00:00:15
Speaker 1: Pushkin.
00:00:31
Speaker 2: I'm Jacob Goldstein, and this is What's Your Problem? My guest today is Catherine Nakalembe. Catherine is an assistant professor at the University of Maryland. And she runs the Africa program at NASA Harvest, which is a project that uses satellite images to track agriculture around the world. And it wouldn't be wrong to say that Catherine's problem is using satellite imagery and GoPro cameras and AI to predict seasonal food shortages in sub-Saharan Africa. But in fact, her work, the problem she's trying to solve, is problematic. ultimately more human than that, ultimately more interesting, I would say, because the problem she's really trying to solve is, how do you use evidence to convince leaders to act to prevent people from starving? Catherine grew up in Uganda. She went to the University of Maryland to start a PhD, and then she went back to Uganda to do her graduate research. And the project was combining satellite data and on-the-ground research to to predict when the crops were likely to fail, leading to widespread hunger. And she was working on this for a few years. And then in the summer of 2015, when she was still a graduate student, there was a particularly bad drought in the region she studied. And she knew, she knew from her work that the people there would not have enough to eat in a few months. And she decided she had to get that information to someone with the power to do something about it.
00:01:58
Speaker 1: I realized really quickly that I knew more about the severity, basically the distribution of drought across this region more than anybody because I was able to look at it from space. And one year I finished my work and I was like, I can't just go back with this information because it had been an extreme drought year. I went to the office of the prime minister to share my My data and evidence in that year had taken lots of photographs and videos, et cetera. So I brought that. And then they had a meeting with the prime minister that led to— Wait.
00:02:28
Speaker 2: Can I just pause? So you're, what, a grad student or a young professor.
00:02:33
Speaker 1: At this point? No, I'm a grad student.
00:02:35
Speaker 2: And so it's an unusual thing for a grad student to be like, oh, my God, I have to tell the prime minister about this. What happened? How did that happen?
00:02:45
Speaker 1: So I've been coming for the last three years. People are trying to do the same things and crops fail. And I'd already started to kind of figure out how to, you know, predict what's going to happen, you know, early in the season. And there's always news around this issue, you know, Karamoja drought again, etc.
00:03:03
Speaker 2: So Karamoja is the region in, I guess, northern Uganda where you were studying.
00:03:09
Speaker 1: Yeah. And so it's always contentious. Is it happening? Are they, you know, just pretending? And this year, I had so much evidence, you know. First, because I'd taken so many photographs, I'd recorded video. And I wanted this to be sort of complementary to my PhD research. So I went and I shared it to the commissioner. And then he asked me, can you stay an extra few days? Because we have this food security meeting on Thursday with the prime minister and his steering committee, which are the other ministers. And I was like, of course. So I show up and I share the evidence and they asked me to say a few words. And I was like, you see this photograph? You see all of this thing that's dead here? This map basically represents the same thing everywhere. And it's eminent and people will be hungry pretty soon. And three days later or two days later, they sent food trucks. I have this photo that was sent to me from the guy who was the secretary to the prime minister or personal assistant. He was like, the trucks left this morning. I have this email, and I was amazed that I did that, and it went that far, basically.
00:04:23
Speaker 2: And that fast. So when you say people will be hungry pretty soon, that is an understated way, it seems, of saying what's going on. What was going on?
00:04:34
Speaker 1: You have subsistence farming where you produce what you eat. Very little is left over. And at the end of the season, you're supposed to sort of start picking up or eating what will be available to you through to the end of the next season. So this is typically around September. And so by August, everything had failed. And so there wasn't going to be anything harvested. But if you've kind of been out in the field a collecting data, and you run into kids eating sorghum shoots, for example, you know, the middle part of a.
00:05:10
Speaker 2: It's kind of like eating grass.
00:05:15
Speaker 1: Exactly. So like the middle part of a corn stem, it's sugary. And so, you know, you find kids eating that and they don't have anything else. They don't have livestock. Their family does not have any other source of income. It's pretty dire. So I think the other thing that might have made a huge difference in this particular instance is the photo evidence that I had. I take thousands of photos. I think I have over 300,000 photos.
00:05:41
Speaker 2: That's such a low-tech thing. Do you know what I mean? It's not like you were doing some wild satellite AI something. You were just walking around and taking pictures.
00:05:54
Speaker 1: Mm-hmm. And I had my map, my drought analysis map, that was very simple. And then we called it NDVI, Normalized Difference Vegetation Index, the anomaly of it.
00:06:06
Speaker 2: Basically, how much drier than usual. Exactly.
00:06:09
Speaker 1: Yeah.
00:06:10
Speaker 2: So, okay, so you do this, and the result in that first year is that they get food aid to people sooner.
00:06:16
Speaker 1: Right. Almost immediately. I have this text message from the commissioner. He said, this is the first time that nobody argued with the evidence.
00:06:28
Speaker 2: Did they then integrate your work in an ongoing way? How did that play out in the years after that?
00:06:36
Speaker 1: So the project that's called Disaster Risk Financing was basically being designed the following week. So I stayed that extra week and was like, I can not only develop the methods, but I can train the office how to utilize the data that's available in order for them to do that early warning, as well as write summaries, reports, et cetera.
00:06:56
Speaker 2: Take pictures. It sounds like one big lesson is take pictures.
00:06:59
Speaker 1: It's a validation for sure, but they'd already understood. So I kind of created this method to understand, compared to the history, if we hit this threshold, things are going to fail. Considering what has happened in the last 30 years or last 20 years, If we hit this threshold, vegetation won't recover and we need to start planning.
00:07:19
Speaker 2: Basically a threshold of drought. If rain is below some level.
00:07:23
Speaker 1: So initially it was like if vegetation conditions drop below this level within this period, a few months, months within the growing season, then we need to do something.
00:07:35
Speaker 2: So it's interesting. I mean, I might naively have thought that the constraint, the fundamental constraint was just money.
00:07:43
Speaker 1: Right.
00:07:43
Speaker 2: That there's just not the money to get the food to the people or infrastructure or something. But it's information at some level, like reliable evidence.
00:07:52
Speaker 1: Reliable evidence that is accessible. But I think money and the willingness to use that money is also really important. So that region where I was studying, even though it had the lowest economic indicators, you know, household income was very low. Infrastructure is very poor. The cost of food was so much higher than it is in the capital. Yes.
00:08:20
Speaker 2: Well, that goes with poor infrastructure, right? Exactly. When it costs a lot to move food, the food costs a lot.
00:08:27
Speaker 1: Exactly. Now, add the fact that there is a drought and traders can do whatever they want. And so it becomes much more expensive. And when the government is trying to acquire this water and food, et cetera, they end up paying a lot more. Uh-huh. And so if you're doing it in anticipation, it means that then you can acquire things at a much lower cost.
00:08:49
Speaker 2: Oh, so if they know in August that there is going to be great need in September, they can buy before there's a run.
00:08:56
Speaker 1: Exactly. And the other thing was, because it was in anticipation, what ended up happening is they would have projects that could hire people to work on that were community development projects. It could be tree planting. It could be road construction, it could be improving wells, and then they will be paid for those jobs. Yeah, that then they could use that money to buy food and other things.
