Transcription
So, Bill, we've known each other for some time now. We probably don't want to quite date it, but it's been a while. One of the things that I love about doing podcasts with my friends is that I actually, in research, learn some things that I didn't know before. Apparently, there's this kind of trifecta of three criteria: will it have a big impact, will you learn something, and will it be fun? What are some of the projects that immediately come to mind that are the most interesting and meet that trifecta for you? What goes, "Ah, that's something that met all three and is fun and exciting for you"?
Well, until age 20, I got to read about lots and lots of things. I ranged quite broadly, including auditing a lot of courses at Harvard that I wasn't even signed up for. Weirdly, then when I got into software, I had to suppress my sort of normal desire to be polymathic and be monomaniacal. So, from age 20 to 35, I didn't stay up to date on geology. By the time I was 30, I started cheating, reading some other things, in particular when I turned over the CEO role. So, it's nice now that, partly because the foundation touches on a lot of things, I do get to range pretty widely. There are a few topics, like climate; you have to learn about weather, materials, and energy. So, it's a great excuse for learning things nowadays. Global health gets a lot of my attention, and I do put a lot of energy into that because it really fits all of those criteria. It's such an underinvested field that you can invent tools that save millions. You can save lives for less than $1,000 per life. But, you know, now you've got climate, you've got AI—no shortage of interesting topics in the world today for somebody like me. My ability to know people who can help educate me and the online tools mean you don't have to worry about getting confused because I know someone who will straighten me out.
One of your projects that actually hit all three for me is your recent Netflix series. I loved it! What's next, "The Future with Bill Gates"? I will humbly say I think it has a lot of DNA in common with this podcast. It's about the future; it's about what could possibly go right. Can you just tell us a little bit about how that experience was making that show? Is there any memorable moment, maybe an outtake that didn't make it into the final cut?
Well, five years ago, I did a documentary with Davis Guggenheim, "Inside Bill's Brain," and he picked things I was working on that could fail: nuclear fusion, polio, and magic toilets that don't need sewer systems. That was an interesting paradigm—why was I putting money into those when essentially no one else was? This one is quite different because it takes topics like misinformation, where I don't know the answer. I mean, literally, that's one of the few problems I say, "Okay, young people, we screwed this one up; you better create around it." You know, is AI going to help? Is AI going to hurt? It was fascinating talking through it with my kids. Did they want to be on the series? Two of them were like, "Ah, that doesn't seem like a priority," and then Phoebe was like, "Yeah, you know, Dad, you're so out of it on this digital stuff. You still try to send me email. Let me straighten you out on these things," which really became the case. I met people that I hadn't talked to much before. I'd never met Lady Gaga before, and that was kind of a privilege. She's a very interesting person.
On global health, we had so much footage. That's the one that I worry will it get the viewership that the others will get because we really do know what to do in that space. It's kind of amazing, and because it's far away, I think people would learn a lot because they're not confronted with 500,000 malaria deaths a year. The fact we see a path to drive that to zero—I mean, I think the tagline is "Come for Lady Gaga, stay for global health." You got it! Absolutely, yeah, Netflix might use that on its rotation.
What are you currently most excited about in terms of the technologies that will make a massive difference in changing what's possible at scale?
Yeah, the current situation is that all the things I'm working on that relate to innovation, whether it's climate or these health issues—malnutrition, infectious disease, or digital tools for, say, teaching or health—the pace of innovation is faster than I would have expected. I have pretty high expectations. I go to product meetings and say, "How come we can't do this twice as fast?" I just do that many times a day. Yet, innovation is even exceeding my best hope, so it's super promising in all of those areas. Take malnutrition: almost half the kids in Africa, their brains and bodies don't develop, and we haven't understood chemically that they're getting enough calories. What are the micronutrients or the mixing in their diet that causes them to be, on average, 5 inches shorter than they would be and 20 IQ points less capable than they should be? For them and their countries, that's pretty profound. Now, with the latest tools of science, looking at these bacteria in the gut, the microbiome, we clearly have a path to solve malnutrition. People should go, "Wow, that is a very, very big deal" in terms of the uplift that comes out of that.
On the energy side, things like either getting fission or fusion to provide both very cheap and constant reliable electricity—that's a longer time frame in the fusion case. But, you know, there are a ton of companies I'm investing in, in five deep understandings of complex science, including all these diseases. This next 20 years is going to be mind-blowing.
Actually, knowing a little bit about malnutrition, can you go to the next level of depth on the bullion to make it a little more tangible so people could say, "Oh my God, we're there now"?
Yeah, well, with malnutrition, if you don't get the right vitamins during pregnancy or in your first several years, you can never catch up. It's a sad fact; you don't get to go back and say, "Okay, eat your Wheaties, and you get the IQ points and the physical capabilities." It's weird that you're just missing small amounts of things like vitamin A and vitamin D. So how do you solve that? Well, you could fortify a food. Like us, breakfast cereals are fortified, but you have to find some food that even the poorest households, who are the most likely to be malnourished because they're not getting eggs and milk and meat, will buy. It turns out these bullion cubes are preferably bought by low-income households because they give them something tasty, and it's very cheap relative to how tasty it is. Now, we're going to put a lot of vitamins, particularly vitamin A, into that bullion cube. It raises the price about 3%.
