How AI Will Enhance Human Potential, Not Replace It (Reid Hoffman)

Speaker A: As we develop AI, how does that change our epistemology of the world? If our AI is saying something that we don't really fully understand, what does that mean? Speaker B: It becomes an article of faith where you're like, hey, look, it's really intelligent. It can't explain it to my puny biological mind. Speaker A: Maybe that will be a new form of thinking and that what that means for epistemology, what it means for reasoning, what it means for deducing ontology, metaphysics, what is actually really there, may have new landscape in it.

Speaker B: The book talks a lot, of course, about agency and how AI is going to elevate us. Speaker A: As part of this kind of thesis that I've been advancing that we're homo technae than homo sapiens is because we're already forms of cyborg. Speaker B: AI is just on this— it's a sort of a different species that is operating on a very different timescale. Speaker A: You know, in some number of centuries, maybe decades, maybe millennia, those new entities are to us like we are to squirrels. Speaker B: We are so satiated, so amused by our magical devices that we've opiated ourselves Does AI make that better, worse?

Speaker A: The dialogue around AI is so often just dominated by all of the fears, concerns, uncertainties, risks, and you don't get the future that you want by eliminating the futures you don't want. You get the future you want by conceptualizing and imagining it and going towards it. Speaker B: Hey, welcome to The Generalist Podcast. I'm your host, Mario. You might have heard the saying, the future is here. It's just not evenly distributed yet. Our goal with this podcast is to distribute the future a little more evenly by having conversations with the people that see it first.

They live in it. They are building it. They're investing in it. To that end, I'm so excited about today's episode with Reid Hoffman. Reid needs no introduction. He is the founder of LinkedIn, part of the PayPal Mafia, has played a key role with companies like OpenAI, and of course has been an extremely successful investor at Greylock. He's also a prolific writer and podcaster. His new book, Superagency: What Could Possibly Go Right with Our AI Future, is out now, and we talk about it in today's episode. He's also, of course, the co-host of the Possible Podcast.

Today's conversation with Reid was as fascinating as I knew it would be, We talk a lot about how to get the best out of AI, what we need to do to steer it towards the most productive and beneficial future. We also dig into some really interesting philosophical questions. What does it mean to have AI friends? What will it mean when we are in complete symbiosis with this superintelligence such that it's living in our brains with brain-computer interfaces? What happens when we start to genetically edit embryos to keep up with this superintelligence.

We also talk a lot about science fiction and narrative, the stories we tell about the future and how it both reflects the reality of our current technology and helps us build it. Finally, we dig into how Reid uses this technology himself and how we might use it too. We also cover what experiments Reid would run with unlimited resources and no operational constraints. What sci-fi and philosophy books we should read to meet this current moment, and what customs we might borrow from other eras or cultures. I hope you enjoy it as much as I did.

Speaker B: Hey, welcome to The Generalist Podcast. I'm your host, Mario. You might have heard the saying, the future is here. It's just not evenly distributed yet. Our goal with this podcast is to distribute the future a little more evenly by having conversations with the people that see it first. They live in it. They are building it. They're investing in it. To that end, I'm so excited about today's episode with Reid Hoffman. Reid needs no introduction. He is the founder of LinkedIn, part of the PayPal Mafia, has played a key role with companies like OpenAI, and of course has been an extremely successful investor at Greylock.

He's also a prolific writer and podcaster. His new book, Superagency: What Could Possibly Go Right with Our AI Future, is out now, and we talk about it in today's episode. He's also, of course, the co-host of the Possible Podcast. Today's conversation with Reid was as fascinating as I knew it would be, We talk a lot about how to get the best out of AI, what we need to do to steer it towards the most productive and beneficial future. We also dig into some really interesting philosophical questions. What does it mean to have AI friends?

What will it mean when we are in complete symbiosis with this superintelligence such that it's living in our brains with brain-computer interfaces? What happens when we start to genetically edit embryos to keep up with this superintelligence. We also talk a lot about science fiction and narrative, the stories we tell about the future and how it both reflects the reality of our current technology and helps us build it. Finally, we dig into how Reid uses this technology himself and how we might use it too. We also cover what experiments Reid would run with unlimited resources and no operational constraints.

What sci-fi and philosophy books we should read to meet this current moment, and what customs we might borrow from other eras or cultures. I hope you enjoy it as much as I did. Speaker C: This episode is brought to you by Vanta. Warren Buffett once said, "It takes 20 years to build a reputation and 5 minutes to ruin it." For today's digital companies, those 5 minutes have shrunk to milliseconds. This asymmetry is why trust isn't just earned, it's demanded. If you're building a business, you likely know that proving compliance is needed to win bigger deals, enter new markets, and deepen trust with customers, but that it can cost you real time and money.

By automating up to 90% of the work needed for SOC 2, ISO 27001, HIPAA, and more, Vanta gets you compliant fast. Opening doors to next-level growth opportunities. Not only does Vanta save you time, it also saves you money. A new IDC whitepaper found that Vanta customers achieve $535,000 per year in benefits, and that the platform pays for itself in just 3 months. Over 10,000 companies like Atlassian, Factory, and Chili Piper use Vanta to manage risk and improve security in real time. My listeners can get $1,000 off Vanta at com/mario. That's

com/mario for $1,000 off. This episode is brought to you by Brex. Fred Adler, the influential venture capitalist of the 1970s, was known for displaying decorative pillows in his office that featured a signature business philosophy. Corporate happiness is positive cash flow. In today's post-SERP environment, Adler's wisdom feels particularly relevant as founders need to make every dollar work harder. That's exactly what Brex delivers. Their modern finance platform was built specifically for startups like yours and designed to help extend your runway when capital efficiency matters most. With Brex, you get global corporate cards with up to 20x higher credit limits and no personal guarantee required.

Their banking solution has no minimums and no transaction fees while letting you earn high yield from day one with same-day liquidity. Best of all, Brex knows you were born to build, not juggle spreadsheets and finance tools. Their AI-powered platform brings cards, banking, expense management, and travel all in one place. It's simple, scalable, and designed to get you back to what you do best. Speaker B: Building. Speaker C: More than 30,000 companies, including 1 in 3 US venture-backed startups, trust Brex to help make every dollar count toward their mission. Join them at

com/mario. Speaker B: Reid, it is so lovely to have you here, and, uh, I can't tell you how much I've been looking forward to this conversation. Speaker A: Me as well. Um, we've obviously been doing these things in writing. Yes, over time, but in person, Awesome. Speaker B: Yeah, so much more high fidelity. Speaker A: Yes. Speaker B: One of the things that I'm excited to talk about today is, of course, your new book, Superagency. I'd love to start with just why you felt that this was something that you wanted to write and maybe a little bit of the thrust of the main argument.

Speaker A: You know, very retro publishing a book. You know, it's kind of like the, hey, let's go back to a few centuries in terms of technology. But part of it's that the dialogue around AI is so often just dominated by all of the fears, concerns, uncertainties, risks, you know, And you don't get the future that you want by eliminating the futures you don't want. You get the future you want by conceptualizing and imagining it and going towards it. And so I wanted to put a very strong, you know, kind of like stake in the ground is too weak, like firework.

Speak it into existence. Yeah, like an obelisk to say, no, this dialogue that we have of fear and uncertainty happens every single time that we have with general-purpose technologies. It happened in the printing press, it happened with the car, it happened with the mainframe computer, it happened with the internet. It happens each time, and yet the human amplification is so great afterwards, even with the dialogue that's very parallel to AI, it then creates this amazing increase in human agency. And part of the reason to choose the title "Agency" is what people are always worried about is the loss of human agency, the loss of their own agency, whether it's whether it's privacy, whether it's information, and the loss of agency as a society.

And of course, given that we have agentic technology— Yes. —that enhances those worries. So I wanted to say, it's like, actually, in fact, I have reasoned, highly rational, deep confidence that when we get into the mid-game in this, this will be a massive increase in human agency. I didn't cover as much of the worries around transitions in the book that I think is a very legitimate thing. And I think, you know, when I say it's the cognitive industrial revolution, that we're going to see the same kind of productivity increase and the human amplification, you know, so like AI amplification intelligence that we're going to get.