00:09:22
Speaker 2: So it's interesting. I thought we were going to be talking about satellites and AI, and I guess we will be. But it's interesting that this conversation starts in such a low-tech way. I mean, obviously, digital cameras were not quite as ubiquitous as they are today. But this is 2016. It's not 2000 or something. Lots of people already had digital cameras.
00:09:44
Speaker 1: Why? Yeah, why? I'd say people believe what they see. And the NASA administrator in 2019, Jim Bernstein, he used to share my photos in his presentations. He'd talk about from NASA to the moon and beyond. But he would still use the photos and not the maps or the images. the numbers because it tells it you know we can see it from above and this is what it means on the ground this is the reality of what it is on the ground and that's i think much more accessible and understandable.
00:10:21
Speaker 2: Um so I feel a little weird now after you're talking compellingly about how important photos are. But I do want to talk to you about satellite imagery and AI.
00:10:31
Speaker 1: But they're photos.
00:10:32
Speaker 2: They're photos. Very good. They're photos. I find them moving in a different kind of way. I mean, you know, I've read this sort of big hand-wavy thing, which is, You know, there's lots of satellite imagery of the Earth now and lots of AI models designed to analyze crops, but those are largely designed for kind of big monocultural agribusiness, and they don't apply so well, certainly, to the kind of smallholder farms in the places you have worked. Like, I get that that's kind of the big picture. But more narrowly, like, tell me about your work to deal with that. Like, first of all, what is it that you want to know from satellite imagery?
00:11:13
Speaker 1: We want to understand what's growing where and how it's doing. The other is the complexity of the agriculture system. So when we're looking for crop fields, we're looking for shapes and then we're looking for some sort of seasonal pattern. So if you had 10 images over a growing season, you should see something like it's cleared now. it browns, it's like brown and wet, and then something green pops up, it matures, and then it's harvested. So it sort of like becomes, you know, dry. And so we should be able to see that signal. However, how low the brownness might be, like how dry something might be, is contextual. And while a reduction might seem like minor in some places, it's actually a very serious problem. And so the first thing was create a cropland map, basically exclude everything that's not cropland, estimate, understand vegetation conditions historically, like what happens every season, what months matter the most in terms of vegetation, in terms of crop production, and then try to understand at what point recovery doesn't happen. So to be able to do that at a really large scale, satellites are a phenomenal tool for this. At the time, even today, the only satellite that allows us to be able to do this at a much higher frequency is MODIS. And MODIS is a 250-meter resolution. That could be multiple crop fields in a single pixel.
00:12:52
Speaker 2: So the farms are too small or too varied.
00:12:56
Speaker 1: Too varied, yep.
00:12:58
Speaker 2: Tell me, what is a typical farm in this region like?
00:13:02
Speaker 1: So in Karamoja, people grow sorghum, millet, a little bit of maize. There's a lot of sunflower, too.
00:13:10
Speaker 2: And will they grow all that in one place, more or less?
00:13:13
Speaker 1: So you will find sunflower mixed in with sorghum, for example.
00:13:17
Speaker 2: And how big is a typical farm?
00:13:20
Speaker 1: So farm plot... On average, less than 2.5 hectares. So a football field, a soccer field is about an acre. That's the average over most of sub-Saharan Africa, but they're much smaller. So like in a single soccer field, you might find maybe four plots as an example.
00:13:43
Speaker 2: Like four different people's little farms?
00:13:45
Speaker 1: Four different people's plots with hedges, and they're not like distinct perfect shapes like you would see.
00:13:51
Speaker 2: Quite small. Hard to see from a satellite, presumably, especially if there's a bunch of different crops there.
00:13:58
Speaker 1: So hard to see their boundaries. So a single modus pixel might include four of those fields that might have different things, for example.
00:14:06
Speaker 2: Yeah. Let's talk about the GoPros. So fun. Tell me about your work with GoPros.
00:14:13
Speaker 1: So one of the issues is in order to get from cropland, where all the crops are growing, to go to crop type. So the USDA, for example, produces a cropland data layer, which if you take a look at it, you can see where corn, wheat, soy is growing. And now that's been changing over the last couple of years. In the US. In the US, yes. This also exists for Europe. They have something similar. There's, of course, some simplicity. There's a single crop. The fields are homogeneous. They're kind of easy to map. In East Africa, in addition to fields being small, I already mentioned the intercropping, mixed cropping, and all the heterogeneity, how people plant at different days. To be able to get a crop type map, you need a lot of training data. So, you know, one of the biggest advances, I think, you know, the reason why we have huge leaps in terms of AI is the availability of training data. You know, same as face detection, all people's faces on Facebook, et cetera, makes that technology a lot easier to develop. And so this is true also for crop type mapping. So it's like face detection, but for crops.
00:15:20
Speaker 2: Why do you want a crop type map?
00:15:22
Speaker 1: There's so many benefits to knowing exactly what crop is growing where. You want this map in order to understand, for example, what kind of fertilizer will be relevant in a particular region. So if you know there's places dominated by rice, just like in Ghana, a particular place in Ghana, where could irrigation infrastructure make a difference? The other and most important, most basic thing is, if you want to understand how much will be produced at the end of the season, you need to know the total area by specific crop. So if we have 25,000 hectares of corn planted... And the average condition or the average yield is 2.5 tons per hectare. You multiply that and you have your production, which will allow you to understand how much corn is available. If it's a critical crop in a particular area, you would know that, you know, food security would be okay. So we had a big issue in Illinois with corn, et cetera. If Germany had a big issue with wheat production, bread being so important.
00:16:27
Speaker 2: Yeah.
00:16:27
Speaker 1: They would have to plan accordingly. So you need to figure out, what do I need to import to have some sort of stability?
00:16:34
Speaker 2: And so for Sub-Saharan Africa in general, or the regions you study, are there crop-type maps now?
00:16:42
Speaker 1: There are some crop-type maps, yes. Some are done on a very small scale, regional scale. But what this project was trying to do was try to, one, just figure out the fastest possible way we can collect data over really large areas. One, because in some countries there are three seasons. So Rwanda has three growing seasons. So you're looking at three crops. So to cover the whole country and get an understanding of did they grow corn more generally, bananas more generally? No. You need to be able to do the survey continuously.
00:17:18
Speaker 2: And the satellites don't have the resolution, essentially, to do it.
00:17:22
Speaker 1: To be able to do it, yeah.
00:17:23
Speaker 2: You can't tell from satellite data.
00:17:24
Speaker 1: So we need to train a model to tell the model, this pixel, this particular spot is banana. Find me banana everywhere else in the image, yeah.
00:17:34
Speaker 2: So you have this idea of like, okay, we need more on-the-ground data. So what do you do? What is the GoPro project?
00:17:44
Speaker 1: GoPros are perfect because they have a GPS. So when we take a photo, the photo has a location. So typically when I'm collecting field data, I need a photo, evidence, but also I need to be clear when somebody labels in the form that maize is in this growing stage and you have to stand in the middle of the field. Because I want to use that location to train my model to predict where else maze might be.
00:18:10
Speaker 2: This is the old school, slow, never going to scale technique that you're trying to kind of quasi-automate here.
00:18:17
Speaker 1: Yes. At this point, even if all I had were images that had a longitude and latitude, and I know where the camera is facing, I could go through all these individual photos in Google Earth Pro and add the label myself.
