Eric Schmidt & Matthew Prince Fireside Chat

Eric, thank you so much for joining us on Cloudflare TV and to celebrate birthday week. We were, it's Cloudflare's 10th anniversary as of September 27th yesterday we launched and it turns out that that's actually the 10th anniversary, or not the 10th, the birthday of Google as well so when we were thinking of people who we wanted to talk to about what the future of the Internet looks like and everything else you were you were top of my list so I really appreciate you taking time to come on board. Well thank you to you and your co-founders, congratulations for an incredible decade and I'm sure that with the growth rate that's going on now this is going to go on for a long time, it's all good. Well I appreciate it and you know we're going to talk a lot about what the future of the But you know one of the things I've really valued in talking to you over the last few months is that you actually have a really great view and insight into what's going on with COVID and so much, there's so much information out there it's really hard to sort of figure out what are real facts and what's not. What are you seeing in that world? I know you've got a lot of funding in that and looked into it, what's, is there good news or bad news or you know what's your kind of perspective over the next few years? I think today it's sort of some good, some bad. I think we have to start by saying that the government has failed us. There is not a national plan and there's not a state-by-state plan and there's not a global plan to get the infection rate down dramatically below one, so-called R-naught, and it's a simple tautology that if you don't get infected you won't get sick and you won't necessarily put yourself at risk of death. We've had more than 200,000 deaths. That number is probably understated by some amount in America which is getting close to the number of deaths in World War II. How is this okay? So whether you view this on a human life perspective or you view it on the question of the number of trillions of dollars that are subsidies that have been required to keep things afloat, it's just not okay. And this is occurring at a time where the science is relatively clear. The disease is spread by aerosol as well as droplets, and the way you're likely to get it is you're likely to be in a room for more than 15 minutes with somebody who's infected at close quarters who has a high viral load. That's roughly the approximation of the science. So all you have to do is avoid that, right? So wear a mask, be outside, and so forth. Don't open the bars, don't go to places that are going to preserve this, and we'll get the number down. What's happened is it seems as though the government has decided to trade off human lives for economic growth. And of course, there are many reasons to have jobs, right? So jobs are important. But if people are afraid to go to work, you can't force them. They're not slaves. There is some good news, and the good news is that there is a large category of essentially rapid test, imaging tests that are in the process of being approved. They're generally available outside the United States and in heavy use. And they're not as accurate as the PCR test, but they're accurate enough. And furthermore, if you were to test positive in an antigen test, you would then get tested by the PCR test to confirm it or what have you. That tool, so give an example. If in the $3 trillion of funding that was released in May and June, if we'd simply put in some number of billions of dollars to enable a very broad and scalable rapid testing infrastructure for America, we would be in much better shape. And so we should do that now. And are there governments around the world that you think are doing this particularly well or that we could model after? Or has it been a failure almost worldwide? Many of the Asian countries have been doing antigen tests. A good example here is South Korea, Taiwan, and Australia. New Zealand managed to eradicate the disease by simply closing the borders. When you travel in and out of America now, there's no requirement for testing. That seems insane. There are all sorts of examples where if you simply adopted the fact that we want to test to see, and when the tests come out, we can use that to deploy our resources. And instead, here's what we're doing. We're taking our most valuable thing, our children, literally the most valuable thing in our entire lives, and we're putting them in schools without knowing whether they're going to come back either infected and so forth. And for people who think that this disease does not affect children, there's a whole bunch of dead children to prove them wrong. You want to take this disease extremely seriously. We don't fully understand the impact of it, but we know it's bad. So over and over again, what we need is we need a national plan to reduce the rate of infection. What happens is people say, well, I'm waiting for the vaccine. Okay, that's fine. That's like waiting for something in the future for an unknown period of time. How is that responsible? I look forward to the vaccine. I look forward to this disease being under control, but it's an unknowable amount of period of time during which people are dying unnecessarily. Do you have any insight into the science behind if you've had it, whether that confers long term immunity and whether or not we are going to be able to have a vaccine at some point? And if so, what the time frame on that may look like? I'll tell you what the scientists have said. First, we think collectively that there is antibiotic resistance to the disease, but we don't know for how long. Furthermore, there is a consensus that there will be vaccines that are safe for humans by the end of this year. And remember, that's the beginning of the process. It's unlikely that they will have been thoroughly enough tested to