Book Club with Jeffrey Sachs
Join world-renowned economist and Columbia University Professor Jeffrey Sachs for lively conversations with the authors of scintillating, inspiring and remarkably important books about history, social justice, and the challenges of building a decent world. Learn more and get involved at bookclubwithjeffreysachs.org. The Book Club with Jeffrey Sachs is brought to you by the SDG Academy, an initiative of the UN Sustainable Development Solutions Network. Produced by Karena Joslin. Audio editing by Matt Rocker, theme song composed by Matt Rocker, performed by Dave Eggar. A kontentreal production.
Book Club with Jeffrey Sachs
Season 4, Episode 9: Ray Kurzweil, The Singularity is Nearer: When We Merge with AI
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Join Professor Jeffrey Sachs and futurist Ray Kurzweil for a compelling conversation on the accelerating pace of technological change and its profound implications for the future of humanity. In his new book, The Singularity Is Nearer, Kurzweil revisits and updates his groundbreaking predictions on AI & AGI, exponential growth, and human evolution and longevity.
Together, they explore a future where AI rivals human intelligence by 2029, nanotechnology rebuilds the world atom by atom, and our minds merge with the cloud to expand intelligence beyond biological limits. They examine radical life extension, the promise of renewable energy, and how exponential technologies are reshaping industries, reducing poverty, and transforming global well-being. But, they also confront the risks while discussing a vision of the future - both awe-inspiring and cautionary - challenging us to rethink what it means to be human in an age of rapid and relentless innovation.
The Book Club with Jeffrey Sachs is brought to you by the SDG Academy, the flagship education initiative of the UN Sustainable Development Solutions Network. Learn more and get involved at bookclubwithjeffreysachs.org.
Footnotes:
- AI
- AGI
- The Singularity is Near
- Dartmouth Workshop
- Martin Kosinski
- Neuromed
- Biotechnology
- Frank Rosenblatt
- Perceptron
- Exponential Growth
- Turing Test
- Longevity
- Humanoid Robots
- Virtual Reality
- Neocortex
- Artificial Consciousness
⭐️ Thanks for listening to Book Club with Jeffrey Sachs!
📚 Join the conversation and go deeper into the ideas shaping our world.
➡️ Watch the conversation on YouTube
➡️ Subscribe to the newsletter for episode updates, readings, and behind-the-scenes insights: https://bit.ly/subscribeBCJS
➡️ Explore past episodes, show notes, and featured books: https://bookclubwithjeffreysachs.org
🎧 Download and subscribe on your favorite podcast platform so you never miss an episode.
📣 Love the show? Leave a rating and review - it helps more listeners find the podcast.
🔁 Share this episode with a friend, colleague, or book club who’d enjoy the discussion!
Hello, I'm Jeffrey Sachs. Welcome to Book Club, a monthly conversation with world-leading authors who have written scintillating, inspiring, and remarkably important books about history, social justices, and the challenges of building a decent world. Hello, I'm Jeff Sachs and welcome to Book Club with Jeffrey Sachs. I am really excited, kind of uh kidna candy store excited today. We're speaking with uh one of the world's most important thinkers, one of the great prophets of our time, prophets of the digital and AI world. That is Ray Kurzweil. Ray Kurzweil is a name that is uh known all over the world. He's principal researcher and AI visionary at Google, and uh that's uh AI visionary is absolutely uh the title, though you could also call him a prophet, you could call him a guru of our time. But he's written a series of remarkable books that at first were deemed, I would say, way out there, kind of science fiction, uh, and now are being read almost as journalism because he predicted the world that we're living in right now, and now we're turning to him uh at every moment to try to understand where we're going. So today we're talking about the most recent book. This is The Singularity is Nearer when we merge with AI, that is with artificial intelligence, and it is uh a book written, uh published in 2024, 19 years after a book with uh basically the same title, The Singularity is Near. And in 2005, when The Singularity is Near was written, uh Ray Kurzweil essentially said, we are on an accelerating path to a new age of uh smart machines, artificial intelligence, and many, many other things. And at the time it seemed uh utterly futuristic. In the new book, the singularity is nearer. Ray Kurzweil basically documents maybe he was uh even expecting too long a period in the first book because things have accelerated so dramatically just along the lines that he predicted. And we're gonna turn to that. First, let me add that in addition to being a principal researcher and AI visionary at Google, Ray Kurzweil has been inventing things, I think, for if I counted correctly, around 60 years. He and I both read Tom Swift uh science fiction books uh in our childhood, but he uh actually took it on and created that world. I I just uh uh read those books. Uh but just to mention some of the things uh that he has invented the first uh CCD flatbed scanner, the first uh omnifont optical character recognition, the first print-to-speech reading machine for the blind, the first text-to-speech synthesizer, the first music synthesizer capable of recreating the grand piano and other orchestral instruments, the first commercially marketed large vocabulary speech recognition software. And he's won a Grammy Award, which is unbelievable, uh recipient of the National Medal of Technology, induction into the National Inventors Hall of Fame, and on and on and on. So we're speaking with someone who uh not only understands what he's talking about, but made it happen. Ray Kurzweil, thank you for joining us.
