🌍 AI in Africa - Agriculture

We are trying to build a localized early warning system for food and nutrition security. And the idea is that existing early warning systems tend to look at agriculture from a regional perspective and by the time let's say there's a likelihood of crop failure, so what, we could pass them on information in near real time or just in time so that they can change their current agricultural practices. And we are trying to build it for smallholder farmers and of course the policy see makers who work with them.

Hello friends. Jared here to tell you about changelog, our membership program for those of you who want to directly support our work. Your plus plus membership gets you closer to the metal with extended episodes, makes the ads disappear, and takes our audio to the next level with higher bitrate MP3s. You can join today at changelog.com/. Welcome to Practical AI, a weekly podcast making artificial intelligence practical, productive and accessible to everyone. This is where conversations around AI, machine learning and data science happen. Join us at PracticalAI FM Community and follow the show on Twitter. We're at Practical aifm. Thank you to our partners at Fastly for shipping our pods super fast all around the world. Check them out@fastly.com.

Welcome to another episode of Practical AI. This is Daniel Whitenack. I am a data scientist with SIL International and I'm joined as always by my co host Chris Benson who is a tech strategist at Lockheed Martin. How you doing Chris?

Doing very well today, Daniel, how are you?

Doing great because we have a really exciting follow up show. We've been doing these sort of spotlight shows on AI in Africa and we've been really, really pleased to partner with the Open for Good alliance and FairForward the Macarera AI Lab and IDRC on these shows. And it's just been really great to feature some of the amazing AI work going on in Africa. We've got a few guests with us today. We've got Joyce Nabende, who is the head of the Macarera AI Lab, who has been joining us as a sort of co host on these shows. Then we've got Leo Matuku, who is a research and AI lead at the Local Development Research Institute in Kenya. And then we've got Godliver Owo Mogisha, who is a senior lecturer and researcher at Busitama University. Welcome everyone.

Thank you.

Thank you.

Great to have you with us. I'm wondering, I'll maybe pass it over to Joyce to start us out with. I'm wondering if maybe based on your experience with the Macarera Lab, which I know has looked at AI and agriculture, which is what we're going to talk about today. Maybe if you want to kick us off into that subject.

Yeah.

Thank you, Danielle, and thank you, Chris. It's good to be back on the show here with you guys. And I'm very excited as well that we have Leo and Goodliva on the show today. And I believe it's going to be a very interesting discussion that we're going to have today. Yeah. So as you mentioned, the maker AI lab has been doing work in AI for agriculture. It's like our strongest research work that has been going on in the lab. And I feel excited that we are going to hear particularly about AI and from, first of all, Godliva, who has, you know, done her PhD in AI for Agriculture and the experience that she has and the thoughts that she has about the topic, but also from Leo as well, who has done a lot of research in AI for agriculture. So maybe just to start, I will start with Godliva. So, Goodliva, you're welcome to the show once again.

Thank you very much, Dr. Joyce and everyone who is on the talk today.

All right, thank you. So maybe my friend. My first question to you, Goodliever, is can you really just give us a general introduction to AI in agriculture and, you know, specifically for the African context? I believe that the African context is a unique one. Why? Because we have the highest percentage of smallholder farmers in agriculture and people depending on agriculture for their livelihoods. So what is it about AI in agriculture that you think it's very important for the African context?

Thank you very much once again. Yes, Artificial intelligence in agriculture, specifically in Africa, is really an interesting part. When I started doing my research, I could not imagine how many challenges we have here, especially when it comes to when it came to applying some of these applications that we've developed and scaling down to regions or talking to different stakeholders. So you realize there are challenges that you never even imagined to see when you hear a smallholder farmer or talking about things that you imagine maybe you could implement or you could have reached to them. So to me, I think our challenges are very unique. Understanding the challenges that are affecting our smallholder farmers locally will give us the best systems that we need here in Africa.

Yeah. Thanks, Godliva. Maybe to Leo. Well, just thinking about what Godliva is mentioning about the unique challenges that we have in the African context. So maybe can you briefly talk about these, but also looking at the broader context of AI in agriculture, particularly for Africa.