be used broadly. However, people will try with courage to take them. Let's assume that it all works out and let's assume it works out really well. Realistically, you as a citizen in America will have access to it perhaps in the first half of next year. What is the point when enough people will be vaccinated so that you could resume your normal lifestyle, probably at least six months more? So one way to think about that is that while there will be great vaccines and hopefully they'll work well and hopefully there'll be one shot and they won't require a cold chain and all of that, the fact of the matter is it's probably a year away at the best case. At the current death rate, which is 800 people a day, you can do the math. How is that okay? 250,000 more deaths. There's reasons to think that things will get worse in the winter. We now know that in the summer things get worse in the south, probably because of air conditioning. There's so many things we could be doing that we're not doing. Well, I appreciate all your investment in this. I know it's something that you've been spending a lot of the last eight months really leaning forward in and even being able to separate the facts from the myth that are out there. I think it's really important, the work that you're doing. Turning to maybe a more chipper topic, we're supposed to be talking about what the next 10 years on the Internet are and also looking back at what the last 10 years are. But it was about 20 years ago that you got a call to come and interview at what was a little startup called Google. What was that process like? And how did you get the call? And what did you think when you did? Well, in my case, I got a phone call from John Doerr, who was a board member, and he said, you should look at this company. And I said, search won't matter, won't be very important. But with his salesman skills, he convinced me to come visit Larry and Sergey. Once I met with them and I understand how incredibly interesting they were, I figured it was worth the risk. I, like most people, did not foresee any of the explosions that have happened in the last decade or two. What's interesting to me is that the Internet is not done yet and that the leaders, the structures, and the platforms will change again. So when I joined Google, the platform was web. And in fact, I can remember in the first decade, we spent an awful lot of time debating the distinction between mobile apps and web-based apps. And indeed, there were standards for web-based apps that were competitive. Today, we know that the mobile apps and the mobile app architecture dominated. So today, if you founded a startup, you wouldn't worry about that. You would focus on iOS and Android and fast networks and AI on the back end. We didn't know that then. It wasn't obvious. It's important to remember that these things get resolved as a technological platform issue, and then you move on. To me, the next interesting questions about the next decade have all to do with the nature of information. It's probably the case that there are increasing returns to centralization around AI and ML. There's a big data and so forth and so on. And yet, we also know that these tools are democratized. In other words, that they'll be available to anyone. So what I don't know is whether, excuse the technical part of this, people will build a very large neural net, and then people will use transfer learning to subset a piece of it, and then that will be freely distributed. Well, that is a democratizing aspect, but a centralizing aspect is the creation of these large networks. And I don't think anybody knows. What we do know is that the computational requirements of these things are still very expensive. And that fight, the centralization versus decentralization fight, is the fight that we always have on the Internet. Yep. So is the fight then, where are there opportunities to, is the fight around the algorithms and the computing? Or is the fight around where the data gathering and pool of information is going to come from? Well, it's both. And so if you look at TikTok, which is always an easy target, TikTok is both consumer-generated content, 15 seconds of entertainment, but also an AI algorithm that is different from the other algorithms and was deemed so important by the Chinese government that they banned its export to the United States. Do you think that's real? Is the algorithm that, is it that interesting? Or is it if all of a sudden, you know, a Microsoft or a Google or an Amazon or somebody had that same volume of usage and volume of data that they would be able to do recommendation and relatively quickly? Well, it seems simple, right? Instead of using an interest graph based on your friends, use an instant reference graph based on your interests, right? So maybe five or 10 programmers could do that. Well, if that were true, and I'm being facetious, then we would have lots of it. But at the moment we have very few. So there must be something more about the algorithm that was invented that has propelled them to such a role. And what I've learned is that these questions about if you had what they have, then could you do this? But remember that when they started, they didn't have it. Let's use Zoom as another example. Zoom had users and then it took off and it took off by a factor of 100. Now, why Zoom versus the others? Well, I would argue gallery view allowed you to have more participants. You might have frozen there. Can you hear me? You froze. I would argue that gallery view, speaking of Zoom, gallery view of 49 participants, whatever it is, might be a better product that was available at the time. Or maybe it was another thing. Or maybe it was the lack of privacy, which they've since fixed. I don't really know. But the important thing to remember about these algorithms is that they don't start off with infinite data. And it is the conjunction