Eric Schmidt

When you look at some of the biggest ideas that have created some of the biggest businesses and industry in the last two decades, it's cloud services. 15 years ago, 20 years ago, if you would have said, here's my idea, I'm going to go to every company in the world. And you know, your IT budget that you're buying these metal boxes and these software things that you buy. Yeah. Forget it.

When you look at some of the biggest ideas that have created some of the biggest businesses and industry in the last two decades, it's cloud services. 15 years ago, 20 years ago, if you would have said, here's my idea, I'm going to go to every company in the world. And you know, your IT budget that you're buying these metal boxes and these software things that you buy. Yeah. Forget it.

Just give it to one company or maybe two companies, throw away all your computers and they'll do everything in the cloud for you. People would have said, you're nuts. That's a crazy idea. It took a tiny little player, someone that was nothing in the business to business world, a company that was known for selling diapers online, which was called Amazon at the time.

Just give it to one company or maybe two companies, throw away all your computers and they'll do everything in the cloud for you. People would have said, you're nuts. That's a crazy idea. It took a tiny little player, someone that was nothing in the business to business world, a company that was known for selling diapers online, which was called Amazon at the time.

It was basically a mail order catalog that was splashed onto the internet. And they said, what if we try this crazy thing? Everybody wrote the idea off as crazy.

It was basically a mail order catalog that was splashed onto the internet. And they said, what if we try this crazy thing? Everybody wrote the idea off as crazy.

Let's start with where we are right now. Folks are very familiar now with ChatGPT and its competitors, which includes Quad and my favorite, of course, Gemini from Google and a number of others. And people are amazed that this stuff can write better than certainly I can. They can do songs, they can even write code. So what happens next?

Let's start with where we are right now. Folks are very familiar now with ChatGPT and its competitors, which includes Quad and my favorite, of course, Gemini from Google and a number of others. And people are amazed that this stuff can write better than certainly I can. They can do songs, they can even write code. So what happens next?

The next big change is in the development of what are called agents. And an agent is something which is in a little loop that learns something. So you build an agent that can do the equivalent of a travel agent. Well, it learns how to do travel agents. The key thing about agents is that you can concatenate them. You give it an English command and it gives you an English result.

The next big change is in the development of what are called agents. And an agent is something which is in a little loop that learns something. So you build an agent that can do the equivalent of a travel agent. Well, it learns how to do travel agents. The key thing about agents is that you can concatenate them. You give it an English command and it gives you an English result.

And so then you can take that result and put it into the next agent. And with that, you can design a building, design a ship, design a bomb, whatever.

And so then you can take that result and put it into the next agent. And with that, you can design a building, design a ship, design a bomb, whatever.

So are we the dogs to their humanity? Will ultimately AI be our overlords? I certainly hope not. The theoretical argument is that computers are running the world and we're the dogs. That's unlikely. A much more likely scenario, which I do worry about, is that the concentration of power that a dictator type person, sort of an autocrat type person, can accumulate under the guise of efficiency

So are we the dogs to their humanity? Will ultimately AI be our overlords? I certainly hope not. The theoretical argument is that computers are running the world and we're the dogs. That's unlikely. A much more likely scenario, which I do worry about, is that the concentration of power that a dictator type person, sort of an autocrat type person, can accumulate under the guise of efficiency

can also restrict liberty. At the key point is what we call technically recursive self-improvement when it can begin to improve itself. We better have a really good way of watching what this thing is doing. And if you think about it for a while, the only way to watch what it's doing is to have another AI system watching it because people won't be able to follow it fast enough.

can also restrict liberty. At the key point is what we call technically recursive self-improvement when it can begin to improve itself. We better have a really good way of watching what this thing is doing. And if you think about it for a while, the only way to watch what it's doing is to have another AI system watching it because people won't be able to follow it fast enough.

One of the things that's most interesting about the judicial system right now, it's being used to give you some summaries of outcomes. The best one is if you're on trial, which thankfully neither you nor I are, you basically want to be in the morning because by the end of the afternoon, they're so tired of you that they just give you a harder sentence. Now, how was that discovered?

One of the things that's most interesting about the judicial system right now, it's being used to give you some summaries of outcomes. The best one is if you're on trial, which thankfully neither you nor I are, you basically want to be in the morning because by the end of the afternoon, they're so tired of you that they just give you a harder sentence. Now, how was that discovered?

That was discovered using machine learning. I don't think that computers should be judges. Because I think part of the principle of our democracy is that humans make decisions and they're held accountable. You want to make sure that you have human agency over everything. There's nothing wrong with the computer making a recommendation to the judge. What is wrong is if the judge just listens to it.

That was discovered using machine learning. I don't think that computers should be judges. Because I think part of the principle of our democracy is that humans make decisions and they're held accountable. You want to make sure that you have human agency over everything. There's nothing wrong with the computer making a recommendation to the judge. What is wrong is if the judge just listens to it.

Let me give you an example where this doesn't work. So if it's a judge in a courtroom, it's perfectly fine. There's appeals and the judge makes a mistake and so forth. It gets worked out. I mean, it's painful, but it gets worked out. But here we are, we're on a ship. You're the commander of a ship and the system has detected a hypersonic missile coming towards you with some high probability.

Let me give you an example where this doesn't work. So if it's a judge in a courtroom, it's perfectly fine. There's appeals and the judge makes a mistake and so forth. It gets worked out. I mean, it's painful, but it gets worked out. But here we are, we're on a ship. You're the commander of a ship and the system has detected a hypersonic missile coming towards you with some high probability.

And you have 29 seconds to press the button. And the system recommends pressing the button. 28, 27, 26, how many times do you think the captain of that ship will not press the button? They'll press the button. So that's an example where the system is designed to have human agency, but there's not enough time. So the compression of time is very important here.

And you have 29 seconds to press the button. And the system recommends pressing the button. 28, 27, 26, how many times do you think the captain of that ship will not press the button? They'll press the button. So that's an example where the system is designed to have human agency, but there's not enough time. So the compression of time is very important here.

And one of the core issues, and you mentioned this before, is these computers are moving so quickly. Another example that I like to use is, I don't know if you know, but there was a war and the war was that North Korea attacked America in cyberspace. America got ready to counterattack and China shut North Korea down.

And one of the core issues, and you mentioned this before, is these computers are moving so quickly. Another example that I like to use is, I don't know if you know, but there was a war and the war was that North Korea attacked America in cyberspace. America got ready to counterattack and China shut North Korea down.

Oh, and by the way, the entire war took a hundred milliseconds, less than a second. Now, how do you think about that? Now, obviously that war has not occurred yet, but is it possible? Absolutely. How do you do a human agency under the compression of time?

Oh, and by the way, the entire war took a hundred milliseconds, less than a second. Now, how do you think about that? Now, obviously that war has not occurred yet, but is it possible? Absolutely. How do you do a human agency under the compression of time?

Well, there are many such scenarios and they go something like this. At some point, the computer's objective function, what it's being trained against is broad enough that it decides that lying to us is a good idea because it knows we're watching. Now, is this a possible scenario? Absolutely. Am I worried about it? No, because I think I'm much more worried about, I think the positive is clear.

Well, there are many such scenarios and they go something like this. At some point, the computer's objective function, what it's being trained against is broad enough that it decides that lying to us is a good idea because it knows we're watching. Now, is this a possible scenario? Absolutely. Am I worried about it? No, because I think I'm much more worried about, I think the positive is clear.

Human plus AI is incredibly powerful. That also means that human plus AI is incredibly dangerous with the wrong human. I know these are all very interesting, the AI overlords and so forth, and they could take us and turn us into dogs, as I mentioned earlier. It's much more likely that the dangers will be because of human control over systems that are more powerful than they should be.

Human plus AI is incredibly powerful. That also means that human plus AI is incredibly dangerous with the wrong human. I know these are all very interesting, the AI overlords and so forth, and they could take us and turn us into dogs, as I mentioned earlier. It's much more likely that the dangers will be because of human control over systems that are more powerful than they should be.

I'll give you a simple example. The social media algorithms select the most inflammatory statements who are often from the most deranged people. And that's because the algorithm works. And because the algorithm says, oh, this is interesting, and a lot of people are listening to it and so forth, that's not a good way to run a democracy.

