Ex Google CEO: AI Can Create Deadly Viruses! If We See This, We Must Turn Off AI! - Eric Schmidt

Well, first, thank you. I wanted to be on the show for a long time, so I'm really happy to be able to be here in person in London. Henry Kissinger, Dr. Kissinger, ended up being one of my greatest and closest friends. And 10 years ago, he and I were at a conference where he heard Demis Hassabis speak about AI. And Henry would tell the story that he was about to go catch up on his jet lag, but instead I said, 'Go do this.' And he listened to it, and all of a sudden he understood that we were playing with fire, that we were doing something that we did not understand it would have the impact on. And that Henry had been working on this since he was 22, coming out of the army after World War II and his thesis about Kant and so forth as an undergraduate at Harvard. So all of a sudden, I found myself in a whole group of people who are trying to understand what does it mean to be human in an age of AI. When this stuff starts showing up, how does our life change? How do our thoughts change? Humans have never had an intellectual challenger of our own ability or better or worse. It just never happened in history. The arrival of AI is a huge moment in history.

Well, like many of the people you meet, as a teenager I was interested in science. I played with model rockets, model trains, the usual things for a boy in my generation. I was too young to be a video game addict, but I'm sure I would be today if I were that age. I went to college and I was very interested in computers, and they were relatively slow then, but to me they were fascinating. To give you an example, the computer that I used in college is 100 million times slower than the phone you have in your pocket. And by the way, that was a computer for the entire university. So Moore's Law, which is this notion of accelerating density of chips, has defined the wealth creation, the career creation, the company creation in my life. So I can be understood as lucky because I was born with an interest in something which was about to explode. And when sort of everything happens together, everyone gets swept up in it, and of course the rest is history.

The most important thing is to develop analytical, critical thinking skills. I, to some level, I don't care how you get there. So if you like math or science, or if you like the law, or if you like entertainment, just think critically. In his particular case, as an 18-year-old, what I would encourage him to do is figure out how to write programs in a language called Python. Python is easy to use, it's very easy to understand, and it's become the language of AI. So the AI systems, when they write code for themselves, they write code in Python. And so you can't lose by developing Python programming skills. And the simplest thing to do with an 18-year-old man is say, 'Make a game,' because these are typically gamers, stereotypically. Make a game that's interesting using Python.

Yeah, a lot of people have posed this, and that's not correct. It sure looks like these systems will write code, but remember the systems also have interfaces called APIs which you can program them. So one of the large revenue sources for these AI models, because these companies have to make money at some point, right, is you build a program and you actually make an API call and ask it a question. Typical example is give it a picture and tell me what's in the picture. Now, can you have some fun with that as an 18-year-old? Of course. So when I say Python, I mean Python using the tools that are available to build something new, something that you're interested in.

Well, the first and most important thing about critical thinking is to distinguish between being marketed to, which is also known as being lied to, and being given the argument on your own. We've, because of social media, which I hold responsible for a lot of ills as well as good things in life, we've sort of gotten used to people just telling us something and believing it because our friends believe it or so forth. And I strongly encourage people to check assertions. So you get people say all this stuff, and I learned at Google all those years, somebody says something, I check it on Google. Do I? And you then have a question: do you criticize them and correct them, or do you let it go? But you want to be in the position where somebody makes a statement like, 'Did you know that only 10% of Americans have passports?' which is a widely viewed but false statement. It's actually higher than that, although it's never high enough in my view in America. But that's an example of an assertion that you can just say, 'Is that true?' Right? There's a long meme of American politicians where the Congress is basically full of criminals. It may be full of one or two, but it's not full of 90. But again, people believe this stuff because it sounds plausible. So if somebody says something plausible, just check it. You have a responsibility before you repeat something to make sure what you're repeating is true. And if you can't distinguish between true and false, I suggest you keep your mouth shut. Right? Because you can't run a government, a society without people operating on basic facts. Like for example, climate change is real. We can debate over how to address it, but there's no question the climate is changing. It is a fact. It is a mathematical fact. And how do I know this? And somebody will say, 'Well, how do you know?' And I said, because science is about repeatable experiments and also proving things wrong. So let's say I said that climate change is real and this was the first time it had ever been said, which is not true, then 100 people would say, 'That can't be true, I'll see if he's wrong.' And then all of a sudden they'd see I was right and I'd get some big prize, right? So the falsifiability of these assertions is very important. How do you know that science is correct? It's because people are constantly testing it.

