A Conversation with Bill Gates Hosted by Eric Horvitz
Articles,  Blog

A Conversation with Bill Gates Hosted by Eric Horvitz

Well, it’s been a fun and engaging first day of the Faculty Summit. It’s been great enjoying this year’s theme titled “Future of Work,” but all the subtitles that go along with that. And, of course, like all of our Faculty Summits, we take a theme and we also hang many different topics on it, including ones that might not seem exactly aligned with the “Future of Work,” like we’ve done interesting briefings on machine learning and human computer interaction, the intersection of AI and ACI and so on. But the “Future of Work” brings together such a rich intersection of topics — AI, HCI, psychology, human factors, social systems, economics, fabulous. And beyond the topics and the content, I’ve personally had a blast catching up with so many of you, so many colleagues, including long-term friends, former students who are now upstanding colleagues and professors at universities, former interns, and folks I’ve wanted to get to know better over time. I was really impressed to hear this year we have about 250 faculty from almost every continent, representing 125 universities. Fabulous to have you all here with us today. And, of course, people joining from Microsoft Research. I’d like to cap off this wonderful day with a discussion with a man I admire greatly and with whom I’ve had a chance to work with, now surprising to me sometimes, over decades during my time at Microsoft. Please welcome Bill Gates, co-chair of the Bill and Melinda Gates Foundation. [ Applause ] So it’s great to have you here. I love hearing about your top-of-mind, which is, I guess it’s always Microsoft to some extent, but the Gates Foundation and the work you’re doing with Melinda, and on these high pay-off opportunities for making the world a better place. I know you think really deeply where you can have the most impact. I’m curious to hear what’s happening. Well, the Foundation has two big areas of focus: The first is our global work where we go to the developing countries and, starting with health, try to improve the conditions there; and the second is working on education, primarily focused here in the United States. And in the first area, we’re driven by the fact that there’s still a big part of the world where over 10 percent of kids die before the age of five. When we got started, that was the global average. Now, due to some things we and our partners have done, it’s been cut in half. And so we have less than 5 million kids die before the age of five every year. That’s still a lot. The three big causes are still malaria, diarrhea and pneumonia, but we’ve gotten some great vaccines out. I spent all day yesterday on the birthing process and how we make that safer for both the mother and the child. So our global health work really has gone beyond our expectation in what’s been achieved. It’s partly biology has gotten better. There were some really powerful things that had been ignored that needed to be done. And it’s also the case that as you get rid of any health problem, the child is better off; that is, they’re closer to their growth curve, their immune system is better. So you always see a nonlinear effect; when you get rid of something that kills 10 percent of the kids attributed, you’ll reduce about 15 percent of the deaths because whatever priming that disease did to weaken the child to make them more susceptible to dying with the other disease, you’re getting rid of that as well. Now the big thing for us is nutrition, keeping kids on their growth path, where we’ve had to fund all this microbiome work. But it looks extremely promising to have dramatic reductions in malnutrition which will have benefits both for death reduction and for the mental and physical capacity of the survivors. Our education work, in some ways, has been less dramatic in that — although there’s some amazing new curriculum things, some teacher feedback things, student survey things, a thing called the common core which organized the math curriculum in a much, far more logical, rigorous way — the actual achievement in terms of reducing dropouts in high school, reducing dropouts in college, even though the U.S. puts more money into this per student than any country in the world by quite a bit, both on a relative and absolute basis. On an absolute basis we’re about the same as we were 20 years ago. On a relative basis we’re weaker. I still believe that it’s a solvable problem. Part of it is a technology/tools opportunity thing, maybe 30 percent. Seventy percent, you have to create a framework where the measurement of excellence is agreed to. And so you’re in a continuous improvement process. So beyond being co-founder of Microsoft and the Bill and Melinda Gates Foundation, we all view you as a co-founder of Microsoft Research — and what are your thoughts about the contributions that that crazy idea back in 1991 have led to in the world on technology, the company itself, society, more broadly? We’re almost 30 years old, MSR. It was phenomenal that we actually, you know, when we did our early software products, we thought, wow, we’ve got money left over. [ Laughter ] We can do some long-term things. And we were students of Xerox