So, Bill, I think we should start with some news, which is a good way to start because you're announcing today a partnership, an investment with OpenAI around Rwanda in global health. Did you do this for us today for the panel?
Sure. Or for the world. Yeah. So when people think about AI and healthcare, they're deeply aware that AI is going to help on the discovery side because AI can model complex disease processes and protein shapes and so it's going to accelerate new tools being available. But I would say it's even more important that AI will be used on the delivery side. And that's true in rich countries but it's even more true in developing countries where you're never going to have enough doctors or clinicians to deal with the demand manually. And so the AI is entering into the health system but not just into the health system, it's all the way down to the level of the patient. So the patient is able to talk in their local language and describe what's going on. And so in order to make this a reality, to see what works, what doesn't work, we're thrilled that OpenAI and the Gates Foundation are committing an initiative called Horizon 1000 where we'll go into a thousand primary healthcare clinics in Africa and the goal is to make the work there much higher quality and if possible twice as efficient as it is today, taking away the paperwork that needs to be done, organizing the resources, you know, so the patient knows what's available and when to come for their appointment. So there's all sorts of things that can be improved. Rwanda has been a great partner for the Gates Foundation in a lot of things and so we're thrilled that that's actually the first country that this work will go into. We'll certainly also over time work in Kenya, South Africa, Nigeria. And you have different systems that we have to connect into. Although not this initiative, there's obviously a lot going on in India as well. And so over the next couple of years I would expect developing world health may even get ahead of rich world because the need is so great and the governments are embracing this and making sure that it's moving at full speed.
And it's getting attention. I mean the fact that OpenAI is involved. I haven't really seen them do partnerships like this. Do you expect some of the other big AI companies to follow suit?
Yeah. So, you know, the Gates Foundation, we have a great relationship with Microsoft, with Google, with Anthropic. No, but I talk to Sam, Dario, Demis, you know, a lot and Mustafa at Microsoft and, you know, so it's wonderful for me that the help and advice I'm giving on these AI issues makes me well informed about how we're using it for our three big scenarios. And our big scenarios are virtual tutor for education, virtual doctor for healthcare connecting into the health system and finally a virtual agricultural advisor. To some people this is the most obscure but of equal importance so that the poorest farmer in the world gets the same quality of advice as a rich farmer. And so we'll mix and match. OpenAI is a little bit unique in that a quarter of their company is owned by a foundation. And so if you actually say, okay, what's that worth? It almost matches the 200 billion in assets that the Gates Foundation is going to be expending over the next 20 years as we finish our work during that time period.
Yeah, I always thought it was odd that they were structured as a nonprofit and now I guess there's purpose behind that. So, it's...
Yeah, they have a complex structure.
They haven't really operated like a nonprofit.
Well, they're increasing. You'll see more and more from them on the philanthropic side. This is one of the first initiatives. Sam and I have been talking about this. We have regular meetings. They're staffing up the foundation side some and so I'm really pleased that developing world health in a very concrete way is actually the first initiative that is purely philanthropic. And the 50 million commitment here is just the beginning. I believe that people in Africa should have this health advisor without having to pay anything for it, should just be a basic capability available to them. And then as you go into the health system instead of filling out paperwork and redescribing everything, the AI that you've been talking to is summarizing that and so it's kind of a digital EHR getting rid of the paperwork together with all the intelligence.
Minister, a big deal for Rwanda. I mean, obviously, congratulations on being the first with this. You have also helped make this country a leader in both healthcare and AI. How did that happen?