00:18:36
Speaker 2: So at least you would have a bunch of images with a very specific geolocation on latitude and longitude.
00:18:43
Speaker 1: And I know what the crops look like. I know what the crops are.
00:18:47
Speaker 2: So this is phase one. But you don't want to have to label a thousand things corn, whatever.
00:18:54
Speaker 1: Yeah, I would have been okay with that too, though. Considering, so I did this in Mali. I did it in Senegal. In Tanzania.
00:19:02
Speaker 2: Wait, this meaning driving around with the GoPro? No, the phone. Oh, the walking around.
00:19:08
Speaker 1: Walking around, yeah.
00:19:09
Speaker 2: Yeah, okay. I mean, it sounds kind of charming in a way.
00:19:12
Speaker 1: I learn a lot from doing field work. You learn what it means for someone's field to be washed out when you go. But if the idea is to try and create a map that will be useful for this task, there's only so much you can do when you cover a very small area. Yeah. So 2020, I call this project a COVID-safe project. 2020 was like, oh, well, now we have this funding. We have to figure this out. I go buy a GoPro at a store, a fashion sort of amount with my dad because I couldn't order it from Amazon. And then we make this magnetic thing on the car and just drive. I go to, there's an agricultural research station. And the director was like, I described to him what I'm trying to do. He was like, oh, course you can drive around here, whatever. So we try it out. So the first thing was, if we get the images, are they useful? Is the GPS good enough?
00:20:07
Speaker 2: And it was.
00:20:09
Speaker 1: And so then we were like, oh, this could actually work.
00:20:17
Speaker 2: We'll be back in just a minute. That's the end of the ads. We're going back to the show. And as you probably recall, when we left off before the break, Catherine had just tested out a GoPro herself and decided that GoPros could be a good way to gather crop data from small farms.
00:20:42
Speaker 1: So I have a friend in Kenya. I asked him if he can buy, if he can order things on Amazon and to work with a team there to see if they can collect data in Kenya. So we get a team of two. So basically four. So it's a team of two people. One team drives south, the other drives north. And we cover all of Western Kenya in a week.
00:21:09
Speaker 2: Wow. And they're just driving everywhere they can drive, in a car, on a motorcycle?
00:21:13
Speaker 1: So you can see from some of the original data that we have from Kenya that we're kind of covering these big... There are lots of fields, but they're proximal to these big roads, which... It's great, but not good enough. And so it became pretty quickly obvious that a motorcycle might be better because a motorcycle can go to a level two, level three, you know, smaller paths and go through. And in Tanzania, they went a little bit crazy because I feel like they're driving into rice paddies and going like this.
00:21:42
Speaker 2: Well, people live far off of roads that you could drive a car on. Exactly. The kinds of farms you're talking about exist far from drivable roads.
00:21:50
Speaker 1: Exactly. So... But just from the Kenya team, we had over, I think the first iteration might have been maybe 300,000 images or more. Then the team of two in Uganda, I kind of equipped them. And during COVID, there's nothing else you could do and asked them to kind of just go drive around. And they went crazy because then we had over 2 million images from them. Okay. And so then it's just too many images. Even I got tired of just like trying to figure it out. But I had an undergraduate computer science student who worked over this and developed Street to Sat, the first iteration of it. And remove clouds, adjust images automatically so that they're kind of like straight. And then remove faces and ensure that they're actually crops in the image.
00:22:48
Speaker 2: Crop and not crop, you called it in the paper. It reminded me, I don't know if you watched the show Silicon Valley, but there's a guy in that show who invents AI.
00:22:59
Speaker 2: And they show it a hot dog and it says hot dog. And everybody's like, wow, it can detect food. And then they show it a hamburger and it just says not hot dog.
00:23:07
Speaker 1: Yeah.
00:23:07
Speaker 2: And all it does is say hot dog, not a hot dog. This is where you are now.
00:23:10
Speaker 1: This is where we are. It's like, is there a crop in this field? Yes. If there is, then that's phenomenal. And then we kind of did a couple of iterations of what we call labeling. Yeah. Basically, go through a couple of images to create a label dataset, which is, this is, we draw bounding boxes, maize, maize, this is what maize looks like, this is what banana looks like.
00:23:34
Speaker 2: So now you're saying you're, this is training, you're creating training.
00:23:37
Speaker 1: We're creating a training dataset, yeah.
00:23:38
Speaker 2: Training data, yeah.
00:23:39
Speaker 1: And then with the first iteration, then we run it through the initial data from Kenya to do sugarcane maize. Okay. Okay. And it was pretty good. It was pretty good. It was like, you know, when it's actually working and you can't believe that it's working, but it is working and it's unbelievable. So you're like, you're in this phase where you're like, oh my God, is this actually happening?
00:24:05
Speaker 2: What are people doing now with this basic technique of, you know, driving around with GoPro cameras where there are these small farms? Like what is one specific thing someone is doing that is helping them take action in the world in a way they would not have been able to do?
00:24:21
Speaker 1: So I give one example. So in Kenya, when we did the first iteration of it as a test pilot, and then were able to actually work with the Ministry of Agriculture for them to go out and collect the data themselves, they were like, do we get to keep the cameras? I was like, yeah. And I asked, you know, what sorts of things are you thinking about? I was like, well, you know, that would be kind of just Phenomenal photographic evidence because a person can go back whenever and send me those photos while they're out there in the middle of nowhere. So remember going back to the low-tech technology is that when the ministry reports, so we have these WhatsApp groups where extension agents share constantly what's happening. If they share that there was a flood and the flood was serious, everybody knows a flood is serious, but how serious is a different story. when it comes to things like beans, water staying on the surface for a long time and the beans turn yellow is because all the nutrients have leached. And so for the extension agents to be able to go through and collect these images without needing to get off, you know, take a photo, et cetera, et cetera, is something that is used as kind of evidence and as part of their reporting in the ministry.
00:25:38
Speaker 2: So it gives you a baseline.
00:25:40
Speaker 1: It's like a baseline, but also people believe what they see.
00:25:44
Speaker 2: It goes back to you walking around taking pictures, but just a lot more pictures.
00:25:48
Speaker 1: Yeah.
00:25:50
Speaker 2: So if you sort of distill it down, what is your big project? What are you trying to do in the world in your work?
00:25:58
Speaker 1: So data is an essential, right? Evidence is essential. But try to use it to improve, you could say, the human condition through better decision making. So If I know for sure that something is coming and it will be this bad, it means that the persons or people responsible for decision-making, they have no excuse to not do the right thing. I think that's kind of what kind of drives me. And I was going to talk about the 2.0 of the disaster risk financing program is because we've gone so far now since then that what I've built out now gives a three-month forecast of what things will be three months from now based on a model that has learned. So because I can process more data, I can combine multiple complex data sets than before, thanks to compute, and increase access to data. But I understand the problem so deeply that that threshold now is specific to the specific research It's no longer one threshold for all of them. It is a district, like a county-specific threshold that could even be lowered to be sub-county that could be zip code-specific threshold that in this zip code, this threshold is it. It doesn't apply to everywhere else.
00:27:28
Speaker 2: It's sort of zip code by zip code in Uganda. You can tell how much drought is going to be really bad. Three months in advance.