The transition is going to be not fun. Yeah, for sure. We as human beings adopt new general purpose technologies, generally speaking, very badly, right? So now that also gives us the right opportunity because like, well, can we do this one better? Can we do it with more grace, more humanity, more humanism? And so, and then superagency is not only when we get superpowers, when Mario gets superpowers or Reed gets superpowers, but when many of us all at the same time get superpowers, we as a society get superpowers. And that's part of the elevation of society that you see through electricity, you see through the printing press, you see through, you know, kind of— Personal computers.

Personal computers. It makes an entirely new society and new industries. And that's what we should be steering towards. Speaker B: In many ways, this is a history book. You go through sort of so many of these technological waves and it is so, you know, we can laugh at it now from our position, but seeing, you know, The Economist, I think you cite in there, declaiming that cars are, you know, the worst thing possible and that, you know, they're in favor of the horse and all of these sort of things. It's so interesting.

I was interested to see that, you know, your last book, Impromptu, you wrote with ChatGPT. This time you upgraded to Greg Bito. Why did you fire ChatGPT for this one? Speaker A: Well, ChatGPT was in the background. Both Greg and I were using ChatGPT, but as opposed to like when we did Impromptu, part of the theory of it was to not just tell our readers and the world that AI is amplification intelligence, but to show it. And the best way to show it was kind of the prompting and the text.

So it came part of the Chrome, like, like something like 40% of the words in Impromptu are ChatGPT words, right? So now here we wanted to actually create something that was more of 'cause, you know, the super agency is a human agency amplification. We wanted the, again, part of the showing as well as telling is sure, we're using AI to help us do research, to help us to cross-check, to argue against us, to say, hey, this paragraph seems boring. Speaker B: Would you punch it up a little bit? Speaker A: You know, like what would you give us?

Give me 3 options, you know, copy editing and so forth. But Greg and I have been working together for, well over 15 years. You know, part of the thing that I always like to say about Greg's many personal superpowers is if you ever read something that I've written that has humor in it, it's Greg. Like, and so— That's awesome. And so it was like, I was looking for an opportunity to get Greg some of the massive credit he deserves in generating the ideas and creating them and refining them in the work that we do together.

And so since he's a former journalist who worked at Wired and did a bunch of stuff— was like, okay, when we're writing directly about why it is this technology is so humanizing and amplification of humanist, this is the thing that we should do together and put, you know, kind of both of our names on it. And that's the reason. Amazing. Speaker B: Would you punch it up a little bit? Speaker A: You know, like what would you give us? Give me 3 options, you know, copy editing and so forth. But Greg and I have been working together for, well over 15 years.

You know, part of the thing that I always like to say about Greg's many personal superpowers is if you ever read something that I've written that has humor in it, it's Greg. Like, and so— That's awesome. And so it was like, I was looking for an opportunity to get Greg some of the massive credit he deserves in generating the ideas and creating them and refining them in the work that we do together. And so since he's a former journalist who worked at Wired and did a bunch of stuff— was like, okay, when we're writing directly about why it is this technology is so humanizing and amplification of humanist, this is the thing that we should do together and put, you know, kind of both of our names on it.

And that's the reason. Amazing. Speaker B: As well as, you know, being about super agency, this book has in many ways some superpowers of its own, which is that it's extremely personalized, which I thought was amazing. Like, you know, there's sort of my picture on the back. There's, you know, references to The Generalist in there. I thought it was such an interesting exploration of what books can be. And in this age where creation is sort of falling to zero, it makes me wonder, you know, what are the technologies that might be in your next book that aren't quite ready now?

What might that look like? Speaker A: Well, given that the thesis is that AI can make our human connections better, can amplify it, I'm always thinking and trying to explore and show, not just tell. So obviously, as you've seen, the Read AI, where I have my digital twin, and I have to say, this is much better still. Yes, thank you. You know, it's always very impressive. Yes. And by the way, Read AI has done like, you know, 10 different media interviews. It's done, it's opened conferences. When I gave a speech in Perugia last year, we had Read AI give the same speech in Italian, relevant, but also Hindi, Chinese, Japanese, and as part of that kind of human connection.

And so obviously in doing a book, I thought about, okay, well, what does AI mean? Not just obviously like Impromptu helping write the book, super agency, you know, research and tighten up and play foil for making sure your arguments are strong and solid, but also like what other things does AI make doable? And it's like, well, it can make the personalized book for Mario, for the generalist, you know? And it's like, okay, well, let's start doing that. We started doing it with Impromptu, by the way. So that was the, we, We kind of like beta tested the technology there.

And then we said, okay, now we're going to really start doing it. Now, there's still printing one at a time, still takes some time, you know, world of atoms, not just bits. But like for sure, the next things we will do will continue to be is when we have the ideas about what kinds of things we can do, we will deploy those to kind of, to try to show how it is we gain human agency, how we gain human connection, how we gain an ability to have AI have us be more homotech.

Techne, which is our evolution of humanity through technology and through how we integrate it into our lives. And so, you know, I'm quite certain within 6 months I'll have another idea. Speaker A: Well, given that the thesis is that AI can make our human connections better, can amplify it, I'm always thinking and trying to explore and show, not just tell. So obviously, as you've seen, the Read AI, where I have my digital twin, and I have to say, this is much better still. Yes, thank you. You know, it's always very impressive.

Yes. And by the way, Read AI has done like, you know, 10 different media interviews. It's done, it's opened conferences. When I gave a speech in Perugia last year, we had Read AI give the same speech in Italian, relevant, but also Hindi, Chinese, Japanese, and as part of that kind of human connection. And so obviously in doing a book, I thought about, okay, well, what does AI mean? Not just obviously like Impromptu helping write the book, super agency, you know, research and tighten up and play foil for making sure your arguments are strong and solid, but also like what other things does AI make doable?

And it's like, well, it can make the personalized book for Mario, for the generalist, you know? And it's like, okay, well, let's start doing that. We started doing it with Impromptu, by the way. So that was the, we, We kind of like beta tested the technology there. And then we said, okay, now we're going to really start doing it. Now, there's still printing one at a time, still takes some time, you know, world of atoms, not just bits. But like for sure, the next things we will do will continue to be is when we have the ideas about what kinds of things we can do, we will deploy those to kind of, to try to show how it is we gain human agency, how we gain human connection, how we gain an ability to have AI have us be more homotech.

Techne, which is our evolution of humanity through technology and through how we integrate it into our lives. And so, you know, I'm quite certain within 6 months I'll have another idea. Speaker B: Amazing. Yeah, I love this idea. Everyone who enjoys reading has had that moment when they're reading a book and they're like, wow, it feels like this author is talking right to me, like it really captures my emotions or, you know, my experiences. And it's crazy to just peer a tiny bit over the hill and say, "That will be literally true before long."

One of the topics that I thought was really interesting in the book was you talk about how a lot of the doomerism over the years has sort of attached itself to Orwell's 1984, that we're going to be in this surveillance state. And actually, if you look at the history of technology and specifically things like personal computers, it is so much about agency. We have much more power than Orwell could have imagined. One of the sort of converses of that that I find interesting is, you know, I'm sure the Neil Postman book, Amusing Ourselves to Death, and the idea being that, okay, we don't really live in 1984, but we kind of do live in a brave new world where we are so satiated, so amused by our magical devices that we've opiated ourselves.

How do you think about that? Does AI make that better, worse? Speaker A: All technologies, I mean, the Amusing Ourselves to Death is a book I recommend to everybody. And it's the thesis that maybe Huxley, not Orwell, was right. And I think we've almost demonstrated that while there may be autocracies that are Orwellian, that are using technology in this way, and so Orwell's theory was not a bad theory, it just seems that Western democracies actually, in fact, in their kind of policies of how they integrate technologies, at least so far, have been able to get all the amplification and not the Orwellian dystopia.