SPEAKER_01My pleasure. It's great to be on your show. I've followed your pronouncements about economics quite diligently. So it's good to have a discussion.
SPEAKER_00I'm trying to figure out the world that we're in, so that's what I want to talk to you about. You have done uh absolutely remarkable things, Ray, and it's an honor to be with you. And I want to start because we have a worldwide audience and a worldwide listenership, people from every continent and countries in many different conditions and so forth. Everybody uses uh AI and everybody uses digital now. And this is, of course, one of the remarkable things about the world. You can go to the most remote villages as I do, and people are on their smartphones, and then they're visited uh, by the way, a very rural community in a very poor part of India recently, and they were in a class with their new drone, which was going to be providing precision fertilization of the field, so they were just preparing their new drones. So these are everywhere now.
SPEAKER_01Educational systems are not adopting it from elementary school up through postgraduate schools. People tell people AI, don't use it.
SPEAKER_00Well, we're gonna get to that. Before we do, since the first book in this uh two-book series, although there are many others that are uh part of this, but was uh the singularity is near and now the singularity is nearer. Can you just tell everybody what is the singularity? Uh, what do you mean by that, and uh in what sense is it nearer and now nearer? Well, it's a complicated question.
SPEAKER_01AGI, artificial general intelligence, I'm saying will be 2029. I have a somewhat difficult definition of it. Basically, AGI would be where AI can do everything that any human can do up through, let's say, postgraduate education. It's almost there. I mean, I talked to a girl like the day before yesterday, and she spent four days uh comparing two books, uh, and the books in her field. Uh, then she decided to ask AI what it would do. So it took her four days. She read both books, compared them, tried to go through it again. The AI took 40 seconds and did a better job than she did. So people are finding that AI is really can do things that anybody can do. And the AI has read a million books, it can compare things that she couldn't do uh outside of fields. So already AI can do things that that people can't do. There's a few things that it can't do. I'm saying AGI will happen in 2029. It'll be able to do everything that a human can do, including being CEO of large corporations, it'll be able to handle that. I do think that, however, it will be positive for our personal income, because m my view is that we're going to merge with AI. It's not like we have ourselves and our natural intelligence, and then we have AI that's out here. We carry it around with us. I mean, I speak to audiences, and everybody has a phone, but we don't consider it part of ourselves, even if it's close to us. Uh it's going to actually be part of our brain. We're going to go to virtual reality where instead of carrying this around, it'll be in our eyeglasses.
SPEAKER_00Aaron Powell, although I think Ray, it's it's also it is already true that it is part of us and that you feel I feel nervous when it's not in my hand, actually. Trevor Burrus, Jr.
SPEAKER_01Absolutely. You can't go out without this. It really is part of your brain. Trevor Burrus, Jr. Yeah, for everything. But it's going to be even more so. So it's not us versus AI. AI is going to be part of our thinking, and we're not even going to be able to tell when a thought comes to us from our natural intelligence or artificial intelligence. And obviously artificial intelligence will dominate, but we're not getting rid of our natural intelligence. So it's going to be a merger. And people don't think of that. They think of us versus AI, and we don't want to use AI too much in educational systems because then we won't learn stuff with our natural intelligence. I remember when I went to college, uh, which was, I don't know, over fi uh half a century. Uh the the big deal was uh calculators. People aren't going to learn l long division if we allow people to use calculators. People today actually don't do long division anymore, but the calculators haven't gone away. It's been part of our thinking, and we understand what the calculators do, and we'll understand what AI will do as well. It'll be part of our thinking.
SPEAKER_00I remember the first moment I saw a calculator, actually, a TI calculator in the store, my dad showed it to me. I couldn't believe it. My my grandfather had a you know a hand-cranked machine, he was a bookkeeper, and uh to see it, so I I know what you're talking about in in uh in that sense. But Ray, actually, before we get to the present and the future, since you've seen this whole evolution of this field, you missed the earliest years, so uh we'll put aside Alan Turing and uh John von Neumann. But starting in the 60s, you saw all of this unfold.