00:27:37
Speaker 1: Exactly. And it's not only in Uganda. The method is largely scalable. I could apply the same methodology to the U.S.
00:27:43
Speaker 2: Do you have the data? Yeah. Can you do it anywhere, or do you need some amount of data collection in some places?
00:27:50
Speaker 1: So for the forecasting component, all I need is satellite rainfall, satellite temperature, satellite vegetation conditions. And then the validation component is where I would need people to actually go and confirm, you know, am I actually predicting the right thing? As an example, you could use auxiliary things, you know, you could use images people are posting on social media about events to sort of as an in-between. But in order for the evidence, you know, to be strengthened, then you want to go, you know, take those photos. And then the other thing is, you know, have a routine component around collecting and gathering that evidence so that it's not whenever you feel like or whenever you have money, but it's just that at this point, we need to check in and see how things are going.
00:28:41
Speaker 2: Like, if I were going to say, Catherine's problem is this, like, what would the sentence be?
00:28:48
Speaker 1: There are lots of quite simple things that we could do already, but we just don't do them because they seem so Maybe we overcomplicate them. I always think about these, a minister will not, you know, read your F1 score table, you know, the confidence of your model, et cetera, but the minister will read a memo that's usually like three paragraphs. What I learned in the risk financing work is that it has to be distilled into a half-page memo. The memo is evidence points that we've hit the thresholds, about 70,000 people will need emergency aid, and we're going to release funding on this day. That's it. And then the minister says, I agree. Right, so whether or not it took me 100 months, $ 5 billion... 25 analysts to create the evidence or whatever it is, usually it's something that is accessible, concise, that helps that decision making. And accessibility to that information is so critical in this case. So I'd spend a lot of time, you know, training and building capacity so people can reproduce things.
00:30:13
Speaker 2: On their own.
00:30:15
Speaker 1: Like you can run the helmet's work all on your own. All the methods, everything you need in the kit. And a lot of people have done this as a team that's done it in Nigeria. People have done it in India on their own.
00:30:26
Speaker 2: Basically building crop-type maps using GoPro cameras?
00:30:30
Speaker 1: Creating crop-type labels, yeah, using GoPros. But also the workflow itself. We have a notebook available that anybody can run with a Jupyter notebook without anything else. So the problem is some things are a lot simpler than we make them out to be. And if the thing is to solve hunger... I think those solutions for solving hunger already exist. And they have nothing to do with my model or my technology. It has to be moving food where it is produced to where it's needed, making it available when it's needed the most, for example. And I can provide the evidence to be like, okay, it's absolutely needed here. So I think that's my problem.
00:31:24
Speaker 2: We'll be back in a minute with The Lightning Round. We're going to finish with The Lightning Round, which is just a little bit more playful than the rest of the show. What's one thing your dad taught you about fixing cars?
00:31:49
Speaker 1: Oh, my God, that guy.
00:31:52
Speaker 2: My dad my dad told me um.
00:31:56
Speaker 1: We had my car battery in 2020 of it was overheating whatever and he was like how do you drive a car without opening the hood and i was like like seriously and then i said this to him uh because his phone is runs out of memory and i was like How do you work with a phone that has no memory? So, you know, I was trying to kind of like paint this problem. My dad, what I've learned from him is you can figure it out. And he has this terminology. It's, I think, a Luganda thing. It says chito chige, which is kill it to learn it. So, you know, in terms of like fixing a radio, you take it apart and then you have to put it back together.
00:32:39
Speaker 2: It's like break it, break it and fix it and you'll understand it.
00:32:42
Speaker 1: Break it to fix it. basically, because then you will understand it. And so he always took apart, you know, car engines, and then he will put them all back together. So I think what I learned is that you can learn, you can figure it out, whatever it is.
00:32:58
Speaker 2: What's one thing you learned from your mom about running a restaurant?
00:33:02
Speaker 1: My mom is a workaholic. She's told me recently that She works, she likes, she keeps working because she likes to work. She's no longer, it's no longer profitable, but she likes to work and she believes she stops working. You know, she'll just like fall over and die. It's just an extreme, it's an extreme statement. But what she means is that whatever the work is, it's important and do it to the best of your ability. And so she does, you know, has done so many different things to kind of sustain it and keep it running. And so... Keep on moving, keep on working, and you can figure it out, I guess, also in this case.
00:33:43
Speaker 2: Who's a better badminton player, you or your sister?
00:33:46
Speaker 1: My sister, she's left-handed.
00:33:48
Speaker 2: Oh, is that a big advantage?
00:33:49
Speaker 1: Of course, you have no idea where her left from her right is. Her backhand is your, what you think is her backhand is her forehand. And then she worked so hard to improve her backhand that her backhand feels like a forehand, right? It's like the one when you hit it and it comes back. You're like, oh my God.
00:34:09
Speaker 2: Yeah. That's the sound it makes too, right? In badminton. What's one thing I should do if I visit Uganda?
00:34:21
Speaker 1: Eat a mango.
00:34:22
Speaker 2: Okay. Are they better there?
00:34:24
Speaker 1: Absolutely. And an avocado.
00:34:26
Speaker 2: And an avocado. Also better?
00:34:28
Speaker 1: Yeah. Never refrigerated. Fresh. Maybe harvested yesterday.
00:34:34
Speaker 2: Um... What about GoPros? Are you dreaming of other uses of GoPros for work or not for work or things other people could be doing with GoPros?
00:34:44
Speaker 1: Yeah. It's funny. I always thought about them. I used to wear one. The first, first iteration of GoPros, when I went to field work, I wanted to record everything. But about GoPros, if I was to design one, I'd include one that has a distance measurement system.
00:35:02
Speaker 2: I would do that.
00:35:05
Speaker 1: So I can know how far my maze is from where I'm standing.
00:35:11
Speaker 2: You're trying to do that with software, essentially.
00:35:13
Speaker 1: Yes. But I'd like for it to give me an estimate of how far something is. It's kind of easy to derive that if you're in a city because you have corners and things, very strict structures. But you can't do that when you're doing it for crops. I did win a GoPro for good, but they sent me a GoPro. which I already have a lot of GoPros.
00:35:37
Speaker 2: You won an award, and the award was another GoPro?
00:35:39
Speaker 1: Yes.
00:35:40
Speaker 2: What do you wish they'd sent you?
00:35:41
Speaker 1: I wish they had a conversation and I'd suggest what I want in the GoPro.
00:35:48
Speaker 2: Yeah, right. A GoPro that could measure distance.
00:35:52
Speaker 1: Yeah.
00:35:53
Speaker 2: I appreciate your time. It was lovely to talk with you. Thank you so much.
00:35:57
Speaker 1: Lovely talking with you, too. Thank you for having me.
00:36:07
Speaker 2: Catherine Nakalembe is an assistant professor at the University of Maryland and Africa program director at NASA Harvest. Please email us at problem at pushkin.fm. Tell us what kinds of shows we should do more of or less of or particular people you think would be good on the show. You can also find me on X and on LinkedIn. Our show is produced by Gabriel Hunter Chang and Trina Menino. Our editor is Lydia Jean Cott and our engineer is Sarah Bruguere. I'm Jacob Goldstein, and we'll be back next week with another episode of What's Your Problem?