PARC, partly because we were beneficiaries in terms of the graphical user interface of the amazing work that people like Butler Lampson, Alan Kay and Charles Simonyi, and, of course, we hired Charles, so we got kind of direct help. Yeah, Chuck, amazing. What they did was amazing. Now, the company they were part of didn’t get profitability from it, but it was still a great set of contributions. So even early on we started to do some natural language stuff with Karen Jensen. And when Nathan Myhrvold came in and he hired Rick Rashid, we got very serious about it. And it’s been a phenomenal asset both, hopefully, to the field; the people here can judge that, but certainly for Microsoft as we tackled — as we tackled search with Bing, as we have gotten into AI, as we’re infusing AI into our office productivity work. More and more the distance between, okay, here’s what MSR is doing and here’s what the product group is doing, it’s gotten smaller and smaller. And even that boundary is a little unclear. It was never — there was always a very close collaboration in Bing, and I’d say in databases that was pretty good. In Office, it’s more a new thing where you have Office Online. And so take PowerPoints as an example, the ability to make suggestions to people and have them try them out, and within a few weeks see, okay, that people are finding that useful, let’s do more of it, it’s been fantastic to see. And so Microsoft Research has been just an incredible investment. Even as we’ve gone and scaled it up and done various sites around the world, every one of those sites has managed to come in at an incredible level of excellence. So we’re here at the 20th Faculty Summit put on by Microsoft Research. I remember the first one, seems like yesterday. This one is on the “Future of Work”. What are some of your thoughts on where jobs work, productivity is going, and maybe even the role of computing platforms and apps moving forward? I remember one great fear I had when we did Microsoft Research, was that if we weren’t careful it would be a block to our view of what was going on in the academic community. That is, we would hire somebody who had an approach and that academia as a whole would have like 20 approaches, a number of which would likely be better than some internal approach. And so I wanted somebody who was open minded enough to say, no, this group here is doing it the right way. And I was worried; would that fluidity exist? A lot of these programs, like having students come here to the Faculty Summit have contributed to the idea that, okay, we’re part of a broader community; we publish great things; people come and go a lot. So I feel super good about that. In terms of work, I would say the — most of the opportunity to make computers improve work is in front of us instead of behind us. Only at kind of a very mechanical level of it’s better than whiteout, are we like, oh, wow, this thing helps me get my work done. For extremely structured tasks, like creating financial statements where the application logic is very opaque and you’re just led through it, or for very horizontal things, where you really understand what you’re up to and the machine doesn’t see, oh, that email’s part of a personnel review; that email is part of planning the salesforce meeting. So we have kind of the low-level horizontal, which in its ways is amazing. And then we have this very structured applications world. The idea of matching what you in your head have as the intention of the overall activity and the semantics around it and, therefore, sort of maximally helping you with those things, I think now that a lot of base level things are done very well, including gathering the data and learning from it, that we should be able to have what the machine thinks you’re up to and what you think you’re up to match each other and improve the productivity quality quite dramatically. But it’s taken a long time to get to this point. I’m sort of chuckling because I’ve been at a number of reviews, in particular product teams at times, where you share your vision, that’s a graceful way of saying it, you share your vision about what you would like to see in terms of deep understanding and knowledge these systems have about people and their tasks. And I know that I’ve listened and I always think, wow, this is like a grand AI challenge, we just can’t do this tomorrow. You might want to share a little bit more about the pursuit that you’ve mentioned to me in front of others about the ontology of tasks and so on. Well, when you write an email, it’s not just some random, sui generis thing that you think, hey, I’d like to write an email let me pick a random person and say some random things. Generally, you’re thinking about your kid’s college application; you’re thinking about the design of a product that you’re engaging people with. And so if you say to somebody, hey, what are you doing, does that email relate to any others? Does it relate to any documents, any lists, any events? Very clearly there will be somewhat of a hierarchy, but just say a graph, representations of a heterogeneous set of things that you’re working on. And you’ll have a notion, okay, if I don’t get a response, I’d like to pay attention to that at some point in the future. And if I do get