The Opportunity Ahead: A Conversation with Bill Gates

We're thrilled to welcome you here today. And it's a pleasure to introduce University of Washington's new president, Michael Young, who will introduce today's speaker. President Young, thank you for being here. It's clear I need to teach Ed an old academic trick, which is you want to quiet a crowd. All you do is whisper very quietly. And the second question on the examination is, has an amazing effect in quieting people down. But thank you for being with us. We are just thrilled to have you here today. This is an exciting day in the life of the university. And I have, I don't know if it's an unenviable task of introducing probably one of only two or three people in America about whom one can honestly say he needs no introduction, particularly to this crowd here. But I am honored to introduce our distinguished lecturer today, someone who has very close ties to this university, but for today's purposes is extraordinary in that there are very few people in the world who changed the world once. And Bill Gates has changed it twice, not only with the creation of one of the most extraordinary computer companies in the universe, Microsoft, but with the Bill and Melinda Gates Foundation has changed our understanding of our relationship to the world and has enabled us in dramatic ways to start addressing a number of global challenges, including global health challenges. It's very few of us who have a chance to change the world twice. Unprecedented to do it twice, I think. But we're very honored to have Bill with us today. They have been long, long time supporters of the university in so many different ways. I think it actually started with Bill's father, who is sitting on the front row, one of our great regents. I think Bill's father and mother probably have the distinction of being put together the longest serving regents in university history. And we are so grateful for your contribution to the university as well. But Bill and Melinda have also been extraordinary. Ed Lazowska, who you heard from a minute ago, actually holds the Bill and Melinda Gates professorship. As one of our great computer scientists here, we have benefited in other ways with his family, enabling us to keep a terrific professor like Ed with us as well. So with that, let me welcome to the stage Bill Gates. Well, good afternoon. Ed, it's great to be here. You know, I used to spend time on this campus stealing computer time. Now, that's a very antiquated concept because now we have personal computers. There's sort of an infinite amount of computer time. But there was a time when, if you were addicted to using computers, it was a very tough addiction to deal with. And this university was very attractive. It had not many computers in the computer science department. They were kind of batch oriented, B5500 systems that were walled off you couldn't get into. But if you roamed around the campus enough, it turned out physics had a PDP-10. Down at the med centers, they had PDP-11s. And if you came at strange hours, you could essentially break in and use computer time. Didn't cause any damage whatsoever. I never did get a degree here or anywhere else. But fortunately for me, my addiction to computers became easier to satisfy because of the invention of the microprocessor. And my friend Paul Allen, who is very hardware-ian, was the one who noticed the very first chip coming out. And we talked about how Moore's law combined with that would do really phenomenal things. Led to the software industry and all sorts of amazing things. But what I want to talk about today is that I think in terms of amazing things that come out of those breakthroughs, we've only seen just the very beginning of it. And particularly when you think of domains outside of classic computer science, things like health and education, the impact to date has been rather modest. But in the decades ahead, I think it'll be rather dramatic. And I think that's an exciting opportunity for all of you. I'm only going to talk for about a little more than 20 minutes or so. And the balance of our time will be available for questions. So as I say things that are wrong or stimulating, please think up questions to go with those. So when we talk about this miracle of availability, one way to look at it is purely in a quantitative sense. Almost any specific number we can look at, it's quite unbelievable how it's changed. The most dramatic of all is the cost of storage. And an example of this, I remember somebody saying to me when we were in a meeting about education about 12 years ago, well, we could record every class. And we could just store up on a server and have everybody look at what the teachers were doing. And because my mindset about how much storage cost was formed back even before this $700 per megabyte data point, when they said that, I thought, no, no, you can't do that. And then I thought about, well, yeah, of course you can do that. And of course now, it's so cheap that it's actually kind of absurd that we're not doing that, that we don't have that data set and we can't analyze. But what's a good teacher do? What's a crummy teacher do? When do the good students start to get bored? When do the bad students start to get bored? How well do people do at calming the classroom down? Well, we should just have this data set and start to understand who does it well, what's going on. And that's driven by a quantitative advance, but it's a qualitative change. Cost of RAM, similar, not quite as dramatic, different level of the storage hierarchy. And then we have processor speed. And this one, of course, is very dramatic over the last 15 years. What's interesting on this one is, if you look at it on a per-thread type basis, we actually probably won't see, short of some surprise breakthrough, we won't see much change here. If we let ourselves have many threads per processor, many processors, then we'll get something. But the nature of the improvement is no longer sort of a brute force, just write your program the way you used to, and you get that benefit. So the numbers are