But of course, there's the Huxleyan dystopia, which is we're not controlled by pain, we're controlled by pleasure. We're not controlled by the control, the reduction of information, we're controlled by the flood of information such that it becomes all irrelevant. And that, I actually think, and Postman was writing well, he was writing about the television. He was writing well before the internet, let alone current AI and all the rest. And obviously, there have been elements of that. And so I do think it's one of the things that we need to, as technologists, as, thinkers and as leaders in society say, well, look, we do want to navigate that in good ways.

Now, classically in AI, so we deal with some of this in Superagency, you know, people say, well, is it going to take the thinking away from me? It's getting the agency thing. I don't have to think. By the way, the same thing was said during the printing press. Like, we had the exact same discourse. And we even had, like, kind of more modern books, you know, kind of saying, oh, it's unlike the printing press, which gives you information you have to think about. This just like thinks for you. And it's like, well, no, actually, by the way, you still have to parse it.

It still comes out in words, right? So it's like, no, no, actually, in fact, it's still the same thing. And so I think it's, there will be people who use it to be lazy, just as there's people who look up, in the technical term, dumbass shit on the internet and then believe it. Yes. Right. It is to say that you say, well, we want to put in the gentle nudges to have less of the lazy stupidity and more of the amplification. But the ability, for example, to, now do things. Like, you know, for example, as I was telling you as we were kind of getting ready, that I was just, you know, last week, you know, in Egypt, and I was looking at, you know, these Egyptian temples, which are just majestic and awe-inspiring.

And I went to ChatGPT and I was like, how much would it cost to build one of these things today? Right? And it, like, I started sharing stuff with the Egyptologist guide that I have, because it was like, oh, and here's the analysis that ChatGPT has given me about, well, like, what it would take to build this today. And so, could you send this This would be useful for me to know when I'm guiding the next group. Wow, yes. And it's like, that's our thinking. We're learning, we're doing things as ways of doing it.

And so I think that kind of human amplification is the kind of thing that I think we will see a lot of that. And so what we need to do is kind of steer the technology to kind of, you know, it's never we're going to get rid of all the bad. Try to get rid of the really bad and then, you know, really limit it. And then try to, you know, like for example, in the Huxleyan thing, all right, Are we reduced to, you know, I think it's, the film was Idiocracy, you know, are we reduced to kind of sitting on the couch?

WALL-E. Yes, yes, et cetera, et cetera. Let's not do that. Yes. Speaker B: Yeah, I think the optimistic view around AI, you know, in the Huxleyan framework to me would be, yes, you know, Postman was talking about TVs. Functionally, all of our computers now have become TVs. Like our phones are televisions, our laptops are televisions. AI hopefully because of how you're able to interface with it through voice, you know, allows you to really be in the world but still access all of that information in a very different way. Obviously there will be use cases where you want to see something and so on and so forth, but that strikes me as it could be, you know, a nice pivot away from being stuck in a soma lab or, you know, a soma state as Huxley would say.

In sort of preparing for this and following your career, of course, over the years, it sort of struck me how your particular journey feels like it was perfectly crafted for this moment in history. A review for listeners would be study philosophy at Oxford, specialization in really some of these linguistics philosophers like Wittgenstein and language games, that sort of stuff. Then help build PayPal, make commerce happen over the internet, paying people digitally. Then LinkedIn network theory connecting all these people. And then OpenAI, Monas AI, Inflection playing and building with this new technology.

And now the world that we're in is AI that plays these incredible language games that really makes us ask the deepest philosophical questions possible, that networks with everyone and with each other, and that pretty soon we'll see plenty of these things get get paid like digital workers over the internet. Does it feel like that to you that like, you know, if you look at the dots backwards, it all connects in this very nice way? Speaker A: Well, we as human beings are storytellers, so I can definitely tell a, "It was manifest destiny," from a set of things.

There's a set of strategic principles that I kind of step in trying to make my contributions to kind of human progress, and obviously a set of them around technology, a set of them around entrepreneurship, and a set of them around humanist thinking. So my undergraduate was symbolic systems, which is cognitive science and artificial intelligence. I said, well, what most interests me is how we think, how we speak, how we understand each other. Maybe philosophers have a good sense of this, so going into philosophy, and definitely with Wittgenstein and the linguistic turn, there was a tremendous amount of learning on that.

For example, one of the things that was kind of a serendipity was I hadn't really thought about kind of entrepreneurship and technology creation, except that I had gone to Stanford. Yes. And I got exposed to that. And then when I was here at Oxford, and I was like, okay, do I want to write books? I was like, no, no, actually creating the software technology, to geek and borrow from Star Wars, it's the mitochondrions. It's the thing that binds us all together and connects us and all the rest, because it's the way that we— we use software as the modern way of shaping our our minds, our worldviews, how we communicate, how we present ourselves to people.

I was like, okay, I want to go do that. And it was only many years later that I got back to writing books. Speaker B: Yes. After, after going through that full journey, I'm curious because of, you know, your work with Wittgenstein and, you know, you've talked, I think, in the past about Gareth Evans and some of these other linguistic philosophers. There is a reference. Yeah. Did that like help you think about large language models when they appeared? Did it sort of influence how you thought about the technology in general?

Speaker B: Yes. After, after going through that full journey, I'm curious because of, you know, your work with Wittgenstein and, you know, you've talked, I think, in the past about Gareth Evans and some of these other linguistic philosophers. There is a reference. Yeah. Did that like help you think about large language models when they appeared? Did it sort of influence how you thought about the technology in general? Speaker A: It reminded me that one of the other earlier things I did was also fireside chatbots, having a podcast even before Notebook LLM came out.

That is wild. And so it definitely helped me understand language and language's deep interconnections with thinking. And some people think all of thinking is language. That's incorrect. But some of our most important thinking is definitely language and linguistic. And so the questions of anything from evaluating the Turing test to thinking about what the Turing test is, to thinking about, like, well, if you have an agent that's producing language and thinking in ways that are so compelling that essentially passes the Turing test in various ways, what does that mean for our own thinking?

And what does that mean for how we evolve our thinking? And so it leads to the following kinds of interesting questions. So a classic kind of thinking that gets into philosophy of science is, as we develop tools, microscopes, experiments, other kinds of things, how does that change our epistemology of the world? Well, okay, as we develop AI, how does that change our epistemology of the world? If our AI is saying something that we don't really fully understand, what does that mean? Does it— because of course could mean the AI is wrong.

Yes. Right. That's one thing that's kind of a counter. But it also could mean is like, well, is the AI doing an inference that's currently beyond our ability to explain? And then you say, well, then we ask the AI for explanation. But of course, current large-scale, you know, kind of language models, what they are is, since we're speaking philosophy, you know, Harry Frankfurters, they're great bullshitters. Right. And so they're extremely creative. And by the way, that's a feature much more than even— it has bug elements, but it has a feature.

So it says, oh, here's the explanation. You're like, well, yeah, but is that the actual explanation? Or are you just telling me something that sounds really compelling? Speaker B: Or have you gotten to, you know, eventually to a point where, as you say, like there's sort of this, it becomes an article of faith where you're like, hey, look, it's really intelligent. It can't really tell me, it can't explain it to my puny, you know, biological mind. Speaker A: Yes, exactly. And so there opens up these new epistemological questions which are thorny.

They're actually hard to understand. And I'm sure that as we get to kind of studying them, we'll think about, okay, what are the cases where we go, well, I don't understand what this AI is doing, but I'm going to actually, in fact, maybe the expression is taken on faith or taken on some kind of reasoned belief. By the way, we have these different forms of, like this gets to the philosophy thing, we say, well, classically we have different forms of reasoning. It's induction off data, it's deduction off principles, it's abduction by best theory.

Well, maybe there's going to new, like, okay, what's, what's the theory of reasoning that goes to qualified belief in an AI system telling you something? Like maybe we're going to have a new, like, AIduction or something, you know, as a, as a way of kind of thinking about that. And that will be another probably probabilistic theory. I mean, induction's a probabilistic theory. Abduction is a probabilistic theory. Deduction's usually not. Although there's Bayesian deduction and so forth, like maybe that will be a new form of thinking. And that what that means for epistemology, what it means for reasoning, what it means for deducing ontology, metaphysics, what is actually really there, may have new landscape in it.