a response, in what context is it worth bothering me with that? If I’m home on the weekend and somebody responds to this, do I need to see that in my list or should it just show up in context for those things? So there is a high-level intention and a taxonomy that the machine and the human should be on the same basis. My executive assistant does a better job of looking at email and deciding which one is important or taking a set of phone calls and decided which one is more important. So she’s been the problem? Yeah. [ Laughter ] So she can take heterogeneous data sources and actually decide what things are worth serving up. But now here we are thinking about how people spend time and what those activities are and not just having it be the super rigid things. You know what we see with application software is often that, like if you send something to a customer and it’s defective, the software can’t capture the richness of that conversation. So you go into email or text and say, oh, I’m so sorry but could you take half of it? We can send the other half. And you have all this unstructured but very free-form dialogue. And then maybe that gets reflected back into the system of record, which is overly structured. The idea that something like personnel review you could actually describe it in a digital form without, say, writing low-level code, just having to say go to C++ and write logic, it seems that that’s a tractable problem. Now, all these things, you have to make sure that the work that the person does pays off, that they don’t want to spend a ton of time structuring some simple thing that, yes, I’m asking my friend to lunch. And, yes, I expect him to answer in 20 minutes. I mean, that would be worth the trouble. So with a lot of the automatic observation and machine learning, the idea of suggesting to people, hey, this looks like work, nonwork; this looks like health; seems like you have these friends, these hobbies; the ability to structure those things and make that part of the UI and learn what types of tasks are out there and have people think, oh, this is similar to this one. Anyway, I think five years from now we will think of the semantics of work as being more modeled software understood than we do today. And we’ll look back hopefully at the way we’re doing it now as far more primitive. And you’ve been privy to, for your review, the reviews we do and the advising that you continue to do at Microsoft to a number of late-breaking technologies and work at product teams at Microsoft, as well as some of the offerings we’ve had out there in world. And what comes to mind as to some things that are exciting for you right now for Microsoft? Well, the whole software-as-a-service thing and the ability to take all the information in the company and create a natural language ability to navigate through the history of the org chart, the history of the project dates, the sales data — the idea that you shouldn’t have to really understand that it’s in this database and this column; you should be able to find patterns and the machine may suggest some patterns that are interesting to you — that is a — a long time ago, in 1991, I gave a speech called, “Information at Your Fingertips”. And we sort of have it in a low-level sense, that it’s all online and if you can write SQL statements, you know it’s Database and you have the right permission, you know, it’s there. Well, now some of it is in other forms of databases. So maybe you need to know three or four query languages. But how far away are we from letting you utter what any information worker might typically want to ask about finding an expert or the history of an account or the sales in this region versus the sales in that region and engage in a dialogue on that? I think we are extremely close to being able to do that. And that’s pretty fantastic. It’s kind of nice to see some of the original dreams come true. There is a dream about really understanding text in a deep sense that I have no idea how quickly we’ll get to that. But I do think in the next five years the semantics of tasks and natural language navigation over all of the company’s information in a rich way and having an agent participate in the meeting that kind of listens and even kind of passably can put things up on the screen that might be of relevance, or can be told explicitly to go get something and bring that into the conversation, it’s keeping the transcript. That kind of demo that’s been done for decades, that thing would be real. People will actually want to use it. And the amount you’ll have to charge them is so little that it might almost be the default way of doing things. I think we called that Info Agent in the early 2000s, and that was a dream we were pursuing back then. Of course, there are many — there’s a range of jobs and roles and types of work, and I thought folks would find this video interesting of your visit, I guess, when you were hanging out with Warren Buffett in his hometown to Dairy Queen — I didn’t realize he had a stake in Dairy Queen. He owns Dairy Queen. He’s a very broad-minded person. [ Laughter ] Right. So, I thought we’d roll this video of your visit to Dairy Queen. [MUSIC PLAYING] Here we are. Hello. Ready to go. Ready to work. We’ve got the wrong nametags