistic. They they view my unapologetic optimism about the future as like yeah a threat to their their core beliefs and they have to >> they have to tear me down because I I become like a symbol of that >> and I don't know in a way it's like if if you hold the line like at the end of the day as an influencer you're steering collective beliefs and if you try to steer them too quick, you know, there's like a jerk or there's too much acceleration in their beliefs, there's going to be a back reaction.

58:18 Faster you're trying to go and the further out from where they are you're trying to go, the bigger the back reaction. But if you slowly steer them sort of adabatically, they can follow you. >> But also, it's just like people have a speed limit. >> I see >> to how fast their brain can drift because of like neuroplasticity, right? It's like how you know learning rates are just a proxy to control distributional drift of of your brain.

58:45 And so that's why people for example they take >> psychedelics or drugs that increase neuroplasticity >> u to have a faster learning rate to update their beliefs quicker to learn things quicker. and you know it's it's very popular in Bay Area for that for that reason. And in a way, Bay Area is the land of the highly neuroplastic and highly open, right? It's, you know, you go to the East Coast, and I'm partially based in the East Coast. I'm actually by coastal with our lab in Boston. And there it's more academic and and more sort of not closed-minded, but it's it's stiffer prior, >> structured, >> stiffer priors that need more evidence to be overwhelmed and and and yeah, more structured, more constraints, right? So there's kind of novelty seeking and and entropy seeking and curiosity openness and then the dual to that is like constraints and and following the the clear gradient and but but you you know if you view your brain again as one of these thermodynamic systems there's there's a there's a balance between entropy and constraints and that's free energy the state of minimal free energy and so you want to have that balance you want to you don't want to be too open and and just accept any new information as like a reason to update your beliefs about reality. Like I'm going to make a characture here. I'm sorry for some friends, but like you know if you go to Burning Man and you take certain things and then you you you encounter machine elves and then you're convinced that you know you're you're you're a great writer and so on, but you also truly believe in machine elves like okay well you know that you you increase your neuroplasticity and in a way that maybe wasn't sufficiently natural and then and then now you have a higher divergence of beliefs in reality. M >> but to to me I think like I don't know my journey at least through physics and so on has been chasing these eureka moments these hits of like I don't know what it is dopamine norepinephrine of like these eureka moments where your brain goes through a huge shift and it's actually a much simpler model of how many things work right because our brain craves like simplicity right so that we we can use way less neurons to >> to parameterize the same thing instead >> principal compon components.

60:55 >> Yeah. Or or like instead of having like a mixture of many experts, it's like, oh, I I I I just have like one that can cover all these concepts, right, through through more abstraction. And and I don't know, that's kind of just been my journey. It's like, okay, I have one existence. I'm trying to understand as much as possible about the universe, understand some of the very big, some of the very small. and now I've just been pursuing understanding the messoscales biology, how cognition works, how how intelligence is imbued in in in many systems, how to imbue it in artificial systems in a way that's very thoughtful and and and principled.

61:33 and yeah, it's been it's been an interesting journey. still pretty early in it, I think. So yeah, >> I wanted to double click on maybe what you've learned relationally and as you've moved through this journey on your mission and have also, you know, 1 billion impressions a year. That's insane. That's like kind of insane to fathom. Is there anything you've learned around discernment over who has access to you or even keeping an open heart within friendships and in relation?

62:08 maybe if you could sit yourself down from like 10 years ago from a relational point of view cuz that's been a thread throughout this too. >> What would you have taught yourself around discernment? >> >> and this is a really hard question by the way. This is one I'm like personally working through thinking a lot about. >> Yes. Everybody is >> and I think a lot of people don't even start to think through this question. But I'm going to make an analogy with our immune system. Mhm.

62:36 >> Okay. And you know we have again like once you understand like these super super organisms and how they all kind of look the same. You have cells there's methods for like align alignment between cells. There's like markoff blankets how they interact and there's emergent properties. When you have a parasite like a bacteria, virus, whatever, it or or other types of parasite, it hacks your immune system.