I think for us it's been very clear that technology was going to play a very central role in the different facets of our development as a country. And so with limited natural resources I think technology becomes a natural go-to and not an afterthought. And so we've embedded that over the last 20 plus years in how we're building some of the foundational digital infrastructure that is enabling and powering some of these advancements whether it's in healthcare, agriculture, education and many other sectors. And particularly when it comes to AI, when we started thinking about how as a country, as Rwanda, we can leverage AI for development, we had to think about where do we have our biggest pain points as a country and where do we think AI is going to play a role in responding to some of these challenges. So access to quality healthcare ranked high, access to quality education as well and making sure because we have the biggest number of our population is within the agriculture sector, making sure that they can have better access to tools that allow for better productivity at the same time addressing food security concerns that we are trying to address as a country. But particularly for healthcare, we're very grateful that the Gates Foundation has really chosen Rwanda to be one of the countries where they start this engagement with OpenAI. And what we're looking to do is creating decision support tools for our 60,000 plus community health workers that provide primary healthcare to our communities across the country. And to Bill's point, there's also a conversation that is going on with Anthropic as well to see how we can have an instant health intelligence platform that then feeds into the entire national health planning systems and allows us to allocate better resources. And so we're very confident that AI is going to deliver these gains. When we started off for Rwanda, we're looking at the fact that when you look at the patient doctor ratio, the numbers were really alarming. I was talking to Peter earlier. We came up with a strategy to have a, we call it 4x4 strategy. And so the idea was to quadruple the healthcare workforce within four years. We're two years down the road. We've already from an enrollment perspective, we are already at 3.8 of the workforce. And so which means within the next remaining two years, we should be able to quadruple our workforce. But then again, they're going to need these tools to support with better care delivery to ensure that some of the administrative tasks that they've been working on, we can use AI to do that. So they are more focused on delivering better and targeted care to our people. And so we're looking forward to creating more partnerships, piloting, but also most importantly, scaling these solutions across the country so that we deliver the gains that we're looking for.
That was a good tease for Anthropic. Now they're going to all compete with each other to try to do this which is great for you. How do you do AI? I mean you guys have data centers, do you use the American models only? Do you use the Chinese models? How does that work?
So for us we've looked at a couple of things. We are a country where when we started this digital transformation journey as a country it's more than 15 years ago. So I did speak early and mentioned we had a lot of data at rest that we're not using. And so the starting point when you talk about AI, what fuels AI is data. And so we had to go back and look at what kind of data sets do we have, building a national data hub, building national data intelligence platforms that help us with that was very critical because then once we build these models they need to be trained on our own data, they need to be context specific and they need to be coming in to address a real problem. And then the second one is really need to look at all different models and see which one is a good fit for what we're trying to address and use it to train. Obviously there's efforts down the road to be able to create our very own models but we don't have to boil the ocean at this point in time. If there are models that are already available open source we can use those as long as we make sure that we have the right data sets and we are removing and paying attention to the biases that could come through if we don't feed the right data into these models.
You have many fans, but Peter Sands is among them who sings your praises. How unique is Rwanda, Peter, in your efforts, you cover all of Africa and the developing world. How unique is Rwanda both in the pioneering of the healthcare and the technology space here?
Well, I'm probably going to embarrass Paula because I do think Rwanda is really kind of a beacon of what can be done. You're talking about a country which has already got to what, 97% connectivity. I mean there are quite a few fairly rich countries that haven't quite got to 97. And also when you think about this point about the 4x4, this is a government that set out an ambition to quadruple the health workforce two years ago and is now almost there two years later. When you think about kind of executional delivery that's quite unusual. I do think that AI is really powerful. I mean, I feel like a kid in a toy shop when I start thinking about all the things we can do with it, but I also know I only have pocket money. And so we have to make choices. And I think one of the things that Rwanda has been able to do is actually be very deliberate in its prioritization. To Bill's point about we could well see lower and middle income countries adopting these kinds of tools faster. In fact this morning I was told it's already happened in the sense that if you look at it in terms of reach and impact what's happened with TB screening where we've invested about 170 million over the last four years doing AI based TB screening. I've been told that that is already the largest single application of AI in health which may or may not be true but it's certainly a very significant one that is delivering very significant impact. I think the only note of caution I would add is we face in many parts of Africa and in lower middle income countries more generally some very basic problems. There are a whole lot of primary healthcare facilities in the continent that don't have internet connectivity and if you don't have internet connectivity it's hard to leverage AI. And indeed some of them don't have power and it's hard to have internet connectivity if you don't have power. So we got some fundamental things to fix. I also think perhaps the hardest challenge, well two challenges. One is we do have to have this, the whole thing has to be framed around problems needing solutions as opposed to a whole bunch of tools needing a problem to fix and there is a little bit of people running around with a whole lot of hammers looking for nails. And we need to locate it very very firmly in what are the big health issues we need to fix and therefore how can we use AI. The second thing and we were talking about it earlier is in a way and this is no disparagement of all the wonderful software engineers out there. The easier bit is the tools. The harder bit is the people who can actually use them and make different things happen. And in many places I think the constraint on the pace of progress we can make in using these usefully is have we got the people who can actually do that particularly because as soon as people are going to be trained in somewhere like Rwanda there's going to be a whole host of interested AI companies and others who will want to hire those people right and unless they can have career opportunities and reasons to stay in their countries, we're going to see the same kind of brain drain as we've seen on the health workforce side. So, I think fixing the people's side of it is a crucial part of it.