SPEAKER_011962, so 63 years ago, I got involved with AI, which is more than anybody else. Yeah. Uh and I contacted Marvin Minsky, who was then actually 29. Uh I was like half that age. Uh and I heard about this field uh of artificial intelligence. So I nobody had heard of it. In fact, most people didn't even hear of computers back in the early 60s. This was about six years after 1956 when they had a Dartmouth conference where uh Marvin Minsky came up with the name artificial intelligence.
SPEAKER_00So he coined the term actually. Yeah. More or less. And uh I wanted to ask you, because uh of course uh I knew him and knew Seymour Papert, his colleague, and they were not as optimistic at some point.
SPEAKER_01Well, in 1953, Minsky uh came up with the idea of neural nets. He actually came up with the idea, which is now what we use, but he was very much against it in the 60s. Uh and he wrote a book with Seymour Papert called Perceptrons, which was the first popular neural net. And he proved that a neural net could not tell if two images were touching each other or not. So, like this image, this is not touching this, so it's not connected. This is connected. It couldn't tell that difference. And this completely killed all funding for neural nets for about 30 years. My God.
SPEAKER_00So when you got to MIT, this was not the thing.
SPEAKER_01Right, but I was actually interested, and I actually went to see Rosenblatt, who was the figure behind the perceptron, and he had the idea that, yes, uh it's true that you that a perceptron could not tell whether or not a figure was connected or not, if it's a single-layer neural net. But if you have multiple layers, it could tell that. Now we have things with a hundred or more layers, and it clearly can do that. So the book really was not accurate. And uh Rosenblatt started to actually create a multi-layer neural net, but he died nine years later, 1971. And multilayered neural nets didn't come until about 10 or 20 years after that.
SPEAKER_00Was Jeffrey Hinton the the first to implement that, or no, that came later. Uh-huh. His contribution was how to calculate the backpropagation, or what?
SPEAKER_01How to apply them to large language models. And he was at Google, we had that that concept. Large language models didn't come until about four years ago. Four years ago, they began to show some indication, but they were filled with errors. It wasn't very accurate. So we've made tremendous progress in the last four years. So this goes back to 1939, the very first computer. It was the ZUSA II. It was shown to Hitler. He was a German, but he was actually not pro-Nazi, but it got shown to Hitler. Hitler didn't see any value in it. One of the big mistakes he made. The third computer on this thing was done by Turing, and Churchill got behind that, and they were able to decode the Nazi error messages with that computer. Anyway, the 1939 computer did.0007 calculations per second per constant dollars. 2024, uh, the Nvidia B200 chip did half a trillion calculations per second per constant dollar. So this is a 75 quadrillion-fold increase in 85 years in the power of hardware. So the price performance of computation per constant dollar has doubled every 18 months since 1939. 75 quadrillion fold increase. Uh the actual uh increase is actually the hardware increase, which is 75 quadrillion fold in 85 years, times the software increase. That depends on what you're doing, but sometimes it can be a million to one. So it could be 75 million quadrillion-fold increase for the same price. That's why we didn't have large language models in 1939, or even four years ago, we had something called large language models, but they didn't work very well. And the the progress is so fantastic. I mean, uh every six months goes by, it's a completely new world, uh, and that's going to continue.
SPEAKER_00Trevor Burrus, Jr. Ray, is it fair to say in 2005 you were extrapolating this curve and neural networks? Trevor Burrus, Jr.
SPEAKER_01In 1999, I wrote The Age of Spiritual Machines. Right. Uh, and I predicted that in 30 years, by 2029, we would have enough power to do AGI. Stanford uh was alarmed by this, and they organized a conference to deal with my prediction. Uh had enough predictions that were correct that they uh had this conference. Several hundred AI experts came and they agreed with me that it would happen, but not within 30 years. They said it would take the the consensus was a hundred years. A century. And that's true of Henton and other people who were there. They were saying it was a hundred years. I was the only person that said it would happen in 30 years. Didn't they understand exponential growth? Some extent they didn't actually plot it out. I mean, exponential growth is really uh quite uh fantastic. 30 steps uh linearly gets you to 30, 30 steps exponentially, 24816 gets you to a billion. And that's the kind of uh progress we've made. Also, uh that plot it goes through different types of technologies, start with electromechanical relays, vacuum tubes, discrete transistors, and finally integrated circuits. Bohr's law only deals with integrated circuits, so there's a much broader kind of concept. Uh and I had about half of this graph 25 years ago and felt that it would continue. And so I was basing this on some real data. Other people were just saying, well, it's going to take about a hundred years. Basically, understanding a little bit of exponential growth, but not really accepting it.