I'll give you a simple example. The social media algorithms select the most inflammatory statements who are often from the most deranged people. And that's because the algorithm works. And because the algorithm says, oh, this is interesting, and a lot of people are listening to it and so forth, that's not a good way to run a democracy.

Maybe we should have a rule that if you make a claim, you have to make a paragraph, right? And actually justify your argument as opposed to, oh my God, the following thing is about to kill and we're all going to die. But that's an example where humans have control, but we've chosen to allow inflammatory speech without the benefit of wisdom. And that's not good.

Maybe we should have a rule that if you make a claim, you have to make a paragraph, right? And actually justify your argument as opposed to, oh my God, the following thing is about to kill and we're all going to die. But that's an example where humans have control, but we've chosen to allow inflammatory speech without the benefit of wisdom. And that's not good.

We don't know the definition of consciousness. My own opinion is that this will not occur in my lifetime. I think that what will be true is that we will coexist with these systems and they'll take on more and more of the drudgery. They'll make the systems more efficient. Efficiency is generally a good thing in economic systems. People will be wealthier. People will be more productive.

We don't know the definition of consciousness. My own opinion is that this will not occur in my lifetime. I think that what will be true is that we will coexist with these systems and they'll take on more and more of the drudgery. They'll make the systems more efficient. Efficiency is generally a good thing in economic systems. People will be wealthier. People will be more productive.

My own view is that in my lifetime, everyone's productivity will double. You can do twice as many podcasts. I can do twice as many speeches. Whatever it is that each of us is doing, Because the tools make us more efficient. And that's the nature of technology invention. It's been true for 200 years. The car made us more efficient. Google made us more efficient and so forth.

My own view is that in my lifetime, everyone's productivity will double. You can do twice as many podcasts. I can do twice as many speeches. Whatever it is that each of us is doing, Because the tools make us more efficient. And that's the nature of technology invention. It's been true for 200 years. The car made us more efficient. Google made us more efficient and so forth.

I think that will continue. Because we can't define consciousness, we can imagine that the system can itself imagine consciousness. But it's highly unclear that one, it could detect it. And second, how would we know? Because it could have just decided to fool us.

I think that will continue. Because we can't define consciousness, we can imagine that the system can itself imagine consciousness. But it's highly unclear that one, it could detect it. And second, how would we know? Because it could have just decided to fool us.

We do not. A simple answer is that the systems will automate a more and more complex world. So if you look at a young person, at the moment I'm at Harvard surrounded by students, they are so comfortable with the world of clicking and moving around. They're in this infinite information space and they're comfortable. Whereas people in my generation find it overwhelming.

We do not. A simple answer is that the systems will automate a more and more complex world. So if you look at a young person, at the moment I'm at Harvard surrounded by students, they are so comfortable with the world of clicking and moving around. They're in this infinite information space and they're comfortable. Whereas people in my generation find it overwhelming.

So people adapt to this explosion of information. But the right system is to have the equivalent of an assistant that sort of organizes your digital world in a way that is net positive for you. Now that has a lot of negative implications, but I don't think that humans will be able to be very productive without their own AI assistant telling them what's most important, reading things.

So people adapt to this explosion of information. But the right system is to have the equivalent of an assistant that sort of organizes your digital world in a way that is net positive for you. Now that has a lot of negative implications, but I don't think that humans will be able to be very productive without their own AI assistant telling them what's most important, reading things.

We have this huge problem around misinformation right now. I just want something, an AI system to say, This is likely to be true, and this is probably somewhat true, and then give me the analysis. And then I can form my own opinions. At the point, going back to your point earlier about agency, which I really liked, is when you give agency to the computer, you're giving up something very important.

We have this huge problem around misinformation right now. I just want something, an AI system to say, This is likely to be true, and this is probably somewhat true, and then give me the analysis. And then I can form my own opinions. At the point, going back to your point earlier about agency, which I really liked, is when you give agency to the computer, you're giving up something very important.

Don't lose your critical thinking. Don't just believe it, even if it's Google.

Don't lose your critical thinking. Don't just believe it, even if it's Google.

Lenny, it was great to be on your show last time. I'm really glad to be back. It's always great to see you.

Lenny, it was great to be on your show last time. I'm really glad to be back. It's always great to see you.

Well, the biggest one would be things like access to weapons. What I mentioned, recursive self-improvement, where the system can actually learn on its own and we don't know what it's doing. I worry about those, the misuse in biology. There are plenty of people working on what are the capabilities of these models and to make sure that they can't produce pathogens.

Well, the biggest one would be things like access to weapons. What I mentioned, recursive self-improvement, where the system can actually learn on its own and we don't know what it's doing. I worry about those, the misuse in biology. There are plenty of people working on what are the capabilities of these models and to make sure that they can't produce pathogens.

Take the equivalent of smallpox and make it even deadlier. And so far, we had a long conversation in the industry about this a few weeks ago. The consensus was that the models that cost less than $100 million don't have this capability. But the ones that are going to cost more than $100 million might have this capability in the future. This is what everybody said. So that's today's idea.

Take the equivalent of smallpox and make it even deadlier. And so far, we had a long conversation in the industry about this a few weeks ago. The consensus was that the models that cost less than $100 million don't have this capability. But the ones that are going to cost more than $100 million might have this capability in the future. This is what everybody said. So that's today's idea.

So if the cost of models drops down, we're in trouble. If the cost of models goes up, then we're good. So you see how the answer is dynamic based on what happens to the technology.

So if the cost of models drops down, we're in trouble. If the cost of models goes up, then we're good. So you see how the answer is dynamic based on what happens to the technology.

In my industry, there are open source people, of which I'm one, who basically believe that the proliferation is net positive because it allows for creativity, it allows for expansion of human knowledge, it empowers everybody. This is a great position.

In my industry, there are open source people, of which I'm one, who basically believe that the proliferation is net positive because it allows for creativity, it allows for expansion of human knowledge, it empowers everybody. This is a great position.

There are plenty of people who disagree, arguing that the tool is so powerful that if you put it in even one evil person's hands, by the time you discover the evil, harm has occurred. That debate is an age-old debate in my industry, and it's not obvious to me how it will play out. I'm an optimist, but I worry about this one.

There are plenty of people who disagree, arguing that the tool is so powerful that if you put it in even one evil person's hands, by the time you discover the evil, harm has occurred. That debate is an age-old debate in my industry, and it's not obvious to me how it will play out. I'm an optimist, but I worry about this one.

Well, a simple answer to any business is if you're not using AI in your business, your competitor is and you're going to get screwed. Excuse my language. It's a serious problem because it's happening very quickly. For most businesses, AI begins by customer service. So, for example, chatbots to replace India call centers, things like that. Very mild improvements in efficiency.

Well, a simple answer to any business is if you're not using AI in your business, your competitor is and you're going to get screwed. Excuse my language. It's a serious problem because it's happening very quickly. For most businesses, AI begins by customer service. So, for example, chatbots to replace India call centers, things like that. Very mild improvements in efficiency.

You see targeted marketing now. The real change is going to be in generative AI. Generative AI, think of it as making pictures. So why do I have to spend a million dollars for a photo shoot for the product?

You see targeted marketing now. The real change is going to be in generative AI. Generative AI, think of it as making pictures. So why do I have to spend a million dollars for a photo shoot for the product?

Why don't I just have the system generate that video and not just generate one, but generate a million versions of it that are targeted to a million different kinds of customers, some of the humans can't do. So I think if you think about business efficiency, AI is the first one, that's the tactics. And then it's basically customer adoption. And then eventually it's business planning.

Why don't I just have the system generate that video and not just generate one, but generate a million versions of it that are targeted to a million different kinds of customers, some of the humans can't do. So I think if you think about business efficiency, AI is the first one, that's the tactics. And then it's basically customer adoption. And then eventually it's business planning.

Over time, there's a lot of evidence that a lot of programming can be replaced by computer. Now I'm a computer scientist and a programmer by trade for almost 50, well, more than 55 years at this point. And I don't wish my trade to go away, but I do acknowledge that the computer can probably write code equal to or better than a lot of programmers. There are plenty of examples of that today.

