The Exchanges

Q And, and that was celebrated at the time, right? Because we were excited about sort of more integration with China. So that wasn't seen as such a problematic thing at the time, right?

A Oh, it's, it's absolutely, it's even more than that. Like, this was like hailed as like, this is the modern way to do business. Like, uh, like a very An example of this that's just super close to home about how things have changed was that in 1999, the World Trade Organization gathered in Seattle to, um, ratify China as a member of the World Trade Organization. And this is the organization that, like, that sort of navigates tariffs and trade rules between countries. All the cool countries were in it, but not, like, communist, socialist China, you know, without With, you know, a totalitarian government and all that, and so the Clinton administration had really pushed for China to come in, and it actually split the Democratic Party. People today, their heads would explode if you tried this out, and it turns out the Republicans were split too. Yeah. Because half the Republicans were like, free trade, free trade, free trade, and the other half were, were like, we hate communism, we hate communism, and, and, and so the whole world, if you look back in 1998, it look, it's like completely upside down. Yeah. From today. And, and there were huge riots in Seattle. Like, the city was like a mess over this moment, because it was viewed as this basic, like, this, this pro-business, anti-labor kind of thing, because it just meant cheap outsourcing in China. Now, the reason for that was becau…

AI assessment note: “Oh, it's, it's absolutely, it's even more than that.”

Q Anish, Steven, we were having such a good conversation offline that I wanted to, to, to get this on the podcast. There are a few topics we wanted to, to discuss. First, we were all fascinated by Andre Karpathy's talk at, at, at startup school. Steven, what did you find so interesting or what were your takeaways or reactions from it?

A Well, I totally love the, the talk. He did an unbelievable, like, a philosopher king version of, of where we are, and I just found his, his metaphors really compelling. In fact, what, what I might do is even take it further back and just say, since he used an analogy of, like, where we are in computing, and lots of people in talking about the Windows three era and stuff like that, and having lived through all of them, I tend to think we're at the It was like, okay, we have 64 K IBM PC era of the microcomputer. okay, we have 60 And the reason I think that is, is, is actually a technical one, which is that we're at the point where people are still trying to figure out how everything works. And all the coding and all of the energy is working around like these very basic working problems. Like with the PC, it was like, okay, we have 64 K of memory and our programs are all too big. No display and all these problems. And with AI, you know, people are like, it's going to replace search. It's going to replace Excel and it's going to replace all these things, but it doesn't add very well. It gives you a lot of errors. Like the thing that you say it's going to do, it just doesn't even do yet. So I feel like we're at a point that is just so, so early and he did a fantastic job of sort of making that arc.

AI assessment note: “I just found his, his metaphors really compelling.”

Q lead both, uh, Office and Windows. Um, and you helped introduce the Surface PCs, which is relevant to the, the conversation we're, we're about to have. So, um, we're here to talk about Apple. We're here to talk about, you know, inspired by the book, Apple in China, but first let's, let's briefly cover, uh, WWDC. Uh, what, what, what reactions or how it takes do we have from it?

A WWC is, you know, their big yearly developer event. It's a remarkable platform that's been doing yearly releases for 26 years. Uh, it blows my mind, the, what they've done and the scale that they operate at. But I think, look, there are three big things that came out of it. The, the first was, you know, obviously the big push. It was also the big reaction, which is, was called Liquid Glass, and it's the new user interface design. So having gone through multiple big user interface redesigns of widely used products, although to be humble about it, not one used by a billion people, the reaction is extremely predictable, which is a bunch of people, their heads just explode and they, they don't want it. They don't like it. You can always tell they, they dive in immediately to what problem does it solve? So if there's change, it must solve a problem, but there are lots of reasons that designs change. And then there's the reality. This was the developer conference. So it's not done. And, and Apple is very, very good at finishing these things. And all the bugs and all the issues, I can't read it, it's blurry, it's noisy, th- those will all get fixed. You know, and what's left is, is it, is it great, and is it unique enough? And time will tell. I, I basically am reserving judgment and not joining in the hysteria on, on either side. I have no need to defend it, and no need to say it's th…

AI assessment note: “there are three big things that came out of it”

Q something of like a fifteen billion dollar acquihire of Alex plus the top talent at scale? And does, if you're Apple, if you're advising Apple or, you know, do they or even some of the other big incumbents need to do something similar? Obviously Microsoft, you know, is what we'll say with opening ad to begin with. But if you're Apple, how do you think about how to play here?

