The Singularity Is Nearer featuring Ray Kurzweil | SXSW 2024

Alright, I'm really excited to be here with you Ray, it's great to be here, it's great to see everyone together, yeah, beautiful audience, so what I like most about that presentation of yours is that you've been working in AI for the longest time. than any other living human being. It means that if you live forever and we get to that, you will always have that distinction. I think it s true. Marvin Minsky was actually my mentor. If he were alive today, he would actually be over 61 years old, so he's us. We're also bringing it back, so maybe I'm not sure how we're going to count the distinction, so okay, we're going to fix the audio, but this is what we're going to do with this conversation that I'm going to start.

Asking Ray some questions about where we are today, we'll do that for a few minutes and then discuss what needs to happen to reach the

singularity

over the next 20 years and then discuss what the

singularity

is. what does it mean how does it change our lives and then at the end we will talk a little bit about how if we believe in this vision of the future what does it mean for us today ask your questions they will come I will ask them as they go in the different sections of the conversation but let's go with Kraken, what? can you hear me?

You can't hear Ray. Well, this will be recorded. You will all live forever. There will be plenty of time. Everything will be fine. I'm just going to start. I guess the audio will be fixed. They do a fabulous job here at South. I think they should be able to hear me and that's okay. We have fixed this with correct audio. Engineers, are we okay? to go, we're ready to go, okay, first question, Ray, so you've been working in AI for 61 years. Wait, can you hear me? It's not so that everyone in front can hear you, but no one in the back can hear you.

Listen to me now, okay, I'll speak louder, first question, so you've been living in a revolution for a long time, you've made many predictions, many of which have been remarkably accurate, we've all been living in an extraordinary situation. two-year transformation with large language models a year and a half what surprised you about innovations in large language models and what happened recently. Well, I finished this book a year ago and I didn't really cover a big language model, so I delayed the book to cover that, but I was hoping for this to happen a couple of years later.

I mean, I made a prediction in 1999 that it would happen in 2029, and we're not there yet, but I'll do it, but it looks like it's maybe a year or two ahead of schedule, so maybe there was a bit of a surprise. . Did you predict in 1999 that a computer would pass the Turing test in 2029? Are you checking that out for something closer? Not to this day I still say 2029 uh the definition of the spin test is not precise we are going to have people claiming that the TR test has been solved and people saying that gp4 actually passes it uh some people, so it's going to be about two or three years where people start affirming and then they continue affirming and eventually everyone will accept it, so it's not like it happens in one day, but you have a very specific definition of the test of Turing, when do you think we will do it? passes that definition well, the shift test is actually not that significant because that means that you can, uh, a computer will pass for a human being and what's much more important is automatic general intelligence Agi, which means that it can emulate to any human being, so you have a computer and it can do everything that any human being can do and that's also 2029, everything happens at the same time, but no one can do that, I mean just take an average language model big today, you can ask him anything, uh and he will answer you quite convincingly, no human being can do all that and he does it very quickly, he will write a very nice essay in 15 seconds, and then you can ask again, he will write another essay and no human being can. it actually performs at that level, so you have to simplify it to have a convincing spin test.

To have a spin test, you have to simplify it. Yes, let me ask the first question from the audience, as I think it is quite relevant. where we are, who is Brian Daniel, the Curtz curve is still accurate, I repeat, the Curtz wild curve is still accurate, yes, in fact, you can see we uploaded the slides on the first slide, so this is a history of 80 years, this is exponential growth a straight line on this curve means exponential uh curvature uh if it were somewhat exponential but not quite it would curve this is actually a straight line uh it started with a computer that did 0.007 calculations per second per constant dollar that is the bottom left In the top right corner there are 65 billion calculations per second for the same amount of money.

That's why large language models have only been viable for two years. PRI, we actually had big language models before that, but it didn't work very well. Well, and this is an exponential curve, technology moves on an exponential curve, we see that, for example, having renewable energy coming from the sun and the wind, it is actually an exponential curve, it is in, it has increased, it has disappeared, uh, we have decreased. the price by 99.7%. We've multiplied the amount of energy coming from solar a million times, so this kind of curve really drives all kinds of technology. And this is the reason we are making progress.