Speaker 1: Pushkin.
00:00:31
Speaker 2: I'm Jacob Goldstein, and this is What's Your Problem? My guest today is Catherine Nakalembe. Catherine is an assistant professor at the University of Maryland. And she runs the Africa program at NASA Harvest, which is a project that uses satellite images to track agriculture around the world. And it wouldn't be wrong to say that Catherine's problem is using satellite imagery and GoPro cameras and AI to predict seasonal food shortages in sub-Saharan Africa. But in fact, her work, the problem she's trying to solve, is problematic. ultimately more human than that, ultimately more interesting, I would say, because the problem she's really trying to solve is, how do you use evidence to convince leaders to act to prevent people from starving? Catherine grew up in Uganda. She went to the University of Maryland to start a PhD, and then she went back to Uganda to do her graduate research. And the project was combining satellite data and on-the-ground research to to predict when the crops were likely to fail, leading to widespread hunger. And she was working on this for a few years. And then in the summer of 2015, when she was still a graduate student, there was a particularly bad drought in the region she studied. And she knew, she knew from her work that the people there would not have enough to eat in a few months. And she decided she had to get that information to someone with the power to do something about it.
00:01:58
Speaker 1: I realized really quickly that I knew more about the severity, basically the distribution of drought across this region more than anybody because I was able to look at it from space. And one year I finished my work and I was like, I can't just go back with this information because it had been an extreme drought year. I went to the office of the prime minister to share my My data and evidence in that year had taken lots of photographs and videos, et cetera. So I brought that. And then they had a meeting with the prime minister that led to— Wait.
00:02:28
Speaker 2: Can I just pause? So you're, what, a grad student or a young professor.
00:02:33
Speaker 1: At this point? No, I'm a grad student.
00:02:35
Speaker 2: And so it's an unusual thing for a grad student to be like, oh, my God, I have to tell the prime minister about this. What happened? How did that happen?
00:02:45
Speaker 1: So I've been coming for the last three years. People are trying to do the same things and crops fail. And I'd already started to kind of figure out how to, you know, predict what's going to happen, you know, early in the season. And there's always news around this issue, you know, Karamoja drought again, etc.
00:03:03
Speaker 2: So Karamoja is the region in, I guess, northern Uganda where you were studying.
00:03:09
Speaker 1: Yeah. And so it's always contentious. Is it happening? Are they, you know, just pretending? And this year, I had so much evidence, you know. First, because I'd taken so many photographs, I'd recorded video. And I wanted this to be sort of complementary to my PhD research. So I went and I shared it to the commissioner. And then he asked me, can you stay an extra few days? Because we have this food security meeting on Thursday with the prime minister and his steering committee, which are the other ministers. And I was like, of course. So I show up and I share the evidence and they asked me to say a few words. And I was like, you see this photograph? You see all of this thing that's dead here? This map basically represents the same thing everywhere. And it's eminent and people will be hungry pretty soon. And three days later or two days later, they sent food trucks. I have this photo that was sent to me from the guy who was the secretary to the prime minister or personal assistant. He was like, the trucks left this morning. I have this email, and I was amazed that I did that, and it went that far, basically.
00:04:23
Speaker 2: And that fast. So when you say people will be hungry pretty soon, that is an understated way, it seems, of saying what's going on. What was going on?
00:04:34
Speaker 1: You have subsistence farming where you produce what you eat. Very little is left over. And at the end of the season, you're supposed to sort of start picking up or eating what will be available to you through to the end of the next season. So this is typically around September. And so by August, everything had failed. And so there wasn't going to be anything harvested. But if you've kind of been out in the field a collecting data, and you run into kids eating sorghum shoots, for example, you know, the middle part of a.
00:05:10
Speaker 2: It's kind of like eating grass.
00:05:15
Speaker 1: Exactly. So like the middle part of a corn stem, it's sugary. And so, you know, you find kids eating that and they don't have anything else. They don't have livestock. Their family does not have any other source of income. It's pretty dire. So I think the other thing that might have made a huge difference in this particular instance is the photo evidence that I had. I take thousands of photos. I think I have over 300,000 photos.
00:05:41
Speaker 2: That's such a low-tech thing. Do you know what I mean? It's not like you were doing some wild satellite AI something. You were just walking around and taking pictures.
00:05:54
Speaker 1: Mm-hmm. And I had my map, my drought analysis map, that was very simple. And then we called it NDVI, Normalized Difference Vegetation Index, the anomaly of it.
00:06:06
Speaker 2: Basically, how much drier than usual. Exactly.
00:06:09
Speaker 1: Yeah.
00:06:10
Speaker 2: So, okay, so you do this, and the result in that first year is that they get food aid to people sooner.
00:06:16
Speaker 1: Right. Almost immediately. I have this text message from the commissioner. He said, this is the first time that nobody argued with the evidence.
00:06:28
Speaker 2: Did they then integrate your work in an ongoing way? How did that play out in the years after that?
00:06:36
Speaker 1: So the project that's called Disaster Risk Financing was basically being designed the following week. So I stayed that extra week and was like, I can not only develop the methods, but I can train the office how to utilize the data that's available in order for them to do that early warning, as well as write summaries, reports, et cetera.
00:06:56
Speaker 2: Take pictures. It sounds like one big lesson is take pictures.
00:06:59
Speaker 1: It's a validation for sure, but they'd already understood. So I kind of created this method to understand, compared to the history, if we hit this threshold, things are going to fail. Considering what has happened in the last 30 years or last 20 years, If we hit this threshold, vegetation won't recover and we need to start planning.
00:07:19
Speaker 2: Basically a threshold of drought. If rain is below some level.
00:07:23
Speaker 1: So initially it was like if vegetation conditions drop below this level within this period, a few months, months within the growing season, then we need to do something.
00:07:35
Speaker 2: So it's interesting. I mean, I might naively have thought that the constraint, the fundamental constraint was just money.
00:07:43
Speaker 1: Right.
00:07:43
Speaker 2: That there's just not the money to get the food to the people or infrastructure or something. But it's information at some level, like reliable evidence.
00:07:52
Speaker 1: Reliable evidence that is accessible. But I think money and the willingness to use that money is also really important. So that region where I was studying, even though it had the lowest economic indicators, you know, household income was very low. Infrastructure is very poor. The cost of food was so much higher than it is in the capital. Yes.
00:08:20
Speaker 2: Well, that goes with poor infrastructure, right? Exactly. When it costs a lot to move food, the food costs a lot.
00:08:27
Speaker 1: Exactly. Now, add the fact that there is a drought and traders can do whatever they want. And so it becomes much more expensive. And when the government is trying to acquire this water and food, et cetera, they end up paying a lot more. Uh-huh. And so if you're doing it in anticipation, it means that then you can acquire things at a much lower cost.
00:08:49
Speaker 2: Oh, so if they know in August that there is going to be great need in September, they can buy before there's a run.
00:08:56
Speaker 1: Exactly. And the other thing was, because it was in anticipation, what ended up happening is they would have projects that could hire people to work on that were community development projects. It could be tree planting. It could be road construction, it could be improving wells, and then they will be paid for those jobs. Yeah, that then they could use that money to buy food and other things.