Over time, there's a lot of evidence that a lot of programming can be replaced by computer. Now I'm a computer scientist and a programmer by trade for almost 50, well, more than 55 years at this point. And I don't wish my trade to go away, but I do acknowledge that the computer can probably write code equal to or better than a lot of programmers. There are plenty of examples of that today.

And there are plenty of startups that are working on automating most software development. I was talking to one scientist, and one of the questions I like to ask scientists is, what system and programming language do you use? And he said, it doesn't matter. And I said, it matters to me. And he said, it matters to you, but it doesn't matter to anyone else. I said, why?

And there are plenty of startups that are working on automating most software development. I was talking to one scientist, and one of the questions I like to ask scientists is, what system and programming language do you use? And he said, it doesn't matter. And I said, it matters to me. And he said, it matters to you, but it doesn't matter to anyone else. I said, why?

And he said, because as long as I understand what I'm trying to do, I don't care how the computer gets me there. And because he's a scientist and because he knows exactly what he wants, he'll just keep generating code. He doesn't care what the language is as long as it gets him to his outcome, which in this case was a very complicated science question.

And he said, because as long as I understand what I'm trying to do, I don't care how the computer gets me there. And because he's a scientist and because he knows exactly what he wants, he'll just keep generating code. He doesn't care what the language is as long as it gets him to his outcome, which in this case was a very complicated science question.

So the fact that the innards of the system don't matter anymore is like a big deal in my little Eric world.

So the fact that the innards of the system don't matter anymore is like a big deal in my little Eric world.

Think about the millions of people who have either, or they're not good programmers, or they don't have a programmer, who all of a sudden, lots of my friends say, well, if I only had a programmer who could adapt the following, well, now they will.

Think about the millions of people who have either, or they're not good programmers, or they don't have a programmer, who all of a sudden, lots of my friends say, well, if I only had a programmer who could adapt the following, well, now they will.

It's a general rule that innovation always comes from small teams and this next generation of agents and so forth will come from both. I would expect the answer, unfortunately, to your question is both. I think the big companies are so well run and so focused on this area that they will do interesting things.

It's a general rule that innovation always comes from small teams and this next generation of agents and so forth will come from both. I would expect the answer, unfortunately, to your question is both. I think the big companies are so well run and so focused on this area that they will do interesting things.

And I also think that the startups are sufficiently specialized and sufficiently important that they'll do so. I'll give an example. Microsoft did a very good job with something called Copilot, which is a programmer's assistant. And there are now at least five startups that I'm aware of that are trying to do a product which is far, far better.

And I also think that the startups are sufficiently specialized and sufficiently important that they'll do so. I'll give an example. Microsoft did a very good job with something called Copilot, which is a programmer's assistant. And there are now at least five startups that I'm aware of that are trying to do a product which is far, far better.

Now, that competition is good, keeps both on their toes. I'm not going to predict the winner, but I can tell you that it's very competitive.

Now, that competition is good, keeps both on their toes. I'm not going to predict the winner, but I can tell you that it's very competitive.

I'll make a blunt statement as I can. It's clear to me that the US and China will dominate. The US because we invented it and because we're inventing it as we speak. China because they have made a decision to focus on this regardless of cost, and they're good, they're catching up. And I'm worried about that.

I'll make a blunt statement as I can. It's clear to me that the US and China will dominate. The US because we invented it and because we're inventing it as we speak. China because they have made a decision to focus on this regardless of cost, and they're good, they're catching up. And I'm worried about that.

Well, thank you for that. So let's start with where we are right now. Folks are very familiar now with ChatGPT and its competitors, which includes Claude and my favorite, of course, Gemini from Google and a number of others. And people are amazed that this stuff can write better than certainly I can. They can do songs. They can even write code. So what happens next?

Well, thank you for that. So let's start with where we are right now. Folks are very familiar now with ChatGPT and its competitors, which includes Claude and my favorite, of course, Gemini from Google and a number of others. And people are amazed that this stuff can write better than certainly I can. They can do songs. They can even write code. So what happens next?

I'm sure that the UK will be part of the success of the US, but what about all the other countries? What do the European countries that are busy regulating themselves to death and perfectly happy, what do they do when AI is imposed on them from a foreign power? in this case, probably US values. That's a loss for them and that's a mistake. I've told them this, they just aren't listening to me.

I'm sure that the UK will be part of the success of the US, but what about all the other countries? What do the European countries that are busy regulating themselves to death and perfectly happy, what do they do when AI is imposed on them from a foreign power? in this case, probably US values. That's a loss for them and that's a mistake. I've told them this, they just aren't listening to me.

What happens when Africa, which will have the majority of population in the next 50 years, population growth, what happens when the systems that are built reflect US or Chinese values and not local values? And there are differences in culture and values that matter a lot. One of the things about this technology that's important is it's very expensive.

What happens when Africa, which will have the majority of population in the next 50 years, population growth, what happens when the systems that are built reflect US or Chinese values and not local values? And there are differences in culture and values that matter a lot. One of the things about this technology that's important is it's very expensive.

It takes large teams and, again, billions of dollars of hardware. Elon has done his new data center. The press reports 200,000 GPUs. A GPU costs about $50,000. So for purposes of argument, that's $10 billion right there in the ground. There aren't that many ELONs, but how many people can afford that?

It takes large teams and, again, billions of dollars of hardware. Elon has done his new data center. The press reports 200,000 GPUs. A GPU costs about $50,000. So for purposes of argument, that's $10 billion right there in the ground. There aren't that many ELONs, but how many people can afford that?

We did a survey of the number of 100,000 GPU clusters and larger, and there are seven or eight, two of which appear to be in China or related to China, and the rest appear to be in the West, but mostly under US control. What happens to every other country? If you thought about it, let's say you're in Germany. The first thing you would say is we need one of those two.

We did a survey of the number of 100,000 GPU clusters and larger, and there are seven or eight, two of which appear to be in China or related to China, and the rest appear to be in the West, but mostly under US control. What happens to every other country? If you thought about it, let's say you're in Germany. The first thing you would say is we need one of those two.

Well, Germany is in terrible financial strait. They're having a whole identity crisis because of the energy cost and the Chinese problem and the Russia energy problem and on and on and on. It's not on their list, but they have to do it right now if they want to stay a player.

Well, Germany is in terrible financial strait. They're having a whole identity crisis because of the energy cost and the Chinese problem and the Russia energy problem and on and on and on. It's not on their list, but they have to do it right now if they want to stay a player.

France, under President Macron, is doing a good job of focusing on this, but the training is all being done outside of France because of the cost of electricity, which is subsidized in the case of the trainers. So again, these are very complicated problems.

France, under President Macron, is doing a good job of focusing on this, but the training is all being done outside of France because of the cost of electricity, which is subsidized in the case of the trainers. So again, these are very complicated problems.

And in the reordering question, which is under... What I'm most interested in is as general intelligence is invented, I want to make sure it reflects Western values, and I want it to be for more than just the US. I want it to benefit the entire world.

And in the reordering question, which is under... What I'm most interested in is as general intelligence is invented, I want to make sure it reflects Western values, and I want it to be for more than just the US. I want it to benefit the entire world.

I call this innovation power. We've all been indoctrinated that there's soft power and hard power. I'm arguing that there's innovation power. Innovation power is inventing new things with all this new technology. And I'm gonna argue that if you can get to scale in these new technologies quicker, you're gonna have a lot more military power.

I call this innovation power. We've all been indoctrinated that there's soft power and hard power. I'm arguing that there's innovation power. Innovation power is inventing new things with all this new technology. And I'm gonna argue that if you can get to scale in these new technologies quicker, you're gonna have a lot more military power.

So for the U.S., what this means is very complicated national security systems, very complicated vision systems, very complicated military management systems, and the adoption of autonomy and drones, which is occurring in Ukraine, but not in the U.S. yet.

So for the U.S., what this means is very complicated national security systems, very complicated vision systems, very complicated military management systems, and the adoption of autonomy and drones, which is occurring in Ukraine, but not in the U.S. yet.