I mean, we. I knew how to make great language models years ago, but we depend on this curve, um, and it's pretty amazing, and it started by increasing the speeds of relays, then vacuum tubes, then integrated circuits and each year it achieves the same amount of progress approximately, uh, regardless of where. you're on this curve, we just added the last point, and it's again, we basically multiply this by two every 1.4 years, and this is why computers are cool, but it actually affects all kinds of technology and we just added. last point about two weeks ago, okay, so let me ask you a question.

You know you wrote a book on how to build a mind. You have a lot about how the human mind is constructed. There are many advances being made in AI systems. built on what we understand about neural networks, very clearly, our understanding of this helps with AI over the last couple of years by looking at these big language models. Have we learned something new about our brains? Are we learning about the inside of our skulls while doing this? It really has to do with the number of connections, the brain is actually organized quite differently, things near the eye for example deal with vision, and we have different ways of implementing different parts of the brain that remember different things You don't really need that MH in a big language model, all the connections are the same, we have to push the connections to a certain point if it roughly matches what the brain does, which is about a trillion connections.

It will work like the brain. we're almost at that point wait so gp4 is 400 billion uh the next ones will be a trillion or more so building these models they are more efficient in building them than our brains we make them to be as as efficient as possible, but it doesn't really matter how they are organized MH and we can create certain software that will further expand the number of connections for the same amount of computation, but it really has to do with how many. The connections are what a particular computer is responsible for, so as we approach AGI, we're not looking for a new understanding of how to make these machines more efficient.

The Transformer architecture was clearly very important. We can really get there more, but software and learning is important too, I mean you could have a billion connections, but if you didn't have something to learn from it wouldn't be very effective, so we have to be able to collect all of these data, so we do it on the web and So I mean we've been collecting things on the web for several decades. That's really what we depend on to be able to train these big language models and we shouldn't really call them big language models. because they deal with a lot more than language, I mean, it's language, but you can add images, you can add things that affect diseases that have nothing to do with language, in fact, now we're using simulated biology to be able to simulate different . ways to affect the disease um and that has nothing to do with language so they should really be called large event models.

Do you think there is something happening inside our brain that cannot be captured by calculation and mathematics? No, I mean, what would that be? I mean, okay, quick audience poll, raise your hand if you think there's something in your brain that can't be captured by computing or math like a soul, okay, so convince them they're wrong. I mean, awareness is very important, but it's actually not scientific there's no way you can swipe someone and the light will turn on oh this one is conscious no this one isn't uh it's not SCI it's not scientific but it's actually extremely important uh and another question why am I how come what's wrong with me am I aware and I'm not aware of what's happening to you uh these are deeply mysterious things but they're not really not aware so Marvin Minsky, who was my mentor for 50 years, said that he is not scientific and therefore we should not bother with it and any discussion of Consciousness he would dismiss, but in reality he did.

His reaction to people depended entirely on whether he felt they were conscious or not, so he actually used that, but it's not something. that we're ignoring because there's no way to know if something is conscious uh and that's not just something we don't know and we'll find out that there's really no way to know if something is conscious or not, what do you know? saying like that is not conscious and you know, the gentleman sitting there is conscious, well how do you prove that? I mean, I mean, we agree with humans that humans are conscious, some humans are conscious, not all humans, um, but what about animals? we have a big disagreement some people say that animals are not conscious uh and other people think that animals are conscious maybe some animals are conscious and others are not there is no way to prove that Okay, I want, I want, I want to run for this consciousness . question, but before I do that I want to make sure I have correctly understood your previous answer so that the feeling I have of being in love or the feeling of whatever emotion I have can eventually be represented in mathematics in a large language model.

Yes, I mean certainly the behavior, the feelings that you have if you are with someone you love, it definitely depends on what the connections do, you can tell if that is happening well or not, and going back to, is everyone here convinced, not quite right, close enough? You don't think it's worth trying to define consciousness. I mean, you spend quite a bit of time in your book giving different arguments about what consciousness means, but it sounds like you're arguing on stage that we shouldn't try to define it. way to show it, I mean, we have certain agreements.