SPEAKER_00Aaron Powell Can I ask you one question about the early days? Something that a little bit puzzles me, I have to say, because I don't find it quite intuitive. Turing, who in several ways was uh a father or the father of the whole story, proposes uh the test of uh something like AGI, the so-called Turing test. He did think exponentially.
SPEAKER_01He thought that would happen by uh 2000.
SPEAKER_00Oh, is that right? I didn't know that.
SPEAKER_01This paper was in 1952, uh so about half a century. And given the perspective from the the middle of the 1950s to year 2000, he was about right.
SPEAKER_00So my question was uh the Turing test, if you could explain it, what it really meant to people like you in in the midst of this, to me, it's too easy to pass. Uh it's not really the measure of AGI, it's it's a measure of a very clever piece of that. But I wonder why. I mean, is it something that he threw out there, or is or do you regard that as, boy, that was a brilliant way to assess where we are. And could you explain for everybody what it means?
SPEAKER_01Aaron Powell I mean, I have a bet with Mitch Kapoor that we would pass uh a sophisticated form of Turing test by 2029, also. Uh it's actually about as difficult as AGI to really match human beings and get all the c context, and when you pause and be convincing as a speaker.
SPEAKER_00So can you explain to people what the Turing test is? Because not not everybody will know it.
SPEAKER_01Uh you have human judges who are not enhanced watching uh both a computer and uh humans answer questions, and it can basically have a conversation like you and I are having now uh with both a computer and a human, but it's over teletype lines, so it can actually tell if they look like a human or not. We're getting pretty close. Aren't we there already, by the way? Well, people have used GPT-4 and passed a Turing test. He didn't actually define it very well. Uh it's like one page out of an essay, and there's different ways to interpret it. I felt there would actually be a five-year period where people would start to claim credibly that we've passed a Turing test, but not everybody would accept it, including me. I don't really accept it today. Uh I I will anticipate it by 2029. I think it'll might happen sooner, but it's got to get all the different kind of periods in which a person pauses and tries to convince people that they're really a human being. We're very close to that, but not quite there.
SPEAKER_00Aaron Powell So, Ray, given where we are right now, is the next big thing the humanoid robots, or is the next big thing further breakthroughs in in the power of AI to do things? Is it all of the above?
SPEAKER_01Well, right now you can get your large language model to almost answer anything. It's not quite there. Uh but recently there's been breakthroughs, for example, in doing higher math problems, so I can actually do math at the level of graduate students in math. Now, how many people can do answer questions that at the graduate level of math, but very few people can do that, and I can actually succeed. It's not as good as the very best mathematicians that have ever lived. Um but uh it will get there. Uh that's actually post-AGI. And also, I mean, I believe large language models should be called large event models, because they deal with more than language. For example, we use the same idea to control robots, but they're not dealing with uh language, they're dealing with the movements of robots. That's coming. I mean, lots of people are creating humanoid robots, and they can do actually pr pretty impressive things. The idea is actually make them affordable so they can actually, for example, clean up your kitchen table, which is actually not that easy. You gotta decide exactly what goes in the refrigerator and what goes in the garbage and deal with it intelligently. We we can do that, but it's coming over the next few years.
SPEAKER_00The answer should be no about what I'm about to ask, but I don't know. To me, when I chat with ChatGPT or ask Perplexity a question or DeepSeek, I can't believe it. I mean, I'm stunned by how good it is. And when I I like watching the scrolling of explaining the logic, you know, in uh some of these programs that they now tell you the steps. And it's pretty much impeccable, at least according to normal standards. So I'm amazed. Are you uh kind of shrugging your shoulders and saying I told you so, or at some level are you amazed?
SPEAKER_01Well, both. I mean I've expected it, but it's still quite remarkable to see it actually happening. Uh and it it shows that the kind of thinking that we do can be brought to computers, but we can use it. I mean, we've always used our technology, and it's made our ourselves wealthier. Our personal income on a per capita basis, the average income, is ten times more than it was a century ago. So literally, we had very little money a century ago. And of course, many things you couldn't buy at all. Uh there was no computers and no electronics, but things you could buy, uh our money was literally one tenth of what it w was today. And this is because of technology. And it's going to continue. When we fully merge with AI, uh, we're gonna have a great increase in our income. In about 20 years, we're gonna merge with it. We're not gonna be able to tell. The difference between ideas that come to us from an AI or a personal environment, and we're going to be a million times greater. And that's such a dramatic change that we borrow this metaphor from physics. The singularity in physics is a black hole where something goes in it and you can't really tell what's going on. Here we can't really tell what's going to happen with human beings when we're a million times smarter. But it it's good to be smarter. I sometimes imagine going to a mouse and telling it, well, gee, we can make you as smart as a human. Would you like to do that? And he might say, Well, I don't know. We might develop nuclear power. That could be dangerous. And but actually, we're we're happy that we're smarter. We don't want to be as smart as a mouse. We have dangers that technology's created, but we're able to overcome them, and we have a much better lifetime than we had uh hundreds of years ago.