of the data aggregation and the algorithm improving in that sort of beautiful cycle that goes on in tech that moves so quickly. And that becomes the barrier of entry. And is there some part of that coming at a problem from a different perspective that helps you think about it in a different way? I mean, I think of Google's entry into some of the search platforms that you tried with Google Plus versus Facebook. But then on the other hand, I mean, YouTube is this incredible social network that exists. How much of that kind of naive mind and not having preconceptions is a factor in being successful in these things? It's a component. And the other component is the ability to fail quickly. So you try something and you iterate. So the key aspects of all these platforms is that they are measuring every second and they're tuning. Certainly during Google's initial rise, our rule was we couldn't ourselves know what a good UI was, but we could measure it. And I remember this young woman came in. We're in the break room in the original building, a tiny company. And she's all upset because she's done a test. And all she's been able to do is increase the click positivity of a particular ad by changing colors, but from two percent to two and a half percent. And she almost looks like she's crying because she's been working really hard. And I said, oh, my God, do you have any idea how much revenue that is? And she's not thinking about revenue, but I am the CEO. I'm thinking about revenue. And so she ran back to the office to work on it harder. Right. So that beautiful cycle where you can make these improvements and measure them and then it drives revenue, drives hiring and so forth. That's the key element. So if you look at the most successful companies, they'll have the property that they have engineering teams like this that move quickly. They'll be mobile platform based with fast networks, AI in the back and the content is generated by others. And there's more content to come in. You can learn, you can measure and so forth and so on. What I don't understand yet is how does the ability to do new kinds of information, video, GANs, that sort of thing, how is that going to change that? It's reasonable to expect that our experience in the next decade in the Internet will be much more video component. In other words, the next decade will be a rise of video everywhere and all sorts. We also know for AI that the ability to have real conversations with the AI is now possible. Look at the success of OpenAI's GPT-3 as an example. There's plenty of evidence that you'll be able to converse and see things which are video. What I don't know is what entrepreneurs will do to use that to both create entertainment, but also create misinformation, you know, interference and all the things people worry about. I think it's too new. What was it when you first met Larry and Sergey? What was it that stood out with it that you said that they just were very interesting? What was interesting about them? Well, my first meeting with them, I had been working when I was at Novell on a thing called proxy caches and I explained to what they were doing and they said it was the stupidest idea they'd ever heard. You didn't need these proxy caches. This is a technical matter, but it's, you know about this because you're at Cloudflare, but basically they are caches that accelerate your end user experience. And so I left thinking I hadn't had such a good experience. What's interesting is that we didn't need them until we purchased YouTube. And because the amount of bandwidth that YouTube was generating, YouTube nearly keeled over until we put a Google-based proxy cache in the front end of the network that everybody saw. So what I told Larry and Sergey at the time is they were right and then I was right. So it sort of worked out and that's how our relationship would work. Inevitably, they were right and eventually I would figure it out too. What were some of your favorite moments behind the scenes or things that the world doesn't know about, kind of the internal workings of Google that were really part of the secret to its success? I figured out that I spent a third of my time managing the founders in the first years because they were busy doing important things. And my strategy was to make them successful because if you get at opposition with the founders, it sort of doesn't work out for the new hire. And I sort of studied the John Sculley experience with Apple and we got along very well, but they had a sense of humor. So one day they had a shared office and they called me into their office and they said, we've been thinking about something. And I said, okay, what? And they said, we've been thinking about getting into the hardware business. And I said, well, I'm not so sure about the economics of hardware. And they said, no, you didn't understand. We want to get into the refrigerator business. And then they had a sort of about a 10-minute dialogue between the two of them, which was escalating about the power of what we now know as IoT and the power of refrigerators and all the things that they would do with it and so forth. And then I realized that they were joking, that the entire thing was a put on. And so they would do that once a week to me. I'll give you another example. When I first started, we had one off-site because they didn't really like off-sites, but I was a traditional manager and it was at a place called Quadris, which is not very far from you. And I showed up, I brought in the team and Larry of course showed up an hour late on rollerblades. Nevertheless, they come to the off-site meeting. And at the end, we have a conversation about what aspirations are. And I said, well, I think we should try to be a $100 billion company. And they looked at me and said, what kind of $100 billion? And I said, do you not know? And they said, I said, probably revenue. I