I agree that you are all aware that you actually came to this room, so that is a very good indication that you are aware, but that is not proof, and there may be humans. beings that don't seem very conscious at the time are either conscious or not and animals, I mean I think elephants and whales are conscious, but not everyone agrees with that, so at what point can we then determine how much How long will it be until we can? essentially downloading the entire contents of your brain and expressing it through some type of machine which is actually an important question because we're going to talk about longevity, we're going to get to a point where we have a longevity escape velocity and it's not that far away, I think that if you are diligent you will be able to achieve that by 2029, that is only five or six years from now, and that right now you go through a year, you use one year of your longevity, but recovering from scientific progress at this moment about four months, but that scientific progress is on an exponential curve that will accelerate every year and by 2029, if you are diligent, you will consume a year of your longevity with one year. passing but you will return a full year and after 2029 you will return more than a year, then you like to go back in time, that is not a guarantee of infinite life because you could have a 10 year old child. and you could estimate his longevity as many decades and he could die tomorrow, but the important thing about capturing everything in your brain we can't do it today and we won't be able to do it in 5 years, but you will be. able to do that because of The Singularity which is 2045 five and then at that point you can go into the brain and capture everything that's there.

Now your thought will be a combination of the amount you get from the calculation, which will be added to your thought and that is automatically captured, I mean right now, anythingthat you have on a computer is automatically captured today and the kind of additional thinking that we will have by adding it to our brain will be captured, but the connections that we have in the brain that we started with we will still have that is not captured today, but it will be captured in 2045. We will be able to enter the brain and capture it as well and therefore we will actually capture the entire brain, which will be backed up, so even if you are erasedWhen you enter towards a bomb and it explodes, we can actually recreate everything that was in your brain.

By 2045, that is one of the implications of the singularity. Now, that doesn't guarantee absolutely because, I mean, the world could explode and all the uh, the computer, uh, all the things that contain computers could explode and you still wouldn't be able to recreate it, so we'll never get to a point where the one that we absolutely guarantee that you will live forever, but most things. things that right now would bother capturing that, it will be surpassed at that time, there are a lot of things there Ray, um, let's start with escape velocity, so do you think that anyone in this audience in their current biological body will live to be 500 years old? wondering, yeah, absolutely who I'm talking about, if you're going to be alive in 5 years and I imagine all of you will be alive in 5 years, not five, okay, if you're alive for five years, you'll probably live up to 500 years. old man if they're diligent and they think the people in this audience will be diligent, so, well, you can drink whatever you want as long as you don't get run over tonight because you don't have to worry about declining everything.

Well, let me ask a question that I want to, I want to understand, we're going to spend a lot of time on what the singularity is, what it means and what it's going to be like, but I want to ask some questions that take me away. We're up there, so I'm going to take this question from Mark Sternberg and modify it slightly to the time frame that AI will be able to do or that computers sophisticated enough in your argument can do everything that the human brain can do, what won't they do? ? will be able to do in the next 10 years, well, it is one thing to do with being creative and some people believe that they will be able to do everything that a human can do, but they will not be able to create new knowledge, that is actually wrong. because we can simulate, for example, biology and the Madna vaccine, for example, we did not do it in the usual way, that is, someone sits down and thinks well.

I think this might work and then try it. It takes years to prove it. and several people uh and it's one person's idea of ​​what might work, they actually listed everything that might work and there were actually several billion different mRNA sequences and they said let's try them all and they tested each one of them by simulating biology and that took two days. So one weekend they tested several billion different possibilities and then chose the one that turned out to be the best and which was actually the Mna vaccine. Until today, they have now tested it on humans.