Something I love about what you do is you identify the problem, you bring data to it, you test it, and figure it out. Especially because you're working with populations who are low income in Africa, the poorest of the poor, even if you're raising the price 3%, you have to build that into everything that's happening. Now we want to get into all of the different areas that you're focused on, so let's start with climate. I was lucky enough to have dinner with you, Reed, and a few others a few weeks ago, and one of the interesting things you were talking about is how one of the interventions we need to do is around cows. I feel like everyone sort of knows that maybe cows contribute to climate, but you were giving me very specific interventions that could move the needle. Could you talk more about those?
Yeah, so there are basically two ways to help with cows. Cows are about 5% of global emissions, which is pretty unbelievable. If your goal is to get to zero, you don't get to skip the cows or the steel or the cement or any of those big areas. There's a whole class of solutions for making meat without cows. Today, it doesn't taste as good, and it costs too much. It's going through a little bit of a lull, but those companies—Impossible, Beyond, Memphis, and others—are pursuing that. In terms of the cows, we have pursued many solutions. One is to vaccinate the cows in a way that their gut bacteria emit the methane, which is also called natural gas or stage 4, which is the second most important greenhouse gas. You can vaccinate them, and that species of bacteria isn't there. Their stomachs are very special because they can eat grass; it's a three-stage fermentation process.
Basically, there's another way you can change what they eat. You could either put that in their water or their feed. There is a drug to change the microbiome—not a vaccine, but a drug—that looks very promising. Then there's a solution where you stick a sort of metal thing into the skin of the cow, and it actually burns the methane. All of these look to be quite cheap and implementable, even in Africa. This is one where I wasn't hopeful when I got started a decade ago, and now it's just a question of which solution for which country ends up being the best.
Some of it's similarly amazing to what you did with the toilet and all the rest. It doesn't look like it's possible; now it is, and it makes a huge difference. One of the other funny things when we were talking about the discovery of the workaround for cows over the dinner table was once you focus on it, it's not just the systemic climate change, but it's also the questions about, for example, what this can do for quality of life in Africa. Right? So say a little bit more about milk production and cow breeds and other things because I never thought of you as the cow expert, but here we are.
Well, I didn't grow up knowing much about cows. Protein is a very important part of a good diet, and foods with proteins tend to be very expensive. So if you can make chickens and cows live longer and be more productive, then that's super beneficial. The West has taken these cows, these Holsteins, and driven them up so they make 30 liters of milk a day, whereas the normal cows make less than 3 liters a day. You have this factor of 10 productivity through the genetics of that cow. Now, you can't just fly a Holstein down to Africa because of the heat and the diseases; it's not adapted. But if you do the crossbreeding properly, only giving up a little bit of the productivity, say down to 20 liters, still a factor of six better, you can improve those cows.
You also take the idea of grazing, where the cows are going out into areas that are now being fenced off, and you can change it so the cows are largely stationary, and the food, which they call fodder, is brought to the cows. You avoid these incredible conflicts between the grazers and the other farmers. It's one of the most exciting foundation things. Once we do the R&D and get those good cows in, it's private market sustainable. We're further along with chickens. I was in Ethiopia a few weeks ago, seeing that we've cut the price of chickens in half. Women who do this make extra money; they give some of the extra eggs to their kids as well as selling them. It's taking, but it's leveraged off of the West that spent the last 100 years doing selective breeding of both chickens and cows.
So, obviously, when we're thinking about chickens and cows, it's climate change, it's malnutrition, it's potentially war and conflict. If we zoom out on climate change, what is the sort of statistic that for you represents the gravity of the situation, or what are you working towards to decrease when you think about climate in general?
Well, the size of the emissions, which is over 50 billion tons of CO2 equivalent per year, and the pie chart of, okay, one of the big five areas—like transport is one of the big five—but then, you know, cars, planes, boats, trains underneath that. Electricity, which is coal, natural gas, industry, buildings, agriculture. I want everybody to have that pie chart in their head. The theory of change is that we need to make all those things without them costing more. I always say there are two numbers: 50 billion and zero. Zero is what we want emissions to be, and zero is what we want the extra cost of green cement, green beef, green rice, green cars. We want it even at the low end where you park on the street because even if the world should pay a lot for green things, they won't because it's a global problem.
The real problem is for future generations. The negative impacts have been somewhat overstated for the current generation, actually, but because it accumulates and gets worse over time, if it causes us to help those future generations, it's not a terrible thing. The damage is mostly in poor countries. It's almost like people didn't realize weather has always been a problem, and they're like, "Oh, every bad weather thing that's climate." No, there was bad weather before, and it's slightly getting worse, but it's near the equator where your absolute temperatures defeat outdoor work and the current crops that we have. That's where you're getting into conditions that humans have never thrived in.