said, well, there's no difference. And I realized that they were joking, right? And in fact, they'd set up the entire day to convince me that they did not understand the difference between earnings, revenue and valuation, which they deeply understood. So in the dynamic of the partnership, they would often have fun because they were so quick and it was very fun, right? Very clever. Was, what were some of the things that weren't fun? Like you guys, you found yourself in a lot of difficult situations or did it, and for a while it seemed like Google, I mean, it still is such an incredibly loved brand, but it does feel like something has sort of changed where tech has gone from a world where it can do kind of no wrong to one where it can do no right. Well, you know, we had our share of internal disagreements and political disagreements as we built the company. They largely let me sort of build it. And as long as the strategy was right and as long as the technology was right, they were pretty happy. And of course, it became quite a good revenue maneuver and evaluation and going public and all of that. I think that the errors that were made start from, like everything else, you can't get everything right. And so the social media explosion, we saw later, but we didn't see it early enough. I hold myself responsible for that. And there's a lot of consequences for that. Another example is that a company like Google has a lot of the talent in it. And eventually, the talent would go to other places. And with that talent would come the knowledge that we had built. And that's how Silicon Valley works. In hindsight, we should have worked harder to keep those people. So we did well. We could have done even better. Yeah. What, you know, as you look back over the last, I mean, 20, more than that, 30 years because of, you know, Novell was at the center of this. How much of, when you were at Novell, how much of sort of the future of the Internet did you see? Or what surprised you about the development over that period of time? Well, I've been involved in the Internet since NCP, which was 1977. So that's 40 years. And I think that the people who built the Internet all had a relatively naive model that if you connected everybody, the right thing would happen. And I, for example, when I built my systems, it never occurred to me to build secure password systems and, you know, phishing and the kinds of things that you defend against. So I think we all grew up understanding that the world was different from our idealized model. And today things are much more robust. I don't think we understood two things. The first is we didn't understand that there would be concentration of power in network effects. In other words, we didn't understand network laws. And we also did not understand how the information networks would differ. All of us, maybe as academics, were brought up under the assumption that computer scientists have, which is information is open, you browse it, you know, so forth. And we all benefit from that. But in many cases, information is not free. And in fact, it's manipulated. And I think nobody was prepared for that. This generation of computer scientists understands this all very well. Right. So it's literally a generation difference. The inherent techno -optimism of our field has hit the reality of how society works. And I'm not saying this with bitterness. I'm just saying that we were naive. And certainly I was naive. How much of that lesson did you learn when you were at Novell? Because you really did see the power of network effects with sort of the fights with Microsoft and everything that went on there. What was the lesson that you took from that when you went to Google? What I learned at Novell is how to run a company that was in crisis. And during my time at Sun, I was a thoughtful and nice person, according to everyone. But I wasn't particularly decisive. And I think one of the best things about doing a turnaround, which is what Novell was at the time, and for the viewers, Novell was doing a specialized protocol. And I got them to move over to TCPIP and things like this. But it was an infrastructure networking company is the way to understand it. Anyway, I think the thing that I learned really the hard way is there are some decisions that matter. So the benefit that I had when I showed up at Google is we didn't have any revenue. And so my position was more revenue solves all problems. And once we got the advertising system working, we ran it that way. And I don't think I would have been able to do that without the Novell experience. You know, we were talking before we went on the air about how the sort of view from tech is distributed systems and scale and measure and making things work that way. And that there's a tension sort of between the tech world and the non-tech world. And sometimes that doesn't come together. And sometimes that's where the sort of gears mash. What's your advice to sort of solve that from kind of either perspective? Because you have straddled both of those lines, both on sort of the tech side and then also on the government and policy side and seeing where those gears do seem to be mashing more these days. What I've learned when I explain these sorts of things to people is when they're going up and down with their head, that means they're listening to me. That doesn't mean they know how to act against it. And what's happening now is that the principles of Silicon Valley, broadly speaking, the disruptive nature of it, the broad use of software, the ability to build distributed systems and hardware that are attributable. In other words, they don't all have to work together. In other words, decentralized computing, decentralizing systems, decentralized networks, the power of the network, the power of influence, the power of information. All of these things run straight into traditional practices, whether