A Well, I think, you know, the, the two plays are gonna, or three. One is gonna be the sort of the weird partnership dance play that Microsoft and OpenAI have. I, I, it seems obvious that at some point, you know, Microsoft will have much more first party software. There's the, we have always made our play by just having everything, which is, which is, um, Amazon. And so you could, anything that comes out, they're gonna, End up supporting, which is just how retail works. Like, you don't, you get all the cereals, not just Froot Loops, you know, and, and then I, I think there's just the, where we're gonna go at our own way, in a sense, which is Gemini taking their approach, or OpenAI itself, or Anthropic itself, and I, I think that Apple needs to pick one. You know, they did announce like, Hey, you can call any model through our API, but the real thing that Apple needs to do is figure out what model is going to be tuned to run on their unique hardware on the edge and their unique offering or constraint, depending on, of privacy and how that plays out.

AI assessment note: “the real thing that Apple needs to do is figure out what model”

Q And, and that was celebrated at the time, right? Because we were excited about sort of more integration with China. So that wasn't seen as such a problematic thing at the time, right?

A Oh, it's, it's absolutely, it's even more than that. Like, this was like hailed as like, this is the modern way to do business. Like, uh, like a very An example of this that's just super close to home about how things have changed was that in 1999, the World Trade Organization gathered in Seattle to, um, ratify China as a member of the World Trade Organization. And this is the organization that, like, that sort of navigates tariffs and trade rules between countries. All the cool countries were in it, but not, like, communist, socialist China, you know, without With, you know, a totalitarian government and all that, and so the Clinton administration had really pushed for China to come in, and it actually split the Democratic Party. People today, their heads would explode if you tried this out, and it turns out the Republicans were split too. Yeah. Because half the Republicans were like, free trade, free trade, free trade, and the other half were, were like, we hate communism, we hate communism, and, and, and so the whole world, if you look back in 1998, it look, it's like completely upside down. Yeah. From today. And, and there were huge riots in Seattle. Like, the city was like a mess over this moment, because it was viewed as this basic, like, this, this pro-business, anti-labor kind of thing, because it just meant cheap outsourcing in China. Now, the reason for that was becau…

AI assessment note: “Oh, it's, it's absolutely, it's even more than that.”

Q Um, Okay, so the, when we look at Apple's dependency on, on China, when does this go from sort of, okay, um, this is great to, okay, this is actually an existential risk for the, for the company, uh, or, or would you say we, we, we are there yet, or how is this developed, you know, into the, into the current era?

A The easiest Way to view the wake up call is, is through the lens of COVID. Right. I, I think, which is all of a sudden the entire world learned that this global system, while great for prices, great for in search of excellence, focus on what you're knitting and what you do well, it turns out it's, it's actually pretty fragile. Yeah. And it's fragile because there's single points of failure all over the place. Yeah. And, and in this case, the single point of failure turned out to be like, well, you, Can't go all, if this city is closed, no one can go into the factory. Yeah. And then if, if transit is closed, then the stuff can't make it. And so, I, I just think COVID was the wake-up call for the whole system. And of course, at a national security level, it's, it's massively important now. Because, you know, you can't afford to have a, a nation shut down.

AI assessment note: “The easiest Way to view the wake up call is, is through the lens of COVID.”

Q Is there anything else you want people to take either from, from this book or from this, this conversation kind of in summary?