So, you said we have to find the problems and then figure out the solutions, the nails before the hammer. What are say top three problems on your mind that you'd love to figure out solutions for?
Well, one of the reasons I think the TB example has worked so well is we've known for a long time that if we can find the people with TB, we can treat them. The trick is how do you find the people with TB? So, we had this glaring problem and that's been I think led to very focused use of AI to solve that particular issue. There are a host of other problems and this is where we need a kind of dialogue between health people and the software engineers because in a sense the health people don't quite know what the tools can do and the software people don't necessarily know what the problem is. But I'd give you another example. We have increasing instance of resistance, malaria being resistant to current first-line treatments, artemisinin-based treatments. And in somewhere like Rwanda where we have pretty very sophisticated disease and genomic sequencing of malaria cases and so on we know what's going on. The trouble is if you go to Burundi or DRC next door we don't know what's going on. And understanding the patterns across the continent of where resistance is and what type of resistance is occurring seems to me like a problem where AI could be incredibly useful. And it's really important because to put this in perspective, in the early 60s our main treatment was chloroquine. We had chloroquine resistance, we had a massive surge of deaths and the whole kind of malaria project was put back. We know something similar is happening now with artemisinin resistance. We just don't know enough about it and that to my mind is a good example of a problem I'd love to use AI to help us address.
Can you fix that with AI, Bill?
Well, you can understand looking at all that genetic data where you look at the mosquito genetics and the parasite genetics. Yes, you can. Fortunately, in this case, we do have a whole new class of drugs that we've done in a partnership with Novartis, a thing called Kafartem. And it's a bit more expensive. And so the question, you don't want to just roll it out. And in fact, as soon as you roll it out, the clock starts ticking on that you'll get resistance to that. And so, as Peter says, figuring out where in the continent this is emerging and being willing to spend a little bit extra there while we get the price of this down. Kafartem as when it first comes out it's more expensive. In about three or four years we hope to have a form of it that's less expensive and so we'll apply it there. You know, there's vertical AI like TB diagnosis or the thing that we use an ultrasound to scan a pregnant woman understand is this going to be a complex delivery so should the woman make sure to get to a higher level health facility. And then there's this horizontal use where the AI is literally available to the patient even when they're not at healthcare and then when they walk into healthcare it's listening to that. Right now there's a lot of paperwork. You have to write down this patient. You have to write down that they received this vaccine. That paperwork is very imperfect in terms of how you identify the patient and how is it readable and filled out in this digital world. And yes, you got to have the electricity and the trained workers, but the idea of creating all of that data in a very reliable form without using the time of the clinician, it's pretty dramatic. The ability to say to a patient, come in at this time, as opposed to just everybody coming in the morning, you know, and waiting. This is a system that digitization is going to lead to lots of efficiency. And you know that's why getting it out and learning, this is the first year that we're really going to see the horizontal application.
I love the topic talking about this and hearing about the real world cases of what you can solve with AI, everything from the mosquitoes to what you were just talking about. Minister, it's interesting that you are the Minister of Information, Communication and Technology and not healthcare but working very closely on this. I mean how do you put these real world scenarios to work in what you do?