00:09:22
Speaker 2: So it's interesting. I thought we were going to be talking about satellites and AI, and I guess we will be. But it's interesting that this conversation starts in such a low-tech way. I mean, obviously, digital cameras were not quite as ubiquitous as they are today. But this is 2016. It's not 2000 or something. Lots of people already had digital cameras.
00:09:44
Speaker 1: Why? Yeah, why? I'd say people believe what they see. And the NASA administrator in 2019, Jim Bernstein, he used to share my photos in his presentations. He'd talk about from NASA to the moon and beyond. But he would still use the photos and not the maps or the images. the numbers because it tells it you know we can see it from above and this is what it means on the ground this is the reality of what it is on the ground and that's i think much more accessible and understandable.
00:10:21
Speaker 2: Um so I feel a little weird now after you're talking compellingly about how important photos are. But I do want to talk to you about satellite imagery and AI.
00:10:31
Speaker 1: But they're photos.
00:10:32
Speaker 2: They're photos. Very good. They're photos. I find them moving in a different kind of way. I mean, you know, I've read this sort of big hand-wavy thing, which is, You know, there's lots of satellite imagery of the Earth now and lots of AI models designed to analyze crops, but those are largely designed for kind of big monocultural agribusiness, and they don't apply so well, certainly, to the kind of smallholder farms in the places you have worked. Like, I get that that's kind of the big picture. But more narrowly, like, tell me about your work to deal with that. Like, first of all, what is it that you want to know from satellite imagery?
00:11:13
Speaker 1: We want to understand what's growing where and how it's doing. The other is the complexity of the agriculture system. So when we're looking for crop fields, we're looking for shapes and then we're looking for some sort of seasonal pattern. So if you had 10 images over a growing season, you should see something like it's cleared now. it browns, it's like brown and wet, and then something green pops up, it matures, and then it's harvested. So it sort of like becomes, you know, dry. And so we should be able to see that signal. However, how low the brownness might be, like how dry something might be, is contextual. And while a reduction might seem like minor in some places, it's actually a very serious problem. And so the first thing was create a cropland map, basically exclude everything that's not cropland, estimate, understand vegetation conditions historically, like what happens every season, what months matter the most in terms of vegetation, in terms of crop production, and then try to understand at what point recovery doesn't happen. So to be able to do that at a really large scale, satellites are a phenomenal tool for this. At the time, even today, the only satellite that allows us to be able to do this at a much higher frequency is MODIS. And MODIS is a 250-meter resolution. That could be multiple crop fields in a single pixel.
00:12:52
Speaker 2: So the farms are too small or too varied.
00:12:56
Speaker 1: Too varied, yep.
00:12:58
Speaker 2: Tell me, what is a typical farm in this region like?
00:13:02
Speaker 1: So in Karamoja, people grow sorghum, millet, a little bit of maize. There's a lot of sunflower, too.
00:13:10
Speaker 2: And will they grow all that in one place, more or less?
00:13:13
Speaker 1: So you will find sunflower mixed in with sorghum, for example.
00:13:17
Speaker 2: And how big is a typical farm?
00:13:20
Speaker 1: So farm plot... On average, less than 2.5 hectares. So a football field, a soccer field is about an acre. That's the average over most of sub-Saharan Africa, but they're much smaller. So like in a single soccer field, you might find maybe four plots as an example.
00:13:43
Speaker 2: Like four different people's little farms?
00:13:45
Speaker 1: Four different people's plots with hedges, and they're not like distinct perfect shapes like you would see.
00:13:51
Speaker 2: Quite small. Hard to see from a satellite, presumably, especially if there's a bunch of different crops there.
00:13:58
Speaker 1: So hard to see their boundaries. So a single modus pixel might include four of those fields that might have different things, for example.
00:14:06
Speaker 2: Yeah. Let's talk about the GoPros. So fun. Tell me about your work with GoPros.
00:14:13
Speaker 1: So one of the issues is in order to get from cropland, where all the crops are growing, to go to crop type. So the USDA, for example, produces a cropland data layer, which if you take a look at it, you can see where corn, wheat, soy is growing. And now that's been changing over the last couple of years. In the US. In the US, yes. This also exists for Europe. They have something similar. There's, of course, some simplicity. There's a single crop. The fields are homogeneous. They're kind of easy to map. In East Africa, in addition to fields being small, I already mentioned the intercropping, mixed cropping, and all the heterogeneity, how people plant at different days. To be able to get a crop type map, you need a lot of training data. So, you know, one of the biggest advances, I think, you know, the reason why we have huge leaps in terms of AI is the availability of training data. You know, same as face detection, all people's faces on Facebook, et cetera, makes that technology a lot easier to develop. And so this is true also for crop type mapping. So it's like face detection, but for crops.
00:15:20
Speaker 2: Why do you want a crop type map?
00:15:22
Speaker 1: There's so many benefits to knowing exactly what crop is growing where. You want this map in order to understand, for example, what kind of fertilizer will be relevant in a particular region. So if you know there's places dominated by rice, just like in Ghana, a particular place in Ghana, where could irrigation infrastructure make a difference? The other and most important, most basic thing is, if you want to understand how much will be produced at the end of the season, you need to know the total area by specific crop. So if we have 25,000 hectares of corn planted... And the average condition or the average yield is 2.5 tons per hectare. You multiply that and you have your production, which will allow you to understand how much corn is available. If it's a critical crop in a particular area, you would know that, you know, food security would be okay. So we had a big issue in Illinois with corn, et cetera. If Germany had a big issue with wheat production, bread being so important.
00:16:27
Speaker 2: Yeah.
00:16:27
Speaker 1: They would have to plan accordingly. So you need to figure out, what do I need to import to have some sort of stability?
00:16:34
Speaker 2: And so for Sub-Saharan Africa in general, or the regions you study, are there crop-type maps now?
00:16:42
Speaker 1: There are some crop-type maps, yes. Some are done on a very small scale, regional scale. But what this project was trying to do was try to, one, just figure out the fastest possible way we can collect data over really large areas. One, because in some countries there are three seasons. So Rwanda has three growing seasons. So you're looking at three crops. So to cover the whole country and get an understanding of did they grow corn more generally, bananas more generally? No. You need to be able to do the survey continuously.
00:17:18
Speaker 2: And the satellites don't have the resolution, essentially, to do it.
00:17:22
Speaker 1: To be able to do it, yeah.
00:17:23
Speaker 2: You can't tell from satellite data.
00:17:24
Speaker 1: So we need to train a model to tell the model, this pixel, this particular spot is banana. Find me banana everywhere else in the image, yeah.
00:17:34
Speaker 2: So you have this idea of like, okay, we need more on-the-ground data. So what do you do? What is the GoPro project?
00:17:44
Speaker 1: GoPros are perfect because they have a GPS. So when we take a photo, the photo has a location. So typically when I'm collecting field data, I need a photo, evidence, but also I need to be clear when somebody labels in the form that maize is in this growing stage and you have to stand in the middle of the field. Because I want to use that location to train my model to predict where else maze might be.
00:18:10
Speaker 2: This is the old school, slow, never going to scale technique that you're trying to kind of quasi-automate here.
00:18:17
Speaker 1: Yes. At this point, even if all I had were images that had a longitude and latitude, and I know where the camera is facing, I could go through all these individual photos in Google Earth Pro and add the label myself.