A Well, I, I think, I would sort of just drill down on one issue that I just think is, is super important for people to really consider in the competition between the US and China, and also how it relates to AI, which is the difficulty we're all gonna have collectively to navigating the world of intellectual property. Yeah. You know, I, I just think, I think it's very, very easy to say China doesn't respect intellectual property, so they can't be part of the world stage, flat out. I mean, they've crushed the pharmaceutical industry that way. You know, what they've done with BYD and Tesla, Arguably pretty rude and stuff like that. But on the other hand, it's just as easy to say, well, because of AI, all knowledge, all content, all information just needs to be free to train and to reuse. Neither of those are realistic positions. And, and I, I think that this issue is far more subtle and far more important than, than I think either China thinks or the, sort of the Digerati in the U.S. think. And I, I believe that that's one that's gonna need a lot, and I worry about wild cards. Like, I worry about the European Union going crazy in one direction or another and passing some rule, then all of a sudden that market is weird. Or Japan, which was like, went the other way because they have a language challenge, which is Japanese Like, they could just, they're in control of Japanese, so they…

AI assessment note: “I would sort of just drill down on one issue that I just think is, is super important”

Q Yeah, and so, Let's get, now get back to when did that start to change? I don't know if it's going to the book, or someone said, hey, China has Apple by the balls, or has us by the balls. Let's talk about that, and when did we realize that happened?

A At one point, it's the point of no return. So the point of no return was basically, I would say, two years into the iPhone, the scale was such, there was nowhere else. I'm sure Apple would just argue, oh, there was no way we could even have started the iPhone anywhere else. And I can tell you, we, like, Surface, there was no chance. We were, and that's a low volume product. Like, there was no, no chance. And most things are very low volume. Which makes the, the non-recurring engineering costs much higher, so even more impossible to build. All the startups, you know, you can't even consider doing that. So, I, I, like, the point of no return happened so long ago that it's sort of a weird thing to think about fifty billion dollars a year in investment, because that's not what did it. What did it was, was this unique blend of socialism or totalitarianism and entrepreneurship. Yeah. And that's the thing that all the experts got wrong. All the experts in 1999 were like two things. Yay for global trade. This is what we all want. And also, don't worry. China is, is, they're gonna stay a third world dictatorship forever. They couldn't have been more wrong. And I, frankly, that's how I felt living there.

AI assessment note: “two years into the iPhone, the scale was such, there was nowhere else”

Q Yeah, and so, Let's get, now get back to when did that start to change? I don't know if it's going to the book, or someone said, hey, China has Apple by the balls, or has us by the balls. Let's talk about that, and when did we realize that happened?

A At one point, it's the point of no return. So the point of no return was basically, I would say, two years into the iPhone, the scale was such, there was nowhere else. I'm sure Apple would just argue, oh, there was no way we could even have started the iPhone anywhere else. And I can tell you, we, like, Surface, there was no chance. We were, and that's a low volume product. Like, there was no, no chance. And most things are very low volume. Which makes the, the non-recurring engineering costs much higher, so even more impossible to build. All the startups, you know, you can't even consider doing that. So, I, I, like, the point of no return happened so long ago that it's sort of a weird thing to think about fifty billion dollars a year in investment, because that's not what did it. What did it was, was this unique blend of socialism or totalitarianism and entrepreneurship. Yeah. And that's the thing that all the experts got wrong. All the experts in 1999 were like two things. Yay for global trade. This is what we all want. And also, don't worry. China is, is, they're gonna stay a third world dictatorship forever. They couldn't have been more wrong. And I, frankly, that's how I felt living there.

AI assessment note: “the point of no return was basically, I would say, two years into the iPhone”

Q something of like a fifteen billion dollar acquihire of Alex plus the top talent at scale? And does, if you're Apple, if you're advising Apple or, you know, do they or even some of the other big incumbents need to do something similar? Obviously Microsoft, you know, is what we'll say with opening ad to begin with. But if you're Apple, how do you think about how to play here?