Well, on the electricity side, and we'll get to nuclear in a minute because you and I share a passion for that being a great source, and there are all kinds of things. One, the need for electricity. Second, you have good scalable, clean, cost-effective power. You can solve all kinds of ills. What are the other energy sources of renewables and green energy that are capturing your attention outside of nuclear?
Well, of course, we want to keep driving solar to be cheaper. That's gone way better than was expected, and there are some new things using perovskites that will drive the efficiency up of that. We want to drive wind costs down, including offshore wind, which is still quite a bit at a premium. We want to improve energy storage, but it's not realistic to think we'll completely solve that problem, which is why you need nuclear in the mix as well. Geothermal actually looks like it might play a role. The western half of the United States has pretty good hot rocks, and then there are geothermal companies that want to dig really deep holes. That's more early stage, but FVO and one other company are showing that they can actually get reasonably good pricing, and now they're scaling up. Google just did a purchase agreement at a premium to help them scale up, which all the tech companies are very oriented towards not raising their emissions.
They will take advantage of that to get these products onto a learning curve. Eventually, we want to have a zero green premium, but somebody has to help get us there, which solar was very subsidized, and then under certain definitions has now gotten to a zero green premium. There are a few things like tidal that probably are pretty limited. Solar panels in space? Maybe. Some people have even talked about putting the whole data center up there. It's just bits. Actually, moving bits from space to ground is easier than moving energy from space to ground. Particularly because launch costs are down, you can at least dream of those things. That's kind of a far-out thing, but it should be in the portfolio of innovation.
One of the things that I think is useful to highlight for most people is, for example, the current discussion of AI is always going to be about all this electricity. But the investment that all the hyperscaler companies are putting into how we make clean data centers, how we have clean power, and all the rest is like that subsidy for the R&D. These advanced purchase agreements are exactly the kind of thing where you kind of like different ways of conceptualizing public-private good.
Yeah, I mean, I say that rich countries, rich companies, and rich individuals should bootstrap the market for these green products. We should buy clean aviation fuels. Some nations may mandate that for private aviation, which that would definitely be a good thing because it would get the volume to go up so that eventually we can get something with either a zero or very low green premium into commercial aviation, which is another 6% of the emissions we have to get all the way to zero.
So, yeah, and it's important to remember that all that data center demand—it's big numbers, and it's coming quickly over the next six or seven years—but it's not as much as electric cars or electric heat pumps. Our climate solution, because you can't avoid using the energy somehow, the only non-hydrocarbon way we really know how to make energy in a portable form is electricity. So we get rid of coal and natural gas, but you have to make a lot more electricity to replace the heating or industrial capacity that those direct hydrocarbons provided.
Can you tell us a little more about solar? You were telling me about the amazing increase in the ability of solar panels and what percentage of the sun they capture for energy, which was fascinating. But you also said that you don't think batteries are going to be here in time, or is it we just need more time to get the batteries where they are? The cost is too high, and I feel like not having the batteries we need is what is inhibiting solar from being the energy source that can save us all.
Most places, we should be adding solar as fast as we can, and we're actually limited by the grid capacity. So I love solar. The efficiency started out at like 10%; it's in the 20s now. It could get as high as 40% with new approaches over the particularly next decade or so. But it's not just a 24-hour storage problem. Lithium-ion batteries are now sodium, and others will solve your 24-hour problem. But you have periods of time, like the Midwest gets a cold front where you get 10 or 12 days. All the batteries ever made in history for every car, every computer wouldn't store a day of electricity. If you're only using a battery once a year, that is, you charge it, and it's sitting there for this unusual thing, that's super expensive electricity. Instead of getting the capital value out 365 times a year, you get it out once.
So various seasonal and bad weather things, where you don't want to shut off power, particularly if it's being used to heat homes, it's way more complicated than people think. Electricity doesn't move long distances today; it's mostly coal or natural gas plants that are fairly near to the usage. Yes, there's innovation in transmission—actually, fairly exciting stuff—but we're going to have to have a mix. In particular, if you look at a country like Japan, where there's essentially almost no solar potential there, and even the wind has periods where you have too much or too little. The U.S. happens to be very blessed; we have incredible wind and solar resources. These open-source models that are now modeling out, okay, what does that energy system look like, are part of seeing when can we get there. There are a lot of these goals that are not well thought through. It's going to be harder than we wished it would be.
One of the things that you already mentioned, gesture at before, is that part of the reason why we need nuclear fission and fusion is because all these awesome renewables—solar, wind, hydro, etc.—well, hydro can do a little bit more 24, but yes, they have limitations on when they generate and when they don't. Battery storage is a challenge. You and I have invested in some fusion things together. Say a little bit about what's the hope and the possibilities for fusion.