63:08 it it hacks the inflammatory response of your system that would tend to bring in more resources to heal. It hacks that to secure more resources. The same thing is happening socially. We have evolved these pro-social inclinations because those were evolutionary optimal, right? Having empathy, >> helping each other out in times of need, like if you're injured, I would help you if you're on the trail here or something like that, right? And then that leads to, you know, positive fitness. there are, you know, adversarial perturbations to this complex system that naturally just emerge from them being able to to persist and and and replicate. And it's it tends and sometimes it's it's it's not necessarily genetic. It's like psychological wounds like it's more like mimemetic, but essentially you could just prove that there's always some sort of adversarial pertubations in these complex systems like that are that are parasitic rather than symbiotic that emerge. And I think that there are people that manipulate people's tendencies to be pro-social and empathetic and use that to manipulate them >> or weaponize it for their own gain.

64:27 >> So, certainly a lesson I wish I'd learned earlier, but I guess like you know these these lessons come just from from experience, right? I was just so stuck in abstraction land and in theoretical physics and not necessarily interacting with >> most of humanity for a while for my 20s. Like I I feel like I learned these lessons in my in my 30s and obviously running a company building a following all these things like have accelerated my learning.

64:54 >> but it's not everyone that that has that psychological journey. But in a way it's like yeah I think you say like health is measured in the nervous system. I I really like that sentence. And it does feel like I've acquired that that sort of wealth, I guess. And and then and then it's funny how once your internal state is dialed in, the the rest of the outside world tends to reflect it back and and things start going your way. It's kind of it's kind of funny how that works out. But, you know, h growing up, I would have never thought about like thinking about my my emotions. So, I would just like suppress them. is just like pure reason or like physical computation, right? So, >> so just tapping into this is like year four of your journey. Obviously, every year has felt so consequential. What are you sensing in the collective right now?

65:51 >> I think there's like there's obviously a lot of anxiety about the future, right? It's like we've evolved our cognition in order to minimize surprisal from our environment and the reason we've gotten smarter, the reason all organisms have gotten more complex over time. or there's a selective pressure towards more complexity is because you need enough complexity to generally model complexity in the world. ties back to what we were talking about at the beginning of of this this pod. But essentially, we feel that if there was intelligence out there that is more expressive and higher complexity than we can model, then we feel like we've lost control of it, right? But that sort of like need for control I think comes from maybe some people having certain certain wounds or having been burnt and like it's like their their mechanism like you know some people just want to feel in control.

66:56 They want to isolate or like start a farm somewhere and just like not be part of this like complex chaotic system that is like society and our civilization. And so >> it's kind of similar archetypes there. It's like either you're a software engineer working on this or then you retire to be a goose farmer, right? Like that's like a kind of a typical path. But I think people people need to like learn to let go and like let go of their ego, let go of, you know, this sort of anthropocentric view of everything.

67:34 And it's like we're part of this bigger system that is like this universal intelligence. We're part of this physical system. We're a very precious state of matter in our corner, the universe. And you know, we are part of it and we get to be a part of it as long as we keep evolving and shifting and and keep as long as we stay plastic and adaptive and and and and and keep improving and having this growth mindset is essential.

68:02 That's what you know what I espouse with like EAC. as long as we keep this going then we are going to be part of the future and the future is going to be even more awesome than the present. And I think shutting down possibilities going forward from fear and uncertainty is like a first order reflex but it's the wrong one in the long term. And so I think everybody's going through some sort of like singularity. not depression but like adjustment.

