The Exchanges

Q Is this stuff working, by the way, yet in ways that are useful for, for quantum sensors, for gravity and other sensors, or is it, is this stuff that we're actually using?

A Uh, I would say quantum sensors and quantum communication are much shorter term technology than, than, uh, quantum computing, and that's why I moved towards, uh, those fields, uh, after subfields after a couple years in quantum computing. Uh, I, I do think, uh, there are very solid prototypes that can resolve, um, regions of parameter space in terms of, uh, acceleration, uh, uh, positioning and timing. Uh, that are not possible with, with classical sensors. And now it's a question of miniaturizing and robustifying those systems. But you can imagine a world where, you know, obviously if you, the first thing that happens in a conflict is GPS goes down, right? Uh, you know, how, how can you precisely position yourself? Um, if you have a very, uh, fine tuned accelerometer, you can just keep track of how much you've moved.

AI assessment note: “there are very solid prototypes that can resolve, um, regions of parameter space”

Q even more visible or standing up strong. They're trying to get them. Unfortunately, we're still in a free country. They haven't taken them down yet, although they're, they're trying. Let's go one step deeper on the tech, just so we understand it. Uh, for some of our, you know, for some of our listeners who are more technical, like what, what, what are you actually doing? What are you building?

A Yeah. So, so essentially we're, we're building, uh, computers for neural information processing, um, and they harness, they really harness fluctuations of nature and nature's, uh, noise, um, uh, essentially electrons tend to jitter around, uh, and we can harness that jitter in order to make, uh, systems that are far more energy efficient, far more data efficient, and, and essentially do the machine learning as a physical process, right? So we're translating between The algorithms that usually run on a digital computer, and we're just instantiating them physically, and that leads to a sort of system that's not alive, but, you know, it's, it's, it's, uh, adapting by itself, and it's, it's, it's quite interesting, and, and, you know, obviously for all the players today that are spending billions and billions of dollars on compute, this could be a big game changer.

AI assessment note: “we're building, uh, computers for neural information processing, um, and they harness”

Q know, putting them in charge. So maybe a society, just as an example, a quote-unquote evil society that just, like, cut its bottom might actually grow a lot faster in some ways, because those people are very problematic, whereas maybe a more moral society Would still grow, but take care of their bottom. Like how does that fit into EAC and how do you, how do you think about that?

A Yeah. So I, I would say like, um, religions and, and subcultures, you know, really they're, they're just highly, uh, evolved memes, right? You have memetic competition for, for subcultures, ways to live your life and different subcultures will confer an advantage or disadvantage to, to, to, to Darren and Darren's in terms of Uh, the growth, uh, of their, their, their subculture. Uh, and so these religions that have evolved are very well fine-tuned heuristics that on a long time scale ensure sort of robust growth, right? Like, who knows, like, if, if someone may not be productive now in society, who knows, maybe things change and, and, and, and, uh, their, their family becomes, like, super highly productive, right? Like, down the line. So it's not worth, uh, you know, cutting out the, the bottom 10%. And so, um, really like all we're saying is like, we're not saying a particular culture is the way to go, and it's not just about, uh, maximal strength, uh, at this moment, uh, it's about, uh, search, maintaining, uh, uh, broad search over, uh, cultures, not enforcing a single monoculture that's top-down prescribed, letting people search over ways that they'd like to live their lives and maintaining that freedom to explore. Uh, uh, that's the important thing because that's always gonna Yield a sort of, uh, high, uh, fitness, cultural, uh, uh, heuristic, uh, at any given point in tim…

AI assessment note: “So it's not worth, uh, you know, cutting out the, the bottom 10%.”

Q Tell us a little more about that. I'm curious to hear how, how does, how does that work in terms of actually information being compacted in a black hole? I know it squeezes all the stuff together, you know, and there's all sorts of interesting formulas at the, at the, at the kind of boundary of the singularity. How does this tie to information theory?

A Yeah, yeah. I mean, you know, information theory is, is, is originally a theory by Claude Shannon, which was, you know, the, laid the foundation for computing, but it's also used in, in, in physics to describe, uh, thermodynamics of various systems, and, you know, Stephen Hawking, uh, very famously worked on the entropy of black holes in their temperature, um, and there, the information, or the entropy of a black hole actually scales with, with the area of the black hole, so somehow it's compressing Information of what's inside the black hole into a hologram that scales as the area, and that was very interesting to me, and so I, I basically studied that mechanism and other mechanisms that are similar for, ah, how does nature actually encode and compress information, and, and can we create algorithms that, that decompress that information, and that's actually what got me into, ah, considering what would be, A way to, to learn this compression code, inspire ourselves from nature, and that got me into studying a quantum mechanical form of computing and, and trying to define a quantum mechanical form of AI. So quantum mechanics, for those not familiar, is the physics usually of the very small things are in superposition. They're very much not like, uh, day-to-day life. Um, and at the time, you know, it was around 2015, uh, I was seeing that, uh, you know, quantum computing Uh, was …

AI assessment note: “the information, or the entropy of a black hole actually scales with, with the area”

Q So it's fascinating though, because it does seem that in the last. 2030, 40 years in the West, like we've been able to grow and like more energy efficient ways. Right. Or is that, is that, is that fake? Is there's nothing like growing without energy or how do you think of that?

A Yeah. So, so it's both like, I'm not saying like, uh, let's grow, uh, and just burn energy for the sake of burning energy. The point is that it's, it's, it's growth on a long timescale. So it's not just the current instantaneous amount of energy we're burning. It's how much we're burning on a long timescale and hopefully that exponentially grows. And so if we're, you were not utilizing Our current energy in a clever way that leads to further growth, then we're doing something wrong, right? Um, and you know, a lot of things we do, like, let's say developing better technologies, using that energy to develop better technologies gives us optionality for, for sort of later growth and, and helps us, uh, unlock that, that, that new scale. At the end of the day, sort of, um, uh, how much, how many humans can we sustain? How big of a civilization can we sustain is determined by our level of, of, of technology. And the more, uh, intelligences that we have, the more humans we have, uh, uh, the more the rate of technology progresses, and we want to sort of keep that, that, uh, that going, that sort of feedback loop going.

AI assessment note: “The point is that it's, it's, it's growth on a long timescale.”

Q It's like, cause there's like, it's like science fiction used to be the most inspiring thing in America, 50 years ago. How do we bring that back? Do we see it coming back? Are there ways to turn this around?

A I mean, that's what we're trying to do with, with EAC. It's like, let's like, let's imagine all sorts of potential awesome futures. Uh, let, let's, like, picture them in our, our mind's eye. Let's, let's, let's figure out how, let's backpropagate for those, from those ideal futures towards today. What actions can we take that maximize the likelihood of the advent of those futures, and let's do those actions, and that's a really great algorithm to sort of amplify, you know, uh, Yeah. The likelihood that, that we make it, um, you know, Elon is very much the same, you know, has grant grand visions of the future and he works relentlessly towards them in a sense. Like once you, once you identify a positive reward in the future and you, you really model it, your brain just like rewires itself to figure out how to get there.

AI assessment note: “that's what we're trying to do with, with EAC. It's like, let's”