Fusion is where you take the big atoms, like uranium, and as they split, you get energy. Fusion is where you take the small atoms, primarily hydrogen, and as you put them together, you release energy. The middle of the periodic table is the most stable, and so you're getting relativistic energy through that mass decrease. Fusion is very difficult; it involves temperatures that are like the center of the sun—millions of degrees. That's plasma physics, which we know a lot more about. We're using AI tools to study those things, and now there's a variety of techniques. Tokamak, which the Common Fusion System is using, is the one with the most credible schedule for, say, 10 years from now. Most of the others are probably more like 15 years away. But at some point, fusion energy will be extremely cheap, and it doesn't have the same waste problems that fission does. I think those are solvable problems, and I'm investing in that because that's more like a six-year time frame if everything went perfectly.
So, society as a whole, even though a lot more money is coming into it, we're still under-investing in fusion. Given that the value of cheap electricity specifically is so fundamental to society, I mean, if somebody says, "We have a water shortage," no, there's a lot of water; it just takes energy to move that water and desalinate it. If energy is cheap, you have infinite wonderful water everywhere in the world. But the energy is too expensive right now to do that at scale. Even recycling things, you know, why can't we? Atoms don't leave the planet—tiny bit of hydrogen, but it's all there. The reason we don't is that the energetic costs of restoring things to their original state are too high.
Well, one of the, for our listeners, a terawatt super awesome way of kind of actually using nuclear fusion waste as fuel and so getting a compounding effect is one of the things that you guys have been working on for a number of years. Maybe even 2006 is when the fission company Terawatt gets started.
Yeah, exactly. Now, one of the things that always amuses me about some of the current public dialogue around AI is like, "Oh, is this going to accelerate climate change because of the electricity cost?" I think what, you know, most of these people are not realizing is how we can also use AI to help us with climate. Because, like, if we can get a lot more intelligence applied to various problems, that can help us with climate. Say a little bit about how you're thinking in that arena.
You know, the extra electricity load is there, but it's like a 10% add-on. It'll challenge the way that we do green accounting a little bit. I wish fission and fusion were sooner because this sort of gold rush for AI backend capacity is kind of the next few years. Even fission will only be able to make a modest contribution in the 2030 time frame to that electricity supply. The value of AI in solving the scientific problems of, okay, how do you grow? How do you make plants more productive? Okay, you model photosynthesis, and you model how you change the plant genetics in order to double the activity. That's a very profound advance—improving photosynthetic efficiency. In fact, because it's kind of a far-out thing, the foundation is the primary funder of that. As we show that it can work, other people will come into that.
But anyway, AI for material science and biology is a gigantic accelerator. So take whatever green product you think is going to be the hardest to get to a zero green premium. Rethink how hard that's going to be because the AI tools are so phenomenal at accelerating all of these paths of innovation.
Yeah, and actually, one of the things I've been thinking about is, while it's a big electricity cost for training these scale learning machines, once you have that intelligence—like, that's how we've made everything is through intelligence. Once you have that intelligence to amplify across the board, applying that to climate change has got to be a multiplier effect. We get this much actual savings in carbon and other kinds of things through the application of this electricity. I don't know what the multiplier will end up being, but I'm certain it's there.
No, it's absolutely there. There are some goals, like not going above 1.5 degrees, that even with AI being a net positive contributor, because of the difficulty of scaling in all the areas in all the countries, some of those goals we will miss. But we will avoid the level of heating that would be disastrous, and we will need to do some adaptation, particularly in poor countries.
So, I want to switch gears. Another area where I think you're probably best known is global health, and I think it's an area where AI can do so much. My husband is a public health data scientist, so he's particularly excited about this area of the interview. You have focused on the eradication of disease, and I think—fact check me—in 1980, the WHO declared that smallpox was eradicated, and that's the first and only disease that we have eradicated. You've said, "Let's tackle polio. Let's tackle malaria." How do you pick what is the next disease you're going to tackle? What an amazing ambition! And then how do you go after it?
Yeah, so most diseases we're going for a burden reduction only. Very few diseases should you try to go for eradication because it's very, very hard to get to zero. Right now, with polio, we're in Afghanistan, we're in Gaza, we're in Somalia, we're in DRC, and we're having to execute high-coverage vaccination campaigns against misinformation and violence in the toughest places in the world. So it's very, very hard. Polio is close. There's one called guinea worm, which is confined to Africa, where President Carter just got to 100. He's been a champion of that, so we're hoping he'll be alive not only to vote in the election but also to come to the guinea worm celebration party. It's going to take a couple of years, so he's going to have to hang on a little longer.
The magic thing that happened at the turn of the century was people got serious about global health—about really measuring, okay, kids die of diarrhea, but what caused that diarrhea? They die of pneumonia, malaria—okay, it's more clear what that is. But let's, even though there's no market, the people who die of malaria—half a million kids a year—it's not like you can make a business case of, "Hey, go to Silicon Valley and do a malaria startup," and look at that spreadsheet. The line that says "life saved" will look good, but the line that says "profit" will have a lot of red numbers because they can't afford these tools.