SPEAKER_00Ray, you you wrote a book about how to create a mind and uh the secret of human thought revealed. And I'd like you to reflect on what you think the capacity of these large language models and deep neural networks in general are telling us about our own intelligence. Are they ultimately analogs of our intelligence or are they what we are?
SPEAKER_01Well, it's a it's the same thing, although it's organized a little a little bit differently. We have uh about 10 billion neurons. Each neuron has about a thousand connections. In order to simulate that, we'd have to do about a trillion calculations per second. That's what we're getting to. I mean, the the uh Nvidia chip is now half a trillion calculations per second per constant dollar. The things that are coming out now are trillions folds, so they're they have the right hardware capability to be able to emulate human intelligence. But it can actually be a lot of our intelligence is not really used. Some of it's used to control our organs and so on. So by the time we get to AGI, we'll we'll be actually much more intelligent. It can do everything that any human being can do. No human being can do that. I mean, Einstein could think great thoughts in physics, but he was and actually played violin, but he wasn't a very good violinist. So we'll be able to do everything that anybody can do today and go beyond that. And we're using it for creativity. So, for example, uh in uh medicine, we can actually try every diff different possibility and try it out. Moderna actually had several billion different calculations about what could be causing COVID, and using different AI techniques, they're able to solve the COVID vaccine in two days. Usually it takes years. We then spent 10 months uh just proving it, but it was it was actually done in two days. Uh and that's how we're going to create solutions to all different types of medical problems and other types of problems uh using AI.
SPEAKER_00Aaron Ross Powell We have plenty of uh friends from college days who spent most of their professional careers on a protein, uh, trying to understand how it folded.
SPEAKER_01Yeah. Well, Alpha Fold II found the shapes of 200 million proteins in 2022. Prior to that, through all of history, human beings had tried this, and we only did about 200,000 proteins. So this was a thousand times as much in one year. Uh, and then we've gone beyond that. Alpha Proteo can design new proteins that didn't exist before, uh, that can bind to a target protein. So we're using AI to solve many different problems that were impossible for us to solve before.
SPEAKER_00I uh once asked uh Gary Kasparov, the uh former world chess champion, about whether he had played uh Alpha Zero or whatever the chess version was called. And uh he he's not a modest man, he's a very nice person, uh, but modesty about his chess is not one of the attributes, uh, naturally, because he was the world's greatest. And he said to me, There's no point in my playing. It would be I would lose every time. It would be like me playing someone at a club. Uh, you know, the gap is so huge. Right.
SPEAKER_01But this is thinking of AI as something different from us. So I play and Alpha Go Zero plays, and it's me against the AI. Yeah. Uh but I think we're gonna bring it into ourselves, it's gonna be part of our capability, and that's gonna be a major change. That's possibly why we call it the singularity, because it's very hard to imagine that.
SPEAKER_00So let's turn to uh the future uh because you you spend a lot of time on jobs, longevity, well-being questions, and uh those are obviously of uh huge interest for everybody and huge interest for me as an economist. One version of this is essentially the machines will do everything, the robots will do everything, we'll do the things we want to do, leisure time, culture, whatever is the human part. Is that your thought? Or you're expressing something somewhat different about this, that it's not machines versus the people, it's gonna be all of us, but we're gonna be doing whatever we do differently. What will we be doing?
SPEAKER_01Well, think of AI again, something different, and they're doing all the work, leaving us with nothing to do, is thinking of us as different from AI. But AI isn't something that came from Mars that's uh an alien invasion of intelligent machines. We created them, and it has human thoughts in it. We're just finding a way to use it. I mean, we did that with books. I mean, I could read a book, it was slow, but I could actually absorb, you know, your thought uh in a book, and then I can articulate it. Uh, this is much more direct. Uh, it will actually appear in my brain, uh, but that's where we're going. So it's not gonna be us versus AI. AI comes from humans and it's gonna make us more intelligent, uh, ultimately a millionfold more intelligent, uh, and that's only 20 years away. But we're gonna be excited about it, just the way we think of our own thoughts.