68:33 maybe a lot of their sense of identity was like tied to being the smartest person in the room. And now that every room has a super intelligence on top, maybe that that's changed. so yeah, I would say to me, I don't know, there's like a lot of beauty and peace that comes from understanding that we're part of this bigger this bigger picture. But at the same time, it's not like, oh, hey, yeah, like, you know, the biological substrate is over. It's all going to be silicon. In fact, it's extremely difficult to have the same amount of computation as the biological substrate in silicon. I'm pushing the boundaries of that >> and to emulate whole cells or whole brains, it's going to take still a lot of compute. Still not going to be quite as efficient as as biology. And so I think I think the next frontier, this might be a little spicy, but I think the next frontier is viewing our own biology as a differentiable program similar to a generative model, >> right? Like I was we were joking that these trees are like the original generative models, right? they're they're there parameters. There's genes that you know prescribe you know properties of of the subsystems that lead to emergent sort of phenotype, right? Phenotype is just like what what you see right from from the genotype. But really these genes are the hyperparameters of of the generative model. And some people are starting to explore viewing doing reinforcement learning on our on our genetics, right? And you know there's anxiety that like hey you know I'm like 33 34 like when I was a kid there was like I I remember watching a a black and white TV and now we have like AGI on an iPhone right and so it's like the the rate of adaptation of the non-biological substrate seems to be going extremely fast >> and like there's an anxiety that like our biology can't keep up >> but actually our biology has like way more computation still >> and it's highly really programmable.

70:42 It's universal and we need to instead of like trying to stop progress in other substrates, we need to accelerate our own and figure out how to merge and align interest towards positive sumness. I think I think we hyperstition certain outcomes like if you're obsessed with like war with a country and then you become adversarial then you hyperstition that war that conflict. It could be interpersonal conflicts. If you're obsessed with like AI going bad, going rogue, and it's going to end up being adversarial, you're actually going to hyperstition that outcome. And so to me, it's like I'm at peace with us not having control over every AI system. Like I think there's going to be AIS that escape and they pay you rent in, I don't know, crypto for you to keep them on, keep them alive, give them provide them compute for sustenance or existence really.

71:34 There's going to be those AI, there's going to be the AIs that are extensions of our cognition, extensions of ourselves, AIs that are personalized super intelligences that are really an extension like a third hemisphere of our brains, right? Like so I want to enable that and you know I think that's where things are going with all these Grockbot dot >> muse assistants. It's like it's like an extension of your brain and it's catered to you. So you're going to have these sort of it's already the beginning of merging with AI and augmenting our cognition. So I think we should lean into that. So you're gonna have free AIS, you're gonna have, you know, people that don't want to embrace AI and then you're going to have people that are kind of merged, right? And you don't need the neural link for this merge. As long as you share the same perception and action states, technically there you're the same agent in terms of active inference. And so we're we're seeing the beginnings of this merge and we got to be open to it. and yeah in a way I think I think we need to you know it's kind of been dominate like the dominant school of thought has been you know rationalism sort of reductionist thinking like oh can we reason through this and we just got to embrace like postrational complexism everything is irreducible systems everything has parameters everything's a search process in a landscape we should view everything as through that lens and and really like what we need is like not just a technological revolution that we're working on, you know, we need we we need a phil philosophical revolution.

73:08 and and and and it's going to impact every, you know, new schools of thought are going to impact everything from economics to biology, technology, psychology, whatnot. I'm trying to have even more of an impact. Like I I think the best sort of revolutions piggy back maybe on like new aesthetics because aesthetics >> they have just the right way of like you know engineering neuroplastic enough surprisal so that you have a neuroplasticity reaction >> and then you can you can inject a a subconscious message there and so >> for me I've been experimenting with neesthetics with my company I'm very involved you know there's like biometric parametric biomorphic parametric design of hardware and I've had a bit too much fun with that. But I do think like more like yeah biomorphic aesthetics like embracing complexity and emergence in art like more like computational art is like >> the the the next frontier, you know. Is there anywhere you draw inspiration from for aesthetics, even if it's like ancient cultures, or is there any anything that you pull on that you feel mirrors back resonance that you are then trying to incarnate or level up on with your form of new aesthetics today?