Well, the generals want the following. They want a battlefield management system that shows all the sensors and shooters. So they have sensors and they have things that shoot. And they want the AI to assemble all of that. They've wanted this for a decade. And various people have promised it to them. I don't think that's how it's going to actually work.

Well, the generals want the following. They want a battlefield management system that shows all the sensors and shooters. So they have sensors and they have things that shoot. And they want the AI to assemble all of that. They've wanted this for a decade. And various people have promised it to them. I don't think that's how it's going to actually work.

I think what's really going to work is that every aspect of the battlefield will be reimagined to be more autonomous. And autonomy, a simple example is, why do you need a soldier with a gun? Why don't you have an automatic gun? Why do you need a jet fighter with a bomb? Why don't you just have an automatic drone with a bomb?

I think what's really going to work is that every aspect of the battlefield will be reimagined to be more autonomous. And autonomy, a simple example is, why do you need a soldier with a gun? Why don't you have an automatic gun? Why do you need a jet fighter with a bomb? Why don't you just have an automatic drone with a bomb?

The next big change is in the development of what are called agents. And an agent is something which is in a little loop that learns something. So you build an agent that can do the equivalent of a travel agent. Well, it learns how to do travel agents. The key thing about agents is that you can concatenate them. You give it an English command and it gives you an English result.

The next big change is in the development of what are called agents. And an agent is something which is in a little loop that learns something. So you build an agent that can do the equivalent of a travel agent. Well, it learns how to do travel agents. The key thing about agents is that you can concatenate them. You give it an English command and it gives you an English result.

And there will be a supervisory system, and the supervisory system will do the planning. But ultimately, as we discussed before, the human control is essential. So under no circumstances should we give up human control to these machines.

And there will be a supervisory system, and the supervisory system will do the planning. But ultimately, as we discussed before, the human control is essential. So under no circumstances should we give up human control to these machines.

But it will change war in the sense that the general will sit there and there'll be a button saying, do you approve of my battle plan based on these autonomous systems? And with that, a war is started or a war is ended.

But it will change war in the sense that the general will sit there and there'll be a button saying, do you approve of my battle plan based on these autonomous systems? And with that, a war is started or a war is ended.

It could. I just watched a play in London two days ago, which is about Dr. Strangelove. And you remember in the movie from 1963, the Russians, this came out of a RAND study in 1960. In this story, the Russians had secretly created a doomsday machine, but had not bothered to tell the U.S. this. And that if they were attacked, the doomsday machine would kill everyone.

It could. I just watched a play in London two days ago, which is about Dr. Strangelove. And you remember in the movie from 1963, the Russians, this came out of a RAND study in 1960. In this story, the Russians had secretly created a doomsday machine, but had not bothered to tell the U.S. this. And that if they were attacked, the doomsday machine would kill everyone.

So it's the best possible story for why you don't want automatic systems that just decide on their own. Because you can get into all sorts of situations where there was a misunderstanding and terror occurs.

So it's the best possible story for why you don't want automatic systems that just decide on their own. Because you can get into all sorts of situations where there was a misunderstanding and terror occurs.

We actually, the three authors all disagreed on this. I'm quite sure humanity will survive and I am an optimist more so than Henry was. And we miss him by the way. But the reason to be on my side of this is that we as humans have faced all of these challenges before. And in all cases, we have survived at various levels of pain. So we will survive.

We actually, the three authors all disagreed on this. I'm quite sure humanity will survive and I am an optimist more so than Henry was. And we miss him by the way. But the reason to be on my side of this is that we as humans have faced all of these challenges before. And in all cases, we have survived at various levels of pain. So we will survive.

Let's reduce the possible pain and let's certainly avoid conflict using these new tools at the scale that we're discussing. It would be horrendous.

Let's reduce the possible pain and let's certainly avoid conflict using these new tools at the scale that we're discussing. It would be horrendous.

I think the first point, which I cannot emphasize enough, is that this stuff is happening much, much faster than I expected and that almost anyone understands. I have never in my almost 50 years career doing this had a situation where there's a surprise every day. And almost all of the surprises are to more power, more insight, better than human performance.

I think the first point, which I cannot emphasize enough, is that this stuff is happening much, much faster than I expected and that almost anyone understands. I have never in my almost 50 years career doing this had a situation where there's a surprise every day. And almost all of the surprises are to more power, more insight, better than human performance.

And as that arrives, it changes huge human systems, because humans have organized themselves in various ways. We need to have a map of the arrival and the impact. I'd say the second point is that there's a set of questions that we don't know. And one of them is, where is the limit of this kind of intelligence? Let me give you an example.

And as that arrives, it changes huge human systems, because humans have organized themselves in various ways. We need to have a map of the arrival and the impact. I'd say the second point is that there's a set of questions that we don't know. And one of them is, where is the limit of this kind of intelligence? Let me give you an example.

It's clear to me that these things will be fantastic scientists, whatever you want to call them. Fantastic scientists, they can analyze things, they can figure stuff out, they can do math better, all that kind of stuff. Much of human behavior is really invention, strategy, and responses. We haven't seen that yet emerge in these systems.

It's clear to me that these things will be fantastic scientists, whatever you want to call them. Fantastic scientists, they can analyze things, they can figure stuff out, they can do math better, all that kind of stuff. Much of human behavior is really invention, strategy, and responses. We haven't seen that yet emerge in these systems.

So is there a limit to where the current technology will go compared to humans? In other words, will the system do 90%? But the 10% that the humans do are the stuff that humans are particularly good at, strategy, thinking about impact, understanding subtleties of humans and things like that, or will that go too? My own opinion right now is there will be a space for us for a long time.

So is there a limit to where the current technology will go compared to humans? In other words, will the system do 90%? But the 10% that the humans do are the stuff that humans are particularly good at, strategy, thinking about impact, understanding subtleties of humans and things like that, or will that go too? My own opinion right now is there will be a space for us for a long time.

And so then you can take that result and put it into the next agent. And with that, you can design a building, design a ship, design a bomb, whatever. So agents look like The next big step.

And so then you can take that result and put it into the next agent. And with that, you can design a building, design a ship, design a bomb, whatever. So agents look like The next big step.

That is my opinion. And I think the third thing I would mention is this question of proliferation. In all my work here, everyone likes to compare this to the nuclear proliferation problem of which Dr. Kissinger was the world's expert because he did it. And his view, which is also Craig's and mine, is that it's different.

That is my opinion. And I think the third thing I would mention is this question of proliferation. In all my work here, everyone likes to compare this to the nuclear proliferation problem of which Dr. Kissinger was the world's expert because he did it. And his view, which is also Craig's and mine, is that it's different.

But the proliferation problem for this is it's so much easier to make this technology broadly available that we have to really think about how we're going to make sure that evil people don't get access to it. And that's an issue for everyone, not just for you and me.

But the proliferation problem for this is it's so much easier to make this technology broadly available that we have to really think about how we're going to make sure that evil people don't get access to it. And that's an issue for everyone, not just for you and me.

Thank you very much, Lynn. Thank you for everything and thank you for your support.

Thank you very much, Lynn. Thank you for everything and thank you for your support.

Once agents are generally available, which will take a few years, I expect that we're going to see systems that are super powerful, where the architect can say, design me a building I'll describe roughly and just make it beautiful. And the system will be capable of understanding that. That's not AGI. That's just really powerful AI.

Once agents are generally available, which will take a few years, I expect that we're going to see systems that are super powerful, where the architect can say, design me a building I'll describe roughly and just make it beautiful. And the system will be capable of understanding that. That's not AGI. That's just really powerful AI.

AGI, which is the general term is general intelligence, is what we have, the ability to essentially have an idea in the morning and pursue it that you didn't have the day before. The consensus in the industry is that that's well more than five years from now. There's something I call the San Francisco school, which says it will be within five years.

AGI, which is the general term is general intelligence, is what we have, the ability to essentially have an idea in the morning and pursue it that you didn't have the day before. The consensus in the industry is that that's well more than five years from now. There's something I call the San Francisco school, which says it will be within five years.

I think it's more like eight to 10, but nobody really knows. And you can see this with the most recent announcement from OpenAI of something called 0.1, where it can begin to show you the work that it does as it solves math problems.