So medical science is very distorted towards rich world conditions, and even amongst rich world conditions towards cancer and a few other things. The incentive system could be improved, but the Gates Foundation—that's our place we come to fill in—is that the things that aren't market-driven, like getting diarrheal vaccines cheap enough for all the kids of the world, not just the rich kids who don't die of diarrhea. That also used to be half a million; now it's down to 100,000. As we went from 10 million under-five deaths per year at the turn of the century to 5 million, diarrhea was one of our best. Pneumonia—we got a vaccine out for that, which was a very expensive vaccine that we worked with all the vaccine companies, Western and Asian, to get those prices down.
So we're basically driven by the inequity where we say, "Why do mothers die in childbirth 20 times as much in Africa? Why do kids die 50 times as much in their first five years in poor countries, particularly Africa but also Southeast Asia?" We're taking all of those and saying, "Okay, let's find the best science. Let's understand the field conditions. Is this stuff deliverable? Will it be accepted?" We have crazy ways of killing mosquitoes that alone doesn't get rid of malaria, but if you treat a lot of humans and the reinfection rate has been massively reduced, then you can get to the point the U.S. got to because we had malaria, where you've cleared during the low season—that's the winter here and the dry season there. You've actually cleared the parasite, so there are no humans, so there's no reinfection taking place.
In the next, hopefully, less than 20 years—we have five years now—we need to do the tools. Our goal is to finish polio during these five years and then, with new tools, get the credibility to get the world to fund a malaria eradication starting in 2030.
Yeah, totally awesome. One of the other projects we work on together is AI and drug discovery, and this is one of the ones I think we may actually even begin to see some of the earliest global benefits from. So where in what areas of AI drug discovery are you kind of most focused on and think can make a global health difference?
Understanding protein and molecular shape space is a perfect AI problem because we have databases—the protein database that we have, 150,000 molecules, we know the shapes. We've trained AI on those. Their ability to predict the shapes and therefore the druggable sites in these proteins is accelerating medical discovery. There was actually a company called Schrodinger that was doing it pre-AI, but now there might be 20 times as many people progressing much faster because AI is very, very well suited to this. Eventually, AI won't just model the low-level what the shape is, but it'll model the cell and the organ and the organism. Even complex disease dynamics—it's beyond human understanding to map out all the different things that go on. The AI models, as you gather the data—which will be the limiting factor—will help you understand overnutrition and malnutrition way better than we do today.
So, you know, most things put aside neurological. In the next 10 to 20 years, I would say the likelihood of dramatic medical advances, even in the neurological one, like Alzheimer's, I'd say those would be solved. I love talking to these companies about, okay, which part of the problem they're solving.
I mean, one of the things that strikes me is I feel like it's super fashionable today to be, "Oh, the world is terrible; the world's a dumpster fire; it's getting worse. The past 30 years have been horrible; the next 20 are horrible." Yet talking to you, I'm like, "No, stop! If we actually look at the data, things have been getting better, especially in global health over the last 30 years," and often because of AI. The future's bright; there's so much that can happen. You just sort of touched on not just the technological progress but also some of the bureaucratic, some of the other things that perhaps AI or else can unlock. Are there things that AI can unlock that are just boring administrative or helping healthcare workers that aren't about cutting-edge technology but are other ways that AI can help in the public health field?
You know, in the same way that the microcomputer revolution allowed me at a young age to think, "Okay, computing will be free; therefore, what would an individual do with free computing?" Paul Allen and I kind of saw that and said, "Okay, software is the only thing that will hold that back," whereas older people kind of thought, "Oh, computers are expensive." The idea of, "Oh, it'll be doing spreadsheets or word processing," they'd be like, "Are you kidding me? That's just too expensive." Here, it's even more mind-blowing that you could say white-collar worker capability and eventually, although robotics is still very specialized, eventually horizontal blue-collar productivity will be very inexpensive.
So, you know, I take an MRI diagnosis that a friend has, and ChatGPT does such a good job of explaining it and showing where it got that material. The creativity, the fluency is kind of mind-blowing. All of us should have this crutch of, yes, if you just want to know what fertilizer is, Wikipedia was good. But if you want to know what a trip with a 16-year-old for four days with a budget of $4,000 to Italy in August looks like, nobody wrote that thing. But the AI, if it's connected up properly, is mind-blowingly good at those things.
So we're already all, you know, in our life of writing poems or speeches or understanding or having complex material summarized. We're already getting a huge benefit, and a lot of white-collar work should already be either more productive or drive towards higher quality.
So actually, one of the things this conversation is reminding me of that I haven't actually ever asked you is, you know, there was this decade or so of a lot of focus on personalized medicine. Where are we at currently on that? Is that improving? Is the promise of that turning out at all?
I always am put off by people's fascination with n equals one medicines. You know, half a million kids die of malaria. Millions of people are diagnosed every year with Alzheimer's. So I'm just the impersonalized medicine guy. I'm like, "You know, the world does not have the resources to do n equals one solutions." If some super-rich person funds that, maybe it helps the scientific discovery path, but I couldn't bring myself to be involved with that because it's so unjust to take finite resources. Eventually, yes, understanding everyone's genetics and saying, "Okay, your drug doses are different because of this." So, you know, I'm kind of taking a provocative position on this. The science that goes under that name is very good science; the people who work on that are very well-intended. It's just, you know, we have rare diseases. We've created such an incentive for them versus widespread conditions. We're not really allocating that effectively, particularly—I mean, the insane stuff is the diseases of the poor countries.