very dramatic, a factor of a million in the smallest case. And so it means you really need young people whose imagination about what all this power can do hasn't been distorted by the scarcity of resource that sort of gets embedded in anyone who's been around thinking about these problems for enough time. And so what this leads to is some areas that I think will be at the center of computer science in the next 20 years. Natural user interface, vision, speech, ink, sensors everywhere. I got a chance to see today some really phenomenal digital sensor stuff going on here. And not super expensive stuff. I mean, stuff that's so cheap it almost just comes for free, and yet it can do amazing things. Cloud computing, we've got Microsoft, Amazon, Google, so many people going after that. And it's really just at the beginning, sort of that infinite server that eventually you'll have quality guarantees that are associated with that. And all you'll be fighting is the latency problems about where the local input and processing should do the work versus going out into the cloud. I'll be talking a lot today about modeling of different types. It's a very broad term, but it basically means being able to determine what would happen in the real world by creating the computer-based model. And a lot of the activities I'm involved with now really have modeling at the center. I'm going to talk particularly about disease modeling, because that's so important to drive the strategies of the foundation work I do. People probably know I'm full time at the foundation. I still go to Microsoft, very interested in the work there. But my primary thing is thinking about the two key focuses of the foundation, which is helping poor people, mainly with health interventions, but some others as well, and then helping improve the education system in a broad way. Machine learning, we're just at the start of that. Now that we can gather so much data and analyze that data, maybe some of those dreams of the past, which were over-optimistic, weren't realized, maybe they can be delivered on. A final one here is robotics. That is deeply underestimated. It's one of these fields where if you say to people enough, it's coming in two years, and in two years it doesn't come, and then you say it's coming in two years and it doesn't come, then eventually they know not to pay any attention. You completely attenuate their interest in hearing you say that. Well, here we are. And in fact, in a pretty meaningful sense, as you get very cheap vision, as you get very cheap navigation, as you get very cheap ability to model environments, you really do start to get robotics. And there's a phenomenal number of ways that those change the economy, change the job market. And so it's too bad that we've kind of fooled people into thinking this might not come, because even people outside the computer science domain ought to be thinking about some of the challenges and opportunities that that will create. So it's the intersection of these qualitatively different ways of changing computers and the quantitative that really intersect. In fact, if you think of a previous boundary in computing, which was going from big computers to personal computers, I always was confused why IBM, or even digital equipment, they made the medium-sized computers, and their chief executive says nobody wants a personal computer. Well, why didn't they see that? Well, it's partly that he saw the quantitative change, but what that meant qualitatively, what the demand side, the use side, he wasn't able to see. So I want to take an example of this human-computer interface, an example of some work being done at Microsoft Research on vision, and just give you a glimpse of how quickly that's come along. OK, this is, I'm going to come down to where my Kinect camera is here. This is a Kinect camera. Hopefully, you guys have heard of Kinect. It's a video camera that goes to the Xbox, and so that's it right there. If I take, this is, we just took some objects that I normally have on my desk and put them here. Nice little plant and wood thing. So I'll just take this off. That's key. Sometimes I forget to take my hand off. And we just move around. And you can see, OK, in the upper left, you get the depth map. Then you get the CCD camera view. Then you get where it's finding surfaces over on the right. Then finally, you're seeing that it's creating a construction. Based on all the information, it's building a 3D model of what these objects are. So the one in the lower left is the one to look at. Now, if I add the textures, it gets very, very realistic. And so this fairly cheap camera, not with a single frame, but with all the different frames it's gathered together, has built a pretty good model of the environment. In fact, except for these edges where it doesn't get very good depth information, an almost exact model. Now, let's real quickly just give you a sense of how this might be used. In order to do that, I'm going to ask Ed Lazowska to come and sit here. And so I'm going to go, I hope I do this right. Shouldn't cause any damage if I don't. So I'm going to use this camera. And I've got, you can see Ed there, sitting in the chair. Now what I'm going to do is I'm going to have the software, the model, put balls on him. And he can shake these balls off. So we can put the balls back on. He's got to do it again. So anyway, it gives you a little bit of a sense that this really is a model that could be fun to work with. All right, thanks Ed. A lot of great examples of this kind of modeling work are being done here at the University of Washington. And this is not just pure pandering. These are really state of the art, exciting things. I got a chance to talk to David Baker about the protein folding work he was doing about almost a year ago. And I was very interested to see, was that work actually going to contribute to the state of the art? And in fact, it was just about three months ago that, in terms of the HIV protein folding shape, that they actually came up with something that is a real help to people who are trying to block the activity of