00:18:36
Speaker 2: So at least you would have a bunch of images with a very specific geolocation on latitude and longitude.
00:18:43
Speaker 1: And I know what the crops look like. I know what the crops are.
00:18:47
Speaker 2: So this is phase one. But you don't want to have to label a thousand things corn, whatever.
00:18:54
Speaker 1: Yeah, I would have been okay with that too, though. Considering, so I did this in Mali. I did it in Senegal. In Tanzania.
00:19:02
Speaker 2: Wait, this meaning driving around with the GoPro? No, the phone. Oh, the walking around.
00:19:08
Speaker 1: Walking around, yeah.
00:19:09
Speaker 2: Yeah, okay. I mean, it sounds kind of charming in a way.
00:19:12
Speaker 1: I learn a lot from doing field work. You learn what it means for someone's field to be washed out when you go. But if the idea is to try and create a map that will be useful for this task, there's only so much you can do when you cover a very small area. Yeah. So 2020, I call this project a COVID-safe project. 2020 was like, oh, well, now we have this funding. We have to figure this out. I go buy a GoPro at a store, a fashion sort of amount with my dad because I couldn't order it from Amazon. And then we make this magnetic thing on the car and just drive. I go to, there's an agricultural research station. And the director was like, I described to him what I'm trying to do. He was like, oh, course you can drive around here, whatever. So we try it out. So the first thing was, if we get the images, are they useful? Is the GPS good enough?
00:20:07
Speaker 2: And it was.
00:20:09
Speaker 1: And so then we were like, oh, this could actually work.
00:20:17
Speaker 2: We'll be back in just a minute. That's the end of the ads. We're going back to the show. And as you probably recall, when we left off before the break, Catherine had just tested out a GoPro herself and decided that GoPros could be a good way to gather crop data from small farms.
00:20:42
Speaker 1: So I have a friend in Kenya. I asked him if he can buy, if he can order things on Amazon and to work with a team there to see if they can collect data in Kenya. So we get a team of two. So basically four. So it's a team of two people. One team drives south, the other drives north. And we cover all of Western Kenya in a week.
00:21:09
Speaker 2: Wow. And they're just driving everywhere they can drive, in a car, on a motorcycle?
00:21:13
Speaker 1: So you can see from some of the original data that we have from Kenya that we're kind of covering these big... There are lots of fields, but they're proximal to these big roads, which... It's great, but not good enough. And so it became pretty quickly obvious that a motorcycle might be better because a motorcycle can go to a level two, level three, you know, smaller paths and go through. And in Tanzania, they went a little bit crazy because I feel like they're driving into rice paddies and going like this.
00:21:42
Speaker 2: Well, people live far off of roads that you could drive a car on. Exactly. The kinds of farms you're talking about exist far from drivable roads.
00:21:50
Speaker 1: Exactly. So... But just from the Kenya team, we had over, I think the first iteration might have been maybe 300,000 images or more. Then the team of two in Uganda, I kind of equipped them. And during COVID, there's nothing else you could do and asked them to kind of just go drive around. And they went crazy because then we had over 2 million images from them. Okay. And so then it's just too many images. Even I got tired of just like trying to figure it out. But I had an undergraduate computer science student who worked over this and developed Street to Sat, the first iteration of it. And remove clouds, adjust images automatically so that they're kind of like straight. And then remove faces and ensure that they're actually crops in the image.
00:22:48
Speaker 2: Crop and not crop, you called it in the paper. It reminded me, I don't know if you watched the show Silicon Valley, but there's a guy in that show who invents AI.
00:22:59
Speaker 2: And they show it a hot dog and it says hot dog. And everybody's like, wow, it can detect food. And then they show it a hamburger and it just says not hot dog.
00:23:07
Speaker 1: Yeah.
00:23:07
Speaker 2: And all it does is say hot dog, not a hot dog. This is where you are now.
00:23:10
Speaker 1: This is where we are. It's like, is there a crop in this field? Yes. If there is, then that's phenomenal. And then we kind of did a couple of iterations of what we call labeling. Yeah. Basically, go through a couple of images to create a label dataset, which is, this is, we draw bounding boxes, maize, maize, this is what maize looks like, this is what banana looks like.
00:23:34
Speaker 2: So now you're saying you're, this is training, you're creating training.
00:23:37
Speaker 1: We're creating a training dataset, yeah.
00:23:38
Speaker 2: Training data, yeah.
00:23:39
Speaker 1: And then with the first iteration, then we run it through the initial data from Kenya to do sugarcane maize. Okay. Okay. And it was pretty good. It was pretty good. It was like, you know, when it's actually working and you can't believe that it's working, but it is working and it's unbelievable. So you're like, you're in this phase where you're like, oh my God, is this actually happening?
00:24:05
Speaker 2: What are people doing now with this basic technique of, you know, driving around with GoPro cameras where there are these small farms? Like what is one specific thing someone is doing that is helping them take action in the world in a way they would not have been able to do?
00:24:21
Speaker 1: So I give one example. So in Kenya, when we did the first iteration of it as a test pilot, and then were able to actually work with the Ministry of Agriculture for them to go out and collect the data themselves, they were like, do we get to keep the cameras? I was like, yeah. And I asked, you know, what sorts of things are you thinking about? I was like, well, you know, that would be kind of just Phenomenal photographic evidence because a person can go back whenever and send me those photos while they're out there in the middle of nowhere. So remember going back to the low-tech technology is that when the ministry reports, so we have these WhatsApp groups where extension agents share constantly what's happening. If they share that there was a flood and the flood was serious, everybody knows a flood is serious, but how serious is a different story. when it comes to things like beans, water staying on the surface for a long time and the beans turn yellow is because all the nutrients have leached. And so for the extension agents to be able to go through and collect these images without needing to get off, you know, take a photo, et cetera, et cetera, is something that is used as kind of evidence and as part of their reporting in the ministry.
00:25:38
Speaker 2: So it gives you a baseline.
00:25:40
Speaker 1: It's like a baseline, but also people believe what they see.
00:25:44
Speaker 2: It goes back to you walking around taking pictures, but just a lot more pictures.
00:25:48
Speaker 1: Yeah.
00:25:50
Speaker 2: So if you sort of distill it down, what is your big project? What are you trying to do in the world in your work?
00:25:58
Speaker 1: So data is an essential, right? Evidence is essential. But try to use it to improve, you could say, the human condition through better decision making. So If I know for sure that something is coming and it will be this bad, it means that the persons or people responsible for decision-making, they have no excuse to not do the right thing. I think that's kind of what kind of drives me. And I was going to talk about the 2.0 of the disaster risk financing program is because we've gone so far now since then that what I've built out now gives a three-month forecast of what things will be three months from now based on a model that has learned. So because I can process more data, I can combine multiple complex data sets than before, thanks to compute, and increase access to data. But I understand the problem so deeply that that threshold now is specific to the specific research It's no longer one threshold for all of them. It is a district, like a county-specific threshold that could even be lowered to be sub-county that could be zip code-specific threshold that in this zip code, this threshold is it. It doesn't apply to everywhere else.
00:27:28
Speaker 2: It's sort of zip code by zip code in Uganda. You can tell how much drought is going to be really bad. Three months in advance.