A Well, I think, you know, the, the two plays are gonna, or three. One is gonna be the sort of the weird partnership dance play that Microsoft and OpenAI have. I, I, it seems obvious that at some point, you know, Microsoft will have much more first party software. There's the, we have always made our play by just having everything, which is, which is, um, Amazon. And so you could, anything that comes out, they're gonna, End up supporting, which is just how retail works. Like, you don't, you get all the cereals, not just Froot Loops, you know, and, and then I, I think there's just the, where we're gonna go at our own way, in a sense, which is Gemini taking their approach, or OpenAI itself, or Anthropic itself, and I, I think that Apple needs to pick one. You know, they did announce like, Hey, you can call any model through our API, but the real thing that Apple needs to do is figure out what model is going to be tuned to run on their unique hardware on the edge and their unique offering or constraint, depending on, of privacy and how that plays out.

AI assessment note: “the real thing that Apple needs to do is figure out what model is going to be tuned”

Q then to build, uh, help build Tesla, you know, which are phones on wheels, and then led to other sort of, um, you know, technologies used for the military, which we'll, we'll get to. But, but first, I just want to ask, Was this inevitable or was this, did this make sense at the time? How do we make sense of this? 20 years later, how did this all happen?

A It is crazy that how it all happened. I mean, I think you go back, like, of course, Apple, we all know it was in the garage in Palo Alto. They, they famously built the computers there. Um, but the thing was, from the very beginning, Apple itself was all about super tight control. I mean, first you had Woz, you know, basically creating works of art on the motherboard, and then they built a factory. Yeah. You know, there was the factory right here, and that's where they built computers. And, and they, they just, Kept doing more of that. Tight control. Build the whole computer. Then, um, Steve left Apple. Macintosh, of course, was famously the same thing, but, and Steve pushed and pushed on costs, and the shape of it, made the famous toaster that you carry around, and things like that. But after Steve left, the company had enormous difficulty competing with the IBM, uh, PC. One of the things that it did, sort of, um, Um, later, like in the, in the late 19 nineties, was a super innovative laptop. Now, everybody's like, what does an innovative laptop mean? Well, it turns out in the late 19 nineties, laptops, nobody bought them. They were, they were slow, expensive, bad computers. They were just portable. That was their big thing. And if you had to have one, you did. But they built this thing that was the original portable for Mac called PowerBook. And They tried to cram a lot of, th…

AI assessment note: “I think you go back, like, of course, Apple... from the very beginning”

Q then to build, uh, help build Tesla, you know, which are phones on wheels, and then led to other sort of, um, you know, technologies used for the military, which we'll, we'll get to. But, but first, I just want to ask, Was this inevitable or was this, did this make sense at the time? How do we make sense of this? 20 years later, how did this all happen?

A It is crazy that how it all happened. I mean, I think you go back, like, of course, Apple, we all know it was in the garage in Palo Alto. They, they famously built the computers there. Um, but the thing was, from the very beginning, Apple itself was all about super tight control. I mean, first you had Woz, you know, basically creating works of art on the motherboard, and then they built a factory. Yeah. You know, there was the factory right here, and that's where they built computers. And, and they, they just, Kept doing more of that. Tight control. Build the whole computer. Then, um, Steve left Apple. Macintosh, of course, was famously the same thing, but, and Steve pushed and pushed on costs, and the shape of it, made the famous toaster that you carry around, and things like that. But after Steve left, the company had enormous difficulty competing with the IBM, uh, PC. One of the things that it did, sort of, um, Um, later, like in the, in the late 19 nineties, was a super innovative laptop. Now, everybody's like, what does an innovative laptop mean? Well, it turns out in the late 19 nineties, laptops, nobody bought them. They were, they were slow, expensive, bad computers. They were just portable. That was their big thing. And if you had to have one, you did. But they built this thing that was the original portable for Mac called PowerBook. And They tried to cram a lot of, th…

AI assessment note: “they ultimately were hitting the limits of manufacturing in the US.”

Q And just to evaluate one of those paths, if Satya could go back in time, knowing what he knows now, do you think he might've done something or would you have considered doing something different? Um, how, how should we think about how this has worked out for Microsoft?