We will also be able to overcome that because we will also be able to test using simulated biology. In fact, they decided to try it. It's a little difficult to stop testing on humans. We will do it so you can try each one. you just pick the best one and then you can test it by testing on a million simulated humans and do it in a few days too and that's actually the future of how we're going to create drugs for diseases and there's a lot of the things that are happening now with the cancer and other diseases that are using that so it's a whole new method that's actually starting now it started with the Mna vaccine we made another cure for a mental illness that's actually now in stage three trials that's how we'll create drugs , from now on, but what are the borders?

What can't we do? That's where the computer is being creative and it's not just trying something it comes up with, it's making a list. of everything that is possible and everything is creativity or it is just brute force with maximum capacity, it is much better than any other form of creativity and yes, it is creative because you are trying all the possibilities and you are doing it very quickly. and you come up with something that we didn't have before, I mean, what else would be nice creativity? So, let's cross the border of creativity. What are we not going to cross?

What are the challenges that will be pending? Well, the next 10 years we don't know everything and we haven't gone through this process. It takes some creativity to imagine what might work and we also have to be able to simulate it in a biochemical simulator so we actually have to figure it out and we will be using people for a while to do that so we don't know everything I mean, being able to do All a human being can do is one thing, but there are so many things we don't know. I know we want to find out and that requires creativity which will require some kind of human creativity working with machines.

Well, let's get back to what's going to happen to get us to The Singularity. So clearly we have the graph that you showed. computing power has been very constant, you know, moving up, you know, on a logarithmic scale, in a straight line, there are a couple more elements that you think are necessary to get to the singularity, one is the emergence of the Nanobots and the other is the rise of brain-machine interfaces and both have been slower than AI, so convince the audience that it would be slow because every time you perfect the human body, a lot of people will worry about it, yeah do something with computers we have a new algorithm or we increase their speed um no one is really worried about that you can do that no no one cares about the dangers in it uh I mean that's the reality there are some dangers that worry people yes Yes , but it goes very, very fast, that's one of the reasons why it goes so fast.

But if you're affecting the body, we have all kinds of concerns that it could negatively affect the body and that's why we want to test it on people, but. but the reason brain-machine interfaces haven't moved on an exponential curve isn't just because you know a lot of people are worried about the risks to humans. I mean, as you explain in the work of the book, they just don't work that well. How could they, if we could test things without having to test it, it would go a lot faster, I mean, that's the reason it goes slowly, but now it's thought that we could find out what's going on inside the brain and put it. things in the brain without actually going into the brain, we wouldn't need something like a brain link, uh, we could just, uh, I mean, there are some tests where we can tell what's going on in the brain without actually putting something inside the brain and That might actually be a way to do this much faster, but your prediction about the singularity depends, maybe I'm reading this wrong, not only on the continued exponential growth of computing, but also on solving this particular problem correctly, Yes, because we want to increase the amount of intelligence that humans can control and that's why we have to be able to match the best computers with our real brain and why do we have to do that?

Because right now, here I go, I have my phone somehow, this increases my intelligence, it's wonderful. but it's very slow, I mean, if I ask you a question, you're going to have to write it or speak it and it takes a while, I mean, I ask a question and then people play with a computer, it can take 15 seconds or 30 seconds it's not like it will just go directly into your brain uh I mean they are very useful these are brain extenders that we didn't have a while ago uh usually in my talks I ask people who have their phone here, I bet There might be one or two people here, but everyone here has their phone.

That wasn't true five years ago, it definitely wasn't true 10 years ago and it's a brain extender, but it has some speed issues so we want to increase it. speed, a question could come up where we're talking and the computer would instantly tell you what the answer is without having to mess around with an external device, and that's almost doable nowadays, and that and something like that would be useful. do this, but couldn't you get a lot of the good stuff you talk about if we kept the problem of connecting our brains to machines? Suddenly you're in this world with complicated privacy issues where things have been injected into my brain.

What's going through my brain is that you know it's going somewhere else, like you're opening up a whole bunch of ethical, moral, existential problems. Can't you just upgrade phones? Well, that's the idea that we can do it without having to go inside. your brain, but being able to know what's going on in your brain externally without going into the brain, you know, with some kind of device, okay, let's keep moving into the future, so moving into the future, we have exponential growth on computer, we figure out a way for you to know, ideally, figure out how to communicate directly with your brain to speed things up.