I think it's more like eight to 10, but nobody really knows. And you can see this with the most recent announcement from OpenAI of something called 0.1, where it can begin to show you the work that it does as it solves math problems.

And the latest models are good enough to pass the graduate level exams in physics and chemistry and computer science and material science and art and political science. At some point, these things in the next, say, five years are going to be super brilliant, but they're still going to be under our control.

And the latest models are good enough to pass the graduate level exams in physics and chemistry and computer science and material science and art and political science. At some point, these things in the next, say, five years are going to be super brilliant, but they're still going to be under our control.

At the key point is what we call technically recursive self-improvement, when it can begin to improve itself. And at that point, I think we're in a different ballgame. And it goes something like this. I say to the computer, learn everything, start now, don't do any serious damage. That's the command. Okay. And the system is programmed to be curious, but also to aggregate power and influence.

At the key point is what we call technically recursive self-improvement, when it can begin to improve itself. And at that point, I think we're in a different ballgame. And it goes something like this. I say to the computer, learn everything, start now, don't do any serious damage. That's the command. Okay. And the system is programmed to be curious, but also to aggregate power and influence.

What would it do? We don't know. So that strikes me as a point where we better have a really good way of watching what this thing is doing. And if you think about it for a while, the only way to watch what it's doing is to have another AI system watching it because people won't be able to follow it fast enough.

What would it do? We don't know. So that strikes me as a point where we better have a really good way of watching what this thing is doing. And if you think about it for a while, the only way to watch what it's doing is to have another AI system watching it because people won't be able to follow it fast enough.

Well, the most important thing right now is don't put them anywhere near human life or mission-critical infrastructure. Don't use it to do hard operations. Don't use it to fly an airplane, those sorts of things. The systems today can best be understood as fantastic advisors. In my case, I had a complicated question. I used one of the LLMs and it sorted it out.

Well, the most important thing right now is don't put them anywhere near human life or mission-critical infrastructure. Don't use it to do hard operations. Don't use it to fly an airplane, those sorts of things. The systems today can best be understood as fantastic advisors. In my case, I had a complicated question. I used one of the LLMs and it sorted it out.

My complicated question had to do with what was the best and cheapest of the speakers I wanted to buy that had the highest of a particular tonal range. Something which I could do with Google by looking at all the documents, I had AI answer that question. And by the way, gave me the right answer and saved me. I could have done it, but I let the system do it for me. That's the next step.

My complicated question had to do with what was the best and cheapest of the speakers I wanted to buy that had the highest of a particular tonal range. Something which I could do with Google by looking at all the documents, I had AI answer that question. And by the way, gave me the right answer and saved me. I could have done it, but I let the system do it for me. That's the next step.

So the question is, at what point does the thing become to actually run my life? Like in the morning, I say, organize my life. Look at my schedule. I have to fit Joe, Bob and Harry in. Do you figure it out? And by the way, Harry, I don't like. And Joe, you should just sort of be mean to him. And I love, you know, whatever the way people actually work, the way they organize their lives.

So the question is, at what point does the thing become to actually run my life? Like in the morning, I say, organize my life. Look at my schedule. I have to fit Joe, Bob and Harry in. Do you figure it out? And by the way, Harry, I don't like. And Joe, you should just sort of be mean to him. And I love, you know, whatever the way people actually work, the way they organize their lives.

And we don't have that yet. But I think the agentic revolution, as it's called, is for sure going to happen. Governments are watching this so far. They're not doing anything too foolish, except that the Europeans who love to regulate are regulating it more than they should. And President Trump has indicated that he's going to scrap the minimal regulation that was in place under Biden.

And we don't have that yet. But I think the agentic revolution, as it's called, is for sure going to happen. Governments are watching this so far. They're not doing anything too foolish, except that the Europeans who love to regulate are regulating it more than they should. And President Trump has indicated that he's going to scrap the minimal regulation that was in place under Biden.

So I think it's going to be pretty open for a while. If you look at China, China, which obviously has an interest in not allowing, given their monopoly on power, they don't want threats. So China will undoubtedly have a rule that means you can do anything you want to as long as you don't threaten the state. which is sort of how their speech works.

So I think it's going to be pretty open for a while. If you look at China, China, which obviously has an interest in not allowing, given their monopoly on power, they don't want threats. So China will undoubtedly have a rule that means you can do anything you want to as long as you don't threaten the state. which is sort of how their speech works.

So I think everyone will adapt, but the race is on. One of the things that I want to say to you and your listeners is this race is very intense and it's happening very fast in the next few years. So get ready. I'm not sure that societies... If I can just be blunt, I just don't think our political system and societies are ready.

So I think everyone will adapt, but the race is on. One of the things that I want to say to you and your listeners is this race is very intense and it's happening very fast in the next few years. So get ready. I'm not sure that societies... If I can just be blunt, I just don't think our political system and societies are ready.

The concepts of governance are prehistoric compared to, and we ourselves are as analog devices, as humans, biological unions are prehistoric compared to what is possible with the technology.

The concepts of governance are prehistoric compared to, and we ourselves are as analog devices, as humans, biological unions are prehistoric compared to what is possible with the technology.

The question you're asking is known as the explainability problem. And today, the systems cannot explain why they know something. If you say, how do you know that? It doesn't have the ability to say, I learned it from Lynn or whatever. Because it learned it in complicated ways. There's a lot of progress on explainability, but it's not solved.

The question you're asking is known as the explainability problem. And today, the systems cannot explain why they know something. If you say, how do you know that? It doesn't have the ability to say, I learned it from Lynn or whatever. Because it learned it in complicated ways. There's a lot of progress on explainability, but it's not solved.

I personally think that the systems will be relatively explainable. One of the thought experiments is what happens when the system designs itself to be non-explainable. And my answer is pretty simple. If you can't figure out what it's doing, unplug it. Just unplug it and then think about it for a while.

I personally think that the systems will be relatively explainable. One of the thought experiments is what happens when the system designs itself to be non-explainable. And my answer is pretty simple. If you can't figure out what it's doing, unplug it. Just unplug it and then think about it for a while.

This is generally known as the proliferation problem, which he spent a lot of time on in the book. And the question here is, if there are 10 or 20 or 30 of these things in 10 years, they'll be regulated. The governments will have all sorts of rules. The Europeans will over-regulate. The Chinese will regulate in a Chinese way. The US will under-regulate, but they'll fundamentally be regulated.

This is generally known as the proliferation problem, which he spent a lot of time on in the book. And the question here is, if there are 10 or 20 or 30 of these things in 10 years, they'll be regulated. The governments will have all sorts of rules. The Europeans will over-regulate. The Chinese will regulate in a Chinese way. The US will under-regulate, but they'll fundamentally be regulated.

So what happens when an evil terrorist gets full access to one of these things? We need to prevent that. Now, one way that could occur is technically called exfiltration, where you take the model and you literally steal it and put it on a hard drive and then copy it, put it on the dark web. That would be really a bad thing.

So what happens when an evil terrorist gets full access to one of these things? We need to prevent that. Now, one way that could occur is technically called exfiltration, where you take the model and you literally steal it and put it on a hard drive and then copy it, put it on the dark web. That would be really a bad thing.

The industry is very focused on the security of these models for that reason. But it's important that if China releases... So here's an example. China just released two incredibly powerful models this week based on open source, and they fully released them. I'm not suggesting that they're dangerous, but I'm suggesting that a future such strategy could be dangerous.

The industry is very focused on the security of these models for that reason. But it's important that if China releases... So here's an example. China just released two incredibly powerful models this week based on open source, and they fully released them. I'm not suggesting that they're dangerous, but I'm suggesting that a future such strategy could be dangerous.

We have to be very careful here about proliferation. We understand proliferation can be used to harm a lot of people.

We have to be very careful here about proliferation. We understand proliferation can be used to harm a lot of people.

Well, let's start with climate change. I don't think we'll get to climate change solved without very, very powerful new energy sources and materials. All of that will come as a result of AI applied to science. Let's think about drugs, drug discovery, the alpha's fold and the Revolution in proteins, single-cell proteins, all of this kind of stuff is happening very, very quickly.