Well, partly because AI will allow for perfect misinformation. So let's use an example of TikTok. TikTok can be understood, it's called the bandit algorithm in computer science, in the sense of the Las Vegas one-armed bandits. Do I stay in the bandit machine and I keep on this slot machine, or do I move to another slot machine? And the TikTok algorithm basically can be understood as, I'll keep serving you what you tell me you want, but occasionally I'll give you something from the adjacent area. And it is highly addictive. So what you're seeing with social media, and TikTok is a particularly bad example of this, is people are getting into these rabbit holes where all they see is confirmatory bias. And the ones that are, I mean, if it's fun and entertaining, I don't care. But you'll see, for example, there are plenty of stories where people have ultimately self-harm or suicide because they're already unhappy and then they start picking up unhappy, and then their whole environment online is people who are unhappy and it makes them more unhappy because it doesn't have a positive bias. So there's a really good example where, let's say in your case, you're the dad, you're going to watch this as the dad with your kid and you're going to say, 'You know, it's not that bad, let me show you some, let me give you some good alternatives, let me get you inspired, let me get you out of your funk.' The algorithms don't do that unless you force them to. It's because the algorithms are fundamentally about optimizing an objective function, literally mathematically maximize some goal that has been trained to. They just, in this case, it's attention. And by the way, part of why we have so much outrage is because if you're a CEO, you want to maximize revenue. To maximize revenue, you maximize attention. And the easiest way to maximize attention is to maximize outrage. Did you know? Did you know? Did you know? Right? And by the way, a lot of the stuff is not true. They're fighting over scarce attention. There was a recent article where there's an old quote from 1971 from Herb Simon, an economist at the time at Carnegie Mellon, who said that economists don't understand, but in the future the scarcity will be about attention. So somebody now 50 years later went back and said, 'I think we're at the point where we've monetized all attention.' An article this week: two and a half hours of videos consumed by young people every day. Right? Now, there is a limit to the amount of video you can, you know, that because you have to eat and sleep and to hang out. But these are significant societal changes that have occurred very, very quickly. When I was young, there was a great debate as to the benefit of television. And my argument at the time was, 'Well, yes, we did rock and roll and drugs and all of that, and we watched a lot of television, but somehow we grew up okay.' Right? So it's the same argument now with a different term. Will those kids grow up okay? It's not as obvious because these tools are highly addictive, much more so than television ever was.

I personally do, because I'm inherently an optimist. I also think that society begins to understand the problems. Typical example is there's an epidemic of harm to teenage girls. Girls, as we know, are more advanced than boys at those, you know, below, and the girls seem to get hit by social media at 11 and 12 when they're not quite capable of handling the rejection and the emotional stuff. And it's driven, you know, emergency room visits, self-harm, and so forth to record levels. It's well documented. So society is beginning to recognize this now. Schools won't let kids use their phones when they're in the classroom, which is kind of obvious if you ask me. So developmentally, one of the core questions about the AI revolution is what does it do to the identity of children that are growing up? Your values, your personal values, the way you get up in the morning and think about life is now set. It's highly unlikely that an AI will change your programming, but your child can be significantly reprogrammed. And one of the things that we talk about in the book is what happens when the best friend of your child from birth is a computer. What's it like now? By the way, I don't know, we've never done it before, but you're running an experiment on a billion people without a control, right? And so we have to stumble through this. So at the end of the day, I'm an optimist because we will adjust society with biases and values to try to keep us on a moral high ground, human life. And so you should be optimistic for that because these kids, when they grow up, they'll live to 100, their lives will be much more prosperous, I hope, and I pray that there'll be much less conflict. Certainly lifespans are longer, the likelihood of them being injured and in wars and so forth are much, much lower statistically.

So I have talked to them and to the others as well, and I think it's pretty straightforward. There's sort of good revenue and bad revenue. When we were at Google, Larry and Sergey and I, we would have situations where we would improve quality, you know, we would make the product better, and the debate was, do we take that to revenue in the form of more ads, or do we just make the product better? And that was a clear choice. And I arbitrarily decided that we would take 50% to one, 50% to the other, because I thought they were both important. And the founders, of course, were very supportive. So Google became more moral and also made more money. Right? All of the, there's plenty of bad stuff on Google, but it's not on the first page. That was the key thing. The alternative model would be, say, let's maximize revenue, we'll put all the really bad stuff, the lies and the cheating and the deceiving and so forth that draws you in, it will drive you insane, and we might have made more money, but first it was the wrong thing to do, but more importantly, it's not sustainable. There's a law called Gresham's Law, it's a verbal law obviously, where bad speech drives out good speech. And what you're seeing is you're seeing in online communities, which have always been present with bullying and this kind of stuff, now you've got crazy people in my view who are building bots that are lying, right, misinformation. Now why do you do that? There was a hurricane in Florida and people are in serious trouble and you sitting in the comfort of your home somewhere else are busy trying to make their lives more difficult. What's wrong with you? Like, let them get rescued, you know, human life is important. But there's something about the human psychology where people talk, there's a German word called Schadenfreude, you know, there's a bunch of things like this that we have to address. I want social media and the online world to represent the best of humanity: hope, excitement, optimism, creativity, invention, solving new problems, as opposed to the worst. And I think that that is achievable.