it's in the government, in the military, in education, in people's expectations. So, for example, people believe that the hierarchy is more important than it really is. And I'm much more interested in getting a model where you have empowered individuals, which is precisely not how the government works because it's all very centralized. In my work for the military, it's the same thing. I'd much rather have fewer aircraft carriers and 1,000 more smaller boats from a defense perspective, same argument for satellites. So, I've been trying to apply this general tech principle, if you will, to everything. And it works sometimes and it doesn't. But I'll give you an example with education. We should be able to get an enormous breakthrough in education because we should be able to speak directly to the student, determine how the student learns best and with his or her teacher have a better outcome. It should be possible to get through all of the barriers to make that happen. And that's not unlike getting through the CIO barriers in the 1980s by the PC revolution. So, again, if we can find the path to get to the customer or the consumer or the constituent and really improve things, we can win. And that, I think, is a great quest for all of us. No, it's interesting. I've been thinking a lot. I've actually been thinking a lot about elections. And if you think of one of the things which has been really powerful about the U.S. election system is it actually is a massively distributed system where you have individual counties that are setting up their own procedures and own policies. And that was actually pretty good at defeating kind of the election challenges of the sort of 1700s through 2000s. But now when you've got, you know, an individual election official who might be up against the PRC or something else, and that there is this tension continuously between the power of being distributed and then the lack of resources that you potentially have in those cases. And this is a good example of where architecture matters. So I, for example, would like to have every county to have bulletproof election software. I'm quite concerned that the existing election machines are breakable. There's some evidence that they're breakable, although there's not evidence they've been broken in by foreign powers yet. But we have to assume that that will occur and therefore they need to get fixed. Having a common platform is not the same thing as having common control. One of the principles here is we could have national standards for election counting and spreadsheets and recording and so forth, while still allowing the way the counties want to work to operate. And they can have different rules if they care. So we have to find that balance, because what happens is that whoever is the constituent writing the rules, they write the rules to put themselves into power. What I'd like to do is have the government precisely put themselves not into power, but have standards, common standards by which they can measure what goes on. And then if you've got a problem, you can deal with it. That's the genius of decentralization. You do, in fact, have, it's not completely decentralized. You do have common standards and you have common measurement. In Google, which is highly decentralized, trust me, we have lots of common measurement. We know what's going on. And in particular, we know if one group over here, which we thought was doing something, is doing something different. So we've got about two minutes and 30 seconds left, and we could talk for a long time. But the question I'm supposed to be asking you is, what do you see as you look in your crystal ball for the Internet over the course of the next 10 years? And what are you particularly excited about? Well, it seems to me that the gains in machine learning and AI and the investment that everyone, including your company, Google, et cetera, is putting in it are going to yield a whole new set of applications. I mentioned TikTok earlier, not because I particularly like TikTok, because TikTok is a surprise, a Chinese company that managed to do something that was unexpected. We should expect more of the unexpected because of the level of investment. And so the people who sit there and say, oh, you know, it's Apple and Google and Amazon and Microsoft and so forth, and it's all done. They're missing the narrative. The narrative is that there's a new platform emerging, which the big guys and the new guys, the new little guys, are going to compete over. And that competition will generally be incredibly helpful. It will produce some very significant large companies as they figure out a way to monetize. But more importantly, it'll have an impact on society, both in terms of entertainment, as we saw with TikTok and its predecessors, but also in terms of information and productivity. I would like to see a huge, successful education company. I would like to see a large set of sort of new, essentially synthetic biology companies, because I think that's an industry that looks like software to me now. And I've also gotten interested in what are called digital threads. Digital threading is a concept in advanced manufacturing, where you basically build a prototype, you build a digital twin, and then you build the entire manufacturing system and software, and you simulate it, and then you can produce these things with much greater accuracy. The revolution in advanced manufacturing, the revolution in the way we interact and so forth, and the revolution of these companies will then shape the Internet. Well, I really appreciate you being good counsel to us over the years and talking through these things, and especially right now, all of the investment that you're making in COVID. And I hope you stay safe, and I look forward to our next conversation. Thank you so much for making time. Appreciate it. Okay. Thank you so much.