SPEAKER_00Aaron Powell What do you think we are gonna do with all of that capacity and ability that we don't do now? One thing I can observe just to mention, you know, a point that I've made for a long time, which I think is right, it has been said that past automation, for example, uh which is uh you know a basic form of things we're talking about, didn't really eliminate work. Uh it just led to different kinds of work and different kinds of skills and so on. The fact is it it did eliminate a lot of what we call work, in that the time that people actually spend at work, whether it is their chores that is non-directly remunerated or in the workplace, really has declined remarkably. I mean, there is leisure time means something, actually. And the average American works about three hours, 15 minutes uh a day to support this uh very wealthy economy. So what will we do with our time, uh, whatever it is, however we're doing it?
SPEAKER_01If you look at our past, in 1800, 80 percent of our efforts was to create food. Now that's 1.4 percent, and yet we have more people working and are making 10 times as much as we did 100 years ago. I mean, it's gonna affect everything we do. Take, for example, longevity. Right now you go through a year and you use up a year of your longevity, so you're older. However, scientific progress is also advancing and coming up with new cures and new ways of treating disease. And right now you're getting back, if you're diligent, which I'm sure both of us are, you get back about four months. So you actually only age eight months. You lose a year, but you get back four months from scientific progress. However, the scientific progress is advancing exponentially. It's getting faster and faster. In about seven years, by 2032 approximately, you'll get back a full year. So you lose a year, but you get back a full year, so you really don't age. Past that point, you'll actually get back more than a year. So you you're not going to die from aging. It doesn't mean you won't die. You I mean, you could have an accident tomorrow, although we're making progress in accidents. Self-driving cars, for example, have virtually no accidents. Uh, once we uh convert all of our driving to self-driving cars, uh the 40,000 people we lose in the United States each year from driving accidents will be close to zero.
SPEAKER_00Trevor Burrus, Jr. Isn't it, by the way, interesting? Of course, this is uh uh the day the daily life, but if there's one accident in a self-driving vehicle, it's lawsuits, scandals, and so forth. If there are thousands of human errors of old people, grandmothers and so forth driving into things, no news. That's standard. Trevor Burrus, Jr.: Teenagers who go out and drunk or whatever.
SPEAKER_01Um they're very aware of that. So if there's a single accident, uh it'll create a lot of publicity. So they're really trying to make it be almost perfect. If you drive the Waymo cars in in San Francisco, there haven't been any accidents at all, and you see them everywhere.
SPEAKER_00Yeah, is that right? No accidents from Waymo so far. Yeah. Uh-huh. I didn't I didn't realize that. I thought our regulations would uh actually prevent this from happening in our cities. Uh well, it's spreading very quickly. Yeah, but some are now taking it up because it it does uh it does make sense. So one vision is that for a lot of people, there really won't be what to do because uh whether it is AI systems, but now if it's uh humanoid robots that uh embody this, so they do the physical as well as the uh as well as the mental process.
SPEAKER_01And as you say, we're gonna we're gonna have the AI within ourselves, so we're gonna be much more capable than we are today.
SPEAKER_00What does that mean within ourselves?
SPEAKER_01We'll be doing uh coming up with new uh cures for diseases, new music, uh new art, new novels. We'll be able to sing like our favorite pop star, go through all the things we that humans can do now. You'll be able to do all of it and much more being enhanced this way with AI within ourselves.
SPEAKER_00Aaron Ross Powell But within ourselves, uh you mean uh physical interface uh as well as I mean what do you mean actually? What what what is coming?
SPEAKER_01Aaron Ross Powell Well, the next thing is going to be virtual reality, which a number of uh companies are trying out. We haven't quite got there, but I think within a few years, uh we'll replace having to carry this. I mean, this is an inconvenience because I mean, how many times have you lost it? Like, where the hell is my phone? So it will be that's an hourly question. We'll be with you all the time. Uh it will actually we'll be able to actually tell what's going on in our neocortex. You don't have to interface with the entire neocortex. The neocortex is like a pyramid, and at the very top you've got the key ideas. That's the where you need to connect to. Right now you can do that with surgery, but you'll be able to do that without surgery by sending things that can sense the neocortex through the bloodstream. There's experiments of that, but that'll happen in the 2030s. So we'll actually merge with it. We won't actually carry all the AI with us. It'll go to the nearest cloud. I mean, this is actually not that intelligent unless it interfaced with the cloud. I mean, that's where it gets all of its intelligence. And cloud can be actually very uh secure. We use encryption, that's provincially unbreakable. So your thoughts will actually go to the cloud. There'll be AI there, it will be as if it's in your mind, but you'll be connecting to it through the cloud. And the cloud will be much more sophisticated than it is today. It's not gonna go down as much and so forth. Aaron Ross Powell, Jr. Ray, one basic question.