So the first rule I've learned is that you need a truly brilliant person to build a really brilliant product. And that is not me. I work with them. So find someone who's just smarter than you, more clever than you, moves faster than you, changes the world, is better spoken, more handsome, more beautiful, you know, whatever it is that you're optimizing, and ally yourself with them because they're the people who are going to make the world different. In one of my books, we use the distinction between divas and knaves. And a diva, and we use the example of Steve Jobs, who clearly was a diva, opinionated and strong and argumentative and would bully people if he didn't like them, but was brilliant. When he was, he was a diva, he wanted perfection, right? Aligning yourself with Steve Jobs is a good idea. The alternative is what we call a knave. And a knave, which you know from British history, is somebody who's acting on their own account, they're not trying to do the right thing, they're trying to benefit themselves at the cost of others. And so if you can identify a person in one of these teams that they're just trying to solve the problem in a really clever way and they're passionate about it and they want to do it, that's how the world moves forward. If you don't have such a person, your company's not going to go anywhere. And the reason is that it's too easy just to keep doing what you were doing, right? And innovation is fundamentally about changing what you're doing. Up until this generation of tech companies, most companies seemed to me to be one-hit wonders, right? They would have one thing that was very successful and then it would sort of, it was typically follow an S-curve and nothing much would happen. And now I think the people are smarter, people are better educated, you now see repeatable waves. A good example being Microsoft, which is, you know, an older company now, founded in basically '81, '82, something like that. So let's call that 45 years old, but they've reinvented themselves a number of times, right, in a really powerful way.

Well, I'm not sure I agree with the founder mode argument. And the reason is that it's great to have a brilliant founder, and it's actually more than great, it's like really important. And we need more brilliant founders. Universities are producing these people, by the way, they do exist and they show up every year, you know, another Michael Dell at the age of 19 or 22. These are just brilliant founders, obviously Gates and Ellison and sort of my generation of brilliant founders, Larry and Sergey and so forth.

So Larry Page and Sergey Brin met at Stanford. They were on a grant from, believe it or not, the National Science Foundation as graduate students. And Larry Page invented an algorithm called PageRank, which is named after him. And he and Sergey wrote a paper which is still one of the most cited papers in the world, and it's essentially a way of understanding priority of information. And mathematically it was a Fourier transform of the way people normally did things at the time. And so they wrote this code, I don't think they were that good a set of programmers, you know, they sort of did it. They had a computer, they ran out of power in their dorm room, so they borrowed the power from the dorm room next door and plugged it in. And they had the data center in the bedroom, you know, in the dorm, classic story. And then they moved to a building that was owned by the sister of a girlfriend at the time, and that's how they founded the company. Their first investor was the founder of Sun Microsystems, whose name was Andy Bechtolsheim, who just said, 'I'll just give you the money because you're obviously incredibly smart.' How much did he give them? $100,000? Yeah, maybe it was a million. But in any case, it ultimately became many billions of dollars. So it gives you a sense of this early founding is very important. So the founders then set up in this little house in Menlo Park, which ultimately we bought at Google, you know, as a museum. And they set up in the garage and they had 'Google World Headquarters' in neon made, and they had a big headquarters with the four employees that were sitting below them. And the computer that Larry and Sergey had built, Larry and Sergey were very, very good software people and obviously brilliant, but they were not very good hardware. And so they built the computers using corkboard to separate the CPUs. And if you know anything about hardware, hardware generates a lot of heat and the corkboard would catch on fire. So eventually when I showed up, we started building proper hardware with proper hardware engineers. But it gives you a sense of the scrappiness that was so characteristic. And you know, today there are people of enormous impact on society, and I think that will continue for many, many years.

Well, Larry said to me, now these were, they're very young, he looked at me and says, 'We don't need you now, but we'll need you in the future.' We'll need you in the future. Yeah. So one of the things about Larry and Sergey is that they thought for the long term. So they didn't say Google would be a search company, they said the mission of Google is to organize all the world's information. And if you think about it, that's pretty audacious 25 years ago, like how are you going to do that? And so they started with web search eventually. And Larry had studied AI quite extensively and he began to work and ultimately he acquired, with all of us obviously, this company called DeepMind here in Britain, which essentially is the first company to really see the AI opportunity. And pretty much all of the things you've seen from AI in the last decade have come from people who are either at DeepMind or competing with DeepMind.