on. This is just a sample of our competence. You think this is bad; wait till we really get into action. Whoa. Whoa. Keep it going. You’ve got to get two balls on there. I’ll see if I can do better. Oh, my God, they’re going to order from you from now on, not from me. Who had the bad luck to draw me? [ Laughter ] If you want anything later on, it’s on me. [ Laughter ] The cone is okay. [ Laughter ] You guys want to learn how to help customers? You bet. A vanilla cone. Is that all you want today? That will do it. Let’s upsell her. [ Buzzer ] You didn’t do anything wrong, it’s just taking some time. That barks at us a little. I’m a slow customer. It’s okay. Thank you. Have you been to Dairy Queen before? You want a dip or not dipped? Can I have a triple chocolate brownie? That is really one of our best items. In fact, if you don’t eat it, we will. All right. What else? I’ll keep it simple and go banana split, please. Okay. Double banana split? Yes, sir. You’re a real man. What would you like? I want an Oreo Blizzard. That’s one of our most most popular. What else can we do? I want a medium Reese’s Blizzard. That’s very popular. Is it? Absolutely. I like that peanut butter. Extra peanut butter for him. [ Laughter ] Three most important things about training someone new. Number one is definitely patience. I’ll put the hand out for the money and Bill can do the high tech stuff. You still have to ask her what she wants for the dipping sauce. You dip chicken, too? Wow, you guys deep a lot of stuff around here. Number two, make sure you smile, so they should smile. Kind of pass the smiles along. We got it. Smile. Have a good time and treat a customer like they’re somebody that’s in your own home. Wow! You get two free ones for that. How did Bill and Warren do today? Could have done better. Bill was good on the menu board. Warren was good at taking credit. But all around, they did all right. I’d hire them. Hi, welcome to Dairy Queen. Can I take your order? I’ll have a small Oreo Blizzard. That’ll be $3.23. Can you pull forward to the next window? Okay. Wow. That is for you. Good job on this one. Love to have you here. Thank you. See you. [MUSIC PLAYING] So, it’s really fun seeing video of you and Warren spending some time experiencing jobs and tasks in the fast food industry. Do we need to think about special solutions for different kinds of workers, including blue-collar, skilled, unskilled, all the way up to company executives, or should we be thinking broadly about platforms that work across the spectrum of jobs? I mean, they have this — that way you put the ice cream in the cone, that they’re very picky about it swirling and not dropping out. And then they do that thing where it’s upside down just to show that you got really great ice cream, that it’s not like all water or something. Anyway, it didn’t work for Warren and I. Had they given us a HoloLens, maybe we could have been guided slightly better. Yesterday when we were doing our maternal mortality thing, they gave me a HoloLens and I got to do a delivery on a model where you actually saw how to get the baby’s shoulder to come through. Anyway, it was quite impressive. And something that will actually be deployed. So for these physical tasks, there are special tools, like augmented reality, that I think will help a lot. My personal desire is more for desk work because that’s 98 percent of what I do is just shuffling the digital equivalent of paperwork all the time. But certainly through my foundation work, when I think about farmers or people out in primary healthcare where you’re going to live and die without meeting a doctor, the idea of remote assistance on complex tasks, that is something I’d like to see us be able to help raise the quality bar. Speaking of that, I had a chance to play with the HoloLens, too, with Jamie Shotton and team at MSR Cambridge last week, I was blown away by how — the experience with the HoloLens too. And I know we’re already doing field tests and actual use cases with people using the mixed reality in a number of job settings. So it’s very promising, I think, in some way that the technology — do you see that really expanding into more general usage? Yeah, it’s surpassing, they’ve got a strong roadmap, and it’s nice that we’re able to bootstrap the volume with tasks, like inspecting an airplane engine, that’s very high-value. Some people thought that, okay, the initial incarnation will be a child walking around the room and seeing little fairies flying around, which of course that’s very important. But we couldn’t quite hit the price point for that with our first generation. That tiny little field of view, it doesn’t work; you always have to look for that little fairy everywhere. So it’s like Generation N that will get to that original consumer version and we’ll have quite a variety of the things. But in the meantime, the idea of the interaction — one of the paradigmatic applications is, would you rather be in a meeting, which is a sort of Skype, camera representation with rectangles for each of the sourced locations, or eventually can you have, let’s say pure VR-type meeting where your Avatar — in terms of your expression, laughing, agreeing, who are you looking at, who are you talking to — do we make that in terms of