00:27:37
Speaker 1: Exactly. And it's not only in Uganda. The method is largely scalable. I could apply the same methodology to the U.S.
00:27:43
Speaker 2: Do you have the data? Yeah. Can you do it anywhere, or do you need some amount of data collection in some places?
00:27:50
Speaker 1: So for the forecasting component, all I need is satellite rainfall, satellite temperature, satellite vegetation conditions. And then the validation component is where I would need people to actually go and confirm, you know, am I actually predicting the right thing? As an example, you could use auxiliary things, you know, you could use images people are posting on social media about events to sort of as an in-between. But in order for the evidence, you know, to be strengthened, then you want to go, you know, take those photos. And then the other thing is, you know, have a routine component around collecting and gathering that evidence so that it's not whenever you feel like or whenever you have money, but it's just that at this point, we need to check in and see how things are going.
00:28:41
Speaker 2: Like, if I were going to say, Catherine's problem is this, like, what would the sentence be?
00:28:48
Speaker 1: There are lots of quite simple things that we could do already, but we just don't do them because they seem so Maybe we overcomplicate them. I always think about these, a minister will not, you know, read your F1 score table, you know, the confidence of your model, et cetera, but the minister will read a memo that's usually like three paragraphs. What I learned in the risk financing work is that it has to be distilled into a half-page memo. The memo is evidence points that we've hit the thresholds, about 70,000 people will need emergency aid, and we're going to release funding on this day. That's it. And then the minister says, I agree. Right, so whether or not it took me 100 months, $ 5 billion... 25 analysts to create the evidence or whatever it is, usually it's something that is accessible, concise, that helps that decision making. And accessibility to that information is so critical in this case. So I'd spend a lot of time, you know, training and building capacity so people can reproduce things.
00:30:13
Speaker 2: On their own.
00:30:15
Speaker 1: Like you can run the helmet's work all on your own. All the methods, everything you need in the kit. And a lot of people have done this as a team that's done it in Nigeria. People have done it in India on their own.
00:30:26
Speaker 2: Basically building crop-type maps using GoPro cameras?
00:30:30
Speaker 1: Creating crop-type labels, yeah, using GoPros. But also the workflow itself. We have a notebook available that anybody can run with a Jupyter notebook without anything else. So the problem is some things are a lot simpler than we make them out to be. And if the thing is to solve hunger... I think those solutions for solving hunger already exist. And they have nothing to do with my model or my technology. It has to be moving food where it is produced to where it's needed, making it available when it's needed the most, for example. And I can provide the evidence to be like, okay, it's absolutely needed here. So I think that's my problem.
00:31:24
Speaker 2: We'll be back in a minute with The Lightning Round. We're going to finish with The Lightning Round, which is just a little bit more playful than the rest of the show. What's one thing your dad taught you about fixing cars?
00:31:49
Speaker 1: Oh, my God, that guy.
00:31:52
Speaker 2: My dad my dad told me um.
00:31:56
Speaker 1: We had my car battery in 2020 of it was overheating whatever and he was like how do you drive a car without opening the hood and i was like like seriously and then i said this to him uh because his phone is runs out of memory and i was like How do you work with a phone that has no memory? So, you know, I was trying to kind of like paint this problem. My dad, what I've learned from him is you can figure it out. And he has this terminology. It's, I think, a Luganda thing. It says chito chige, which is kill it to learn it. So, you know, in terms of like fixing a radio, you take it apart and then you have to put it back together.
00:32:39
Speaker 2: It's like break it, break it and fix it and you'll understand it.
00:32:42
Speaker 1: Break it to fix it. basically, because then you will understand it. And so he always took apart, you know, car engines, and then he will put them all back together. So I think what I learned is that you can learn, you can figure it out, whatever it is.
00:32:58
Speaker 2: What's one thing you learned from your mom about running a restaurant?
00:33:02
Speaker 1: My mom is a workaholic. She's told me recently that She works, she likes, she keeps working because she likes to work. She's no longer, it's no longer profitable, but she likes to work and she believes she stops working. You know, she'll just like fall over and die. It's just an extreme, it's an extreme statement. But what she means is that whatever the work is, it's important and do it to the best of your ability. And so she does, you know, has done so many different things to kind of sustain it and keep it running. And so... Keep on moving, keep on working, and you can figure it out, I guess, also in this case.
00:33:43
Speaker 2: Who's a better badminton player, you or your sister?
00:33:46
Speaker 1: My sister, she's left-handed.
00:33:48
Speaker 2: Oh, is that a big advantage?
00:33:49
Speaker 1: Of course, you have no idea where her left from her right is. Her backhand is your, what you think is her backhand is her forehand. And then she worked so hard to improve her backhand that her backhand feels like a forehand, right? It's like the one when you hit it and it comes back. You're like, oh my God.
00:34:09
Speaker 2: Yeah. That's the sound it makes too, right? In badminton. What's one thing I should do if I visit Uganda?
00:34:21
Speaker 1: Eat a mango.
00:34:22
Speaker 2: Okay. Are they better there?
00:34:24
Speaker 1: Absolutely. And an avocado.
00:34:26
Speaker 2: And an avocado. Also better?
00:34:28
Speaker 1: Yeah. Never refrigerated. Fresh. Maybe harvested yesterday.
00:34:34
Speaker 2: Um... What about GoPros? Are you dreaming of other uses of GoPros for work or not for work or things other people could be doing with GoPros?
00:34:44
Speaker 1: Yeah. It's funny. I always thought about them. I used to wear one. The first, first iteration of GoPros, when I went to field work, I wanted to record everything. But about GoPros, if I was to design one, I'd include one that has a distance measurement system.
00:35:02
Speaker 2: I would do that.
00:35:05
Speaker 1: So I can know how far my maze is from where I'm standing.
00:35:11
Speaker 2: You're trying to do that with software, essentially.
00:35:13
Speaker 1: Yes. But I'd like for it to give me an estimate of how far something is. It's kind of easy to derive that if you're in a city because you have corners and things, very strict structures. But you can't do that when you're doing it for crops. I did win a GoPro for good, but they sent me a GoPro. which I already have a lot of GoPros.
00:35:37
Speaker 2: You won an award, and the award was another GoPro?
00:35:39
Speaker 1: Yes.
00:35:40
Speaker 2: What do you wish they'd sent you?
00:35:41
Speaker 1: I wish they had a conversation and I'd suggest what I want in the GoPro.
00:35:48
Speaker 2: Yeah, right. A GoPro that could measure distance.
00:35:52
Speaker 1: Yeah.
00:35:53
Speaker 2: I appreciate your time. It was lovely to talk with you. Thank you so much.
00:35:57
Speaker 1: Lovely talking with you, too. Thank you for having me.
00:36:07
Speaker 2: Catherine Nakalembe is an assistant professor at the University of Maryland and Africa program director at NASA Harvest. Please email us at problem at pushkin.fm. Tell us what kinds of shows we should do more of or less of or particular people you think would be good on the show. You can also find me on X and on LinkedIn. Our show is produced by Gabriel Hunter Chang and Trina Menino. Our editor is Lydia Jean Cott and our engineer is Sarah Bruguere. I'm Jacob Goldstein, and we'll be back next week with another episode of What's Your Problem?