Well, the first is to think about scale. I think a current example is look at Elon. Elon is an incredible entrepreneur and an incredible scientist, and if you study how he operates, he gets people by, I think, sheer force of personal will to overperform, to take huge risks, which somehow he has this brilliance where he can make those tradeoffs and get it right. So these are exceptional people. Now, in our book with Genesis, we argue that you're going to have that in your pocket, but whether you'll have the judgment to take the risks that Elon does, that's another question. One of the other ways to think about it is an awful lot of people talk to me about the companies that they're founding and they're a little widget, you know, like I want to make the camera better, I want to make the dress better, I want to make book publishing cheaper or so forth. These are all fine ideas. I'm interested in ideas which have the benefit of scale. And when I say scale, I mean the ability to go from zero to infinity in terms of the number of users and demand and scale. There are plenty of ways of thinking about this, but what would be such a company in the age of AI? Well, we can tell you what it would look like. You would have apps, one on Android, one on iOS, maybe a few others. Those apps will use powerful networks and they'll have a really big computer in the back doing AI calculations. So future success companies will all have that, right? Exactly what problem it solves, well, that's up to the founder. But if you're not using AI at every aspect of your business, you're not going to make it. And the distinction as a programming matter is that when I was doing all of this way back when, you had to write the code. Now AI has to discover the answer. It's a very big deal. And of course, a lot of this was invented at Google, you know, 10 years ago. But basically, all of a sudden, analytical programming, which is sort of what I did my whole life, you know, writing code and, you know, do this, do that, add this, subtract this, call this, so forth and so on, is gradually being replaced by learning the answer, right? So for example, we use the example of language translation. The current large language models are essentially organized around predicting the next word. Well, if you can predict the next word, you can predict the next sequence in biology, you can predict the next action, you can predict the next thing the robot should do. So all of this stuff around large language models and deep learning that has come out, the Transformer paper, GPT-3, ChatGPT, which for most people was this huge moment, is essentially about predicting the next word and getting it right.

So I'll give, well, it's almost always set, company cultures are almost always set by the founders. I happen to be on the board of the Mayo Clinic. Mayo Clinic is the largest healthcare system in America. It's also the most highly rated one. And they have a rule which is called the 'needs of the customer come first,' which came out of the Mayo brothers who've been dead for like 120 years. But that was their principle. And when I initially got on the board, I started wandering around, I thought this is kind of a stupid, you know, stupid phrase and nobody really does this. And they really believe it and they repeat it and they repeat it, right? So it's true in non-technical cultures, in that case it's a healthcare for service delivery, you can drive a culture even in non-tech. In tech, it's typically an engineering culture. And if I had to do things over again, I would have even more technical people and even fewer non-technical people and just make the technical people figure out what they have to do. And I'm sorry for that bias because I'm not trying to offend anybody, but the fact of the matter is the technical people, if you build the right product, your customers will come. If you don't build a product, then you don't need a sales force. Why are you selling an inferior product? So in the How Google Works book and ultimately in the Trillion Dollar Coach book, which is about Bill Campbell, we talked a lot about how the CEO is now the chief product officer, the chief innovation officer. Because 50 years ago you didn't have access to capital, you didn't have access to marketing, you didn't have access to sales, you didn't have access to distribution. I was meeting today with an entrepreneur who said, 'Yeah, you know, we'll be 95% technical.' And I said, 'Why?' He said, 'Well, we have a contract manufacturer and our products are so good that people will just buy them.' This happened to be a technical switching company, and they said it's only 100,000 times better than its competitors. And I said, 'It will sell, unfortunately it doesn't work yet.' Yeah, it isn't the point, but if they achieve their goal, people will be lined up outside the door. So as a matter of culture, you want to build a technical culture with values about getting the product to work right. And working means, another thing you do with engineers is you say, they make a nice presentation to you and they go, I said, 'That's very interesting, but you know, I'm not your customer. Your customer is really tough because your customer wants everything to work and free and work right now and never make any mistakes. So give me their feedback. And if their feedback is good, I love you. And if their feedback is bad, then you better get back to work and stop being so arrogant.' So what happens is that in the invention process within firms, people fall in love with an idea and they don't test it. One of the things that Google did, and this is largely Marissa Mayer back when, is one day she said to me, 'I don't know how to judge user interface.' Marissa was the previous CEO, she was the CEO of Yahoo, and before that she ran all the consumer products at Google. And she's now running another company in the Bay Area. But the important thing about Marissa is she said, 'I can't, I said, well, you know, the UI, the user interface is great at the time, and it certainly was.' And she said, 'I don't know how to judge the user interface myself and none of my team do, but we know how to measure.' And so what she organized were A/B tests. You test one, test another. So remember that it's possible using these networks to actually kind of figure out, because they're highly instrumented, dwell time, how long does somebody watch this, how important it is. If you go back to how TikTok works, one of the things, the signals that they use include the amount of time you watch, commenting, forwarding, sharing, all those kinds of things. And those you can understand those as analytics that go into an AI engine that then makes a decision as to what to do next, what to make viral.