So can you actually say more about—you've traveled recently, Africa, Asia, all over—about the conditions with healthcare workers on the ground? What have you seen there, and how can we help improve them?
Well, people should understand that most people in sub-Saharan Africa never meet a doctor—not when they're born, not when they're sick, not when they die. This is not a doctor's thing. Your image of healthcare for most people in those countries is primary healthcare—a modestly trained person who can give you some antibiotics, bed nets, vaccines. Very importantly, to give a pregnant woman their prenatal checks. Now, we'll get an ultrasound evaluation in; we'll see, okay, which 10% of pregnancies might be complicated, and then the woman does need to get to somewhere where there's trained personnel who could do a C-section.
We can see with this AI-trained ultrasound, is it going to be a complicated pregnancy? The predictions are stunningly accurate. You go to all that trouble, which you couldn't afford to do for all pregnancies. The greatest shortage of doctors in the world is there. The idea that, in native language, through your smartphone—which, okay, not everybody has, but time's on our side—even in the poor countries, you will get health advice, and a lot of the diagnostic tools will be available at a point of care where individuals can take. A lot of people experience this with lateral flow COVID tests. Now we're trying to convert those to be point of care but molecular tests and still super cheap but also more sensitive, more accurate.
Healthcare is, you know, it's in a—we only have $100 per person per year versus in the U.S. where we spend $15,000 per person per year. So it's a 150 times difference. In fact, there is triage involved in, okay, which things should you treat now? Even things like blood pressure, cholesterol, obesity. My hope is that not only are the costs of those treatments going down, we're also going to put them into forms where it's like yearly dosing. The cost in the delivery system of getting GLP-1 to all Africans, you know, 10, 15 years from now, will be able to do that.
So, in addition to the elevation of humanity through all forms of global medicine, what is actually, in fact, healthcare? Education is another area that you and the foundation work on intensely. Obviously, people have encountered AI a lot with this. It's kind of like, you know, what does it mean? What are some of the kind of maybe more surprising or other kinds of uses of AI in education, whether it's kind of a globally available tutor or other kinds? What are some of the AI for education for the world that's kind of captured some of your attention?
Well, I think the first thing to admit is that tech lovers like myself have talked about the benefit of technology being used in education for our whole career, and the actual benefit for the average student has been very, very modest. If you're a motivated student who can get on Khan Academy for two hours a night or watch YouTube videos about photosynthesis, wow! People like ourselves, we are able to learn in a way that's unprecedented. There's a company called The Great Courses where, when I'm on the treadmill, I love watching those things. Just so much great stuff out there. But the current math achievement of a high school graduate in the U.S. is not better than 100 years ago. It's not like medicine, where there's new tools and new understanding. If I said, "Oh, in 1900, the best math teacher was then," you couldn't contradict me because it may be true.
So we've done a lot, and I'm a believer. But AI, because of the fluency and personalization, I think we can have very high aspirations of how we mix social experience and the cloud classroom experiences with the teacher and working on your own. Correcting your pronunciation as you read things immediately, telling you, "No, you didn't get this math problem right." Not turning in homework two days later after the poor teacher spent all this time. You're like, "Well, was that a manipulation error, or was that a conceptual understanding error?" The AI, and Kigo, is on the cutting edge of this, but there are others like CK12—many that it'll say, "Yes, those two minuses you didn't cancel properly," as opposed to, "No, you set the problem up wrong because these two trains, you didn't get the when they pass each other algebraic equation properly."
The idea that it will learn how a tutor keeps students motivated, using the domain like sports or health or construction that the student can relate to, the promise of having fantastic personal tutors in the inner city U.S. and in poor countries is super exciting. Talk about an area that's underfunded. Global education is a very underdeveloped field, but particularly now with AI, I'm encouraging philanthropists to get in and showing that there are things like structured pedagogy, where the teacher is given a very clear way of teaching that we are seeing some very good results.
I really appreciate—I think some people are like, "Oh, the boy who cried wolf." It's like you told me 20 years ago that MOOCs were going to do it or that whatever. People are skeptical: "Are we finally there?" Can you tell us more about the First Avenue Elementary School in New York? Obviously, you're super concerned about equity. Is this something that is replicable that could scale?
Well, I love Khan Academy, but it was mostly used by motivated students. For the last, I think it's like eight years, they've been saying, "Okay, how do we get into the classroom? How do we work with the teachers, explain the stuff?" Yes, the computers and the internet stuff is getting more pervasive. The pandemic actually helped a bit with that. Saul and I were amongst the first two people who OpenAI was nice enough to let us mess around with early ChatGPT-4. A lot of the cool things, like having it write songs and poems, actually Saul taught me that stuff. I was like, "I wouldn't have known to ask it can write like Shakespeare. Wow!"