SPEAKER_00You have supported, I believe, over the years, about I don't want to put words in your mouth, but you've supported the idea that we're gonna be so efficient and proficient uh that everybody can be brought along, that there should be uh universal basic income and basic standards, which I would say as an economist makes complete sense to me, and I agree with that. It's not the way that we're going right now, uh somehow it seems. We don't even have that rhetoric at the moment. We're going in a different way. And just to put it in a not personal, but in a in a semi-personal way, Silicon Valley seems to be led by libertarians who say, we don't want anything of uh bringing everyone along. You're on your own and it'll be fine. How do you find that part of this story? I mean, if you if you look at economics, uh liberals want to go a little bit more advanced, and conservatives don't want to go quite as advanced.
SPEAKER_01But if you actually look at it, whether there's a liberal or conservative person or organization in charge of politics doesn't really seem to matter. I mean, a hundred years ago we had absolutely nothing from the government. Social Security was put in in 1935. Right. And that was actually very uh spare compared to today's Social Security. Today we have a million things that try to move it back a little bit. It doesn't really matter. Uh if you look at what's affordable, it moves along regardless of the economics of the person in charge. And we're gonna have to provide some kind of money somehow, otherwise we won't have a market, and that's not very good for economics.
SPEAKER_00Uh can you tell your colleagues out there in California, if you're out there, that this is reality? Because Well, it's gonna happen regardless. I hope you're right. You know, one of the things that happened, I attribute a lot of the changes of American society of the last 50 years to kind of the pre-AI but still digital revolution, the automation, the robotics, and so forth as as it occurred and the assembly lines. And it did leave people behind. We became much richer on average, but not much more generous, I would say, uh, and not really acknowledging I mean, we have things that we didn't have before.
SPEAKER_01I mean, now you can get a a wristwatch, you know, for less than twenty dollars.
unknownYep.
SPEAKER_01It gives you more capability than we could have like a year ago. I mean, this cell phone would cost billions of dollars in order to uh generate the computation it has in the 1960s, millions of dollars in the 1980s, and yet it costs, you know, a few hundred dollars. And everybody has it. And that wasn't true 15 years ago. So prices are we talk about inflation, but things are coming down dramatically, uh, and it's gonna affect everything we we're doing.
SPEAKER_00By the way, I agree that we measure everything wrong. We don't measure all of the free goods and services every time we look at a map or are able to find our friend quickly because uh all connected. And this would have been immeasurably impossible, and nothing in our national accounts or our GDP or our consumer price index remotely reflects any of this. So our systems of measurement have completely broken down.
SPEAKER_01Yeah, and we're gonna be that much more smarter as we go along. I mean, that's the whole idea of the singularity. It's driven by this exponential growth in both hardware and software. 85 quadrillion fold in the last 80 years, uh, just from hardware alone.
SPEAKER_00I'm with you. What I don't understand is why we're almost engulfed in a sense of panic right now rather than a sense of plenitude. So we have capacities that were unimaginable. Every child could be in school, everyone could be on healthcare, and yet we seem to be gripped with this uh almost sense of panic rather than uh what should be true. Hey, we just went through how many quadrillion uh uh increase of capacity. Why do you feel that is, by the way?
SPEAKER_01I mean, it's two different levels of looking at things. The kind of panic that people feel is really not real. I mean, compare ourselves today to 80 years ago. We had a hundred million people killed. I mean, we're concerned now about wars that kill thousands of people, and that's terrible. And I don't want to see that. But compare that to 80 years ago, we had literally a hundred million people in Europe and Asia lost their lives, plus many more injured. We don't have problems like that anymore.
SPEAKER_00Although I think one of the real issues, and maybe it it is an appropriate question also for you, which is that one of the reasons people are nervous about the wars is that we have the capacity to end, because of technology, we have the capacity to end humanity. It's it's not a not a myth. And so there's anxiety about something which was also unimaginable even during the period of those massive wars. If I turn that to uh AGI, we hear from Elon, for example. And so, you know, people who have a lot of credibility in these things. This is terrifying, this is the end of the world, this could end the world. Lots of dystopian talk, not so much clarity about what is their dystopia exactly. Is it runaway machines? Is it autonomous weapons? Is it the machines launch World War III? Is it uh something else? What what do you make of that kind of seems to be real fear about AI?
SPEAKER_01You might remember when we were in grade school, we would have these nuclear war protections. We'd get under our desk, we put our hands behind our head to protect us from nuclear war. Uh it seemed to work. Um we were not blown up by nuclear war. Uh in fact, if you were to ask people after the two atomic bombs that we used in anger, what's the likelihood we're going to go the next 80 years without this happening again, people would have said zero.
unknownYes.