So when I go back to Google to visit, and they were kind enough to give me a badge and treat me well, of course, I hear the echoes of this. I was at a lunch where there was a lady running search and a gentleman running ads, you know, the successors to the people who worked with me. And I asked them, 'What's it going?' And they said, 'The same problems, you know, the same problems have not been solved, but they're much bigger.' And so when you go to a company, I suspect, I was not near the founding of Apple, but I was on the board for a while, the founding culture you can see today in their obsession about user interfaces, their obsession about being closed and their privacy and secrecy. It's just a different company, right? I'm not passing judgment. Setting the culture is important, the echoes are there. What does happen in big companies is they become less efficient for many reasons. The first thing that happens is they become conservative because they're public and they have lawsuits. And a famous example is that Microsoft after the antitrust case in the '90s became so conservative in terms of what it could launch that it really missed the web revolution for a long time. They have since recovered, and I of course was happy to exploit that as a competitor to them when we were at Google. But the important thing is when big companies should be faster because they have more money and more scale, they should be able to do things even quicker. But in my industry anyway, the tech startups that have a new clear idea tend to win because the big company can't move fast enough to do it. Another example: we had built something called Google Video. I was very proud of Google Video. And David Drummond, who was the general counsel at the time, came in and said, 'You have to look at this YouTube people.' I said, 'Like, why? Who cares?' And it turns out they're really good and they're more clever than your team. And I said, 'That can't be true,' you know, typical arrogant Eric. And we sat down and we looked at it and they really work quicker, even though we had an incumbent. And why? It turns out that the incumbent was operating under the traditional rules that Google had, which was fine, and the competitor, in this case YouTube, was not constrained by that. They could work at any pace and they could do all sorts of things, intellectual property and so forth. Ultimately we were sued all over all of that stuff and we ultimately won all those suits. But it's an example where there are these moments in time where you have to move extremely quickly. You're seeing that right now with generative technology. So the AGI, the generative revolution, generate code, generate videos, generate text, generate everything, all of those winners are being determined in the next six, 12 months. And then once the slope is set, once the growth rate is, you know, quadrupling every six months or so forth, it's very hard for somebody else to come in. So it's a race to get there as fast as you can. So when you talk to the great venture capitalists, they are, they're fast, right? 'We'll look at it, we'll make a decision tomorrow, we're done, we're in,' and so forth. And we want to be first because that's where they make the most amount of money.

So harvesting and hunting is a good metaphor. I'm interested in entrepreneurs. And so what we learned at Google was ultimately if you want to get something done, you have to have somebody who's entrepreneurial in their approach in charge of a small business. And so for example, Sundar when he became CEO had a model of which were the little things that he was going to emphasize and which were the big things. Some of those little things are now big things, right? And he managed it that way. So one way to understand innovation in a large company is you need to know who the owner is. Larry Page would say over and over again, 'It's not going to happen unless there's an owner who's going to drive this.' And he was supremely good at identifying that technical talent, right? That's one of his great founder strengths. So when we talk about founders, not only do you have to have a vision, but you also have to have either great luck or great skill as to who is the person who can lead this. Inevitably those people are highly technical in the sense that they can and very quick moving and they have good management skills, right? They understand how to hire people and deploy resources that allows for innovation. Most of the, if I look back in my career, each generation of the tech companies failed, including for example Sun at the point at which it became noncompetitive with the future.

There are almost no examples of doing it simultaneously in the same building. The Macintosh was famously, Steve in his typical crazy way had this very small team that invented the Macintosh, and he put them in a little building next to the big building on Bandley Drive in Cupertino. And they put a pirate flag on top of it. Now, was that good culturally inside the company? No, because it created resentment in the big building. But was it right in terms of the revenue and path of Apple? Absolutely. Why? Because the Mac ultimately became the platform that established the UI, the user interface, ultimately allowed them to build the iPhone, which of course is defined by its user interface. Why couldn't they stay in the same building? It just doesn't work. You can't get people to play two roles. The incentives are different. If you're going to be a pirate and a disruptor, you don't have to follow the same rules. So there are plenty of examples where you just have to keep inventing yourself. Now, what's interesting about cloud computing and essentially cloud services, which is what Google does, is because the product is not sold to you, it's delivered to you, it's easier to change. But the same problem remains. If you look at Google today, right, it's basically a search box and it's incredibly powerful. But what happens when that interface is not really textual? Right? We'll have to reinvent that. Working on tech, the system will somehow know what you're asking, right? It will just, it will be your assistant. And again, Google will do very well, so I'm in no way criticizing Google here. But I'm saying that even something as simple as the search box will eventually be replaced by something more powerful. It's important that Google be the company that does that. I believe they will.

