The Exchanges

Q Do you feel a need to be liked by other people?

A I think everybody has some element where they need to be liked. And so, yes, I would like to be liked by everybody. Emotionally, yes. Intellectually, I understand that that's not going to happen. One of the curious things, and this is, I don't think I wrote this down, but this is something I've learned over time, is that with enough numbers, with an audience of enough numbers, um, no matter what it is you do, somebody will not like it. And, and vice versa, like, let's take, I have this book of advice, there's 450, when you ask people their favorite one, the best ones, there's no agreement on the best ones. Everybody's list of the favorite ones is, there's none, there's no overlap. And so, I, I think part of what I've learned is that there's going to be people who, some people who don't like what you do all the time, but they're not going to be the same people, and when I was running a magazine called Co-Evolution and Whole Earth Review, I was the editor, and we kind of ran a little, maybe provocative kind of articles, and my goal was to piss off One fifth of the audience, but keep rotating. Make sure it wasn't the same.

AI assessment note: “And so, yes, I would like to be liked by everybody. Emotionally, yes.”

Q How do you find that balance between in a world of leverage, right? Like where an internet article can go viral and reach a hundred million people, or it can reach 10 people. How do you find that point at which I've done enough? I'm comfortable with this. Versus the trade-off of, do I raise the bar? Do I make this better? What is the advantage to doing that?

A So there's a couple things, a couple bits of advice buried in a book about that. One is this rule, which was actually based on some research in various different fields of life, that when you're trying to optimize something versus trying to explore, where you have something that works and you just want to make it kind of Better and go, you know, and optimize it versus doing the inefficient thing of going out to try and try something new and comes down to something like when you go out to eat at a restaurant, do you get your favorite thing that you know works or do you try something that is a new dish that may not work? And so the ratio is actually One to three, two to three, one to three. So you, you actually want to spend, um, two thirds of your time trying to optimize things that you want to go deeper and better, and then a third trying, trying new things, and that's been shown in many, many ways to, to be roughly true. So that's one answer is, yes, lots of times you're going to just try and make it really, really better or not thing, but a third of the time you want to be taking a chance. The second thing is this idea that, um, I learned from doing art. There was actually a book called Writer's Time about how to write a book, and it said that basically, look, the, the, the, the, the amount of work to write a book is infinite. It just could go on and on, and that's true of al…

AI assessment note: “the only thing you control is your time. You say, I have this much time”

Q Do you think we're at the hockey stick inflection point of AI finally? I mean, we've been talking about AI for 20 years.

A No, I don't. I, I think we're still at day one. And, um, so here's the thing about AI, artificial intelligence, is, is that most people think of, um, intelligence, including a lot of AI researchers, as a single dimension, like a decibels IQ. It's multidimensional. There's multi records. It's all gradient. None of it's binary. There's, there's, there's probably hundreds of different dimensions and different types of cognition in our own minds. And, um, so far we've synthesized basically two varieties, a pattern recognition and a little bit of kind of generative ability. There's this logical deduction, um, You know, symbolic reason. None of these things have we come close to synthesizing. What's shocking to us is how far we can go with just these two things. We've, and then, and they're very flat and kind of bottom up into this connectionist idea, which is like, yeah. And what I've learned from that is that that bottom up Will take you much further than you thought you could go, but will never take you all the way to where you want to go. So Wikipedia, this classic bottom-up thing, over time, there's a more and more top-down editorial control, and eventually, over time, there'll be even more to get where we want, to get that kind of encyclopedia that's totally reliable, that does what we want to do. There's going to be Some more top down control given to that bottom up power, but…

AI assessment note: “No, I don't. I, I think we're still at day one.”

Q How do you find that balance between in a world of leverage, right? Like where an internet article can go viral and reach a hundred million people, or it can reach 10 people. How do you find that point at which I've done enough? I'm comfortable with this. Versus the trade-off of, do I raise the bar? Do I make this better? What is the advantage to doing that?