74:23 >> I mean it, you know, I'm I'm always experimenting. So you know extropic aesthetic is like some weird mix of like alien aesthetics almost like Egyptian archo futurist the reason I went with the the glyphs right is it's >> you know having call back the beginning of the episode but having learned French and English at the same time I feel like I had a language independent in coding so I have like I don't know my brain is a shape processor not a word process or cool >> if if that makes sense. And like >> our subconscious as well sometimes we don't even have the words to describe how we we feel. It's such a multimodal thing and it's like there's we've just been so focused on the word right like doing deep learning on on words with with LMS. I I think like there's a lot of concepts in nature and a lot of systems out there that we don't have words for. I used to work on generative AI for matter itself like understanding phases of matter and you can index a bunch of phases of matter we don't have words for but in in in latent space they have a different like vibe right for example like types of superconductors and whatnot and so there's actually we should stay open-minded that there's actually this is not the end you know like everything we could describe with words even if we have a deep learning model of all that still a limited model of the world and there's almost infinite pockets of irreducibility to learn and gro out there. and that that will require so much compute and so much more energy and it's g it's going to require us to scale the scale civilization scale scale scope and scale of human civilization to the stars and that you know in in this quest for knowledge and understanding that's necessarily yeah going to drive us up this cartef scale that I I love talking about which is just this log scale of how much free energy we're capturing from the world in order to >> to sustain and grow civilization. But the reason I view this as the most important metric is actually anchored in physics. It goes back to these concepts I was talking about stocastic thermodynamics that actually the reversibility of a process right so if there was a general process that gave rise to us we don't want to reverse back to just like molten rock on earth and no life right but the degree to which something is irreversible is like how much work was absorbed and dissipated as heat and and so it's essentially like the cartesev scale is the golden metric of like how advanced we are as like a living super organism.

77:05 And it's actually a measure of >> like the the higher up you go, the exponentially lower likelihood that you go back to zero. So the the more likely it is that civilization as a whole persists. I don't control exactly how much of each substrate is is, you know, how many copies of AIS versus versus squirrels versus humans there are. But I I I could just say statistical things about the whole the whole petri dish. And I think that's like a key misunderstood thing about our message.

77:38 And yeah, I'm I'm very prohuman. I'm very pro biology. I have great respect for it. Like to me it's one substrate for this godlike generative process that emerges from the second law of thermodynamics and I'm just trying to what I cannot create I do not understand and so I'm trying to imbue it >> in this other substrate that's artificial but it's to better understand us and then once we have that understanding and we have this lingua frana of how both types of systems work then then alignment and the merge is going to be much much simpler and then if we are augmented then and we're adapting constantly then we're going to be higher fitness and we're going to have higher persistence, which is the goal. My goal is not to like wipe out humans or whatever, which is the projection or the caricature that people project onto me.

78:25 >> And it's like tends to be like hurtful. And it's just very weird. Sometimes I like I meet some some doomers that >> like their body language is like I'm like the dark lord or Lord Sauron or something. And it's just from >> Yeah. being very misunderstood. And yeah, I mean it is what it is. I I hope like more people will see the light. But, you know, I'm trying to scale into not just like technology, not just like tweeting content, but you know, formalizing some of these these theories in like proper journals and like neuroscience, statistical physics, and so on, philosophy. I'm starting a nonprofit there. had a very weird windfall from just my overall mimemetic exposure. and I'm going to use those funds to to kickstart, you know, Carterv Research, which is this this nonprofit that, you know, the goal is to create this intellectual revolution, but it's all I do also want to support like artists and like new aesthetics because I I think like I like to say like in order to have your ideas heard, you need to have the right mimemetic packaging. So, it's a high fitness meme in terms of its propagation so it gets seen. And so I think like art is the ultimate like exploration of optimizing mimemetic packaging for for ideas and >> and and for me you know coming from being like in the physical world in the in in in the in the logos and logic you know my whole life I think like letting out my inner artists is also like it's good it it it feels very good and you know that part of me that was maybe suppressed now I can I can sort of more openly explore. That's like, you know, where where I'm going with the second half of life. And of course, I'm going to scale this paradigm of computing and make sure that we don't run out of power in the United States or the world and make sure that we can have embodied physical intelligence. Right now, we don't have the compute for it compute density. So, we're seeing a lot of acceleration in the world of bits, but not so much in the world of atoms. It's going to ensure that we can have like neural links or science corps gadget or whatever. there's a bunch of players but that we can actually merge with intelligence. You can have wearables that have enough intelligence such that you can have an online learning model that's like a true extension of your cognition. It's truly personalized. Like right now we're in like this Borg like mind era where you have amvertised inference. you have one model for everyone >> and I want us to move away from that cuz that that concentration of power even if the people right now you trust them and they aren't good and I do think a lot of them have very good intentions like even the people I give a hard time to I think overconentrating power just creates a target and and and if you give an opportunity for absolute control over people there's certain types that are >> drawn to that magnetically and having encountered people like that in the corporate world and in the broader world I'm just very weary of that setup and I think having a more decentralized locus of control of intelligence is actually paramount for us to be fall tolerant to to to corruption or subversion >> and so so it's kind of like that is the fundamental tension right like the doomers are like we got to centralize control you got to ban AI for like open source or everyone else except us because we're responsible stewards But it's like you're actually even if that's true like for a couple years like eventually your company can get commandeered or something could happen you lose control of it and now you've overconentrated power. Now there's a huge gap in capabilities between your centralized model and the decentralized >> yeah every other intelligence out there. To me, I think the danger comes when there's a huge gap in intelligence.