latency, emotional sharing, richness competitive? And we’re not even close. The demos are getting better and better, and HoloLens is a piece of that. But I still — and sometimes I’m overoptimistic on these things, I’d put that, the idea of that type of ad hoc meeting in virtual space, that we can see your reaction and who you’re talking to and get the audio and video right on that. But I think that should be within a five-year timeframe. And once you get it to be equivalent, there’s a lot of things about it that are vastly superior in that you’re not partitioned into these, oh, three people at that location, one at that location. It is much more free-flowing. And then eventually get to the point of thinking, okay, will that actually compete with traveling to do activities in some cases. So then they have a lower carbon footprint. So, speaking of this idea of distribution and free-flowing, we’re seeing the rise of gig economy, freelance work apps like Uber having big effects on the world and including effects on labor and how work works. Any thought on what this might mean moving it to the future and the freelance economy? The key element is education. And it is interesting in a certain way how now that we have all courses about almost everything, you still have to pay a little bit for the learning company to get the very best. But in math and science, even the free stuff is extremely good that’s out there. And yet — and for early level math we have Khan Academy and things, that’s had a truly modest effect on actual skill levels. And so the motivational aspects — why do you learn? Many people learn just to get a degree because that is how you get a job. But how do you make things interesting? That’s kind of an unsolved problem. So I do view retraining and education that in order to deal with the dynamism of this fast-paced innovation, that you also want to make equivalent breakthroughs, not just in automating things, but improving education. And if you could have one wish to improve one thing in the economy, education is kind of the master switch of your capability. There’s a few things like energy and stability and trust that are important, too, but education would be right up there. And yet if I say to you, oh, the best math teacher lived 50 years ago; he’s better than any math teacher since then, you couldn’t disprove that was right. That is, our understanding of how you draw kids in and how you understand their misconceptions is it’s very ad hoc. The average math teacher today is probably about the same as 20 years ago. And yet if we can do agents, we’re talking about the general issue of being able to read and form knowledge, the idea of being as good as a one-on-one math tutor and seeing what types of conceptual mistakes people make, and, even more importantly, understanding how you motivate them — do you keep it at the right level, or do you create the right sort of problem domain that they might be interested in, or some sort of contrast or draw a social aspect into it. That motivational piece is where you see most of the variance. It’s not domain knowledge that makes the difference, as we see, because it’s online and yet we don’t see much change. So, I am enthused about that. Whether that means that your learning will be more episodic and you continue to do it in adulthood, that would be ideal so that even if the job market is switching, you’re qualified. Do you have a sense that that will be needed in the future more than it is today? Somewhat. I mean, as society gets more productive, your production frontier moves out. And so you can choose to say, okay, we still have shortages. We’re not at the production frontier where every handicapped kid has a full-time aide and every older person has somebody visiting them and helping them. And particularly if you look at poor countries where people live and die without ever meeting a doctor and there’s still malnutrition of 40 percent of the kids, some day society will get to the point where our sense of purpose won’t be based on work and avoiding disease and getting enough food. And we’ll have to have sort of a religious, philosophical view that binds us together and has us collaborate towards goals other than the human basics. In many cases we aren’t meeting what we all would define as those human basics. So, yeah, some degree of retraining will be very important. It’s skills like driving become less in demand, or certain manufacturing-type activities become less in demand, you still have value. And society, even beyond the economics of it, giving people a sense of purpose, something that they’re connected to, is a very — helps organize society. In particular you don’t want large numbers of young men who have nothing that they feel engaged in the sense of a purpose in. So, yes, I do think that there should be an equal amount of innovation and discussion about helping people learn new things, motivating them. If you could supercharge education, then you, by many generations of where the labor market is well-matched to what those demands will look like. So on this topic of the evolution of tasks and jobs, sets of tasks, there’s been a lot of chatter about the rise of new forms of automation, coming with reaching what is largely perceived as an inflection