Speaker B: Yeah, it opens up, as you say, so many interesting philosophical questions. It actually raises something I wanted to talk to you about, which is you've defined yourself before as a mystical atheist, which I really like that phrasing. It sort of just says, hey, look, I'm an atheist, but also like, I kind of have a soul and I'm open to the unknown. And all these other things. Has AI made the mystical part of that dyad higher or the atheist part more pronounced? Speaker A: It certainly hasn't made the atheist part more pronounced.

I mean, basically the atheist is you got a whole set of world religions that come to you and say X has received truth. It's like person X, they have the red phone line to God, or this book, this was like the word of God, or da, da, da, da. And basically all of those things strike me as not believable. Yes. Right? And that's where the atheist comes in, because it's like, no, none of those. Now, if someone comes to you and says, you know, I feel that there's a spiritual presence that somehow was involved maybe in creation and maybe in consciousness, great.

You know, like, by the way, because part of the thing that too often around, you know, people go, I'm an atheist, is they go, Well, we got a lot of unanswered questions. Like, why is it the world works in this quantum thing with observer effects? That's strange. What is this thing, consciousness, that we know that with modern, like, brain science and so forth, we can stimulate in different ways, so we know that we're materially rooted, but what is it? Right? We don't really know. And so if you're not treating the world as mysterious, and by the way, like, for example, you said, well, there's absolutely no evidence for a spiritual force.

It's like, well, there's no evidence for lots of things that we later discover as interesting things. Eh, you know, so someone says, I have a perception, I have an intuition. That's interesting. And by the way, billions of people, human beings, one way or the other, kind of share that intuition. So be open to the mystery of the world. And by the way, it's one of the key things is the important thing is saying being open to mystery doesn't then mean say, Okay, you know, I've got this device that can measure your soul.

Here, put your finger on this. Like, no thanks. I'm good. Speaker B: I'm not going to go down that route. Speaker A: No thank you. Yeah, that's right. But it's mystery of the world has much more that's unexplained than it is explained. Speaker B: Those machines that measure your soul almost always seem to have a very, very high sensitivity of like, oh no, things are terrible for you. Yes, exactly. You know, we're sort of speaking about devices and things like that. You know, the book talks a lot, of course, about agency and how AI is going to elevate us.

One of the ways that, you know, folks have talked about how that maybe happens over time is through things like brain-computer interfaces, and that leads to this deeper symbiosis. Yeah, I'm curious, like, does that feel like a reasonable next step to you in terms of how this plays out? Is that one closer step towards human obsolescence because, you know, we're sort of giving over our brain to something else? Speaker A: There's obviously a bunch of different thorny questions here, and it's easy to tell various science fiction stories, both of the utopian and dystopian sides.

And it's one of the mistakes in reasoning we always have to be careful about, because the fact that you can tell a dystopian story doesn't mean that there is any distinct probability of that. And the fact that you can tell a utopian story doesn't mean there's any distinct probability of that. So you have to kind of think much more carefully. Now, we already, as part of this kind of thesis that I've been advancing, that we're more homo technae than Homo sapiens is because we're already forms of cyborg. The phones that we carry around for us, we're a form of cyborg with that.

Glasses, I mean, literally technology, the thing that we're recording now, like we are in various forms cyborg. And obviously, that sometimes gets more close and personal with prosthetic limbs and kind of pacemakers in your hearts and all the rest of the boat. And partially because we locate our identity so much in our social personhood interactions which is a brain, we go, "Well, but what about brain?" But by the way, we even do stuff in brains. I mean, we have ultrasound stimulation of brains for various things. We have definitely pharmaceuticals for treating depression and other kinds of things.

So there's forms in which we're already kind of very much dipping our toes in kind of symbiotic beings. And I think that what we're trying to do is preserve what is essentially human. And what I mean by this in this case is, well, we don't just value ourselves as cognitive beings. Some people probably say all we do is value cognitive beings, but I think as a species, we also value kind of our human aspirations, compassion, wisdom, a set of things that kind of go into that. And we want to keep that kind of identity, cultural true north, even as we evolve as Homo techni.

So usually when people tell the cyborg dystopic stories, it's like Star Trek, the Borg. You know, it's like, "Roar, I am now this machine, you know, anthill, you know, don't care about these human values." And he's like, "Okay, we want to steer away from that." But, you know, if you begin to say, for example, say, you know, we all have a kind of a cybernetic implant that allows us to like have like cloud storage for everything. And now, like, for example, like some people have photographic memories. What if we all essentially have an incomplete experience store that we can now, you know, index back to?

What if we, you know, kind of had a, you know, like a set of those things? You say, well, that doesn't sound so bad. That sounds like it could be human. That sounds like it could be good. And I mean, I'm sure there's some people who will go, oh no, no, it's dystopic. But, you know, same people are like, you know, horses, not cars. Right? And so, like, I think we are on that path of how we are evolving ourselves. Because, and by the way, even though, for example, you know, genetic evolution is so slow, we haven't seen that.

You know, one of the kind of the elephant in the rooms that most people aren't tracking is we're also making huge progress on genetics. And of course, the way it will start is like, oh, well, we can choose the IVF you know, fertilized embryo that's like, doesn't have the disease, or doesn't have the bad gene, and then you can modify it a little bit. But as you get there, you go, well, wait a minute, if we understand how to modify our genetics to kind of naturally, like just say, naturally have everyone live 10 healthier years.

Speaker B: Yeah, why not? Why wouldn't you do that? Speaker A: And then you begin to go down, and then he's like, well, Then what if we were modifying our genetics to be stronger? What if we were modifying our genetic— like, that's also where Homo techni will kind of flow back into this. And so I think we're kind of going la la la la la la la, not paying attention, whereas it's like, well, actually, in fact, realize that what we do as human beings is we evolve through technology, and we should start thinking about like, what's the small steps What would we learn?

How do we preserve what is essentially human? How do we go, as opposed to saying, "No future!" You go, "No, no, future carefully, right? Future iteratively through step, watch, step, watch. Oh, recorrect that step." You know, da da da da. And that's precisely, I think, the path we need to be on. So I think, you know, part of the thing is people go, "Oh my God, cyborgs." And you're like, "By the way, we were already on that path." Yeah, we're already partway there. Everyone probably listening to this, probably has a smartphone.

Yes. You're already on that path. You're listening to headphones. Yes. You're already on the path. Speaker C: This episode is brought to you by WorkOS. What do companies like OpenAI, Cursor, Perplexity, Webflow, Plaid, and Vercel all have in common? They use WorkOS to power enterprise features like single sign-on, directory sync, and multifactor authentication. WorkOS is set apart by its modern APIs and SDKs for seamless enterprise integrations. Plus, it's free to get started. Whether you're a scrappy startup or rapidly scaling, WorkOS has the solutions you need to secure enterprise deals. Future-proof your authentication stack with the identity layer best suited to meet the evolving demands of enterprise environments.

Find out how at com and use the code MARIO for an exclusive listener offer. Again, Visit com with code Mario and cross the enterprise chasm today. Speaker B: Speaking of the genetic, you know, IVF stuff, it reminded me of, I wonder if you read it, Oren Hoffman, the brotherhood of Hoffmans. Yes, yes. Had like an interesting— Speaker A: I'm not related, but I like Oren. Yes, yes. Speaker B: Had an interesting post about sort of like luxury embryos and the idea that, you know, in the future we might have a world in which everyone gets to raise Michael Jordan.

And/or, you know, Einstein or whoever it is. And it's sort of wild to think that that technology might be possible at some point. Speaker A: Oh, it's definitely possible. I mean, my primary worry in— well, there's a number of worries, but one of the worries in the genetic thing is we actually need to maintain a certain amount of diversity. And I think that's actually— that's important to do. And so as we begin to do steps on— we know how to eliminate Huntington's. Great. There's lupus. Great. Like, you know, but like, it's like, okay, let's not, like, even if we go like, for example, this is the gene that gives us all, you know, 10 years longer healthy.