Well, let's start with climate change. I don't think we'll get to climate change solved without very, very powerful new energy sources and materials. All of that will come as a result of AI applied to science. Let's think about drugs, drug discovery, the alpha's fold and the Revolution in proteins, single-cell proteins, all of this kind of stuff is happening very, very quickly.

There are huge companies being set up to essentially identify drug candidates and test them in computer rather than in humans. And people believe that the gains will be massive in terms of human health. What about education? People don't learn the same way. Why are we still sitting in front of a teacher with 30 kids in a row?

There are huge companies being set up to essentially identify drug candidates and test them in computer rather than in humans. And people believe that the gains will be massive in terms of human health. What about education? People don't learn the same way. Why are we still sitting in front of a teacher with 30 kids in a row?

Wouldn't it be better if they had their own self-supervised learning with the same teacher in the room saying, you know, Johnny, how are you doing? And Mary, how are you doing? And they're different and the system adapts to their education. What about health care? A lot of people in the world have very poor health care.

Wouldn't it be better if they had their own self-supervised learning with the same teacher in the room saying, you know, Johnny, how are you doing? And Mary, how are you doing? And they're different and the system adapts to their education. What about health care? A lot of people in the world have very poor health care.

We're very fortunate here in the US to have very good health care, although we complain about it all the time. Can you imagine if you had the equivalent of a nurse practitioner that was better in doctor stuff than most doctors? There was an article today, for example, that there are many, many cases in cancer

We're very fortunate here in the US to have very good health care, although we complain about it all the time. Can you imagine if you had the equivalent of a nurse practitioner that was better in doctor stuff than most doctors? There was an article today, for example, that there are many, many cases in cancer

where the system can actually detect the cancer quicker and more accurately than the cancer doctor. So these are huge systemic changes. They will affect billions of people to the positive. So please don't shut down what we're doing. Just watch what we're doing and keep an eye out on us. And Dr. Kissinger was very concerned that people like me not be put in charge.

where the system can actually detect the cancer quicker and more accurately than the cancer doctor. So these are huge systemic changes. They will affect billions of people to the positive. So please don't shut down what we're doing. Just watch what we're doing and keep an eye out on us. And Dr. Kissinger was very concerned that people like me not be put in charge.

He wanted the society as a whole, people who were humanists and political leaders and artists and humanists and things like that. He did not trust that the tech people alone would get it right. And I agreed with him.

He wanted the society as a whole, people who were humanists and political leaders and artists and humanists and things like that. He did not trust that the tech people alone would get it right. And I agreed with him.

So are we the dogs to their humanity? Will ultimately AI be our overlords? I certainly hope not. The theoretical argument is the computers are running the world and we're the dogs. That's unlikely.

So are we the dogs to their humanity? Will ultimately AI be our overlords? I certainly hope not. The theoretical argument is the computers are running the world and we're the dogs. That's unlikely.

A much more likely scenario, which I do worry about, is that the concentration of power that a dictator type person, you know, sort of an autocrat type person can accumulate under the guise of efficiency can also restrict liberty. I'll give you an old example.

A much more likely scenario, which I do worry about, is that the concentration of power that a dictator type person, you know, sort of an autocrat type person can accumulate under the guise of efficiency can also restrict liberty. I'll give you an old example.

Google engineers design a car that is perfect for New York City and they optimize the traffic so you have maximum occupation of the roads all the time. And every car is controlled by the single computer. So a pregnant woman or a man who has an emergency gets in their car and there's no button to say, I have to drive faster than everyone else because I'm in a real emergency.

Google engineers design a car that is perfect for New York City and they optimize the traffic so you have maximum occupation of the roads all the time. And every car is controlled by the single computer. So a pregnant woman or a man who has an emergency gets in their car and there's no button to say, I have to drive faster than everyone else because I'm in a real emergency.

Yeah, she's in labor or something. My point is that the human systems tolerate flexibility, and sometimes that flexibility comes at the cost of total efficiency. And yet all of us would agree that that pregnant lady about to go in labor should get priority. So if you're going to automate systems, you better have them be flexible to human conditions, the good and the bad.

Yeah, she's in labor or something. My point is that the human systems tolerate flexibility, and sometimes that flexibility comes at the cost of total efficiency. And yet all of us would agree that that pregnant lady about to go in labor should get priority. So if you're going to automate systems, you better have them be flexible to human conditions, the good and the bad.

And I worry a lot that the path to power from a leadership perspective involves the restriction of limits. And the best way to do that is by aggressively implementing AI tools to prevent freedom, is my personal view. In the book, what we say is that there's been a long time debate between the good king and rule by many. And we collectively agree rule by many is better for lots of reasons.

And I worry a lot that the path to power from a leadership perspective involves the restriction of limits. And the best way to do that is by aggressively implementing AI tools to prevent freedom, is my personal view. In the book, what we say is that there's been a long time debate between the good king and rule by many. And we collectively agree rule by many is better for lots of reasons.

But what if, in fact, the average person is on the day-to-day basis happier in a benevolent king with an efficient system? We're going to run that experiment. I obviously know what my vote is, and I'm sure I know your vote, which is a rule by many, not by a dictator. But you see my point. I'm using the word dictator in the sense of centralized authority.

But what if, in fact, the average person is on the day-to-day basis happier in a benevolent king with an efficient system? We're going to run that experiment. I obviously know what my vote is, and I'm sure I know your vote, which is a rule by many, not by a dictator. But you see my point. I'm using the word dictator in the sense of centralized authority.

Well, a better way of saying that is that we better give instructions to the AI to preserve human agency. You can imagine a scenario where the agentic revolution, which I mentioned, actually does things so well, humans don't really control it on a tactical basis. In other words, it works so well and that we discover one day we've given something up and that's not good.

Well, a better way of saying that is that we better give instructions to the AI to preserve human agency. You can imagine a scenario where the agentic revolution, which I mentioned, actually does things so well, humans don't really control it on a tactical basis. In other words, it works so well and that we discover one day we've given something up and that's not good.

We want to preserve that freedom for human agency. I think for most people, having the travel agent be automatic and not having to fiddle with the lights in their room and the computer getting it set up because it's a pain in the ass, excuse my language, having those efficiencies is a good thing. Having all the world's information at your fingertips is a good thing.

We want to preserve that freedom for human agency. I think for most people, having the travel agent be automatic and not having to fiddle with the lights in their room and the computer getting it set up because it's a pain in the ass, excuse my language, having those efficiencies is a good thing. Having all the world's information at your fingertips is a good thing.

But when it ultimately prevents you from having freedom, then it's not such a good thing. And I think people will discover that boundary.

But when it ultimately prevents you from having freedom, then it's not such a good thing. And I think people will discover that boundary.

google decided that work-life balance and going home early and working from home was more important than winning and the startups the reason startups work is because the people work like hell and i'm sorry to be so blunt but the fact of the matter is if you all leave the university and go found a company you're not going to let people work from home and only come in one day a week. Ooh.

google decided that work-life balance and going home early and working from home was more important than winning and the startups the reason startups work is because the people work like hell and i'm sorry to be so blunt but the fact of the matter is if you all leave the university and go found a company you're not going to let people work from home and only come in one day a week. Ooh.

Can you show the Starbucks clip?

Can you show the Starbucks clip?

He's the former Google executive chairman and CEO.

Did he move the needle?

It's going to be a long time before humanoid robots, which is what we see in the movies all day, right?

It'll be a very long time before we see that.

What you're going to see is very, very fast mobility solutions, right?

Air-based solutions and also hypersonics, also things underwater.

There's a lot of that going on.

It's a different domain.

If you look at the Maguro and some other boats that the Ukrainians used, they have essentially used USVs to destroy the Russian fleet in the Black Sea.

This was crucial for them because they needed to be able to export the grain from Odessa around and it's like 6% or 10% of their economy.

It's a very big deal and they did that with drones.

Look, you guys know this guy well.

He's really that smart.

I want America to win.

Right?

I am here because the American dream, the people who invested in me, in my case Berkeley and so forth, people took a chance on me.

I want the next generation to have that.

I also want you all to remember, I was just in, as part of the World War II surrender ceremony in Honolulu, and they talked about fighting tyranny.

