The Exchanges

Q I mean, I'm a disciplined guy, but fuck, I don't. Um, what was the most, what was the most surprising truth you revealed in this exploration of yourself as the first true algorithmic guinea pig, respectfully?

A I individually, in the ways of species, are transitioning from being stewards of knowledge, being stewards of knowledge, to not knowing anything, or rather not being the Yeah, moving into a frontier where we are no longer in that role. Where AI is going to be a much better steward of knowledge. It's going to know all things. All things that have been discovered, AI will know it. And AI is also going to do a new discovery. And it's going to do so at a speed that will be, uh, we will be unable to, to, to keep up. In that scenario, AI is knowledge of record. So it's no longer that a person maintains the superior knowledge. And so that calls into question everything you understand about being human is you're going from knowing to not knowing.

AI assessment note: “you're going from knowing to not knowing.”

Q Does it hurt when you believe in the right path and you see them choose a different path, which is natural, especially as you grow up and you learn yourself, does that hurt you as a father?

A I'm happy that they get a go at this specialness of conscious existence. And the way I frame this with them is I say, I'm going to act in the role of your future self. So for my son, Talmadge, I'll say, if you ever want to talk to your future self, if you want to talk to your 46 year old self or 42 year old, I'll fill in for that role and we'll do role play. And we'll talk about, you know, what do you value for decisions made in college? And how would you think about, you know, the following things? And it's a way for him to have this dialogue, because if you think about right now, one of the most sacred Belief we have about existence is that we, the URI, in this moment, get to decide what we do, how we do it, and why we do it at any moment. So if you set your alarm for six in the morning at 10 at night, and you wake up at six in the morning, you want to snooze, six AMU gets exactly what they want. They get to hit the snooze button. So present you always has the authority to do whatever it wants, no matter what it said in the past. And no matter what future you would say, present you dominates. So imagine a future where you are basically a, uh, your present you is one vote among thousands of future yous. You're like a democracy of sort for yourself. Now, that, of course, challenges some of the things we consider most sacred about our existence, but why is that idea not feasible…

AI assessment note: “I'm happy that they get a go at this specialness of conscious existence.”

Q What's the most challenging part of that? Because the, the challenging part to me is the unknown unknowns are infinite. And so it could be this never ending quest that actually leads to inactivity. Do you worry about that? And what's the most challenging part of that segment?

A Yeah. I pressure test with a few questions. Uh, if a new idea lands, I can pose a question. What must remain, what must be true for this to be true? What must remain true for this to be true? And then, uh, what would change which would make this untrue? As you're trying to move throughout time, you're moving horizontally and vertically through time and space, pressure testing it. So you could say a given thing that may be true in this moment, but it may be entirely untrue in three years from now. And so, for example, I was talking to a friend last night, And I was explaining to him, uh, the concept of don't die. And he's a smart person. And he predictably said, well, I disagree with blank because I think super intelligence is blank and ML models of blank. And so he basically put forward the argument of the skepticism around whether or not we will achieve super intelligence. And so I, I wanted to reframe because I, I understand his arguments. They're legitimate. They're well reasoned. I understand them fine. I said, cool. What I'm trying to do is I'm trying to reframe the problem and say, if you look at our moment from the 25th century, that's the framework, which invites how to, how to think about what to do in this moment. Because if you can say, are we going to develop super intelligence? You're going to have hundreds of thousands of different opinions about where algorithmic…

AI assessment note: “I pressure test with a few questions.”

Q Does the only way to succeed, though, not mean everyone agrees? Because you can have one nation that doesn't agree to this and goes, fuck you, and presses a button. Like, does, is it not a world of entire conformity or none of it works?

A I think To, to generate, try to generate clarity of thought. I see the future as a computational mesh. Of goal alignment. And so, for example, if you, like, previously, the analog version of Brian would look at a menu, order something, and eat it. Now, this algorithm runs me, where it's measuring me, look at the evidence, and then the protocol is in place. I can imagine that systems are inside my body and outside my body, and they run like a stock market. You're looking at transactions with my DNA and my proteins in the millisecond, in the second time frame. It's just running without my awareness, and so I'm an autonomous human, and everything's just running. Now, that autonomous reconciliation of how to keep Brian to not die is running in you as well, and everyone else, and then also with the planet, and also with AI, that these systems are going to run faster than our minds, the more efficient than our minds, they'll have more data than our minds, so the future is us in this autonomous network Of computational goal alignment among all intelligence. Now, that sounds unfathomable to us. It's also inevitable. Give me an argument, like, tell me where we won't inevitably do this. Now, an example is, Ozembek is an algorithm. If people are overweight, and they're having problems with, with, uh, overeating, You take a pill, and it turns off hunger. Let's put side effects aside for a …

AI assessment note: “I see the future as a computational mesh. Of goal alignment.”

Q Cause you, you, OS funds was the first that I've seen that's done it in kind of extensive, uh, generous detail. I mean, you know, as you said, it's a playbook and it's quite rare for investors to give out a playbook. Is that, do you think we'll move towards that as an investing community?