Thanks, Joyce, for the introduction When I think about why AI plays a unique role and especially in agriculture and in the sub Saharan context, one thing I want to point out is the issue of climate change. So when we look at the continent, we are being affected by climate change. And as you mentioned, a majority of the agricultural practices are by smallholder farmers. And traditionally smallholder farmers have been relying on intuition and experience to, you know, farm and to take themselves through the planting cycle. However, when you start thinking about climate change and the need to move towards climate smart agriculture, I think AI provides opportunities to help smallholder farmers cope with these effects of climate change, where rain fed agriculture is no longer very reliable as it was in the past. And then again, granted that smallholder farmers are majority of the agricultural producers in our context, they tend to mainly farm for subsidence, rather subsistence farming. And so ideally this means that there are issues when they are not able to have a plentiful harvest when it comes to food and nutrition security. So AI I think supports interventions in the agricultural space that can be even provided at a granular level at the household level, as opposed to typical interventions which tend to just look at a region. But AI can really support those subsistence farmers at their household level.

Yeah, I have a quick follow up for you. For those in the audience who may not have had to experience food insecurity in a direct way, we have a global audience and some people have to deal with that, that issue and some don't. For those who don't, can you kind of describe a little bit about what the implications of that are so that we can kind of understand the problem better before we dive into how AI is helping to remedy that?

So in terms of food and nutrition security, when I mentioned that smallholder farmers in and especially in sub Saharan Africa are mainly subsistence farmers, it means that first and foremost the crops they produce are to feed their families. And then if there is any surplus, that is what is sold. Typically in more food secure nations, you tend to be able to have a large scale farming that provides food and nutrition for the general population. So you purchase your foods, purchasing your food, you are actually producing it, and then what you're selling is a surplus. So ideally, if there is crop failure, let's say due to drought, due to disease outbreak, then it means that you are likely to, as a household that is relying on subsistence farming, you're likely to go hungry if you don't have the income to purchase the supplemental nutrition. So that is why food and nutrition security is a very important issue in this context and why it needs to be addressed with immediate efforts due to issues such as climate change, to think.

About how maybe advanced technology and specifically AI can help with some of these issues. Some people might be thinking, well, how is their data? AI requires data. How is their data related to maybe small farms in the African context? What's available for us to sort of work with and that sort of thing. Maybe I'll pass it over to Godliver and maybe Joyce as well, since I know her lab works in this area. Godliver, could you start us out and maybe just describe kind of generally, what are the categories of what people are trying to do related to agriculture with AI or advanced technology methods, and what data is sitting behind that, what data is enabling that?

Thank you very much, Daniel. Interesting question. Yes. So there is a lot of data that is surrounding us, apparently, to understand some of these barriers in agriculture. Mostly, I could say I live around the crop data that we see, but also so the practices these farmers applying on their farms. I just got to know about it recently. Yeah. So you realize the challenge, for example, we have crop pests and diseases, which we've been trying to handle in a long time. Yeah. But you also realize that that may not be like the major problem or like the nature challenge some people are specifically facing, or they even don't know what it is.

Yes.

So I don't know. Talking about more data, I think it's going to move from what we see in the gardens, but also how these farmers try to relate what the practices they use and where they get resources and which type of resources they get. Maybe that is the farm materials and all that. So if we had to really to intervene in there to understand why we have these many problems in agriculture, I think we have to go back to understand maybe the data sources. Yeah. I think Joyce can add on more here. She's been experienced more in data acquisition and all that. Yeah, thanks.

Yeah, thanks.