It's extremely hard for non-founders to do this in corporations. Because if you think about a corporation, what's the duty of the CEO? Many, there's the shareholders, there's the employees, there's the community, and there's a board. Trying to get a board of very smart people to agree on anything is hard enough. So imagine I walk in to you and I say, 'I have a new idea, I'm going to kill our profitability for two years, it's a huge bet, and I need a billion now.' Would the board say yes? Well, they did to Mark Zuckerberg. He spent all that money on essentially VR of one kind or another, doesn't seem to have produced very much. But at exactly the same time he invested very heavily in Instagram, WhatsApp, and Facebook, and in particular in the AI systems that power them. And today Facebook, to my surprise, is a very significant leader in AI, having released this language called or version called LLaMA, 400 billion, which is curiously an open source model. Open source means it's available freely for everyone. And what Facebook and Meta is saying is, as long as we have this technology, we can maximize the revenue in our core businesses. So there's a good example. And Zuckerberg is obviously an incredibly talented entrepreneur. He's now back on the list of the most rich people. He's figured out, you know, everything he was doing, and he managed to lose all that money while making a different bet. That's a unique founder. The same thing is almost impossible with a hired CEO.

Focus is important, but it's misinterpreted. In Google, we spent an awful lot of time telling people we wanted to do everything, and everyone said, 'You can't pull off everything.' And we said, 'Yes we can. We have the underlying architectures, we have the underlying reach, we can do this if we can imagine and build something that's really transformative.' And so the idea was not that we would somehow focus on one thing like search, but rather that we would pick areas of great impact and importance to the world, many of which were free, by the way, this is not necessarily revenue-driven. And that worked. I'll give you another example. There's an old saying in the business school that you should focus on what you're good at and you should simplify your product lines and you should get rid of product lines that don't work. Intel famously had a, the term is called ARM, it's a RISC chip. And this particular RISC chip was not compatible with the architecture that they were using for most of their products. And so they sold it. Unfortunately, this was a terrible mistake because the architecture that they sold off was needed for mobile phones with low memory, with small batteries and heat problems and so forth and so on. And so that decision, that fateful decision now 15 years ago, meant that they were never a player in the mobile space. And once they made that decision, they tried to take their expensive and complex chips and they kept trying to make cheaper and smaller versions, but the core decision, which was to simplify, simplify to the wrong outcome. Today, if you look at, I'll give you an example, the Nvidia chips use an ARM CPU and then these two powerful GPUs, it's called the B200, they don't use the Intel chip, they use the ARM chip because it was for their needs faster. I would never have predicted that 15 years ago. So at the end, maybe it was just a mistake, but maybe they didn't understand in the way they were organized as a corporation that ultimately battery power would be as important as computing power, right? The amount of battery you use, and that was the discriminant. So one way to think about it is if you're going to have these sort of simple rules, you better have a model of what happens in the next five years. So the way I teach this is just write down what it'll look like in five years. Just try, what will it look like in five years, your company or whatever it is, right? So let's talk about AI. What will be true in five years? That it's going to be a lot smarter than it is. Be a lot smarter. But how many companies will there be in AI? Will there be five or 5,000 or 50,000? 50,000. How many big companies will there be? Will there be new companies? What will they do? Right? So I just told you my view is that eventually you and I will have our own AI assistant which is a polymath, which is incredibly smart, which helps us guide through the information overload that it is today. Who's going to build it? Make a prediction. What kind of hardware will be on? Make a prediction. How fast will the networks be? Make a prediction. Write all these things down and then have a discussion about what to do. What is interesting about our industry is that when something like the PC comes along or the internet, I lived through all of these things, they are such broad phenomena that they really do create a whole new lake, a whole new ocean, whatever metaphor you want. Now people said, 'Well, wasn't that crypto?' No, crypto is not such a platform. Crypto is not transformative to daily life for everyone. People are not running around all day using crypto tokens rather than currency. Crypto is a specialized market. By the way, it's important and it's interesting, it's not a horizontal transformative market. The arrival of alien intelligence in the form of a savant that you use is such a transformative thing because it touches everything. It touches you as a producer, as a star, as a narrative. It touches me as an executive. It will ultimately help people make money in the stock market, people are working on that. There's so many ways in which the technology is transformative. To start you in your case, when you think about your company, whether it's little, you know, itty bitty, or a really big one, it's fundamentally how will you apply AI to accelerate what you're doing, right? In your case, for example, here you have, I think, the most successful show in the UK by far, right? So how will you use AI to make it more successful? Well, you can ask it to distribute you more, right? To make narratives, to summarize, to come up with new insights, to suggest, to have fun, to create contests. There are all sorts of ways that you can ask AI. I'll give you a simple example. If I were a politician, thankfully I'm not, and I knew my district, I would say to the computer, 'Write a program.' So I'm saying to the computer, you write a program which goes through all the constituents in my interest, figures out roughly what they care about, and then send them a video which is labeled, you know, of me digitally, so I'm not fake, but it's kind of like my intention, where I explain to them how important I as their constituent have made the bridge work, right? And you sit there and you go, 'That's crazy,' but it's possible now. Politicians have not discovered this yet, but they will, because ultimately politicians are around a human connection and the quickest way to have that communication is to be on their phone talking to them about something that they care about.