Explain why Nanobots are essential to your vision of where we are going. If you really want to know what happens inside the brain, uh. You have to be able to get down to the level of the particles in the brain so we can know what they are doing and that is doable. We can't actually do it, but we can show that it is feasible. and that's a possibility, we actually hope that you can do this without affecting the brain at all, okay, so we go ahead, we have Nanobots running inside our brain and they are understanding it. out of your head they're pulling thoughts out, they're putting thoughts in, let's get to this nice question that Louise kraa charmingly answers, what are the top five ethical questions that we're going to face while that's happening? um, four is enough four four okay, there could be It could even be six rays, but you can give us four.

I mean, we're going to have a lot more power. Yes we can really with our own brain control computers. Does that give people too much power? I also want to say that we are talking now. about having a certain amount of value based on your U in your talent, this will give talent to people who otherwise wouldn't have talent and talent won't be as important because you will be able to gain talent by simply merging with the right type of large language model or whatever we call them um and Al also seemed a little arbitrary why would we give more power to someone who has more talent because they didn't create that talent, they just have it um and What everyone says is we should give more power to someone who has talents in an area.

This way, you could play with Talent, just like in The Matrix, you could learn to fly a helicopter, simply by downloading the appropriate software. and we're supposed to spend a lot of time doing that, is it fair or unfair? I mean, I think that would fall into the area of ​​ethical challenge, and it's not like we get to the end of this and say, okay, this is ultimately what the singularity represents. It's about and people can do certain things and they can't do other things, but it's over, we'll never get to that point, I mean, this curve will continue, the other curve will continue indefinitely, and we've actually shown, for example , with nanotechnology, we can create a computer where a one-liter computer would equal the amount of energy that all human beings have today, as if 10 to 10 people could fit on a ladder, the computer, that creates ethical problems. some, I mean a lot of the implications go against what we've been assuming about human beings, wait for the talent question, which is super interesting, do you feel like when we get to 2040 everyone will have the same capabilities?

I think we will be more different because we will have different interests and you might like some fantastic kind of music and I might be interested in some kind of literature or something else. We will have different interests and therefore we will excel at certain things depending on what your interests are, so it's not that we all have the same amount of power, but we will all have fantastic power compared to what we have today and if you are in Texas, where there are no regulations, will probably get it first instead of you in Massachusetts exactly, yeah, let me ask you another ethical question while we're at it, so a few minutes ago you mentioned the ability to replicate someone's brain and bring them back, so let's say I do that.

My father passed away six years ago, unfortunately I bring him back and I can create a mind and body like my father. It is an exact and perfect replica of all his thoughts, what happens with all the bills he owed when he died. because that's a lot of money and a lot of bill collectors call me, do we have to pay them or are we okay? Well we are doing something like that with my daughter and you can read about it in her book and it is also In my book we collected everything my father had written, he died when I was 22 so he has been dead for over 50 years , and we put it into a big language model and basically asked the question, uh, of all the things you've ever written. what is the best answer to this question and then you can ask any question you want and then you could talk to him and you would tell him something.Then you go through everything that he had written and you find the best answer that he actually wrote to that question and it was actually a lot like talking to him, you could ask him what he liked about music, he was a musician, uh, he actually liked M's the best um and it was very similar to talking to him, uh, and I report it in my book and Amy talks about this in her book and Amy actually asked the question could I fall in love with this person even though I've never met them and Does she do a pretty good job?

I mean, you really fall in love with this character. that she believes um even though she never met him um so with today's technology we can do something where you can actually emulate another person and I think as we go forward we can do it more and more responsibly and more and more that really matches with that person and really emulate the way they would move and so on in their tone of voice, well, you know, my dad loves Brams too, particularly those piano trios, so if we can solve the problem of back taxes , we will get my dad's Bots and your dad's Bots. hanging out would be cool, well yeah that would be cool, okay we have 20 minutes left.