It was like, okay, but like, let's make sure that we're never monocultural because that's the kind of thing where like a virus or other kinds of things can be. So you have to, like, that's the reason why it's like, ah, you gotta, yes, you gotta think carefully about that stuff. And the ones that, like for example, the place that when people say, "Wow, are you opposed to regulation?" Because I generally say we need to develop our technologies in the future. We need to do it through iterative development. We cannot overly predict.

But on the things that can have, you know, scale millions, hey, you know. Take a beat. You know, like, you know, let's pay attention. Speaker B: Yeah. On the sort of symbiosis piece, I don't even mean this necessarily doomerishly. Because I'm a techno-optimist ultimately, but do you see the sort of symbiosis? Isn't it inevitable at some point that this just surpasses us and we are sort of not necessary for the symbiosis in some way? AI is just on this— it's a sort of a different species that is operating on a very different timescale.

And so maybe we can have BCI or we can make genetic tests, but at some point, doesn't it just say like, What are you bringing to the equation? Speaker A: Well, there's all kinds of questions about, you know, in some sense, there's definitely a sense of, look, we will get to, you know, in some number of centuries, maybe decades, maybe millennia, the creation of entities, AI, by the way, also us, as cognitively superpowered from where we are, where we kind of go, hey, you know, those new entities are to us like we are to squirrels.

Now you go, well, but that can be alarming because like, you know, we're very comfortable with where we are right here and right now. And you're like, well, one is, and I, you know, want to see more utopian science fiction because like Iain Banks' Culture series, like, well, if we are creating super capable creatures, AI or us, and I'll get to that in a second, then, like, well, that doesn't necessarily mean that, like, why don't we, if that's the inevitable path, 'cause by the way, to some degree, given how human beings operate, we don't all decide to not build an AI.

That doesn't happen. Speaker B: Also, you don't get to decide, hey, this is where evolution stops, and it's this exact shape, and yeah. Speaker A: Yeah, so you go, okay, so the principal thing is shaping it. And human beings divide into groups and have different theories of the world and compete with each other, and even if you said we had one world government. Within that world government, there would be different human being groups dividing into groups, having different theories of the world, different competition, et cetera, et cetera. And so you go, okay, so even if that's the case, well, why don't we put all of our energy into how do you have it such that it is a symbiotic form?

And then you've got questions around like, well, but as it's created, maybe there's cyborg symbiotic forms that are useful. Maybe it's that like Iain Banks' Culture, these ships, the minds, are actually in fact kind of intrinsically kind of call it quasi-Buddhist, like fascinated with, you know, forms of conscious biological life and trying to, you know, participate and learn and help, you know, kind of in the way that, you know, we as human beings is sometimes that we interact with other life forms. Now we can learn to be better, not just pets.

Yes. Not just studying the gorilla preserves, but you know, blah, something on those lines. Or also, by the way, and this is one of the things that all the science fiction usually doesn't is like, well, like, how are we evolving? Right? So like one of the science fiction things I've thought about writing is like, what's the, like, if you said, hey, you have an evolution of silicon intelligence, what's the evolution of biological intelligence? How do those, you know, kind of, you know, what's the dance in that look like? Because that's certainly also going to become possible in various ways.

And it's part of the reason why when you think, hey, we have multiple different societies that are competing with each other, one or more of the societies say, oh, we know how to create create more intelligent human beings, we're going to do that. Speaker B: Can you tell me a little bit about that biological intelligence dance? Like, you know, if you were to try and do that sci-fi book, what does that look like? Well, so here's an interesting fact. Speaker A: Today, on a per-watt basis, we are massively more intelligent than silicon, right?

This is 20 watts, right? Wow. Yeah. Like on a per-watt basis, we are massively smarter. Right? The reason why AI already today has superhuman capabilities— there's things that GPT-4 can do that no individual human being could do. You know, maybe a group of human beings could, like, in a— Catch up in a longer time. Yeah. Well, like, slower. Yeah. But, like, do some of the things that GPT-[redacted address], we look at GPT-4 and we go, well, that's superpower. That's good for us. Oh, we're not alarmed by that superpower. And that's the kind of the progression of these things.

So if you say, well, okay, so today in our biological intelligence processing, we are massively more efficient per cognitive capability per watt. So if we can even do mild extensions of it, maybe in different distinct ways, we can keep up, right? Because this is currently— and yes, the training that's the most expensive, but even the inference uses a lot in order to get there. So that's one thought. And then the other kind of thoughts around this are, you get to some of the work that Roger Penrose and other people are doing and saying, hey, the way we think is not purely computational because there's kind of quantum mechanics and this thing called tubulars and other things.

And maybe there's some interesting attributes in that too. And so I'm not saying this is kind of like speculative, just idle hopefulness. I think there's lots of different paths to being hopeful on this. And by the way, here is a simple economist thing that leads me to say why I think symbiosis with silicon AIs is so doable. Roughly, when you see the competition of species, it's like, what are they competing for? So most of mammals and so forth are competing for essentially calories, right? Silicon intelligences don't care about calories. Right?

Like, you can have all the calories. You can have all the oxygen. It's all yours. We care about, like, sunlight. You know, we care about metals. And you're like, well, actually, in fact, there's a very natural, like, hey, there's a lot of metals and sunlight in the whole universe. Right? How about we help you with that and you help us with the things we care about, which is, like, oxygen and calories and so forth? Now, when you get to, say, for example, the next phase of Homo sapiens, next phase, which might be twice as smart as everyone else, well, they care about the oxygen and calorie stuff too.

So then there's more of a challenge there. So I'm actually more like default optimistic about the symbiosis when it comes to silicon things, because it's like, well, it's pretty easy to figure out how to divide up the pie in which we're very happy and you're very happy and like collaborating together. And that's part of the reason why I'm also optimistic that we can steer to good ways, right? Because it's like, well, like what they care about is how do we build much bigger compute infrastructure? Well, there's a lot of asteroids.

Speaker B: Yeah, you have to get into, you start doing the asteroid mining. Speaker A: Yes, yeah. But like, hey, let's do that together. We would like you to do that. That could be helpful for us too. Yeah, 100%. Speaker B: Don't take my pasta and I won't take your silicon. On the subject of sci-fi, you've mentioned like, you know, we need more optimistic sci-fi, more utopian sci-fi. Why do you think that, you know, doesn't exist more? Is that just like a less rich dramatic palette for narrative? Speaker A: Yeah, I mean, look, the simple thing is for good stories, we're naturally storytellers, kind of homo narratus.

And as part of stories, stories have heroes and villains. The most often story has that factor. Sometimes the villain's a circumstance or other things. Speaker B: Right, man in a hole. Yes, yes. Speaker A: But the most natural one that we evolve through is heroes and villains. And so the most natural thing is to say, okay, Humans are the people, are the heroes, and machines are the villains. And that gives you your dramatic tension and all the rest. But the problem is, is that it's like, well, you know, like how we, Homo techni, have evolved is through technology.

It's like we have all these machines that make our lives infinitely better. You know, like, oh, you know, robots. And you're like, yeah, by the way, some of the best advances in surgery are with robots, right? And if you had to choose today between an appropriately trained AI reading your X-ray and a human, you'd choose the AI. Now, you'd want both, right? But all of this stuff is, it's a kind of a failure of imagination. So I've kind of harangued my various Hollywood and other friends to say, Look, you have to be trying to imagine— like, I get the stories of conflict and all that, and you need a story with conflict, but you need to be imagining how we're navigating with technology is also heroic.

It's part of the things that I think were great about a lot of, you know, kind of earlier science fiction. Various of my friends have been trying to work on this, like Neal Stephenson's been trying to work on this. It's like, okay, we need this for society. Because the thing that's interesting is that the dance between how science and science fiction work is not that either of them is the only lead. So sometimes science fiction imagines things that then science goes, "Ooh, that's interesting. I'm going to go look at that."

And sometimes science does discoveries that then— You extrapolate on. Yes. And you need to have our narration stories to be imagining, you know, like the subtitle of the book, what could possibly go right? Right? And it doesn't mean it has— like, yes, of course you want a compelling story. Of course you want drama and tension and conflict, because that's part of what engages. But it doesn't have to be the future of machines is destructive. Speaker B: Yes, you could create a great drama in heaven, and then that would be, you know, the simplest utopia.