Well, remember that Google also, that's the old question of why did you not do Facebook? Well, the answer is we were doing everything else, right? So my defensive answer is that Google has eight or nine or 10 billion user clusters of activity, which is pretty good, right? It's pretty hard to do, right? I'm very proud of that. I'm very proud of what they're doing now. My own view is that what happened was Google was working in the engine room and a team out of OpenAI figured out a technology called RLHF. And what happened was when they did GPT-3 and GPT, the T is Transformer, which was invented at Google, when they did it, they had sort of this interesting idea and then they sort of casually started to use humans to make it better. And RLHF refers to the fact that you use humans at the end to do A/B tests where humans can actually say, 'Well, this one's better,' and then the system learns recursively from human training at the end. That was a real breakthrough, right? And I joke with my OpenAI friends that you were sitting around on Thursday night and you turn this thing on and you go, 'Holy crap, look how good this thing is.' It was a real discovery, right? That none of us expected, certainly I did not. And once they had it, the OpenAI people, Sam and so forth, we'll talk about this, they didn't really understand how good it was. They just turned it on and all of a sudden they had this huge success. Disaster, because they were working on GPT-4 at the same time, it was an afterthought. It's a great story because it just shows you that even the brilliant founders do not necessarily understand how powerful what they've done is. Now today, of course, you have GPT-4o, basically a very powerful model from OpenAI. You have Gemini 1.5, which is clearly roughly equivalent, if not better in certain areas. The Gemini is more multimodal, for example. And then you have other players, LLaMA, the LLaMA architecture, L-L-A-M-A, does not stand for llamas, it's large language models, out of Facebook and a number of others. There's a startup called Anthropic, which is very powerful, founded by one of the inventors of GPT-3 and a whole bunch of people. And they formed their company knowing they were going to be that successful. It's interesting, they actually formed as part of their incorporation that they were a public benefit corporation because they were concerned that it would be so powerful that some evil CEO in the future would force them to go for revenue as opposed to world goodness. So the teams, when they were doing this, they understood the power of what they were doing and they anticipated the level of impact, which, and they were right.

Well, Tim has done a fantastic job in Steve's legacy. And what's interesting is normally the successor is not as good as the founder, but somehow Tim, having worked with Steve for so long and having set the culture, having Steve having, they've managed to continue the focus on the user, this incredible safety focus in terms of apps and so forth and so on, and they've remained a relatively closed culture. I think all of those would have maintained had Steve, you know, tragically died. He was a good friend. But the important point is Steve believed very strongly in what are called closed systems where you own and control all your intellectual property. And he and I would battle over open versus closed because I came from the other side. And I did this with respect. I don't think they would have changed that. And they've changed that now? No, I think still Apple is still basically a single company that's very highly integrated. The rest of the industry is largely more open.

When Steve and I worked together very closely, which was what, 15 years ago for his death, he was very frustrated at the success of MP4 over MOV format files. And he was really mad about it. And I said, 'Well, you know, maybe that's because you were closed and QuickTime was not generally available.' He said, 'That's not true, my team, you know, our product is better and so forth.' So his core belief system, he's an artist, right? And given the choice, we used to have this debate, where do you want to be Chevrolet or do you want to be Porsche? Do you want to be, you know, General Motors or do you want to be BMW? And he said, 'I want to be BMW.' And during that time, Apple's margins were twice as high as the PC companies. And I said, 'Steve, you don't need all that money, you're generating all this cash, you're giving it to your shareholders.' And he said, 'The principle of our profitability and our value in our brand is this is this luxury brand, right?' So that's how he thought. Now, how would AI change that? Everything that he would have done with Apple today would be AI-inspired, but it...

Everything that he would have done with Apple today would be AI-inspired, but it would be in the context of the Apple ecosystem, which is what he built. And I think that's the key difference. He would have found a way to make AI work within the closed system that he believed in, rather than the open system that the rest of the industry is moving towards. And that's a fundamental philosophical difference that I think would have played out in very interesting ways.

Thank you. It's been a pleasure. I've enjoyed the conversation.