HIV. It's not the single thing that, by itself, will solve that problem. But it was a real contribution. And so it was a fascinating way that he, together with Zoran Popovic, built software and got people to play around with it. More recently, a couple weeks ago, I got to see the work that Zoran's doing, based on a grant from the Gates Foundation, the education group there, about making it fun and interesting for kids to develop math understanding. And how can you actually test their knowledge, draw them in, come up with problems that are not too challenging, not too simple for them, and create a progression that's fun and interesting. And so they're doing some really fascinating work, where they generate the game that fits for that student, see how that student does, and then generate the next game so that it's constantly challenging enough, drawing that student in. And then they report that information back to the teacher. So if they're really stuck on a concept, the teacher can intervene with that. If they're ready to move on, the teacher doesn't make them repeat anything. And they have this in a variety of different formats. So again, it's computer science being applied in a way that people might not have expected. And it's non-trivial to do it right, because, of course, the motivation of that human subject, what draws them in, is a little difficult to characterize. And yet, it's ultra-important. I mean, imagine if we can make math fun, make it personalized, what a huge contribution that would be. Well, a lot of the foundation's work I mentioned is focused on helping the poorest people in the world, the sort of $2 billion out of $7 billion. We just are crossing over that $7 billion mark about now. Those who live in the toughest conditions. And a lot of those are small farmers. Overwhelmingly, they're in parts of Asia and in Africa. And the disease burden that they face, particularly infectious disease burden, is about 20 times as much as we do in the rich countries. A way to measure that would be to say, what percentage of children under the age of five die? And in the worst of these countries, it's about 20%. 200 per 1,000 die before the age of five. In the United States, that number would be about seven or eight. And most of that is birth defects that in the first week of life are a problem. Once you get past one week, then you have a 10,000 times greater chance of living to age five as a child born in Africa. Now, one of the key reasons for that is a few particular infectious diseases, of which malaria would be number one. In Africa, malaria is the number one killer of children, followed by pneumonia and diarrhea. But this is the worst of them all. And so I got involved in modeling in a bit of a strange way. I was doing the work at the foundation. And Nathan Myhrvold and some other people and I were doing some work on a nuclear reactor design. And that involves some fairly sophisticated modeling, stochastic modeling, where you look this proposed reactor design that doesn't exist. The way you decide if it works is a very complex model. Well, so as we were looking at this Monte Carlo stochastic modeling, we were talking about, well, couldn't the same approach be used for diseases? Couldn't we model all the different objects, the mosquitoes, the people, the transport of the people, the weather, and sort of look at not only what is going on with that disease, but what interventions would make what kind of difference? In particular, what type of interventions would cause a disease eradication, at least in that local area? And that's the most magic thing of all, if you can do eradication. There's only one human disease that's ever been done for. That was smallpox. It was done by 1977. No other human disease has been eradicated. This disease model, I'm showing it on malaria today, which is where we've done the most work. It's also used for what we hope is the second human disease that will be eradicated, which is polio. That's down to a few thousand cases a year. This is the thing I spend the most time on as an individual project. With luck, within three years, the final 10 countries where we have those cases will have dropped to zero. And so that will become the second human disease to be eradicated. So a little bit about malaria. Malaria, it's Italian for bad smell. What they thought was that those swampy smells must make you sick. And so don't go down near the swamp. Actually, then, a British sergeant hanging out in India, doing basically nothing, was messing around and figured out that mosquitoes were the transmissive vector of malaria. And so if a mosquito has the parasite and it bites you, parasites called falciparum, it will often kill you. And particularly if you're young, particularly if you're malnourished, it will kill you. And so the disease model that was done up until very recently was pretty straightforward, where you basically use differential equations. Why? Well, that was what was easy to do. Now, it doesn't really capture the richness of what goes on. You can't use these differential equations to talk about transport and weather. You can't really model the different interventions. You just have these different elements, different mosquitoes infect human. But the equations are fairly simple. This new model we have, which is the Monte Carlo one, is far more complicated, you can see, because there's a lot of boxes on that slide. Each of those boxes stand for the weather model. We actually model the immune system, the state of the immune system, which is where you're developing. Over time, if you're exposed to malaria enough, you become semi-immune. And that's because you generate certain types of antibodies. And we model all of this, both the vector lifecycle and the interaction of the mosquitoes and the humans. Now, whenever you do a model, you hope that your model and what's known from the past have some relationship to each other. Those of you who look at climate models know that they sit and tune parameters like MAD to try and match past weather records. And in their case, there's a lot of past weather