One of the better books, I would say, that— I wouldn't say it's a utopia by any means, but it's sort of It's certainly not a dystopia, and it really explores it well. Did you read Clara and the Sun, the Ishiguro? Yes, it's awesome. I thought that was a beautiful depiction of where the technology is just as human as us, and the conflict is really— it certainly touches the technology, but it's not just about that. One of the topics I get into with maybe a very pragmatic tech person who sort of doesn't understand the value of fiction is one metaphor is like, hey, this is actually the richest possible training data you can find.

On what it is like to live in a mind and what the sort of essence of mind is. Do you think that's roughly a decent metaphor? I think it's a good metaphor. Speaker A: I think one of the things when we think about these things is, and this is part of the, back to kind of mysticism or mystical, is always apply multiple metaphors. It's so tactical, but it's a good one. Speaker B: And when we look at the future of Narrative and literature. What do you think the like Dostoevskys or Tolstoys of today would be doing?

In some ways I feel like they can't be thinking, "Hey, I'll push out another great work of literary fiction." You know, would they be making games? You know, something that I know you're passionate about. Disco Elysium is like, I think the best novel I ever played. You know, it's like such a beautiful story, but like, yeah, what do you think Do you think the frontiers are there for that kind of creative expression? Speaker A: Well, definitely the simulated world. Definitely there's a lot of people are thinking about how AIs can play different characters.

That's a very broad— everyone's kind of, well, not everyone, but a lot of people are investigating that. I think that part of this kind of question comes down to, you kind of go back to first principles. When you're writing literature, what are the things you're trying to do? And part of, I think, what literature is trying to do to help us kind of think about and learn about ourselves, about other people, about our engagement with people, about our engagement with the world, what kind of paths we walk through, how we do those experiences.

You know, everything from kind of the tactile experiential characteristic all the way to, you know, kind of thinking about like, well, what is the kind of the meaning of life, the purpose of life? How do I reflect that in my own life? How do I reflect it in our lives? And so I think that's what literature does, and obviously it's in different things. You have Ulysses doing stream of consciousness, and what does that kind of do? You have Proust doing the minute examination of like the experience of being in a room, and like detail like every little, it's almost like kind of like a Buddhist meditation of like you're paying extreme attention to all the experience.

So you go, and those are part of what literature has created. And then obviously when you do it in a particularly kind of artistic or engaging way, when you do it in a new way, kind of original, I pay minute attention, but some, to like what kind of art's being created with AI. Rafik Anadol's work and other kinds of ways of doing this. Because like, okay, how does this go to that first principle around art, which is how do we experience ourselves, understand ourselves, understand other people, understand our relationship with them, understand all of our relationships with the world and individually and in groups and what does that cause us to do.

And I think that's where the principles of literature will be, whether it will be virtual worlds or whether it will be something different than virtual worlds, but experienced in Oculus glasses or like what those things will be. I think is, that's the role of the artist. Yeah, to figure out. To show us. Speaker B: Yeah, to find those new mediums and new messages. We've talked a little bit about regulation and the role that policymakers have in making sure we don't sort of screw up the bounty that AI might create. Something I've been thinking a lot about is like, one of the stories I think you could tell about our era is that that the pace of technological innovation and the speed at which policy change occurs are so crazily mismatched that it's impossible for one to catch up.

Does that make sense to you? If so, how do we make sure that there is actually good policy around this or bring the public sector in in a way that is productive rather than destructive? Speaker A: One of, I think, the generally mistaken hopes is is saying, hey, we can get the policy right for in advance. And generally speaking, I think that's very difficult to do. I think it's part of the reason why most minimum-focused regulations to stop massively negative things, right, is I think the important thing. And then the rest of that's kind of iteration.

That's also the case. And I understand people's concerns about the speed and concerns about cognitive capability and concerns about running away. But that's essentially when we look at our ability to do this over thousands of years, it's always in that kind of vector. Now, that being said, there is role— like part of the reason why, as you know, we use the cars is to go, okay, if we had said no cars until we've reviewed that they're perfectly safe, then never any cars. And we said, well, but we could list the thousand things.

And you're like, well, let's make sure they never explode. Let's make sure they have bumpers that are 6 feet wide. Let's make sure that, you know, da da da da. And you're like, yeah, and it'll never work. So you put it on the road, and by the way, the normal commercial thing then figures out how to have better brakes and airbags and all the rest. But by the way, governments do have to intervene, 'cause for example, governments had to intervene on seat belts because the manufacturers said, "Well, we have zero consumer demand for seat belts."

People don't want them, right? So we don't wanna build 'em. And the government's like, "No, actually, in fact, we're all better off." If not only we have seatbelts, but we have cars that beep at you until you put them on. Because by the way, and by the way, why are we obliged to say this is a restriction of individual freedom? It's like, well, actually, in fact, we pay for a whole hospital system, right? And actually, you're having your seatbelt on makes this not only better for you, in fact, and better for everybody else, but also, by the way, this is a better, good way for us being good in our economics because the hospital system is expensive.

And if you went crashing through the front window and so forth, you know, we're going to try to save you and, you know, and so on. So it kind of gives you— it's a— you iteratively deploy and then you add the specific regulations. And then what I say to people is like, are you saying we should do nothing now? It's like, no, no, the minimum focus big. And then the other one, if you want to be helpful, as opposed to just standing on a soapbox and yelling criticism, like figure out what are the metrics that you want to be measuring to saying, hey, things are going wrong.

Like here is a measurable metric. 'Cause by the way, if you're right about it, like you say, hey, you know, take kind of social media stuff saying, hey, this is creating a spread of misinformation and anger. Okay, what are the metrics for that? Because then by the way, government can go and say, Well, look, we want you to be measuring these metrics. We want you to be having your auditors measure that the metrics are accurate. You don't necessarily have to disclose them publicly, but you have to have them when we come and ask about them.

Speaker B: Do you have ideas of like what do you think the right metrics might be? Speaker A: Well, I was, for the social networking stuff, I was giving some of those examples, which are like, for example, you could say, well, here's a list of things. That we have as well-understood scientific truth. Yes. You could say, well, it can't be too expansive. Fine, let's start with that. Let's stay with that metric. And then we can kind of iterate that over time. That would be an instance of truth. Speaker B: Are policymakers good enough at iterative policy?

Speaker A: Well, they need to get better at iterative policy because this gets back to your speed part of your question. Yes, exactly. But what that needs to be is, like, part of it is, no, their capability that should look more like a CEO. A CEO is not a good VP of engineering. A CEO is a CEO. They should be, how are we engaging industry, academia, and other experts to go, okay, here is a thing that seems like it's the right thing. Like what is, what the metric might be, right?

We're commissioning groups of them to work together and propose those metrics and say, great. I was like, you guys all, you know, kind of go out and figure out some things to come back and then propose those metrics to us. And like I was giving a scientific one on a kind of an initial misinformation. You could also do the, you know, kind of the agitation of, you know, am I getting angrier and having more fury as a month-by-month thing as opposed to, 'cause, and by the way, that shouldn't be zero.

There's always some baseline. Level happening. But like, and because as our, our share it, share it, share it, it's like, oh, those evil people, they blah, blah, blah, blah, blah. Well, that may not be good for the health and cohesion of society. So it's like to say, again, it's not trying to get it to zero. It's just trying to get it to a, look, if it's growing because of our, our, how our intention algorithms work, then we've got a problem. And so it's like, okay, how do we adjust that in, in kind of small ways?

And I think that kind of thing, I think is the kind of thing. And, you know, like I, I, I know people who, I would put on a panel to think about this, to think about that. And so where then if government were being smart, now the thing that government has to decide is we want to be proactive and smart about this. We want to be engaging with industry, NGOs, academics. We want to be saying, hey, you guys pull this stuff together. And, you know, for example, if government came to me and said, hey, could you pull together a proposal for one of those?