A I think more people will be interested in solving larger, more complex problems. I think that, I mean, for example, if you look at what people are doing, those who have made an enormous amount of money in the world. And now have the option to do whatever they want. If you consider what kind of endeavors they're pursuing, A lot of them are around trying to create breakthrough scientific endeavors that would benefit the world. And I think that if you look at the underlying motivations of, of why people do this, I mean, there's social cues of like, what's, how do you achieve status? What's cool? What's important? What, what's meaningful? I think these things are good motivators to get people to, to join and do these kinds of things. I think there's a tailwind on people wanting to do things that impact the world in a positive way. And I, so I do think that in the coming years it will become more accessible because more and more people will be trying to figure out how to break down the barriers for what makes it really hard today.

AI assessment note: “I think more people will be interested in solving larger, more complex problems.”

Q The final one. Blueprints in 10 years. Where do you want it to be then, when we revisit this in 10 years? Where is Blueprints?

A Just did a photo shoot where I posed as the, um, Vitruvian man. Da Vinci's Vitruvian man, you know, arms out. And I'm going to share that on social media. I'm going to call it the Autonomous Man. So in that moment, Da Vinci was Charting new territory with dimensions and, and art of the dimensions and sketch of the human body, and you did it with this level of perfection that nobody had done it before. And if you take that and say, if you were to redo that now, what would be the modern interpretation? And I'm suggesting that is the autonomous man, which is where we're walking towards this future where what we do primarily as humans today will be automated by our algorithms. And we are going to love it. And we're going to comment that we can't believe that we humans ever existed in the current state we're in.

AI assessment note: “Just did a photo shoot where I posed as the, um, Vitruvian man.”

Q pander to founders and to LPs. So this is a whole new concept for me, Brian, but, uh, I'm learning. Can I go back to the idea process though? Because you ruminate on it, you sit with it and understand the different personas and how they interact with it. What then determines whether you act upon an idea and really pursue it versus disregarded and put it to one side?

A So there's the idea as the information It itself. And then there's all the things around the idea that can't be seen. And so when I look at the idea, I don't try to identify what can be seen. I try to identify everything that can't be seen. And those are basically the unknown unknown. Those like the zero principle concepts. And so for me, new ideas are not something to accept or reject. New ideas are something to excavate. And you're trying to Fill this thing out as broadly as possible, because yeah, new idea may be just the tip of the spear. You know, it may be something of the iceberg. It may be something much, much more, uh, expansive. And so to me, it's the, the possibility of finding new primitives to work with.

AI assessment note: “new ideas are not something to accept or reject. New ideas are something to excavate.”

Q Does the only way to succeed, though, not mean everyone agrees? Because you can have one nation that doesn't agree to this and goes, fuck you, and presses a button. Like, does, is it not a world of entire conformity or none of it works?

A I think To, to generate, try to generate clarity of thought. I see the future as a computational mesh. Of goal alignment. And so, for example, if you, like, previously, the analog version of Brian would look at a menu, order something, and eat it. Now, this algorithm runs me, where it's measuring me, look at the evidence, and then the protocol is in place. I can imagine that systems are inside my body and outside my body, and they run like a stock market. You're looking at transactions with my DNA and my proteins in the millisecond, in the second time frame. It's just running without my awareness, and so I'm an autonomous human, and everything's just running. Now, that autonomous reconciliation of how to keep Brian to not die is running in you as well, and everyone else, and then also with the planet, and also with AI, that these systems are going to run faster than our minds, the more efficient than our minds, they'll have more data than our minds, so the future is us in this autonomous network Of computational goal alignment among all intelligence. Now, that sounds unfathomable to us. It's also inevitable. Give me an argument, like, tell me where we won't inevitably do this. Now, an example is, Ozembek is an algorithm. If people are overweight, and they're having problems with, with, uh, overeating, You take a pill, and it turns off hunger. Let's put side effects aside for a …

AI assessment note: “the future is us in this autonomous network Of computational goal alignment”

Q What's the most challenging part of that? Because the, the challenging part to me is the unknown unknowns are infinite. And so it could be this never ending quest that actually leads to inactivity. Do you worry about that? And what's the most challenging part of that segment?

A Yeah. I pressure test with a few questions. Uh, if a new idea lands, I can pose a question. What must remain, what must be true for this to be true? What must remain true for this to be true? And then, uh, what would change which would make this untrue? As you're trying to move throughout time, you're moving horizontally and vertically through time and space, pressure testing it. So you could say a given thing that may be true in this moment, but it may be entirely untrue in three years from now. And so, for example, I was talking to a friend last night, And I was explaining to him, uh, the concept of don't die. And he's a smart person. And he predictably said, well, I disagree with blank because I think super intelligence is blank and ML models of blank. And so he basically put forward the argument of the skepticism around whether or not we will achieve super intelligence. And so I, I wanted to reframe because I, I understand his arguments. They're legitimate. They're well reasoned. I understand them fine. I said, cool. What I'm trying to do is I'm trying to reframe the problem and say, if you look at our moment from the 25th century, that's the framework, which invites how to, how to think about what to do in this moment. Because if you can say, are we going to develop super intelligence? You're going to have hundreds of thousands of different opinions about where algorithmic…

AI assessment note: “I pressure test with a few questions.”