So maybe before I respond to that I'll pick on what both Bill and Peter highlighted around malaria. And for I think we've seen a recent surge of malaria cases across Africa and Rwanda is one of the countries that has had to deal with that. And to Peter's point there's this whole resistance to traditional malaria treatment method that's been there. But what we've done and that's why I like the point Peter made around finding the problems and figuring out what's the best way to solve them and which technologies make the most sense for us. What we did was one to use a combination of both AI and drones. And so you're able to use drones in terms of figuring out what are the mosquito breeding sites across the different communities using a bit of AI for prediction and modeling. And then based off that we're also using drones for spraying as well. And so in many ways that reduced the malaria cases that we would be able to get. The second part was where you have our community health workers being the frontline workers for primary healthcare. 70% of the cases they deal with on a yearly basis are malaria cases. And so if you're empowering them with these decision support tools that help with better diagnosis, better prediction, then I think we're able to see how AI along the entire pyramid of care can be used to sort of not just from a prevention perspective, but also to treat and lower these cases and lower the number of deaths that you could see across the board. But coming back to your question, even and I mentioned it earlier for us technology is central to everything that we do. And so the point that was being made earlier, figuring out the problems for us it was knowing what are the biggest pain points in healthcare. If diagnostics is one of them, where do we use AI and other emerging technologies to solve for that. The other one was around demand forecasting for health commodities because you would have stockouts and then slow procurement processes eventually the prices become high and very unaffordable for our patients. And so the idea was how do we use AI enabled market intelligence to better do demand forecasting for these health commodities and make sure that we're reducing on the stockouts that we're having. Bill talked about maternal healthcare where we're using AI enabled ultrasounds to support that. And so what you're seeing is that there's less referrals that are happening across the board. So we've done a similar exercise even in agriculture and education to say what are the pain points we're all trying to deal with from a policy standpoint as a national government and do we see an opportunity for AI and other emerging technologies to help solve for that. And then we're able to work with various partners to co-design what kind of solutions can help to do that. And so to your point today when I see the Ministry of Health as an example because the conversation is on health, they're very much our digital ambassadors because everything they're doing, they're data, evidence-based. They're making sure that they're leveraging technology for all the interventions and that's what we want to do to make sure that it's mainstreamed across the different sectors.
I mean, it really requires a whole of government sort of approach. Bill, do we do this in the United States? I mean, do we have this kind of thinking about technology first and pioneering AI within our healthcare system?
Well, the US, you know, spends over $10,000 per citizen per year and, you know, so people's expectation of the healthcare system are very high. We have a lot of doctors. In most of the African countries that we work in, you hope that they have $100 per person per year. And so, you know, it's a factor of a hundred different. And you have a lot less doctors. Most Africans will never meet a real what the US would call a doctor ever. And so it's really about primary healthcare. And the big advances we've made that have allowed us to cut childhood death by 50%, maternal death by 40%, HIV deaths by over 50%. It really has to do with integrating into this primary healthcare system. It's not about doctors giving malaria medicines, giving HIV medicines, even TB diagnosis, we have to be able to do that out in the rural areas. So, you know, I'm very excited about what the US can do with AI. I told the person who runs the NHS, Wes Streeting, if they use AI properly, that alone might get them reelected because the current citizen expectation that the NHS waiting lists will be cleared and that they'll have cheerful doctors who don't feel overloaded. You know, there's zero expectation for that. And yet I believe if during their mandate there is enough time to really get that going.
You should tell President Trump that.
I don't know, that if they do this right he'll get reelected.
Maybe if he thinks so. Okay. He's not supposed to run again but... I mean the Republicans and he hasn't ruled it out. Anyway, that's a complicated topic. Dedicate a whole AI cluster to working on optimization there. So the things look a little bit different but the potential is absolutely the same. The rich world is going to be more regulated because the bar that you're comparing to but even so, you know, doctors are seeing today that when patients walk in they've already talked to the AI. And so you're going to end up with the AI that the patient talks to and the AI for the healthcare system. In countries like Rwanda we'd like to start off where it's not two systems, that if the patient's talking to the AI then as they come into the clinic that summary goes to the doctor. The transcript comes back so it's always available to them to ask more questions even when they're not in the health system. So, you know, there's an opportunity to show a deep level of integration right from the very beginning.