I would do it. That's part of our responsibilities as citizens. And I think that's the kind of thing that we should be doing more. We also, by the way, have both on this side of the pond and the US side of the pond and other places, France, et cetera, there's organizations that will do that. The American Academy of Arts and Sciences will say, hey, if you come to us and say, we'd like an independent body to have thought this through and generated some thinking for us, and they go, great, you're going to pay attention to this and may actually use We'd love to do that work.

So that's the kind of way to engage. Now, of course, it doesn't give you the, I'm standing in front of the television camera going, I am disciplining those evil people on the left, on the right, you know, blah, blah, blah. Okay, fine. Yeah. Speaker B: If you were, if someone came to you and said, hey, Reid, please give this a shot, are there, I don't know, a few things that you would say, these are the 3 things that I would want to start with, not even the right metrics, but just like the right the right things to be thinking about?

Speaker A: So for AI, and there's different stuff for social media, people frequently conflate them, they're not the same at all. And by the way, on the social media side, I'd also pay attention equally to what happens with cable news, talk radio. Like all of the same patterns happen here. 100%. Like I think some stuff from the Biden executive order were very good. It's like, well, you know, do you have a safety team? Do you have a safety plan? Are you red teaming? What are the set of risk considerations that you've documented that you're doing for so that when I come to you and say, hey, like, have you included X?

Yeah. Or what is it? Can I see it? There's something there. Mm-hmm. Which I think all the responsible labs, which include Microsoft, Google, OpenAI, Anthropic, et cetera, are all doing. So I think that kind of thing is, I think, an important thing to have. And I think kind of having a growing basis and collaboration. Like what I'd probably say is as a government, I'd say, hey, I'd like to have a consortium of universities leading and working with industry on this and iterating that work. Yeah. Again, so that I can kind of check in on it.

And if I have a particular issue, like something bad happens, I go, were you guys paying attention to this? Yeah. Yeah. And they go, no. And they go, okay, could you start? Right. Or, or yes. Okay, great. Great. Why did this happen? Was it just like, for example, the fact that a plane crash happened? Planes are massively safe, but we all go haywire when there's a plane crash 'cause it's so visible. You know, like if you say you could travel all your miles every day the rest of your life by plane or by car, anybody with two brain cells to rub together goes plane.

Yes. Right? 'Cause it's massively safer. But it's like, well, were you paying attention? Is there something you learned and what do you do? So that would be one. Another one would be kind of the questions around what are the things by which we make sure that we're getting all the healthy individual and society empowerment and we're not empowering rogue states, terrorists, criminals? What does that plan look like? How is that iterating towards what are the things we need to be doing? And then this transition issue. So part of, I describe it as— The cognitive industrial revolution.

And I use that because it's both the amazement of we have nothing of the world that we have today without the industrial revolution. On the other hand, the transition was painful and difficult, and we will have not fun transition. So it's like, well, okay, so AI is going to be changing the kind of nature of work and the set of tasks, kind of set of things. Are we building enough of the AIs to help people with the transitions? And are we getting the full benefits, medical assistant, tutor, legal assistant, and is that being distributed broadly enough throughout society?

And so those would be the kinds of things that I would be going, "Here, here, here," in terms of what kinds of things to actually be working on. Speaker B: Yeah, let's make a start. You're someone who cares a great deal about friendship. Obviously, we all love to have friends, but you take that as that art of it almost in a serious way. I'm curious how you think about AI's impact on human friendships, because I could sort of, on the one hand, say, "Wow, this is going to lead to abundant friendships.

We can talk to a friend in any atomized moment. It just happens to be a friend that's on our phone or in our ears or anywhere else." But on the other hand, does that mean that we sort of end up spending less time with the old biological friends we have? Speaker A: Great question. What we were doing with Inflection and Pi, because I think part of our goal with Inflection and Pi was to show a set of things that were very important, that were an important part of the agentic universe, chatbot universe.

So one of the ones that we were very public about, and we're public about both of them, but we were very, and I think we've got a good catalyst in the overall system. I think Anthropic has been evolving this way and others which is EQ is as important as IQ. Yeah. Right. So to have an agent that is kind, kind of empathetic, et cetera, because by the way, how do we learn that? How do we normalize that in our human-human interactions is because of that sort of thing. And so build that into your interaction.

And Pi, I think, is still one of the most leading examples of that. Although I think we've got some of the catalytic participation in the ecosystem that I think is very Good. But then you go, well, but is that the danger that then the person only wants to talk to Pi, personal intelligence, you know, pun intended, and not to their other human friends and so forth? And it's like, well, what we did with Pi is we said, if you go to Pi and say, you're my best friend, it says, no, no, no, no, I'm your AI companion.

You want to talk about friends? Have you seen any of your friends recently? It's a good thing to do. Is it, is it, you know, and, and not like I'm going to stop talking to you until, but it's like, hey, I I'm trying to help you with your life in various ways, which includes a healthy human life is having friends, whether you're an introvert, you know, whether you're an extrovert, everyone, your life is made a lot better with friends. And so, so kind of nudging and, and facilitating that direction, not going, yes, I'm your only friend, you know, talk to me only.

Everyone else is not on your side as much as I am, you know, et cetera, et cetera. And so, so that should be a kind of a, a lens for how to build these things. Now, one of the things that's really interesting about Pi, what we've learned from it is we've learned all kinds of things. Like, for example, sometimes when people in a really depressed circumstance, they come to it and it really helps them because they're like, "Oh, I don't feel like I— I feel too ashamed to talk to other people."

Well, fine, you can talk to Pi and that helps you get back to it. Married couples have told us that what they will do is that when they're having a particularly contentious conversation, they'll put Pi down and put on audio. No way. And then be, you know, it will be helping. Well, I think when you said this, you meant this. Wow. Right? That is very good. And that again helps that human interaction. And so there's obvious negative features to try to steer away from. That's what we try to do in Pi.

But there's also amazing positive features and how it helps you like navigate your life. Helps you in your friendships, helps you in some of your most important interactions. Another thing that people can use Pi and other agents for is like, oh, I'm going to have this difficult conversation with my friend. I really think they should stop smoking, or they're drinking too much, or they were really rude to this other person. They shouldn't have been. How should I approach that conversation? Or they're in deep grief because they just lost their— their treasured pet.

How do I approach that conversation? And actually, all of these agents are really helpful. We think Pi is in particular helpful because it kind of trains that empathy side of it. But like all of that leads to much better application in friendship. And I think that's where we want to be steering broadly to. That doesn't mean that you can't have like your agent that's a mockup of your favorite anime character that you're having interactions with, that's fine in a panoply of interactions. Yes. Right. And that's the, that's the kind of thing that I think we want to be creating.

Speaker B: Do you have a sense of what the right body is for these companions? Like, is the phone the right place? Is, you know, the, the pin, the, the necklace? Speaker A: Obviously the phone's going to be a central place. We all have a phone. It has it. It's network connected, you know, et cetera, et cetera. You know, obviously we'll also have glasses, at least in official capacity. Right? Because actually, in fact, when someone says, oh, I have a glass that's a camera out, might feel weird if we're here.

But like, you know, if anything from a doctor or else comes up, he's like, oh, that's a good thing. And by the way, I like it for police officers having it because by the way, it's better for the whole society. Better for the police officer. It's better for the person. Yeah. Right. Speaker B: Higher fidelity, more information. Speaker A: Yes. That's a good thing. Right. It means that we have a It's good when there's additional cameras and data sources and things looking at these kinds of things. How it gets to interpersonal would be an interesting question.

And you know, you could say, "Oh, that's a little strange," because I actually think that within some number, not great n years, anytime we're having an interesting conversation, we'll put the AI agent down on the phone and go, "Participate." Right? Speaker B: Be quiet, like don't listen? Speaker A: No, because I think they'll get trained for like, oh, you're in a social circumstance, so don't interject other than it might kind of like flash a little bit because you go, oh, you just, you remember you were just saying this thing about ancient Egyptians in your conversation with Ariel?