Welcome to the replacement for Siri. And again, would Steve have done that quicker? I don't know. It's very clear that the first thing Apple needs to do is have Siri be replaced by an AI and call that Siri.

Startups, by definition, are huge risk takers. You have no history, you have no incumbency, you have all these competitors by definition, and you have no time. So in a startup, you want to prioritize intelligence and quickness over experience and sort of stability. You want to take risks on people. And part of the reason why startups are full of young people is because young people often don't have the baggage of executives who have been around for a long time, but more importantly, they're willing to take risks. So it used to be that you could predict whether a company was successful by the age of the founders, and in that 20 and 30-year-old period, the company would be hugely successful. Startups wiggle, they try something, they try something else, and they're very quick to discard an old idea. Corporations spend years with a belief system that is factually false, and they don't actually change their opinion until after they've lost all the contracts. And if you go back, all the signs were there. Nobody wanted to talk to them, nobody cared about the product, right? And yet they kept pushing it. So if you're a CEO of a larger company, what you want to do is basically figure out how to measure this innovation so that you don't waste a lot of time. Bill Gates had a saying a long time ago, which was that the most important thing to do is to fail fast. Right? From his perspective as the CEO and founder of Microsoft, he wanted everything to happen and he wanted to fail quickly, and that was his theory.

Yeah, I do. Fast failure is important. You can say it in a nicer way, but fundamentally, at Google, we had this 70-20-10 rule that Larry and Sergey came up with: 70% of the core business, 20% on adjacent business, and 10% on other. What does that mean? Core business means search ads. Adjacent business means something that you're trying, like a cloud business or so forth. And the 10% is some new idea. So Google created this thing called Google X. The first product it built was called Google Brain, which is one of the first machine learning architectures. This actually precedes DeepMind. Google Brain was used to power the AI system. Google Brain's team of 10 or 15 people generated 10, 20, 30, 40 billion dollars of extra profits over a decade. So that pays for a lot of failures, right? Then they had a whole bunch of other ideas that seemed very interesting to me that didn't happen for one or another, and they would cancel them. And then the people would get reconfigured. And one of the great things about Silicon Valley is it's possible to spend a few years on a really bad idea and get cancelled, if you will, and then get another job having learned all of that. My joke is the best CFO is one who's just gone bankrupt, because the one thing that CFO is not going to let happen is to go bankrupt again.

It's more complicated than that. So Larry and Sergey started TGIF, which I obviously participated in, and we had fun. There was a sense of humor. It was all off the record. A famous example is the VP of sales, whose name was Omid, was always predicting lower revenue than we really had, which is called sandbagging. So we got a sandbag and we made him stand on the sandbag in order to present his numbers. It was just fun, humorous, you know, we had skits and things like that. At some size, you don't have that level of intimacy and you don't have a level of privacy. And what happened was there were leaks. Eventually, there was a presentation, I don't remember the specifics, where the presentation was ongoing and someone was leaking the presentation live to a reporter. And somebody came on stage and said, 'We have to stop now.' I think that was the moment where the company got sort of too big.

There's some. In Google's case, we had a position of why lay people off, just don't hire them in the first place. It's much, much easier. And so in my 10 years, the only layoff we did was 200 people in the sales structures right after the 2000 epidemic. And I remember it as being extremely painful. Right? It was the first time we had done it. So we took the position, which is different at the time, that you shouldn't have an automatic layoff. What would happen is that there was a belief at the time that every six months or nine months you should take the bottom 5% of your people and lay them off. The problem with that is you're assuming the 5% are correctly identified. And furthermore, even the lowest performers have knowledge and value to the corporation that we can take. So we took a very much more positive view of our employees, and the employees liked that, and we obviously paid them very well and so forth and so on. I think that the cultural issues ultimately have been addressed, but during there was a period of time where there were, because of the free-wheeling nature of the company, there were an awful lot of internal distribution lists which had nothing to do with the company.

They were distribution lists on topics of war, peace, politics, so forth.

A distribution, like an email dist, think of it as a message board, okay? Roughly speaking, think of it as message boards for employees. And at one point, somebody discovered that there were 100,000 such message boards. And the company ultimately cleaned that up because companies are not like universities, and there are in fact all sorts of laws about what you can say and what you cannot say and so forth. And so, for example, the majority of the employees were Democrats in the American political system. And I made a point, even though I'm a Democrat, to try to protect the small number of Republicans because I thought they had a right to be employees too. So you have to be very careful in a corporation to establish what does speech mean within the corporation. And what you are hearing as wokeism is really, can be understood as, what are the appropriate topics on work time in a work venue? Should you be discussing? My own view is stick to the business, and then please feel free to go to the bar, scream your views, talk to everybody. You know, I'm a strong believer in free speech, but within the corporation, let's just stick to the corporation and its goals.