He has put a lot of Khan resources in. He gets support from the Gates Foundation and many others. He has created this Kigo, and last school year, he had it in a small number of schools, including this New Jersey school. I went there to meet the teachers, to meet the students, and see. You meet a kid who clearly is ahead of his class, and sort of the factory-based model of, you know, 30 kids in a classroom, you definitely have a problem where you need to do remediation and catch kids up. But you're also pained by that kid who's ahead, you know, and maybe checking out or being disruptive.
Yet, you think, "Wow, we want to drive that." The personal tutor aspect allows that student to sometimes be off on his own, sometimes helping the other students. This Khan dashboard, along with Kigo, is seen. When you walk in, your teacher in the morning, instead of people handing homework, you have to deal with that. You just go to your dashboard and say, "Okay, who connected in last night? How many hints did they need? How far did they get in the progression?" You're giving feedback. You can have the parents connected up to that. Even the thing where a paper gets turned in—you don't turn in the paper; you turn in the AI session. You can just say to the AI, "Okay, how much did the student do? What's your suggestion on how we get them to either help with the first draft or help with the grammar or the logic?"
So it's great to see it being there, and seeing it in person reminded me of embedding it in. Always with teachers, whenever you have some new thing, there's maybe 10% of the teachers that latch on to it, and you get these great results. Then when you tell the other 90%, "You must use this," those results almost always just disappear. So, okay, how do we make this one that scales? The humility of how far we haven't come is even with AI. We'll have to do that, but what I saw made me even more optimistic.
So one of the things that people sometimes miss about the approach that you and the foundation bring to these problems is, you know, not just quantification of, like, okay, cost per life saved, etc., but also the systems thinking. What are some of the kind of non-teaching-related levers for change in education? What are some policy things that either we as Americans or, you know, the world should be thinking about to improve education?
There's a lot of very good data about not having cell phones in the school, and you know, some great work going on there. There's very good data about boys should probably start later than they do. The school day should start later. There's all the learning out of the charter school movement, which is hard to stop because it shows that long school day, long school year are incredibly beneficial. Engaging the parents in, you know, here's where your kid's having challenges and communicating with them, although these digital tools are going to make that far easier for the parents who want to engage in that.
We're seeing now in communities where there are charters, even though most of the kids are in the public schools, those public schools essentially compete. They either adopt those practices or find their own ways. There are places like New Orleans or D.C. or Austin where school performance is up. We ought to make sure that even as we try to put AI into this, some of those learnings are incorporated.
I mean, as a mother of three boys, I'm like, how are they going to compete with the girls? What is happening?
So to switch gears from education, I think I'm going to steal one of your own questions, which is if you had the opportunity to meet with someone from the year 2100, what would your questions be for them?
Wow! How did you deal with the AI challenge opportunity? You know, you said earlier that I have this view that life is improving, which is objectively true. There are always footnotes, like nuclear weapons, bioterrorism, and now AI needs to be added to that list. But, you know, the past 50 years, life in general—if you're a woman, if you're gay—I mean, it's kind of sad in a way that because we're so problem-oriented, we're not very reflective about, "Hey, 5 million kids a year aren't dying." You know, like you mean with climate people, and they say, "Oh no, climate's going to ruin the world." Do you think you're going to go back to 10 million a year dying? No way! You're not going to go back.
It is a super big headwind because of the impacts on agriculture. So, you know, I bring my hay. The world is pretty good, but that 2100 person, I hope to hear how they avoided, I'll call it the four footnotes: AI, nuclear weapons, bioterror weapons, and polarization—people being able to get along and cooperate, including in how governance adapts, and which AI will force governance to come up with different ways of taxing people and regulating things. It's a little scary that it's happening at a time when the broad trust in government is at a very low point, both relatively and absolutely.
What are some of the things that you think people should anticipate coming with AI in the next 3 to 5 years?
Well, it's so mind-blowing. You know, it's sometimes hard to get your head around it. No one expected the white-collar thing to come before the blue-collar thing. In Life 3.0, they have these hills where the computers are doing the easy things, and, you know, warehouse work, which we can't yet do, was down there in the lowlands. Helping diagnose, you know, was way up the hill, or writing legal briefs or code. We're surprised at that order, but the so-called blue-collar horizontal robot that can be told, you know, "Go to this construction site and help. Go to this restaurant, go to this hotel, and clean the rooms." Even if the price is such that in the home it only drops by for an hour, it doesn't live there at first. Those things are, I believe, within easily within the next decade.
Well, and actually going back to your original vision of a PC on every desk with software that would help people do their work and lives and so forth, what do you think is going to come now?
I think a lot of people haven't realized that part of what's happening with current AI is essentially the largest programming language will be natural language, e.g., English, and everyone will have a coding assistant—not just the PC, but a coding assistant. What do you think, going back to those kind of early—what do you think that will transform the world?