records. But there's a lot of knobs on those models. How do clouds work? How does the ocean air boundary work? And eventually, they tune them to get close. And so that's why it's believed they have some predictive value. Here, we're taking malaria history. And we've done this for a lot of different places. And we're matching our model to actual data. There was an effort in Nigeria in the 1970s to eradicate malaria by using lots of sprays and nets. I'll show you that it failed. But here's the garky data, those green dots. And then the nice green line is what we get, which basically matches. Here's where that was based on the villages. They didn't do anything in. These are the villages where they did spraying. And so we put spraying into our model and seen if we stayed pretty much the same. And within the error that's appropriate here, we were able to do that. Now here, we took their most severe intervention, where they were giving drugs, anti-malarial drugs, like MAD to people. And we saw what they saw, which is that it failed. And this is the final garky data point, something that was actually done. It was kind of a sad thing that it failed. Now there are people who said, if you'd just stayed for another three years, you would have gotten it. If you'd done some other thing, you would have gotten it. But there were no models. Those were just guesses, no models that could deal with that. What we did is we added another thing beyond what they did in garky, which is a vaccine that is called transmission blocking. That is, if you have the vaccine, in this case, we modeled that half the people who got the vaccine would no longer allow the mosquito to draw out the parasite from them. So it's a transmission blocking vaccine. In fact, by doing this after, if you do this and stay three years, which is off my chart there, you do get an eradication. And so that's the whole idea of this model, is to say, how intensely do we have to use tools or invent new tools in order to get this eradication? And so this will be the world effort to get rid of malaria will be driven by this model and trying out different things. This is the malaria map. It's far more intense in Africa than it is anywhere else. That's why most of the deaths take place there. But it's basically an equatorial phenomenon. It used to exist. The United States have a lot of malaria. But the nice thing about the US is we have winters. And the combination of killing a lot of mosquitoes, particularly with DDT, because it's an incredible insecticide that lasts for, its effects last for quite a while, and having a winter was that we were able to do a complete eradication. That won't work in the equatorial regions. That is, they don't get, the mosquito population doesn't drop down massively like it does in temperate zone winters. It stays at quite a high level. They have dry season, but that doesn't create the same level of variation. So just to show, I'll show you, this is now Madagascar. And what we're showing is on the left, this is Madagascar as it is. And yellow is bad. That means there's somebody in that grid location suffering from malaria. Here's where we just do bed nets. And here's where we do bed nets plus vaccines. And what this simulation shows is that in Madagascar, just even doing bed nets plus vaccines is not enough to get an eradication. If we add another tool, the transmission blocking vaccine, then we can actually get an eradication. But it's fascinating. What you see, this is more than you ever thought you'd learn about malaria, but in the center of that island is the high part of the island. And that's where most of the people are. And so malaria up high there is seasonal. Unfortunately for them, the island is big enough that the mosquitoes thrive. And very few people live in the lowlands. And the reason is malaria. They don't want to live there. That's why most of the cities in Africa are far inland, is because of the desire to escape from malaria. So anyway, that's the example of the epidemic modeling tool. And now that we have that, it's hard to believe that in any of these diseases, tuberculosis, AIDS, polio, that we didn't have this kind of tool in the past. It makes you so much more rational in terms of what you do. And the model tells you data that you need to go and get about various things. So I think this type of advance, this type of innovation, we'll see it very broadly. Modeling materials, modeling alloys. And when you talk about batteries, one of the things you need are these chemicals with great properties. Well, now we don't have predictive models for that. In the next decade, we will have predictive models for that. It's a matter of understanding both the physics and some computer science work in order to do that. When we look at crops right now, we don't have a model of what makes a crop disease resistant. But between the biology that's being discovered and the computer science, we are going to have that. And it's a phenomenal thing to have, particularly if you have changing weather conditions, to be able to see how bad is that going to be for your wheat, or maize, or the other crops that are more popular in Africa. And when you think about the important things in the world, you can see that computer science is going to affect those things. Hopefully in education. After all, in the United States, if we don't make advances to improve the efficiency of education, then the country, because of the escalating costs, will become less educated, not more educated. In fact, we're one of very few countries who has that characteristic today, that as you get older, you are likely to be more educated. And it's kind of an inversion of the time arrow. It's supposed to go the other way. Well, if we let educational costs skyrocket, we're simply not going to be able to achieve that. And there are a variety of things, but I think technology is the big promise there. I'm sure some of you have played around with this website called KhanAcademy.org, which is Saul Kahn doing nice little 12-minute lectures about a variety of things. There's many people doing work like that, but