A Oh, these are tricky things. You can't ask me questions about my friends, but I, I wouldn't, putting it aside as advice, I, I would say, look, I, I genuinely, to those listening, I do not know anything that's going on. Like, um, but there, Microsoft has been involved in, in AI research since 1993. They were the first hires into Microsoft research when the labs were formed. I was working for Bill way back then. And, you know, Satya worked closely with those researchers when he was the manager of the Bing team for a while, because that was the cornerstone of what Bing was doing, because that's what Google was doing in, say, 2000 six, seven, something like that. And I think that, um, I, I would not be at all surprised, I don't know anything, I swear people, you know, that there's some first party thing that they're just trying to figure out when and how good does it have to be, because if you are Microsoft, it turns out that, like, The strength of Microsoft is that you never have to be best. You just have to be included. Yeah. And, and so it's, it's likely that that will be their play. That's generally how enterprise software works. It's the inclusion and the bundle.

AI assessment note: “You can't ask me questions about my friends, but I, I wouldn't”

Q this, on this innovation. What, you know, we just saw Meta, uh, sort of, um, you know, purchase 49% of, Of, of scale. So, so they're making some big moves. How do you make sense of, of, of Meta's move? And could we see Apple do a similar type of play where they sort of buy the talent in house or buy the company? How do you think about that?

A I, I think what, what Meta is doing is at the very least, it's great for AI. You know, I'll let Meta decide how great it is for Meta or whatever, but it's great for AI. And the reason is, is that if you go back, the way platforms have been evolving, You know, the, the, the era of, you know, in mainframes, IBM had a hundred percent. In PCs, Microsoft had 95%. Then you start getting out there, and then in servers actually ended up being like, at best case, fifty-fifty, and it's been downhill, and it's all Linux now. In phones, you, you have this split, you know, eighty-twenty globally, but you have to really dig down by country where it's eighty-twenty the other way. And in cloud, it's very likely we're gonna end up in a, you know, in a sort of a forty-forty-twenty kind of world. And, and so AI, it's almost inconceivable that there'll be one winner. Right. And there's, the overlay is China, and how that plays out. And the mobile industry is a really good example of how difficult it is. You know, the smartphones, so to speak, Japan had this huge lead in this thing called iMode, but it only existed in Japan. And so, It, you know, like they were leaders in technology and experience. They had all these micropayments, all this stuff, but like it, what, it didn't go anywhere outside of the country. And so the biggest risk to AI is that it, it becomes either it, someone tries to think i…

AI assessment note: “I'll let Meta decide how great it is for Meta or whatever”

Q this, on this innovation. What, you know, we just saw Meta, uh, sort of, um, you know, purchase 49% of, Of, of scale. So, so they're making some big moves. How do you make sense of, of, of Meta's move? And could we see Apple do a similar type of play where they sort of buy the talent in house or buy the company? How do you think about that?

A I, I think what, what Meta is doing is at the very least, it's great for AI. You know, I'll let Meta decide how great it is for Meta or whatever, but it's great for AI. And the reason is, is that if you go back, the way platforms have been evolving, You know, the, the, the era of, you know, in mainframes, IBM had a hundred percent. In PCs, Microsoft had 95%. Then you start getting out there, and then in servers actually ended up being like, at best case, fifty-fifty, and it's been downhill, and it's all Linux now. In phones, you, you have this split, you know, eighty-twenty globally, but you have to really dig down by country where it's eighty-twenty the other way. And in cloud, it's very likely we're gonna end up in a, you know, in a sort of a forty-forty-twenty kind of world. And, and so AI, it's almost inconceivable that there'll be one winner. Right. And there's, the overlay is China, and how that plays out. And the mobile industry is a really good example of how difficult it is. You know, the smartphones, so to speak, Japan had this huge lead in this thing called iMode, but it only existed in Japan. And so, It, you know, like they were leaders in technology and experience. They had all these micropayments, all this stuff, but like it, what, it didn't go anywhere outside of the country. And so the biggest risk to AI is that it, it becomes either it, someone tries to think i…

AI assessment note: “I think what, what Meta is doing is at the very least, it's great for AI.”

Q And just to evaluate one of those paths, if Satya could go back in time, knowing what he knows now, do you think he might've done something or would you have considered doing something different? Um, how, how should we think about how this has worked out for Microsoft?