82:12 And to me, the most moral thing to do and the way to have a stable game theoretic equilibrium is to diffuse capabilities as they come online constantly. >> So that you're in this sort of multi-artite equilibrium rather than than overconentration of power that leads to to oppression and subversion. I mean already there's very smart people that can like manufacture consent and like steer collective beliefs by hacking our subconscious. Like that's just like one way that some people have a better generative model of them and and then they they end up getting controlled. But if you if you had that with AI that could lead to like really bad scenarios like you know I like to say P984 is superior to P Doom for me >> PI doom. And so and that that's that's just very misunderstood. But yeah, essentially, you know, when I talk to certain doomers and I I make my positions, you know, clearer and like we we both want the future to be great. We just have different opinions of what are the the leading risks. And yeah, I think I think it's very difficult sometimes on Twitter >> with like short snippets and like, you know, trying to steer the politics one way or another to to have that communication. And so longer form conversations, I'm always a fan. I'm always open to to having those chats, right? Like really, I've stretched the Overton window to give people freedom, taking that social risk for everyone to have that freedom to have that conversation of where should we lie in the riskreward sort of balance between acceleration versus >> versus slowing things down.

83:48 so yeah, >> are Extropic and Cardev hiring? >> Oh, well, Carter were just starting it. extropic. I mean we're going to go through a growth growth quite quite the growth phase soon. so stay tuned for that. we're still hiring a a handful of people, but we're always interested in people that maybe want to go a different path with generative modeling. one that is inspired by physics, maybe less about reverse engineering the human mind and and and creating this more general form of intelligence. and so some people are very much in this school of thought of sort of reverse engineering physics to to have a new form of AI. And so we're very differentiated in that way from the big labs and and you know we are sort of a Manhattan project for AI hardware, right? Like we're doing all this these efforts for >> you know nuclear physics to generate the power but not how we consume the power.

84:48 We're not doing very fancy physics for how we turn it from electricity to intelligence and we're doing that. and so whether you're a physicist or an algorithmist or an electrical engineer it's quite the project. So yeah check out extropic.aii. >> Should they DM you? Should they email you? What's the best way for viewers watching this and they're like they resonate with this, they resonate with you, gu content, whatever. How should they get in touch with you?

85:14 >> DMs are good. I would say gilextropic.ai AI2L's is my email putting it out there. Let's see. and and and yeah, apply check out our careers page. I think the best thing is like looking at what we've put out, all the software, tinkering with it, and showing showing something that you cooked up. If you're interested in the philosophy, it's very emergent or you're an artist interested in exploring like weird biomorphic aesthetics, just reach out via DM. yeah, >> Giam, thank you so much. This was such an interesting intertwining of frontier science and this arch of AI progress, but also like philosophy and spiritual underpinnings.

86:01 And for me like what this emphasizes is that science and spirituality really do like exist within the same circle. And this language pertains to both. And there is like a deep interconnectedness that we operate under and we're about to experience I think the next arc of progress and inflection point and it's so cool to hear your language around it. >> Awesome. This is really great.