I want to get to what I want to understand most because it's something that, by the way, this book is wonderful. I think you are. everyone will receive signed copies when it comes out, it is truly remarkable as our complete books, whether you agree or disagree, will definitely make you think more. One of the things I don't think you do in this book is describe. what will it be like one day in 2020 2045 when we are all much smarter then it is 2045 we are all a million times smarter I wake up I have breakfast or I don't have a good breakfast the answer to that question is kind Same as it is now, but first of all, The reason it's called The Singularity is because we don't really fully understand that question.

The Singularity is borrowed from physics. The singularity in physics is where you have a black hole and there is no light. can escape and therefore you can't really know what's happening inside the black hole, so we call it a singularity, a physical singularity, so this is a historical singularity, but we borrow that term from physics. and we call it singularity because we can It doesn't really answer the question if we actually multiply our intelligence a million times. How is that? It's a bit like asking a mouse. What would it be like if you had this person's amount of intelligence?

The mouse really wouldn't have it. even understand the question uh he has intelligence he has a good amount of intelligence but he couldn't understand that question he couldn't articulate an answer uh that's kind of what it would be like for us to take the next step in intelligence by adding all the intelligence that the singularity would provide wait, I just want to make sure I understand but, but I'll give you an answer. I said if you're diligent you'll achieve longevity escape velocity at F in five. or six years um and if we really want to emulate everything that happens inside the brain uh, let's go out a few more years, say 2040 245 now there are a lot of things where you talk to a person who has all the connections that they originally had plus all the extra connections we add by allowing them access to computers and that becomes part of their thinking, so can you assume that person explodes or something happens to their mind?

You can definitely recreate everything that is. a computer origin because we do that now every time we create something with the computer it gets backed up so if the computer goes away you have the backup and you can recreate it it says ok but what happens with your thinking in your normal brain that hasn't finished? computers um we don't have some ways to support that when we get to the singularity by 2045 we will also be able to support that uh because we will be able to find out that we will have some ways to actually find out what's going on in that uh in that kind of non-biological brain not mechanical um and so we can backup both his normal brain and the computer brain plus um and and and I think that's feasible by 2045 in Your vision of it, so you can backup his entire brain now, that does not guarantee.

I mean, the whole world could blow up and lose all the data centers, so it's not an absolute guarantee. It would be a shame, um, what I do. I don't understand if we will be completely different people if we share memories and we all upload our brains to the cloud and receive all this information directly to our neocortex. Are we still different? uh, yes, but we could also find new ways to communicate, so that the computers that extend my brain interact with the computers that extend your brain. We could create something that is like a hybrid or not, and it will be up to our own decision whether to do so or not. so there will be new ways of communicating let me ask you another question about this this is what when I was reading the book this is where I got stuck you were extremely optimistic right you are optimistic about where we are today optimistic that technology has been a massive Force to the good You're optimistic that it will continue to be a massive Force for Good, but there's a lot of uncertainty going forward.

What you described well, first of all, I'm not necessarily optimistic about things that can. go wrong uh we had things that can go wrong before we had computers um when I was a kid Atomic weapons were created and people were very worried about an atomic war, we actually went under our desk and put our hands behind our head to protect us against an atomic war and it seemed to work, we're actually still here, but if you ask people, we actually had two weapons that exploded in anger and killed a lot of people in a week and if you ask people, which one is the possibility that we will go another 80 years and this will never happen again?

Nobody would say that's true, but it's happened now, that doesn't mean it's not going to happen next week, um, but anyway, that's a big danger and I think it's a much bigger danger than computers uh yes there are dangers but computers will also be smarter to avoid types of dangers um yes there are some bad people in the world but I mean go back 80 90 years um we If 100 million people died in Asia and Europe because of World War II, uh, we don't have wars like that anymore, we could and we certainly have the atomic weapons to do that, um, and you could also imagine that computers could be involved with that, um, but if you actually look and this happens directly, warn Pece, first of all, if you look at my lineage of computers ranging from a small fraction of a calculation to 65 billion, that's a 20 quadrillion fold increase we've achieved in 80 years and you look at This personal income of the US is made in constant dollars, so this has nothing to do with inflation, uh, and this is the average income in the United States, multiplied by about 100 times, and we live much more successfully if really People say, "Oh, things were great 100 years ago, they weren't, and you can look at this graph, most of them.