Q Does it hurt when you believe in the right path and you see them choose a different path, which is natural, especially as you grow up and you learn yourself, does that hurt you as a father?

A I'm happy that they get a go at this specialness of conscious existence. And the way I frame this with them is I say, I'm going to act in the role of your future self. So for my son, Talmadge, I'll say, if you ever want to talk to your future self, if you want to talk to your 46 year old self or 42 year old, I'll fill in for that role and we'll do role play. And we'll talk about, you know, what do you value for decisions made in college? And how would you think about, you know, the following things? And it's a way for him to have this dialogue, because if you think about right now, one of the most sacred Belief we have about existence is that we, the URI, in this moment, get to decide what we do, how we do it, and why we do it at any moment. So if you set your alarm for six in the morning at 10 at night, and you wake up at six in the morning, you want to snooze, six AMU gets exactly what they want. They get to hit the snooze button. So present you always has the authority to do whatever it wants, no matter what it said in the past. And no matter what future you would say, present you dominates. So imagine a future where you are basically a, uh, your present you is one vote among thousands of future yous. You're like a democracy of sort for yourself. Now, that, of course, challenges some of the things we consider most sacred about our existence, but why is that idea not feasible…

AI assessment note: “I'm happy that they get a go at this specialness of conscious existence.”

Q How do you think about The statement that's often said we will always overestimate usage or adoption in the short term and underestimated in the long term. When you think about implementation, progression, adoption of AI, how do you think about timescales associated? It's a shit question, so forgive me for it, but the timescales associated and whether we are overestimating or underestimating it in one or the other.

A Yeah, this is a question that Again, triggers thoughts of the moment, and to me, it just creates a fog for everybody, because then if you're genuinely interested person about this question, you hear dozens of the top experts in the world all express different opinions, and for good reasons. Like, they all have well-reasoned arguments. And so then you say, what, how do I discern who's correct and who's incorrect and in what ways? It's impossible. And that's when I say, okay, if I can't discern truth in this moment, I have to fast forward a few hundred years because otherwise I can't deduce truth in any way. And so how do you come up with code scale? If you try to basically generate coherence of thought, because the question is basically inviting an answer of what do we do now? Because presumably knowing whether we're one year away from super intelligence of 10 or 20 or 50, a hundred or never, it informs something about reality. Like somehow it informs the next step we have in this present moment. So by playing on these timescales, you can work with yourself to create coherence of thought. And that's, that's ultimately the best tool we have as a species, as an individual, is, um, overcoming, uh, overcoming the noise of the moment and finding signal.

AI assessment note: “to me, it just creates a fog for everybody”

Q Cause you, you, OS funds was the first that I've seen that's done it in kind of extensive, uh, generous detail. I mean, you know, as you said, it's a playbook and it's quite rare for investors to give out a playbook. Is that, do you think we'll move towards that as an investing community?

A I think more people will be interested in solving larger, more complex problems. I think that, I mean, for example, if you look at what people are doing, those who have made an enormous amount of money in the world. And now have the option to do whatever they want. If you consider what kind of endeavors they're pursuing, A lot of them are around trying to create breakthrough scientific endeavors that would benefit the world. And I think that if you look at the underlying motivations of, of why people do this, I mean, there's social cues of like, what's, how do you achieve status? What's cool? What's important? What, what's meaningful? I think these things are good motivators to get people to, to join and do these kinds of things. I think there's a tailwind on people wanting to do things that impact the world in a positive way. And I, so I do think that in the coming years it will become more accessible because more and more people will be trying to figure out how to break down the barriers for what makes it really hard today.

AI assessment note: “I think more people will be interested in solving larger, more complex problems.”

Q How do you think about The statement that's often said we will always overestimate usage or adoption in the short term and underestimated in the long term. When you think about implementation, progression, adoption of AI, how do you think about timescales associated? It's a shit question, so forgive me for it, but the timescales associated and whether we are overestimating or underestimating it in one or the other.

A Yeah, this is a question that Again, triggers thoughts of the moment, and to me, it just creates a fog for everybody, because then if you're genuinely interested person about this question, you hear dozens of the top experts in the world all express different opinions, and for good reasons. Like, they all have well-reasoned arguments. And so then you say, what, how do I discern who's correct and who's incorrect and in what ways? It's impossible. And that's when I say, okay, if I can't discern truth in this moment, I have to fast forward a few hundred years because otherwise I can't deduce truth in any way. And so how do you come up with code scale? If you try to basically generate coherence of thought, because the question is basically inviting an answer of what do we do now? Because presumably knowing whether we're one year away from super intelligence of 10 or 20 or 50, a hundred or never, it informs something about reality. Like somehow it informs the next step we have in this present moment. So by playing on these timescales, you can work with yourself to create coherence of thought. And that's, that's ultimately the best tool we have as a species, as an individual, is, um, overcoming, uh, overcoming the noise of the moment and finding signal.

AI assessment note: “It's impossible. And that's when I say, okay, if I can't discern truth”