SPEAKER_01And yet it never happened again. And we have actually people in charge of nuclear wars that were not enthusiastic about, and yet we have not had that problem again. It doesn't mean we're free from it. And yes, you can imagine AI advancing in ways which are dystopian. So far, though, uh it's actually been beneficial. There are some problems. AI clones can pretend to be a human being, and you're not really sure of it. And I mean, there's there's issues we have to deal with. Uh the ability of AI to protect us from the abuses of AI are also advancing. So sometimes people describe things that will come, and we imagine that we don't have ways of protecting ourselves, but the ways of protecting ourselves will also advance. The question is, well, do we want human beings to have more intelligence? And I sometimes go back to asking a mouse, would you like to have intelligence? And they might say, well, no, I think it's better to say a mouse, but uh I think it's actually good that we've advanced in intelligence, and this is the next stage of our being able to do that.
SPEAKER_00Aaron Powell Ray, what would you like us? And I will leave it to you to define us, whether it's citizens, whether it's the government, whether it is Google. What would you recommend to do that isn't happening right now to get the best outcome of all of this?
SPEAKER_01Well, uh I'm pleased with the way it's going. First of all, we have competition. There's lots of different companies that can put out things that that uh have an impact on the market. Uh I would prefer actually to use AI from a large company and the number of large companies doing this because they are are protective of their reputation as well as their liability. You don't want somebody going saying, well, okay, you've got this disease, you should take this, and turns out to be a problem. So large companies are very exercised about being able to give more responsible advice. Doesn't mean everything is perfect, but uh I mean you've used it. It's actually pretty good. And we use it because it's beneficial, and it's gonna get even more beneficial, and that's good. So I I like the way we're going with a lot of companies advancing this. I th I think that's beneficial.
SPEAKER_00Is there anything uh you would add or you would tell our friend David Sachs in the White House right now, please do this so that we stay on the right course?
SPEAKER_01I mean, I I think there is a tremendous amount of liability of large companies. They're very, very careful because AI can advise you on anything. And so they're they're really just because of the liability and reputation, very motivated to give the right advice. There's a lot of issue on the reliability of online information. I think a lot of this comes from social networks. I think social networks are great, but it's actually created by people who are not enhanced, and not all of it's very reliable. More and more people are actually listening to AI, which is actually beneficial because AI is actually more reliable than social networks.
SPEAKER_00Yeah, it by the way, it gives you both sides of an argument, it describes it, and it can give you a list of evidence too.
SPEAKER_01Yeah. And it can give you citations for the evidence as well.
SPEAKER_00Yeah, I think extremely helpful. Let me uh close with one final question. I think I may know your answer to this because it's consistent with uh what you've been writing. But when we have AGI or when we have AGI plus, will we have machine consciousness? And will that make a difference for us in a fundamental ethical way as well?
SPEAKER_01I mean, that's a philosophical question, which is not to say that it's not a real question, but there's no scientific way to uh answer that. I believe that you're conscious. Uh not all human beings seem conscious, uh but uh if they're engaged in discussion with me, I believe they're conscious. We can actually do that now with machines. Are they conscious? It's really hard to say. People tend to say no, but to actually use it as therapists and so on, and that's just beginning. It's gonna become even more the case in the future. So I think we will uh agree that they're conscious, but there's actually no way of scientifically proving that. We go outside of shared human experience, like for example, animal rights. Are animals conscious? I mean, I believe my cat is conscious, but not everybody agrees to that.
SPEAKER_00I think they haven't actually met my cat, but um I I bet your cat is, uh, would be my guess.
SPEAKER_01But uh I I think they will be conscious, but it's it's really a philosophical issue. Uh and and we'll be talking to other people, and part of their intelligence will be AI, and we won't be able to tell the difference. So we we're gonna accept that they're conscious.
SPEAKER_00Ray, thank you very, very much. I I know there will be uh many books forthcoming. One of them will be called The Singularity Is Here, uh maybe 2029, maybe even uh even faster. So that's uh that's one that we're going to be reading soon. Thank you for your remarkable vision and helping us to to look ahead. My pleasure to talk to you, Jeff. Yeah, really uh most grateful. So we've been speaking with the remarkable Ray Kurzweil uh about his book, The Singularity is Nearer when we merge with AI. And the theme of merging with AI to strengthen our own intelligence and our species intelligence uh is the underpinning of uh Ray Kurzweil's vision. You've been uh listening to Book Club with Jeffrey Sachs in conversation with Ray Kurzweil. Ray, thank you again for being with us and uh hope to see you soon. Thank you. My pleasure. Thank you for joining in the conversation. Please subscribe, rate, and write a review on whatever platform you listen for your podcast.