I have 50 graphs in the book that show the kind of progress we've made, the number of people living in poverty. Extreme poverty has decreased dramatically and we actually did a survey where they asked people living in poverty, Has it increased or decreased? 80% said it has increased, but the reality is that it has actually decreased by 50%. In the last 20 years, what we think about the past is actually the opposite of what happened. things have gotten much better than before and computers are going to make things even better. I mean the kind of things you can do now with a The big language model didn't exist two years ago Do you ever worry that you take it as a fact?

Computers have improved things. Do you ever worry that personal income will continue to rise? big difference in the past. I mean, talk about what effect the railroad had. I mean, a lot of jobs were lost or even the Cuton Jny ​​happened 200 years ago. and people were very happy making money off cotton Jenny and suddenly that disappeared and the machines were doing that and people say, well, wait until this starts, all the jobs will be lost and that's actually what was said at that time, but actually the income was gone. we increase more and more for every person we work for that we create and if you say well, what are they going to do, you couldn't answer that question because it was in industries that nobody had a clue about, like all the electronics, so things are they're getting better, even if jobs are lost now, you can certainly point to jobs like computer programming, um, Google has, I don't know, 60,000 people programming computers and many other companies, at some point, that's not going to be a feasible job.

They can already code many language models they can write code not in the same way that an expert programmer can, but how long will that take? It is measured in years, not decades. um, however, I think things will get better because we eliminate jobs. but we create other ways to have income, uh, and if you really point out something, let's say this machine and the one being worked on can wash dishes, you just have a bunch of dishes, it will pick the ones that need to go in the dishwasher. and clean everything else and that will wash the dishes for you.

If we wanted that not to happen, we would say, well, this is a little disturbing. Let's get rid of it. It's not going to happen and no one would recommend it. We will find things to do, we will have other methods of distributing money and it will continue with these types of curves that we have already seen. It's kind of notable that we had big language models before we had robotic dishwashers. You have grandchildren, do you know what you would say to a young person? Do you know they agree? Do you know how you would tell them to better prepare for what will be, if you're right, a markedly different future?

I would be less worried about what will make money and much more worried about what turns them on. They love video games, so they should learn about that. They should read literature that excites them. Some of that literature will be created in the future. by computers but um and find out what in the world uh it has a positive effect on your mental being and if you know that your child or your grandchild this gets to one of the questions that are asked on the screen here if you know that someone is going to live hundreds. of years, as you predict, how does that affect the way it certainly means they shouldn't retire at 65, but what else does it change about the way they should think about their lives?

Well, I talk to people and they say well, I wouldn't want to live past 100 or what or maybe they're a little more ambitious in saying I don't want to live past 110 um, but if you really look at when people decide that they've already I've had enough and I don't want to live anymore, that never happens unless these people are in some kind of terrible pain, they're in pain physically, emotionally, spiritually or whatever, and they just can't stand to be alive, no one takes their life anymore. that and if we can really overcome many types of physical problems, cancer eliminated, etc., which I hope will happen, people will be much happier to live and will want to continue experiencing tomorrow. and tomorrow it will be better and better uh this kind of progress is not going away um so people will want to live um you know unless they are in terrible pain but that's what the whole medical profession is about, which is The computers of tomorrow will look enormously amplified.

Let me ask you a big question that appeared on the screen. This is from Colin McCab. AI is a black box, no one knows how it was built. How do you prove that AI is trustworthy to the users you want? Trust it, embrace it and accept it, especially if you are going to upload it directly to your brain. Well, it's not true that no one knows how they work. Most people who use a large language model don't know what data meaning goes into. There are things that happen in the Transformer layer that even The Architects don't quite understand, but we're going to learn more and more about that and actually how computers work will be, I think, a very common type of talent that people want to earn um. and ultimately we will have more confidence in computers, I mean, a lot of perfect AR language models, so you can ask it a question, it may give you something wrong, um, I mean, we've seen it recently, uh, the reason.