A So there's a couple things, a couple bits of advice buried in a book about that. One is this rule, which was actually based on some research in various different fields of life, that when you're trying to optimize something versus trying to explore, where you have something that works and you just want to make it kind of Better and go, you know, and optimize it versus doing the inefficient thing of going out to try and try something new and comes down to something like when you go out to eat at a restaurant, do you get your favorite thing that you know works or do you try something that is a new dish that may not work? And so the ratio is actually One to three. Two to three. One to three. So you, you actually want to spend, um, two-thirds of your time trying to optimize things that you want to go deeper and better, and then a third trying, trying new things, and that's been shown in many, many ways to, to be roughly true. So that's one answer is, yes, lots of times you're going to just try and make it really, really better or not thing, but a third of the time you want to be taking a chance. The second thing is this idea that, um, I learned from doing art. There was actually a book called Writer's Time about how to write a book, and, and it said that basically, look, the, the, the, the, the amount of work to write a book is infinite. It just could go on and on, and that's true …

AI assessment note: “the only thing you control is your time. You say, I have this much time”

Q So with, with your thinking of success, I have to ask what is success for you individually?

A Great question. Um, I would say, um, I, I, I ask that in a slightly different way in my own vocabulary, which I ask other people is, what are, what am I trying to optimize? And what I'm trying to optimize is, um, learning for myself and increasing other people learning in the collective learning of the world. And I'm also To my own advice, I, I, I'm aiming so that on the day before I die, I can say I fully become myself. And I want to help that for everybody else is that I want as many people in the world to be able to say on the day before the die, I have fully become myself. And in my mind that requires tools and stuff around us to allow. So, so the story that I like to tell Is just to do the thought experiment of imagining Mozart being born 2000 years before he was born, before they invented pianos or symphonies, and the way that his genius would have been lost. We would not have, he would, you know, maybe he could play some drums or whatever was around, but it would not have, we would not have had his genius exposed that way and illuminated and fulfilled. So we needed to invent those technologies of pianos and stuff, and so Van Gogh before there was oil paint, what a loss that would have been, or George Lucas before we invented cinematography, what a loss that would have been. So that means that there is a Shakespeare today alive in the world, and she's waiting for us. To i…

AI assessment note: “what I'm trying to optimize is, um, learning for myself and increasing other people learning”

Q Uh, I want to come back to something we started talking about a little bit earlier, which is artificial intelligence. Why are you so excited about AI and what do you fear about AI?

A I'm very excited by AI. Incredibly excited. There are, there are books to be written about what's happening right now, and I really, I, I, in some ways, I think we have, we're still under-hyping it for the long term, that, that, that the, the consequences of this are so profound. This is by far our biggest invention after the scientific method, and, um, we humans are incredibly opaque to ourselves. We don't know how we work. We don't know how our brains work. We don't know what consciousness is. We don't know what intelligence is. We don't know what humans are. And the major, the way that we're going to find out about who we are is through AI robots, as we tend to try to make them like us and not like us, and how do we have other kinds of thinking, and all these things are going to come back, and then In addition to that, making other and trying to replicate ourselves is, is that, um, we're also going to use them to come, to become better humans. So it's like one of the things going on right now is you have these language models and chat bots that are trained on all human writing and behavior, and we're kind of, we're shocked that there are racists and ageists and sexists When that's kind of like the average human, but here's the thing is we don't accept that. We're saying, no, they need to be better than us. This thing, this thing that we're making cannot be racist, cannot be …

AI assessment note: “I'm very excited by AI. Incredibly excited.”

Q What would be the signs that AI has been being weaponized? How would we know?

A Oh, oh, oh. Well, there's all different levels of it. There's, in a curious way, because this may seem contradictory, in a curious way, Um, you know, there's this idea that take humans out of the kill decision. So you have, uh, robots, soldiers, and drones that make, that autonomously decide whether to fire the, the gun or not. I actually am okay with that. In a weird way, because one is, I think we, I think it's possible that we could make them better than us, and so they don't do war crimes. So they're actually, so we have the rules of war, which is a weird concept, but it's better than not having rules of war, and that we kind of say, okay, these are going to follow it, and so they do follow it. My hope is that that process of doing that, we realize how absurd it is to have war, because right now we're saying, No, no, we don't want robots killing people. We only want humans to kill other people. It's like, that's not, no, no, we don't want humans to kill people either. And so, I think, I think by having to write the laws of war for the robot soldiers, it will illuminate how Screwed up the idea of, of, of accepting war as, as a legitimate, um, process. So, so, so that's, so that's, that's part of it.

AI assessment note: “you have, uh, robots, soldiers, and drones... I actually am okay with that”