point in AI, machine learning-centric AI these days, but more and more automation coming on the front, frontiers of — and edging into tasks of various kinds that have required human intellect in the past. And so what’s the short and longer term futures of the influence of AI innovations on jobs, tasks and the economy? Well, some things, we’re always just gaining more respect for the human capacity to do the thing. You know, speech recognition, in the ’60s, people thought, oh, we must be close. That must be an easy thing. Now, hey, actually we are pretty good. But that’s 50 years. Image recognition, the software is really amazing. Some things like the dexterity of the human hand, we’re not that good. The mobility piece is a closer to being a solved problem. But if you do combine really incredible vision, really good arms, really good legs — and so the tasks like cleaning up a room or assembling something, if those can be done by robots, eventually those robots will not be super expensive. They won’t make a lot of mistakes; they won’t take a lot of breaks. So that — it’s hard to say what the date for that is; some people working robotics have told us that we’re stupid to think it’s anytime soon. And then other people tell us, oh, it’s all exponential, so just go to bed tonight and we’ll wake up and the singularity will be somewhere. So there’s quite a range. And I’d say I’m more optimistic than Brooks and less optimistic than others. But society will want to be ready for the point where those things happen. We have a little bit made it opaque by taking car driving as the paradigmatic task. And it has so much demand for reliability in dealing with arbitrary situations that are unexpected, like somebody stopped in front of you or the ball rolling out. Shedding that momentum, there’s just no magic way to do it. And so it may actually hide from people that slowly but surely — dexterity, cost of robots. We really are making progress on these thing. And when Amazon said they were going to retrain 100,000 people, what did that mean? I wasn’t sure. But it was nice of them to say that. But sometime — I do think in the next 20 years a lot of things that are physical in nature, even with the complex recognition, cleaning up a room is super hard. You have to have almost human-like vision and arms to do it. So will that be achieved in 20 years? It would be interesting to get people to forecast various things and see how much there is. So, yes, society should get ready for that. Government policies, in terms of the education offerings, the way the safety net works, the way the tax system works which currently discourages labor, you want to have a huge shift since you have this — you really want labor. So things that are more encouraging of labor like income tax credit will dominate and things like Social Security taxes that cost you on labor, those would go away. But we have many decades to get it right. But it is a fairly dramatic thing that you want to get society broadly involved in helping you think about with plenty of lead time. So, we often think about, when we hear the phrase “Future of Work,” I think most people in this world think about the western developed world and don’t think deeply about the largest swath of humanity that’s outside the more developed nations, and the influence of technology potentially, including AI, IoT, different kinds of systems work going on and platform work. I thought we’d show a video of your visit to a farm a little bit east of here on the FarmBeats project under Ranveer Chandra at MSR. Roll that video and we’ll come back to discuss it. More than three quarters of the poor people in the world are farmers. They’re faced with a very tough problem: They have to grow enough food to feed their family every year. When you think of digital technology, you don’t think of measuring soil moisture; you don’t think about helping people know when to plant or understand what’s going wrong on their farm. But if we can make these sensors small enough, cheap enough, then the chance to get this down to more and more farmers, get them additional productivity, is pretty exciting. We’re taking this data, we built machine learning in the AI models to do two things: One is virtual sensor prediction. So we’re predicting things leaf wetness, evapotranspiration, soil radiation. And we’re using the data to customize the model for the farm and we’re getting very accurate results for each one of them. These sensors use a new type of connectivity that’s very inexpensive called TV white spaces. TV white spaces is unused bandwidth in between the TV broadcast channels. Governments are now allowing this bandwidth’s views to be used to transmit data. In this case it’s the data that’s coming from the field that goes back to the computers that helps create the best advice to the farmer. What you’re seeing here is a TV white spaces router. This is the here your Wi-Fi antenna at home powering all of this through solar panels. You just power this on and you get Wi-Fi on demand in the farm. We use it to send drone emission. Once the flight is complete, it will start transmitting the data over the white space to do precision map generation. Bring it back. Any data you can get to farmers can make a huge difference. The