Well, the ability to navigate data, you know, which a long time ago you had to have some IT guy write a thing, and you know, okay, what's the header and footer? There’s a thing called RPG, report generation language, and you know, in COBOL, you have this section in the picture. Anyway, a lot of that stuff is so obsolete it just makes you laugh. The idea you can sit down and engage in a dialogue about data in a very rich way means that our ability to run businesses better, understand PX, adapt to changing things will be so incredible. It won't require custom software, and in fact, the whole complexion of the software market—how many applications will there be?
At first, what we have is everybody adding AI to every application and saying, "Okay, pay me extra because now I got a little AI." But in fact, the number of applications you need—think of a college, which has a scheduling app, a finance app, a support-the-student app. That should all be one thing. Every encounter with the student to the college is all maintained in a rich way. The software applications will be very different, and I'm trying to figure out, okay, how quickly does that happen? It's incredibly beneficial that this software is more adapting to you, including creating user interfaces dynamically than you going, "Okay, I use this software package for this, and then I go to this website and do this."
You're a low-level clerk. Even looking at your email, the email is so stupid that you have to figure it out. It only can time-sort the order; it doesn't know what's important. Then you have your message, and you have to go back and forth, and you're the one who puts it in little folders and things. I mean, I thought the semantic level of interaction with the computer would be higher by now, even without AI. But now, with AI, the idea that the very high-level task that I'm doing—I'm working on my budget, I'm considering buying a new home—that it will be working with you, not at the spreadsheet cell level, but at the, "Oh, let's break this task down in a high-level form. Here's what I can automatically go and do for you." It's super revolutionary. We'll all have an agent that is a utilitarian help you get things done, reads everything you read, but the things you meant to read, it reads. Then, you know, your agent can figure out, okay, which parts of that are important enough to take your time to understand.
We are now switching to rapid fire. I'll let Reed ask the first question.
Is there a movie, song, or book that fills you with optimism for the future?
Well, "The Better Angels of Our Nature," Stephen Pinker sort of documents how violent death, lifespan, and education have gone. There are some lessons of why we've done well. It doesn't guarantee, and he doesn't say that's the future, but you know, if you have one book which should get you back into the mindset of, okay, how far have we come? What should we feel good about? I'd recommend that one.
Fabulous! And what is a question that you wish people asked you more often?
How does malnutrition work? You know, a lot of things I think about are boring, and we should solve with most people having to ever figure out toilets and nuclear reactors and understanding disease. I'm surprised people aren't more curious. You know, when I first said, "What do kids die of?" I had a hard time finding out, and I would have thought, "Well, shouldn't we all be asking that kind of thing?" It's more important than GDP.
Yeah, exactly. I love it. So where do you see progress or momentum outside of your industry—and of course, that's very broad—that inspires you?
Well, India is an example of a country where, oh, there are plenty of things that are difficult there. The health, nutrition, education is improving, and they're stable enough and generating their own government revenue enough that it's very likely that 20 years from now, people will be dramatically better off. It's kind of a laboratory to try things that, when you prove them out in India, you can take to other places. Our biggest non-U.S. office for the foundation is in India, and the most number of pilot rollout things we're doing anywhere in the world are with partners in India. You know, if you go there and you've never been, you might think, "Whoa, this is a chaotic place." You're not used to so many levels of income all being on the street at the same time, but you will get a sense of the vibrancy.
All right, last question. Can you leave us with a final thought about what is possible to achieve if everything breaks our way in the next 15 years, and what's the first step to get there?
The potential positive path is so good that it will force us to rethink how should we use our time. You can almost call it a new religion or a new philosophy of, okay, how do we stay connected with each other, not addicted to these things that will make video games look like nothing in terms of the attractiveness of spending time on them? It's fascinating that we will—the issues of disease and enough food, of climate—if things go well, those will largely become solved problems. The next generation does get to say, "Okay, given that some things that were massively in shortage are now not, how do we take advantage of that?" Do we ban AI being used in certain endeavors so that humans get some, you know, like you don't want robots playing baseball probably because they'll be too good? So we'll keep them off the field. Okay, how broadly would you go with that?
We are so used to this short world that, you know, I hope I get to see how we start to rethink these deep meaning questions.
Bill, a tour de force. Thank you!
Yeah, great talking to you!
"Possible" is produced by Wonder Media Network. It's hosted by A.R. Finger and me, Reed Hoffman. Our showrunner is Sha Young. "Possible" is produced by Katie Sanders, Edie Allard, Sarah Schle, Adrien Bane, and Paloma Moreno Jimenez. Jenny Kaplan is our executive producer and editor. Special thanks to Sya Yalamanchili, SAA Sapa, Ian Alice, Greg Biato, Parth Ptil, and Ben RIS. A big thanks to Aubrey Bonovich, Ian Saunders, Christy Anthony, Alex Reed, Jen Cresc, David Sanger, Larry Cohen, Alicia Salmon, Sha Simons, Denali Wean, Andrea Dramer, John Ryder, the whole team at Gates Ventures, and Little Monster Media Company.