it's really an exemplar of the fact that lectures, testing your own knowledge, personalized challenge, those things should move into the mainstream. And it's very timely because we have a problem that we want to solve. And so although my full-time work is now this foundation work, I find myself coming back to people like you saying, help me, give me some good computer science to solve all of these problems. And I'm optimistic that you'll deliver. Thank you. Okay, thanks for a wonderful talk. We've got 20 minutes for your questions. What I'd like to do is, I'll take the privilege of asking the first one, but I'd like this to be student-dominated, please. There are a set of microphones. So please, if you've got questions for Bill, think of them, line up at the microphones, and please begin by letting us know your name and what your major is, and that'll be a way of introduction. I'm obviously inspired by the notion that computing has a role to play in so many of these things. And you mentioned the seven billion threshold. I read that we're actually gonna cross that threshold on Halloween, sometime on October 31st. And I wonder what you think the particular implications of population growth are on the problems that you described. Yeah, if you go back to the 1960s and 1970s, the prediction was that world population would get up to about 20 billion. And that, from a global perspective, was very scary. And the people, the sort of liberal group at that time, Club of Rome, wrote about the need for forced sterilization and that nothing was going to work. Well, in fact, two factors avoided that. One was, it turned out that as people got healthier, as they got better nutrition, they decided to have smaller families. And so today, the majority of countries in the world are actually below replacement level. Now, we still, so we don't, the peak population in the US FDA model is about 9.3 billion. I think because they underestimate what we're doing with various health things, it'll be more like 8.8, but that's not a key difference, about 500 million. But the problem we have with population today is that it's in some locations that can't afford the population increase. It's exactly where we can't afford the increase. So for example, Africa, which is 800 million today, will be over 2 billion by 2040. That is, it will be bigger than China or India. And if you take it, if you treat it like a country, it'll be by far the biggest in the world. And yet, in terms of stability, education, agriculture resources, that's very troublesome. So it's still very important to very quickly give women who want reproductive health tools, those tools, and to improve health, because that, you have about a 10-year leg. When you get rid of malaria, just getting rid of malaria in Africa, say within 10 years, which is optimistic, it would take the peak population of Africa from 2 billion down to 1.6 billion. Just that one act alone. And so we do have levers that we can pull on this. You know, another lever you could pull, but people aren't very open-minded, is immigration. That is, all those shrinking countries, you know, right across the Mediterranean, you've got all shrinking countries that should welcome those people in. But that, we can't count on it, so. Question for you. Let's start right here up front, please. Hello, I'm Andrew, and I'm a computer science student here. I'd like to thank you for coming. It's certainly a privilege to hear you talk. I'm interested in your work in education, particularly your investments in the Khan Academy, which I'd personally like to thank you for saving me last year in linear algebra. What is being done currently to bring this technology to the mainstream education, where you have, like you were saying, an ability to quantitatively evaluate students, and if there really isn't even a need for a teacher role anymore with games and stuff for the internet for education? Thank you. Yeah, as you go from kindergarten up to college, certainly the need for adult supervision hopefully goes down somewhat. But remember, education to some degree is about motivation. I mean, if you wanna learn physics and you're highly motivated, you know, Feynman wrote a book. It's called The Feynman Lectures on Physics. Read it. You know, you don't need a university. And if you're that motivated that you can sit and read that thing, wow, that's, or, you know, theory of algorithms, just read Knuth. And by the way, do the problem sets while you're at it. So education is not about the unique availability of information. It's about some purine of the information into a form that the student chooses to ingest it. And so how do you create that framework? Well, I think there'll always be a teacher to some degree, but the place that that's most replaceable is in the case of the lecture. And so I think the answer to your question is, to take full advantage, you can take slight advantage of Khan at all, just as a supplement. Mom, I was confused about this. You know, mom doesn't know, mom points you to Khan. You watch the thing, try some things out, and it helps, it's a supplement. But the big impact of Khan so far is where the teacher's been willing to flip the classroom. And what's meant by that is that there are no lectures done in the classroom. You assign the lectures, which are watching the Khan videos, maybe two or three a night, and then you do the basic problem sets associated with those lectures. When you come in that next day, the teacher sees who watched those lectures and who didn't, and sees how well people did on those problems. And so that teacher can then divide the class up into groups and do problems that develop that knowledge, put that knowledge in context, make the knowledge interesting. You know, why would you wanna know about that, which is not the part that Khan is focused on. Actually, we have other people on the web looking at, you know, if I wanna do nursing, why should I learn math? If I wanna do construction, why should I learn math? And complementing that with the Khan stuff. So we have about 20 schools now that have redesigned the entire