weather is always highly variable; deciding to, say, invest in fertilizer; when do they plant; understanding which crop would be the right one at this time. Even 20 percent more productivity means they can afford school fees, save a little bit for a tough year. Climate change is going to make the farmer’s job a lot harder. Just closing that yield gap even a modest amount would make a huge difference for all those farmers. So there’s a suggestion that we could do a lot, we, the broad community of folks interested in technology and society, and applying advances in computing to assist and transform the nature of jobs and tasks and needs, particularly the developing world. I know Ron Bere does work in eastern Washington farms, but does most of his work out in India and Africa these days, looking at parallels and trying to generalize. Any thoughts about technology in developing worlds in the future of work? Certainly there are areas, particularly Africa, where the population growth will be very high. The world at large, the birth cohort, has peaked. And so actually all the global level population increase is people surviving longer so that — filling out the age pyramid. But that masks a very dramatic shift where, over the course of this century, half — the number of children born will go down in the non-Africa parts of the world will go down quite a bit and the number in Africa will go up quite a bit. And so Africa is projected to — Sub-Saharan Africa — to triple, to quadruple their population. And they already have very small farms with very low productivity and they have climate change, and the quality of the governance in terms of getting farmers advice and infrastructure is actually pretty limited. So anything you can do, whether it’s better seeds, getting through their mobile phone, advice to them, helping them if they have three good years be able to get a good price so that, say, they go from one in every 10 years, in fact, one every four years, they can use their monetary savings to get through that without basically having extreme malnourishment; that if they didn’t have the savings, that’s what they’d have to go through. So there’s many ways that we want to attack this problem. There was a thing called Digital Green, actually, out of the Microsoft Research stuff that is being used to get some of those farmers advice. That helps, but we have such a long ways to go, even at the seed level, things like photosynthetic efficiency. And that’s very advanced work to try and get more into their fields and sort of recreate what happened with the green revolution. But I’d say whether or not enough can be done to avoid the living conditions there being even worse than they are today, it really does hang in the balance. So changing topics a bit. It seems at least for me, my family, friends and colleagues, that work and family time, work, recreation, play, start interleaving a bit. I seem to be working all the time. My wife comments on this. And it seems that that technology has been, in part, the source for some of that interleaving. Maybe some of it is a good thing. But your thoughts about that, maybe some personal comments to you in life. Well, everybody has to have a certain sense of discipline. You can’t read all the books you want to read. And so time allocation, it’s maybe — particularly my situation — the most limited resource there is, and so deciding, okay, family activities and taking some time off and then still getting the great enjoyment out of the time I spend on Microsoft work and also the foundation work. I analyze how my spend my time pretty carefully. And then I like the fact that when I’m on vacation I can read books about the latest in AI or biology. And that’s when I get to read super complicated things, and in some cases write fairly good memos. I go back and look at my time and see that I’m super lucky. Experts are often willing to come talk to me about things. Yesterday, when we were talking about birthing, we had people who spent their whole life trying to improve maternal survival. And just phenomenal to hear about how the new technology and their experience can come together on those things. And my kids are now — the last one is in high school. So actually we’re worried that we won’t have the distractions of having kids around, because that’s a big pleasure to do that thing. My wife and I, because we work together at the foundation, we don’t have any shortage of time together, which is great. It’s fantastic. [ Laughter ] Some people, when their kids leave, it’s like, wow, that was our main joint project. So what do we do together now? But I take advantage of the tools. I look at my Oura Ring and see if I’ve had enough REM sleep. And sometimes I worry, am I spending too much time thinking about things like that? But you calibrate after a while how much the tools really help you or not. So you mentioned how important engagement is. One comment, and I’ll end with this, is I — and I think the people in this room and the rest of the world — are so amazed and happy about how engaged you have been with everything, the things you learn, Microsoft, the Foundation. So I’ll give you a big thank you and thank you very much, Bill. All right, thanks, Eric. Thank you so much. [ Applause ]