A Oh, these are tricky things. You can't ask me questions about my friends, but I, I wouldn't, putting it aside as advice, I, I would say, look, I, I genuinely, to those listening, I do not know anything that's going on. Like, um, but there, Microsoft has been involved in, in AI research since 1993. They were the first hires into Microsoft research when the labs were formed. I was working for Bill way back then. And, you know, Satya worked closely with those researchers when he was the manager of the Bing team for a while, because that was the cornerstone of what Bing was doing, because that's what Google was doing in, say, 2000 six, seven, something like that. And I think that, um, I, I would not be at all surprised, I don't know anything, I swear people, you know, that there's some first party thing that they're just trying to figure out when and how good does it have to be, because if you are Microsoft, it turns out that, like, The strength of Microsoft is that you never have to be best. You just have to be included. Yeah. And, and so it's, it's likely that that will be their play. That's generally how enterprise software works. It's the inclusion and the bundle.

AI assessment note: “You can't ask me questions about my friends”

Q And so let's say Trump brings you in and says, hey, I've noticed that, uh, you know, nearly all of computing is, uh, you know, every layer of stack is, is done in China and, and, you know, Asia more broadly, um, and I'm trying to sort of decrease our dependencies, um, Microsoft man. What, what should I do?

A Well, obviously, there are much smarter people who know more and have much more at stake than I do trying to solve this. I, I mean, it, you have to do it. Yeah. Like it, I, the thing is, it goes against all these academics that were experts in 2000 that said it was good. Yeah. But they didn't think about all of these, like, nobody thought that the dependency would be on, on, you know, sort of the replacement for the Soviet Union in terms of long-time enemy. And, and so you, now you, you know, no more would we be dependent on the Soviet Union for something than we would be on, on China. Like, you, there's no choice in, in that matter. Now, hopefully, our detente with China is already vastly more open than anything that up until 1990 that the Soviet Union was, so we're in a very good sort of peaceful situation there, but that doesn't change The posture. Yeah. And so, you know, Apple is, like, doing this experiment. It's obviously more than an experiment to build in India, which solves their problem. It doesn't necessarily solve the, the current administration's view of the solution. Right. I mean, because if you think of it, like, the other part of free trade that inverted, twisted, and, you know, split the party, the political parties in half in the US was NAFTA in, in the early 19 nineties, which was, like, Free trade with Canada and Mexico. What could be more of a no-brainer? …

AI assessment note: “there are much smarter people who know more and have much more at stake”

Q And so let's say Trump brings you in and says, hey, I've noticed that, uh, you know, nearly all of computing is, uh, you know, every layer of stack is, is done in China and, and, you know, Asia more broadly, um, and I'm trying to sort of decrease our dependencies, um, Microsoft man. What, what should I do?

A Well, obviously, there are much smarter people who know more and have much more at stake than I do trying to solve this. I, I mean, it, you have to do it. Yeah. Like it, I, the thing is, it goes against all these academics that were experts in 2000 that said it was good. Yeah. But they didn't think about all of these, like, nobody thought that the dependency would be on, on, you know, sort of the replacement for the Soviet Union in terms of long-time enemy. And, and so you, now you, you know, no more would we be dependent on the Soviet Union for something than we would be on, on China. Like, you, there's no choice in, in that matter. Now, hopefully, our detente with China is already vastly more open than anything that up until 1990 that the Soviet Union was, so we're in a very good sort of peaceful situation there, but that doesn't change The posture. Yeah. And so, you know, Apple is, like, doing this experiment. It's obviously more than an experiment to build in India, which solves their problem. It doesn't necessarily solve the, the current administration's view of the solution. Right. I mean, because if you think of it, like, the other part of free trade that inverted, twisted, and, you know, split the party, the political parties in half in the US was NAFTA in, in the early 19 nineties, which was, like, Free trade with Canada and Mexico. What could be more of a no-brainer? …

AI assessment note: “there are much smarter people who know more and have much more at stake”