So he taught me more stuff than most of any of the employees at Google.

We forget that our ancestors, our great grandparents or whatever,

fought the Great War to keep liberalism and democracy alive.

I want us to do that.

How do we do that as Americans?

We use our strengths.

What are our strengths?

We're chaotic, confusing, loud, you know, but we're clever.

We allocate capital smartly.

We have very deep financial markets.

We have this enormous industrial base of universities and entrepreneurs which are represented here.

We should celebrate this.

And then you left.

We should stoke it.

We should make it go faster and faster.

I spent lots of time in Europe because of the Ukraine stuff.

They are so envious of us.

When you're in Asia, they are envious of us.

Don't screw it up, guys.

That's what I want to work on.

These agents are going to be really powerful, and they'll start to work together.

Elon is a good friend and he's addressing this issue of population directly himself.

One metric of the success of a society is its ability to reproduce.

And so I think this is a legitimate concern of the West.

It's much worse in Asia.

The Chinese number is about 1.0 for two parents.

In Korea, it's now down to 0.78 for two.

So it's really important to recognize that we as humans are collectively choosing to depopulate and the numbers are staggering, right?

And imagine a situation where instead of having growth, you have shrinkage and furthermore, they're getting older.

And so as a business, all of a sudden your revenue is declining and there's nothing you can do because you can't innovate with fewer and fewer customers.

So if you just put it in a business context, ignoring the moral issues, which are all very real, it's just bad, right?

So we have to solve that problem.

I happen to be in favor broadly of immigration because I think immigration helps us solve that problem, but as a global mechanism, we have to address that.

In any case, from my perspective, you're going to have these issues, but America is organized around the concept of American exceptionalism.

And as long as we understand that the way we make progress is we invest in the right people, in the right businesses, we have a strong capital market, we invest in the infrastructure that they need, we'll be fine.

That is my actual opinion.

So first place, the speech that the president delivered about a month ago about AI strategy, which I think you probably wouldn't say it, but you kind of wrote it for him, was exactly right, right?

So thank you.

Yes.

If you didn't write it, then it must have been your twin.

But in any case, you got the emphasis right, which was that investment in research, investment in the kind of stuff that we do is really, really important.

Predicting Sergei's behavior is something I can fail at.

I don't agree with you on this AGI thing because there's this group which I call the San Francisco narrative, because they all live in San Francisco, and their narrative goes something like this.

Today we're doing agents.

The agentic revolution will change businesses, which I agree with.

that what happens is the systems will become recursively self-intelligent.

With recursive self-improvement, as it's called, if you have a scale-free problem, and a scale-free problem, for example, is programming or math, where you can just keep doing it, you get these enormous fast gains if you buy enough hardware, do enough software, so forth and so on.

That is still underway.

The collective of that says that in the next three-ish years, they believe that we will get forms of superintelligence.

I tried for 20 years.

And the way they define it is basically a savant, a chemist savant, a physics savant, a mathematician savant.

I am not in favor of essentially working at home.

I don't agree with the three years, but I do agree that it'll be maybe six or seven years.

It's not general intelligence yet.

General intelligence is when it can set its own objective function.

Right.

There's no evidence right now of the ability to set your own objective function.

The thinking, and I'm writing a paper on this, so I've been studying it, is that the technical problem is non-stationarity of mathematical proofs.

And what you're doing is you're trying to solve against the objective function, but the objective function keeps changing, which is how humans operate.

Your goal changes every day, whereas computers have trouble with that.

As a math problem, we don't have an algorithm yet for LLMs that can do that.

People are working on it.

And the test will be, can you basically, using the information available in 1902, can you derive the same thing that Einstein did with special relativity followed by general relativity?

And the reason, I mean, many of you guys all work at home to some degree, but your careers are already established.

We cannot do that today.

And most people believe that the way this will be solved is through analogy.

So the theory of great geniuses is that they understand one area extremely well and they're so brilliant, the lady or man, can then take their ideas and apply it to a completely different domain.

If we can solve that problem, then I think it's over.

Then we get to AGI and then it's a whole different world.

We're soon going to be able to have computers running on their own, deciding what they want to do.

But think about a 20-something who has to learn how the world works.

What you just said is exactly what's going to happen for the next few years.

that each of us will have assistance which on our command and our prompting will be incredibly helpful to whatever problem we have you know personal you have people who are using these things for relationship advice for you know talking to their kids i mean it's all crazy stuff but the fact that matters that's it to me the real question is when does it cross over to having its own volition its own ability to seek information and solve new problems that's a different animal but have we seen any evidence of recursive self-improvement yet

And, you know, they come out of Berkeley or Dartmouth and they're very well educated.

Not yet.

I've funded a number of startups which claim to be close to it.

But of course these are startups and you never know.

Which tells me it's five, ten years.

Well, I'm not at Google anymore.

Every issue of Gemini is top of the leaderboard.

So 2.5 just overcame everybody, and I'm sure there's another one coming.

When I think about how...

Demis is working really hard on this question about scientific discovery.

how much I learned when I was at Sun just listening to these elder people who were five or 10 years older than I was argue with each other in person.

So that is a path to getting to AGI.

I am so happy to be part of this.

You created this incredible community, and there's all of these smart people that spend all their time listening to you.

Very concerning.

How do you recreate that in this new thing?

And I'm in favor of work-life balance and that's why people work for the government.

Sorry.

Strays.

Sorry, sorry, sorry.

If you're going to be in tech and you're going to win, you're going to have to make some trade-offs.

And remember, we're up against the Chinese.

I honestly believe that the AI revolution is under-hyped.

Now we have the arrival of a new non-human intelligence, which is likely to have better reasoning skills than humans can have.

The Chinese work-life balance consists of 996, which is 9 a.m.

to 9 p.m.

six days a week.

By the way, the Chinese have clarified that this is illegal.

However, they all do it.

That's who you're competing against.

Well, you and I just talked about this as part of your, as your incredibly important work in the White House.

I had thought that China and the United States were competing at the peer level in AI and that the good work that you have done and your predecessors did to restrict chips were slowing them down.

They're really doing something more different than I thought.

They're not pursuing crazy AGI strategies, partly because of the hardware limitations that you've put in place, but partly because the depth of their capital markets don't exist.

They can't raise, based on a wing and a prayer, $100 million or the equivalent to build the data centers.

They just can't do it.

And so the result is they're very focused on taking AI and applying it to everything.

And so the concern I have is that while we're pursuing AGI, which is incredibly interesting and we should talk about, and all of us will be affected by this, we better also be competing with the Chinese in day-to-day stuff.

Consumer apps, this is something you understand very well, Chamath.

Consumer apps, robots, and so forth and so on.

I saw all the Shanghai robotics companies, and these guys are attempting to do in robots what they've successfully done with electric vehicles.

And their work ethic's incredible, they're well-funded.

It's not the crazy valuations that we have in America.

They can't raise the capital, but they can win across that.

The other thing the Chinese are doing, and I want to emphasize this as a major geopolitical issue,

is that my own background is open source.

In the audience, you all know open source means open code, open weights means open training data.

China is competing with open weights and open training data, and the US is largely and majority focused on closed weights, closed data.

That means that the majority of the world, think of it as the Belt and Road Initiative, are going to use Chinese models and not American models.

Now, I happen to think the West and democracies are correct.

And I'd much rather have the proliferation of large language models and that learning be done based on Western values.

It's not clear.

You know, Alex Wang's a good friend.

He's come in.

He's taken over.

He's obviously incredibly capable.

I would not hold...

I would not say that they're going fully closed, and I think also they got screwed up because the deep seek people, R1, did such a good job, right?

If you look at the reasoning model in deep seek, and in particular their ability to do reinforcement learning forward and back, forward and back, and forward and back, this is a major achievement, and it appears that they're doing it with less precision

than numeric precision than the American models.

As a bit of technical things, there's something called FP64, FP32, FP16.

The American models are typically using 16-bit precision for their training.

The Chinese are pushing 8 and now even 4.

Well, a number of the large companies have said that they want to be leaders in open source as well.

Sam Altman indicated that the smallest version of the O3 model would be released, I believe, open weights, and they have done so.