  • G Lasser

    thank your for enlightening us with ideas from the sharpest minds.

    The model for positive impact being to solve large problems elegantly with profitable solutions. ?


  • oh my gosh ponies

    I would love my computer to help with; (1) "I see you are thinking about eating at Restaurant. Remember, you already ate at that restaurant a few months ago and didnt like it. Why not try DQ" (2) "you watched that movie last year and hated it. Can I remind you of this other movie you wanted to watch 2 weeks ago" (3) a scientific paper template and generator. It auto creates the sections with titles. Grabs recomended papers and puts them in background and reference section. Etc (4) using halo lens to ID plants and animals I am looking at. Like an automatic field guide. ALSO doing the same thing for rock identification and showing large geologic structures I am looking at, and helping peel away the strata as I look at it.

  • Coastwalker

    Colour me skeptical but if the software is as clever as he anticipates why would you need a moronic human being to drive it. Just exterminate the human being and the job still gets done.

  • Alexander Yeshua

    JS and C# cannot be together

    EXE and APK (console command) are differences

    so u know whch was actually "artist" as everyone cn see frm demo, and whch one really know how fr wrte…

    u knw if anyone knw how fr write ONLY when u see they use notepad fr write….

    why? EXE… so many 'feature update', less new EXE…i knw fr write, but hey, let jst see them(some artist frm any orgzn act)

    come on, Dot EXE WRITER IS NOT DEAD YET, I AM A WRITER myself, so many web, apk, with already hv tmpltes, similar things: use – navigateActivity tmpltes, UserLogIn templates, and Web html tmplates… many graduate.. less rare exe today…

  • Alexander Yeshua

    Building "a.i"
    Authorize narator to batch tht hs alots of string, bt then, the int Systems.windows.CreateObject() will always start when the narator EventListener matcher="create _ "
    Fr new command, but also, the form will always be duplicated, it nt actually algrorithm, bt ws CreateForm, with ready made format, then string bck to bat., as a copy file….
    Dnt worry, jst watch too much mvies, bt the movies ws nt inspiring… inspired by Lorde(Yeshua Chryse, Bishop He himsele)…hahaha…

  • Alexander Yeshua

    Against ransomware:
    1st download "ollydbg2.0"
    Ready 2 pc and a hrdsk extrnl, use anther pc to cptre the flg, then anther one would be copy frm the blue screen- "directory mode" bfr starting window as whole fr be seen(Graphic mode)
    Also, DISEASE, WAS LIED, should force the news or eductns fr speak beliefe, teache fact n not production tht fail fr filter fake news…
    dangerous archeologies years…
    what i think others should do was nt "maintaining lies abt health" , but ths: stop teaching lies that effect world wide…. when thy worried, they think alots, as elders/educated, should nt lied to Father's children…
    stop being lied, Dr Bill gates… and other Drs… the users(steam platform already hv vctries against money system), should chnge, dyd suggest, fllow Father's teachng and Father's consent

Leave a Reply

Your email address will not be published. Required fields are marked *