classroom experience around those Khan lectures. There's a charter school called Rocketship that's completely redesigned around using online curriculum. And so within a year, we'll have the results from that. Only one school did that last year, Los Altos, fifth grade. And, you know, the numbers appear to be very, very strong. Now, the acid test for education is inner city schools, where the environment, the motivation is the weakest. And that's why half the schools we're trying out right now are inner city schools. Alex Darling, biochemistry major. One of our professors, my math professor, is big into education in third world countries. And he's written a couple articles, including stuff about computer science and whether or not it's necessary. I was wondering, what are your opinions on computer science in education, starting even in elementary school, and whether that's a good idea? Get this right. Well, I think that computers is a great tool and you should be using it as much as you can. If by computer science, you know, we mean understanding queuing theory and, you know, hashing and, you know, how you do database indexes, I'm not sure what, where in the curriculum that could come in. I do think being able to write a program, you know, is a, if you think you understand, say, prime numbers, the idea that you can express that by writing, say, a SIB algorithm, I think that's a very interesting test of whether I know prime numbers or not. But I wouldn't, it's hard to say anything is necessary in the curriculum. I mean, can we say that geometry is necessary in the curriculum? No, it's probably not. People don't get to use arc signs that much once they graduate. There's certainly a level of complex symbolic thinking that is valuable to be exposed to. You know, personally, I might put statistics in instead of geometry. I'd put statistics, certainly, before calculus. I'd have more people take statistics than calculus. Where computer science belongs in that hierarchy, I don't know. So here's what I think our pitch would be on computer science, and that is there's this notion originally from Jeanette Wing of computational thinking. And computational thinking is a lot of the things you've described. It's algorithmic expression, it's problem decomposition, it's stepwise fault isolation, what we call debugging. It's modeling. There are a set of things like this that need to be part of everyone's bag of tricks. And writing a program is perhaps a way to learn and exercise a set of those skills. But the critical thing to me is that this computational thinking is something that every student needs, which is, again, different than understanding hash functions. No, I think that's right. I do find that people who have computer science backgrounds, when given a problem from another domain, the idea that they take the system and they look at the size of various elements, they look at the rate-limiting steps for various elements, you know, and they can say, okay, we need to optimize here. That type of thinking is like, yeah. And what other domain gives you that type of systems thinking? Maybe some parts of science and engineering. But the basic notion of what's an algorithm and that many systems in societies are basically poorly designed algorithms, I think that's very worthwhile. Hi, my name is Ashat, or Ashok, whichever one you wanna go for. I'm a computer science major and also an economics major. So I guess over the last couple decades, we've seen, at least in the United States, a growing disparity of wealth and income. And, you know, we've seen money concentrated in the hands of just a couple people, at least, you know, you among them. I'd say you're doing pretty good. But in addition to this, we've also seen a growing power of money to influence politics, especially within the Citizens United court case. So we have kind of here, we have kind of a cycle where, you know, the money is, in a lot of cases, who has the money influences, you know, what kind of information gets out. So I'm wondering, you know, do you think that this is the kind of money the question of, I guess, getting the, making voters more informed, increasing people's ability to make rational decisions in regards to politics is something that we should be focusing on as a nation? Well, I definitely think that, first of all, the world at large is less inequitable today than in any time in history. That is, the poor countries are getting richer faster than the richer countries are getting richer. So the number of people in abject poverty, percentage of people is at an all time low today. And that continues to go down. And innovation will continue to drive that down. And my favorite metric is the global under five children deaths, which is now somewhat under 8 million a year. You know, 50 years ago, that was over 20 million a year. So I do think that in most problems, time is on our side. You're absolutely right, there are some big fortunes and society can't, it's not good to have a society where you don't have mobility in between different income levels. That is, if you're born in the bottom quartile, education ought to be good enough that you have a reasonable chance of getting into the first or second quartile. And so if you really look at where we're letting people down in terms of the American dream, I wouldn't say, you know, and you can say this is self-serving, I wouldn't say it's because a few people are very rich. I'd say it's because we aren't doing a good job in education to give them an opportunity to move up into the top few percent. In terms of the very rich, the, you know, Warren and I, Warren Buffett and I, we're the two wealthiest Americans, are certainly believing that the rich should be taxed a lot more than they should and the rich should give away more of their wealth than they currently do. And we've certainly been willing to speak out about that. Warren is the only person who's ever had a tax named after him, the Buffett tax, you know, which is sort of about he should pay more than his secretary should. You know, so you can be very frustrated with the political