We have that these computers give you incorrect information. First of all, they don't have the information and they don't really know what they don't know and that's something we're working on, to get it right. I don't know that, that's actually very good if you can actually say that because right now you're going to find the best you know and if you've never been trained on that information and there's nothing there that tells you, it's just going to give you You're the best guess, which might be very incorrect, and we are actually learning to be able to determine when he knows and when he doesn't know, but in the end we will be quite confident when he knows and what he doesn't.I don't know and we can actually trust what he says, so your answer to the question is A, we will understand more and B, they will be much more reliable, so it won't be that risky to not understand them well, okay?

You've spent your life making predictions, some of which, like the Turing test, you've stuck to and been remarkably accurate. As you go from overwhelmingly optimistic to somewhat pessimistic, what is a prediction? Well, my books have always had a chapter. about how these things can go wrong and the dangers, tell me a prediction that you're pondering right now, but you're not sure if you want to fulfill it or not, um, I mean, there's something good... Known dangers in nanotechnology, if anyone outcreate a nanotechnology that replicates well known is if it replicates everything in paper clips uh turns the entire world into paper clips uh that wouldn't be positive no and and you unless you're Staples but then and and that's feasible uh take someone who is um It's a little bit mental to do that, but it could be done, and we'll actually have something that actually prevents that, so we'll have something that can detect that this is actually turning everything into clips and destroy it before it does. that um but I mean I have a chapter in this new book The Singulari Is Closer um that talks about the kinds of things that could happen oh the most notable part of this book is that it does exactly the math on how long it would take. to the Nanobots to turn the world into gray goo and how long would it take for the blue goo to stop the gray goo is an extraordinary book coming out soon you definitely need to read to the end but this leads to maybe let me try to answer The question we asked before is what young people should think about and what they should work on.

You said: your passions and what excites you. Shouldn't they be thinking about how to design and design these future systems so that they are less likely to make us great? I don't know if everyone wants to work on it, but the people in this room with a proper technological mindset, they should all work on not becoming greu, right, that will be on the list, yes, but that leads to another question which is. What will be the role of humans in analyzing that problem when they are only one millionth, one billionth, or one billionth as intelligent as machines?

I say that again, so we're going to have these really difficult problems to solve, yeah, right now. We are together with our machines. You know we can be extremely intelligent, but in 10 years. Within 15 years there will be machines that will be much more intelligent than us. What will be our role? What will be the role of humans in trying to solve first? After all, I see them as extensions of humans and we wouldn't have them if we didn't have humans to begin with and humans have a brain that can think about these things and we have this thumb, it's not really appreciated very much, but like whales and whales Elephants actually have a bigger brain than ours and can probably think deeper thoughts, but they don't have a thumb, so they don't create technology.

The monkey can create it. It actually has a thumb, but it's actually facing down. inch or so and therefore can't really grip very well, you can create some technology, but the technology you create cannot create other technology, so the fact that we have a thumb means that we can create integrated circuits that they can become a big language model uh that comes from the human brain um and it represents and it's actually trained with everything that we've ever thought, anything that human beings have ever thought has been documented and can go into these big language models. language um and everyone can works on these things and it's not true, well only certain rich people will have it.

I mean, how many people here have phones? If it's not 100%, it's like 99.9% and you don't have to be from a rich group. I mean, I see homeless people having their own phone, uh, it's not that expensive, um, and that represents the distribution of these capabilities, it's not something that you have to be fabulously rich to pay for, so you think we're headed toward future in which we will live much longer and we will be much more equal, I repeat, well, do you think we are heading towards a society in which we will live much longer, we will be richer but also with much more equality, yes, absolutely, and we have I've already seen it, well we're on time, but Ry and I will be back in 2124, 2224 and 2324, so thanks for coming today.

Thank you so much. It is an American treasure. Thanks Rch.