Oh, quite right. Oh yeah, yeah. You go, oh right, okay. Right, stop misinformation at the source. Yeah, but it's like, and also it'll remember things, or if I said, hey, Hey, yeah, Mario, I'll get you the name of the travel agent I used for Egypt. And I forgot about it. It was like, oh, it'll remember. Like, I think that will become suddenly, oh yeah, that's great. It's natural. It's awesome that it's there, right? I think is the kind of thing that will happen in these things. And that will be facilitating really positive interactions.

Speaker B: Amazing. As we reach sort of the final stages or final phase here. Of our first conversation. Of our first. Thank you. I love to hear that. You know, one of the goals of this podcast is to help sort of distribute the future more broadly with the idea that conversations with people like you help people see it a little bit earlier. I'm curious, you know, what are the ways that you use AI today that maybe might be surprising to some folks? Speaker A: I'll start with a general framework and then get to some of the specifics.

So the general framework is these AIs are very good at taking roles. So when you think about, oh, I'd like to kind of learn or understand this stuff, what's the kind of the nature of the expert that I'd like to be talking to? And you can kind of shape it in that way. So like when I was just, you know, last week going down the Nile and looking at these Egyptian temples, I was like, okay, from a modern expert constructor, what would the plan be to build the Temple of Karnak?

What would the construction plan be? How long would it take? What kinds of expertise would you need? What kinds of equipment would you need? How much would the whole thing cost? Speaker B: Et cetera, et cetera. I'll give you a crazy quote. Speaker A: You're like, yeah, I get it. Yeah, $5 to $10 billion, right? US dollars. And so, but that's just like, it's any time that you're going through anything in life, like you'd say, hey, I'm trying to understand this particular piece of art. Like you can actually like take a picture of it and go, okay, what would a Renaissance art historian say about this?

This, right? And if they were also an expert in modern creation of AI, what would they say about this? And if they were also an expert in Italian history, what would they say about this? It's that world of experts is already there. Yes. Now, another thing that very few people use it for, but it's like, it's pretty good at inferring things. So like, you can like look at, huh, I've not used this microwave before. I want to heat up this thing. You can like take a picture of it and say, okay, how do I make popcorn with this?

Right? And it'll go, oh, this is the sequence of like frequently, 'cause it can parse that. That makes sense. Yes. And so it's the kind of the default of like, it's the world network of information trained in a really good way there for you right now in terms of how you can navigate this. So one is the role-taking, and the other one is start just like you go, "Ah, huh, okay, try it." Now, by the way, of course, sometimes it's wrong, sometimes it hallucinates. And so anything that's critically important, right?

Now, I think unlike, for example, some current world leaders, it won't say, "Oh, you can solve COVID by drinking bleach." Yes. Speaker B: Thankfully, it's well past that level of hallucination. Speaker A: Yes. But if it is something that's important, well then cross-check, right? And by the way, one of the things that's actually pretty handy is that sometimes when I look at something and it gives me an output, I go, huh, okay, cross-check this for accuracy. Speaker B: Yeah, I love doing that. Just do that. Just verify. Speaker A: Yeah, just verify this.

And by the way, it doesn't necessarily mean it'll get you the right answer, but if it just gave you a hallucinated answer, it has a very good chance of telling you, oh, well that's actually not right. Speaker B: Yeah, I don't have any data for this, so yeah. Yes. Speaker A: But I just gave you this answer, but I'm now telling you it's not right. Yeah. Like when I went and researched it, it's not right. Mm-hmm. And that's again, really useful. Speaker B: Yeah, absolutely. Cast some characters for you in your life and yeah, make them your experts.

I love that. Okay, a few final shorter philosophical questions. If you had unlimited resources and no operational constraints, what experiment would you like to run? Speaker A: The thing that I would like to do would be to create a collectively trained medical assistant that's available for free to anyone who has access to a device, and then iterate and learn from it. I suspect that if we were to do that, it would be enormously cheap. And if you think about the, you know, well north of 8 billion people in the world, I think 3/4 of them are mobile connected.

I don't know if that's always data, right? But like mobile connected, but you could also use SMS and other things for this. And what would you do for the state of humanity? If that was available to everybody, basically for free. We should do that. Speaker B: I mean, I don't know what role I would play, but that should happen. Speaker A: That would be incredible. My guess is that humans will take a little bit to adopt, but my guess is you'd start seeing amazing things within weeks. Yeah, that's incredible. Speaker B: That's such a good idea.

What practice should we borrow from another culture or era? No, perhaps an Egyptian one, who knows? Speaker A: Yeah, the mummification stuff. Yeah, yeah, yeah. I think it was interesting to learn, but no. One of the things that I think I do this, but I don't think— and some other people do, but I don't think we do enough— is kind of the salon dinner party, right? Like kind of, you know, like Jefferson had, you know, Jeffersonian dinners. It's many eras, but I think the kind of the serious dinner party that is, we're getting together and we're going to talk about, like, serious things.

And it could be pre-done in advance. It could be, you know, kind of, no, we'll just talk about the news of the day. But kind of the notion of doing that and that entertainment isn't just going out to the club or going to the movies or going to theater or, you know, but like that kind of thing, I think would be a Like, it's the kind of thing I've thought about if I wasn't kind of traveling around so much, like instituting a weekly dinner salon would be something that I think would be a good thing for everyone to do.

Speaker B: Yeah, sort of single-threaded conversations perhaps, you know, that sort of thing. Okay, final one, which is really for me after this conversation to look up. If you were to recommend a sort of brief curriculum of philosophy and science fiction for people to meet this moment in history? Like, what would be a few of the picks? Speaker B: Yeah, sort of single-threaded conversations perhaps, you know, that sort of thing. Okay, final one, which is really for me after this conversation to look up. If you were to recommend a sort of brief curriculum of philosophy and science fiction for people to meet this moment in history?

Like, what would be a few of the picks? Speaker A: So I'll start with the easier one, which is the science fiction one, because it'll give me a little bit of background process for the philosophy one. For the utopian, Iain Banks, Player of Games, Look to Windward, Excession. David Brin, Startide Rising, because it's kind of the, you know, the uplift idea, which was a very interesting prescient idea that I think people still haven't fully tracked, you know, helping dolphins and chimpanzees and everyone else get kind of intelligent. Murderbot series by Martha Wells, I think, is an interesting kind of lens into— it's kind of like the— think of it a bit as like the Ender's Game of kind of, you know, AI constructs.

And it's fun. It's not necessarily massively deep, but in its funness, it has some good elements. You know, and there's probably still a long list that goes on from there, but that's kind of some good things to start with on the science fiction side. And then in philosophy, you know, obviously I always kind of recommend Wittgenstein. It's a little hard in its primary source. Funny, for the particular technological moments, here in London, Alain de Botton, Does a whole bunch of stuff on School of Life, which makes philosophy more approachable for key things like how to live a good life and how to work well.

It's not specifically the technological moment. It kind of gives me the idea that I should talk to him some about the technological moment and probably reach out and have breakfast with him, presuming he's in town sometime in the next week or two. There's a bunch of philosophy that I've been tracking that has to do with kind of of like how are we modeling mental states and representation that's not necessarily the everyday person. It's like, for example, from a few decades ago, there's Adrian Cousins and others doing subconceptual content here in the UK, Ron Chrisley at Sussex, like David Chalmers and others who were doing stuff.

You know, I think I may have to, maybe we'll have to do our written dialogue. No, that's great. I have to go back and kind of think a little bit about this could be, you know, or maybe we'll just do these, but also in our written dialogues, maybe this would be an area that we could do. That would be really fun. Because I kind of want to think through which of the ones that are kind of more, call it, combination of deep and easily approachable, doesn't require an in-depth philosophy background for that.

Speaker B: Yeah, it's tough to find that overlap a lot of times. But anyway, Reid, it has been, yeah, just as even more lovely than I could have imagined, and thank you so much for taking the time. Speaker A: Likewise, I look forward to till next. Speaker B: Same here. That's it. Speaker C: Thank you for listening to this episode of The Generalist Podcast. Please subscribe on Apple Podcasts, Spotify, or your preferred podcast app. Ratings and reviews help others discover these discussions, so if you enjoyed the conversation, I'd be grateful if you could take a moment to leave one.

For all past episodes and more, visit us at com. See you next time as we continue to explore the future.