system. I certainly am myself right now. I was in Washington, D.C. on Monday and Tuesday meeting with members of the House and talking about things like cutting science budgets, you know, is not the way to keep the country strong. I don't know exactly why politics feel so pressured right now. It's certainly worked very well up to now and so maybe, you know, the system will, you know, realize the problems that it has there. But I don't think, you know, just by getting rid of the wealthiest in the country, that would automatically solve these other problems. I really think you do need to fix the education system. I do think you need to get the cost of education and healthcare to go down. Then I think our society will feel far more equitable just for those reasons alone. Let me ask a question about civic discourse for a second. I was worried 10 years ago that the web and our ability to get information targeted just to us and to communicate with people who are just like us was going to sort of increase the hostility of civic discourse, the fact that I don't have to listen to people who disagree with me because I can find a cluster of friends. It seems to me that that's happened, but it's not obvious to me that technology has played a role. I wonder if- Well, we have a variety of things that may have led to the current polarized situation. We have money politics, which is bad. We have districts where the voters have the, they are designed so they're safe districts so that the primary allows somebody who's fairly extreme to the left or right to be the one who ends up holding that seat. California actually recently went and fixed that. Could that be fixed more broadly? Could you fix money politics? And certainly the media, people do wonder if this polarization partly comes from the Fox News phenomena that you're kind of listening just to those people who agree with you. I don't really know. I do think that underestimating people's ability to go and find things that represent broad viewpoints, I wouldn't want to be too pessimistic about that because the web gives you so many choices. I do think it's far better than broadcast TV, that if you're curious about a topic, your ability to dig in about that topic and learn about the topic. One thing I worry about, and I don't know what the solution for this, is complexity. The US tax code is so complex that you don't know where to be outraged and you don't have time to read it. There's this bill called the Volcker Rule. It's 430 pages long, and the law firm's summary is 41 pages long. Well, I'm gonna read it, but what percentage of the voters will read it? I don't know. The healthcare reform bill, 2,300 pages. I'm not gonna read it. I care a lot about it, but it's too arcane. If it was readable, maybe. So this complexity, the fact that we really can't talk directly about the issues means that we talk about personalities and things. So I think complexity's hurting us. The internet is one of the few things that might help us have explainers who are out there who can talk through, hey, here's why you ought to be pissed off about the tax code, and here's why medical costs are going up and what you ought to do to help those things go down. So I tend still to be optimistic, but some of the current data points are scary. Thank you for the talk, Bill Gaze and Ed. Thank you very much. My name is Tony Weavering. I'm a computer science student. I graduated back here back in 98. I'm back here as a student again. So my question for you is about the healthcare reform that was passed. The question was, so a couple years ago, I was in an accident. I was put in a medically-induced coma in Harborview Hospital for four months. I remain an employee of my company as a computer science company. I remain an employee. My healthcare insurance dropped me like seven days after it happened because it was going to be too expensive. They did not want to take care of me, cover me anymore. So what do you think about, I cannot get insurance for the rest of my life, no matter what. Even on my way back to work, I've been in an internship three times, still cannot get insurance, no matter what. What do you think about the healthcare insurance reform that was passed? What do you think about it? Do you like it? Do you not like it? Well, healthcare is a very complex topic, and I know a lot more about malaria and hash tables than I do. And I like hash tables and I dislike malaria. But in terms of healthcare reform, I do think it's super important because the cost trends are such that we are going to be denying people care. And, you know, that's horrific. And yet the trade-offs to society, the resource costs involved there, you know, it was in the news today that the state legislature, oh, the governor's budget proposal will cut the amount of money to the UW. Well, the reason that, again, fine. You know, why is that happening? Well, I think there's many steps in the causation, but rising medical costs, if you go all the way back, that is a core piece of what's going on there. And so, you know, in order to make healthcare always available to somebody like you and not having to cut education budgets forever, I mean, you get to zero at some point, you really have to take innovation and make it your friend. You have to make it come up with lower-cost ways of doing things. Now, that requires measurement. It may require a framework for competition. You know, we're really not there. The best single book on this is one by Michael Porter called Redefining Healthcare. Again, it's not simple. It's 400 pages, but if you're interested in the topic, I think it's well worth looking at. Great, over here. Hi, I'm Fabrianne Walker, and I'm an electrical engineer major, and you talked about robotics being something that will be happening in the next couple of years, hopefully, and I wanted to know how it'll affect the issues that are facing the world right now, like overpopulation or diseases or just some of the kind of catastrophes that we're facing. Well, robots, you know, they clean the rug today, and they disarm bombs today. You know, the term robot