And he told me anyway that this model is much smaller than 10 to the 26th, it's much easier to train, and it will fit or can fit on your phone.

So one path is to say that we'll have these supercomputers doing AGI, which will always be incredibly expensive and so forth.

But we also have to watch to make sure that the proliferation of these models for handheld devices is under American control, whether it's OpenAI or Meta or Gemini or what have you.

I think you were the first investor or the earliest investor in it.

I was.

I'm sorry.

It's okay.

rockets are really cool and they're really hard I had I'm as you know I'm a pilot and I know lots about jets and I'd assume that rockets were as mature as jet engines they're not it is an art and a science these things are very hard to do the amounts of

In our case, the rocket is 4 million pounds of thrust.

You have to hold the thing down to test it, and you can't even hold it with metal things.

You have to have other things to hold it down as well.

There's so much force, otherwise it will take off.

Another interesting thing about rockets is that a rough number is that 2% of the weight of the rocket is the payload, 18% is roughly the rocket, and 80% is the propellant.

And my reaction as a new person is, you're telling me you can't do any better?

And the physicists say, after 60 years of physics, that's the best we can do to get out of the gravitation of the Earth.

And so I think rockets are interesting and they're challenging.

There's always an opportunity for competition.

In relativity space's area, it's essentially a LEO competitor.

So low Earth orbit, satellites, that sort of thing.

It's like a reunion of all of our former companies, David.

The order book is full.

We just have to launch the rocket.

Well, first of all, I'm a software person, not a hardware person.

I explain to people that hardware people go to different schools than software people.

and they think slightly differently.

So I'm always at a limitation in these new industries.

I had worked for the Secretary of Defense and have a top secret clearance and all that.

I was given a medal, et cetera, for trying to help the Pentagon reorganize itself.

And when the Ukraine war started, I was watching and I thought, well, here's an opportunity to see a country that has no Navy and no Air Force, how they do this with automation.

And indeed, it has been a spectacular success as a matter of innovation and outnumbered three to one with huge differences in kinetic strength, weapons, mobilization and so forth and so on.

Ukraine has held on really quite well.

And what's happening now is you're seeing essentially the birth of a completely new military national security structure.

One way to think about it is that we all, so first place, and I've seen it live, and I will tell you that real war is much worse than the worst movies you have ever seen about war, and that's all I'll say.

It's really horrific, and it's to be avoided at all cost.

And then, right, for obvious reasons.

Now, why is this all important?

And the...

And I love all these people who say, well, the warmongering talk, be careful what you wish for because the other side gets a vote.

When I started working and trying to understand what Ukraine was doing, Russia was pushed back and they've come back with a very, very strong second and third round.

So the enemy gets a vote in this situation.

But to go on, the rough way in which war will evolve is first things will have to be very, very mobile and very much not in fixed places.

This takes out most of the military infrastructure that exists in the world.

Things like tanks, of which we're now building a whole bunch more, even stronger tanks here in America, don't make any sense in a world where a two-kilogram payload from a well-armed drone can destroy the tank.

It's called the kill ratio, and that drone costs retail $5,000, $4,000.

The American tank costs $30 million.

You can send an awful lot of those drones to destroy those tanks.

The likely evolution goes something like this.

So first, people learn that drones are like rifles and like artillery.

So it's more efficient to use drones now than to use mortars, grenades, artillery.

That's clear.

If you just look at the economics, economics in terms of cost or effectiveness as it's called.

The next thing that happens is that both sides develop drone capabilities, which is what you're seeing now, and each then becomes a war of drone against drone.

So you have drone against anti-drone.

And so then the shift moves to how do you detect the enemy drone and how do you destroy it before it destroys you.

So the doctrine ultimately is the drones are forward and the people are behind.

And I've seen operations in, for example, sitting in Kyiv, where the Ukrainians are commanding things over Starlink, I might add, in the distant war, and they're very, very effective.

So we've solved the latency problems, we've solved the timing problems, and so forth in that area.

The ultimate state is very interesting and I don't think anyone has foreseen this.

If you go back to our conversation about RL and planning, which is what you're seeing with AI, let's say that we're on one side and we have a million drones and there's another side over here that has another million drones.

Each side will use reinforcement learning AI strategies to do battle plans.

but neither side can figure out what the other side's battle plan is.

And therefore, the deterrence against attacking each other will be very high.

Today, the way military planners operate is that they count weapons.

They say, well, you have this many, and I have this many, and you can do this kind of maneuver and so forth.

but in an ai world where you're doing reinforcement learning you can't count what the other side is planning you can't see it you don't know it and i believe that that will deter what i view is one of the most horrendous things ever done by humans which is war because unless there's a perfect balance between either side there will be some mutual destruction of the drone supply like there would be with any artillery stock and traditional warfare and whoever's left ends up winning like they're just

Well, it's very important to understand that there's no winners in war.

Eric Schmidt is here.

By the time you have a drone battle of the scale I'm describing, the entire infrastructure of your side will be destroyed, the entire infrastructure of the other side will be destroyed.

These are lose-lose scenarios.

It was my Gulfstream.

Well, I'm arguing that it's not a detente, it's a deterrence.

That as deterrence can be understood as I want to hit you, which I don't, but I want to hit you so much, but that if I do that,

That's right.

the penalty is greater than the value of me hitting you.

And that's how to transform.

He was way too smart.

Well, there are many advantages to moving to drones and automation.

One, they're much, much cheaper.

They're much, much cheaper.

And two, you can stockpile algorithms.

You can essentially learn and learn and learn.

And remember, you can also build training data that's synthetic, so you could be even better than the others.

The final question I've been asked by our military is, what's the role of a traditional land army?

Way too smart.

Way too smart.

And I wish I could say that all of these human behaviors can occur without humans being at risk.

I don't think so.

I think that the way robot war, essentially drone war will occur is there will be these destructive waves

but eventually humans are going to have to cross a line.

They're going to have to... After we've depleted them.

Well, wouldn't that be great to know, Maria? But they destroyed all the documents on their way out. So, again, this whole thing was propped up I JUST THINK YOU HAVE TO LOOK AT THE ARC OF THE STORY HERE. THE JANUARY 6TH COMMITTEE WAS SET UP FOR THE VERY SPECIFIC PURPOSE OF GIVING SOMEBODY LIKE JACK SMITH THE REASON TO GO CHARGE PRESIDENT TRUMP AND TAKE HIM OUT POLITICALLY. THAT'S THE WHOLE THING.

Well, wouldn't that be great to know, Maria? But they destroyed all the documents on their way out. So, again, this whole thing was propped up I JUST THINK YOU HAVE TO LOOK AT THE ARC OF THE STORY HERE. THE JANUARY 6TH COMMITTEE WAS SET UP FOR THE VERY SPECIFIC PURPOSE OF GIVING SOMEBODY LIKE JACK SMITH THE REASON TO GO CHARGE PRESIDENT TRUMP AND TAKE HIM OUT POLITICALLY. THAT'S THE WHOLE THING.

Well, wouldn't that be great to know, Maria? But they destroyed all the documents on their way out. So, again, this whole thing was propped up I JUST THINK YOU HAVE TO LOOK AT THE ARC OF THE STORY HERE. THE JANUARY 6TH COMMITTEE WAS SET UP FOR THE VERY SPECIFIC PURPOSE OF GIVING SOMEBODY LIKE JACK SMITH THE REASON TO GO CHARGE PRESIDENT TRUMP AND TAKE HIM OUT POLITICALLY. THAT'S THE WHOLE THING.

THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES.

THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES.

THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES. THE PRESIDENT OF THE UNITED STATES.

THE PRESIDENT OF THE UNITED STATES. all of it down, and won. Won in a mandate, won the pocket of a vote, and here we go now with an opportunity for a real reform agenda.

THE PRESIDENT OF THE UNITED STATES. all of it down, and won. Won in a mandate, won the pocket of a vote, and here we go now with an opportunity for a real reform agenda.

THE PRESIDENT OF THE UNITED STATES. all of it down, and won. Won in a mandate, won the pocket of a vote, and here we go now with an opportunity for a real reform agenda.