is an unusual term because we tend to apply it for something that can move on its own. You know, if something paints an automobile in a stationary position, or if it helps do surgery, you know, is that a robot or is it not? But when it comes to things like, you know, doing physical security or doing transport, you know, we, having cheap robots really can improve the world, can do fantastic things. There's a, one of the people I brainstorm with over in Nathan Myhrvold's company, you know, talks about how whenever a mother's delivering, there ought to be a robot that comes in that can do a C-section, because it's very difficult to get a doctor in there. And this is less far-fetched than it sounds, sounds very far-fetched. But if you want to cut maternal mortality dramatically, you have to find a way to get out into rural areas and give very sterile C-sections. It just, there's a level of mortality you can't get below unless that intervention is available. And it should be absolutely possible to do. You know, when you talk about making things less expensive, the robot can perform a lot of functions that are now in a fairly expense. When you talk about manufacturing being done in the United States, it happens to be one of the things that might allow resurgence in a fairly dramatic way. I think it's pretty hard to predict, you know, just like if any of us had been asked to say, okay, the internet's going to be cheap and available, which parts of the world of business will move to the internet most rapidly? You know, I think we would have missed, now it seems obvious, but we would have missed that somewhat. So I don't know what the breakthrough applications are, but I know when you get to the right price points, they're extremely broad. I think, you know, precision agriculture is another example where there's a lot of robotics in it, sort of GPS-controlled combines and things like that that's clearly had a big impact on productivity. Yeah, and at the expensive end, and now can we make cheap robots that for African agriculture, smallholders actually would help out. So I'm a computer science and industrial engineering student. I was kind of wondering what you feel like some of the responsibilities of computer scientists and other people who generate this technology are to society? Well, we should always measure the good impacts and the bad impacts of the stuff we have. You know, certainly whenever new technology comes along, there's a lot of fear about what it's going to do. You know, when the printing press came along, there was a great fear that people would just read books and not go out and seek, you know, real experiences. And, you know, I tried that when I was a kid, just reading books, and, you know, maybe I'd be more rounded if there hadn't been books around. So, you know, I think it's hard. I haven't seen any evidence of any significance that socialization has really broken down in some bad way. You know, I know that, you know, with Skype now, I meet my daughter's boyfriend's parents when I walk into her bedroom, and there they are on Skype. So socialization is changing, but I don't think, I'm not sure you can really say that there's a negative vector there. And there's so many positive vectors about if you're curious and you want to learn something. You know, I was young, and I'd ask a question about, hey, what's fertilizer? You know, if the world book didn't have it, I wasn't going to get the answer. Now, of course, you can go online, watch videos, and, you know, my children are always asking questions where if I don't know the answer, we'd just go and it's actually kind of fun to learn it together. So I'd say on balance, you know, I think we're quite, quite far ahead. You know, there are some things about privacy, you know, maybe misuse of, overuse of pornography, you know, letting your kids play too many video games, so you have to set a quota, just like I had to have a quota for how much I was allowed to read. But, you know, I think overwhelmingly, we've got mostly positive things, including the opportunity ahead of us of how the tool gets used in mainstream education. Okay, because I promised one last question, very quickly, please. All right, my name is Xing, and I'm an electrical engineering major. I'm not going to ask you anything about politics or how to be a rich man or something. Just wondering, a really quick question. So from personal computer to internet, and now it's smartphones, well, mobile internet. What will be the next generation of the technology, and what will be the future of technology? Well, in a sense, the only difference between a phone and a PC is sort of the screen size. You have the size of the screen and the input technique. And so, you know, in a classroom, you have a big, big, big wall, and in your pocket, you have this little device. The next generation is either a screen that you can fold out to any size that you want, kind of going back to the papyrus scroll, or more likely, it's simply projecting onto your retina. Because if you want to get a very high projection experience, if I have a projection ability, I can just, and I have a camera that's watching my gestures, I can just say, okay, I want a newspaper this size, and I can get perfect HD resolution right in front of me. And the cost and weight of having that capability is almost zero. So, you know, eventually, we'll laugh that there were these big, flat glass screens that were expensive, and if you dropped them, you broke. All you're trying to do is put stuff on your eye. That's all. So what a weird contraption, all these LCD chemicals and chips and things. You're just trying to project into my eye. Why don't you just go ahead and paint there? So I do think that's the display technology. Now, then you get into the pervasive sensors and the robots and things that mean that the dividing line between this is a computer A, this is computer B, sort of gets lost in that picture. So the pervasiveness and perhaps the projection are, along with the new interaction techniques, are where it takes on a different form. Thanks to all of you. Thank you.