Kodleva. I think you raised very important issues around data. Also, just thinking about what Leo mentioned earlier, that the farmers have the knowledge, they have this information, but maybe it's not captured well, like, you know, when is it going to rain? When am I going to harvest? I think that I always think about that, like that's indigenous knowledge that the farmers know and have. And it would be very good to capture that in a systematic way that that can be incorporated with other data sources that are openly out there, can help the farmers. But also, Godliva, you hinted on the crop pests and diseases. I know that the work that we've been doing has really been around capturing image data for crop pests and diseases and helping to providing the potential for building AI models. But there are also other data sources. Some might be that the farmers don't have access to them, but some that they might have. So for example, if we are looking at the farmers, what is important to them is my soil, Do I have good soil? So information or data data around their soil, health monitoring data, that's something that's very important for them. There's also satellite imagery data, although as a smallholder farmer I might not have access to the data, but I might have access maybe to the output of the data, depending on how advanced the people who are building the technologies can be able to build out the applications that the farmers may have access to. But that's also data that's available that can be used by the farmers in building out AI applications for health. Maybe moving forward to leo. So Leo, in your experience, what datasets do we have, especially in the African context? Because I think this is a unique context that if you were to build an AI application in our context, that means that you probably have to collect the data yourself, right? That if you were to pull a data set out there, that might not work for us in our context. So what is your thoughts generally around the data and using this data for building applications in our context?

Thanks so much, Joyce. I agree with you and Gotlieba on the data sources that you've mentioned that might be available in terms of supporting the building of technologies to support agriculture in our context, we have tried to use mobile phones to basically collect ground truth data, which typically is not available. As you mentioned, that data sets out there might not be relevant for context. So we are trying to see what are the easiest tools and technologies we can use to collect this data directly from the farms themselves, from the farmers. And we found that mobile phones using tools such as the ODK geo apps really help in maybe creating proxies for some of those more advanced data collection tools such as the Grameens, which might be expensive in our context. And we're using this to collect data, for instance, on farm boundaries. It's very important to calculate, calculate area under cultivation and what is being cultivated within these different plots so that we can try and estimate, let's say, the yield production, the access to inputs required to, you know, for bountiful harvest. So farm boundaries are one of the data sets we also try to collect. The other is the access to specific inputs. When we talk about climate smart agriculture, in this case we would like farmers to access hybrids of seeds, for instance, and fertilizers that work well in their context. So access to input data and what inputs are being applied in different parts of farming communities is a very important data set to collect. You mentioned the use of things such as photos to be able to identify pests and diseases as part of farm management practices. But I would also want to talk about another data set that we find very important, which is commodity prices. So what are the market prices of these crops once they are harvested at different times throughout the year? Because if those are predictable in a way, then it's easy even for farmers to know to estimate what their income would be once they take their crops to the marketplace. So I think across the general ecosystem, different stakeholders have a role to play to make these data sets available. So for instance, their market prices might come from government sources and yield estimates. While a lot of projects are trying to see how we can make it easy to collect this ground truth data from farm samples that can be used to train artificial intelligence, algorithms and to.

JS Party is a weekly celebration of JavaScript and the web. So fun is at the heart of every episode. We play games like Front End Feud.

I gotta go with the big O Opera. I know it's a wild card, but I just feel like it might be hanging on.

Show me Opera. Three strikes and you're out. Discuss and analyze the news in the immediate term.

I'm really excited about those hyper developer.

Productivity tools like Copilot that just automate the boring stuff for you.

Explain technical concepts to each other like we're five.

Did that make sense?

Yeah man, no. Magical sawdust Muffin fairies.

So your muffin is your function or variable.

Yeah, your muffin started down in the sawdust because you defined it later.

Yeah, you defined it down there, but actually it got hoisted up.

Sure.

Okay.

I did my best debate hot topics.

Like should websites work without JavaScript?

I'm going to appeal to authority and read some quotes at this time.

Okay, I've lost complete control of this page.

Go ahead. Okay, the first book no Code is.

Faster than Code Interview amazing devs like Rich Harris, Una Kravitz, and many more.

CSS is one of those languages that is very easy to pick up quickly.

And learn things like how to change a text color, but it is very tricky to master.

This is jsparty. Listen and subscribe today. We'd love to have you with us.

Okay, so as as we move forward here, I'm fascinated by this discussion about the data side of things because that's really in the practicalities of, of this. And I want to follow up with, with you, Leo. You were just discussing about the various data sets and as I explore the LDRI website, I, I was learning about this African open data network, particularly struck by this kind of line about unless the right people have the right data in the right forms to help them make the right decisions, our development goals will remain unattainable. So I'm wondering maybe if you could kick off a little bit of discussion about why open data and sort of networking and community is important as it relates to working in this area. And then Joyce and Godliver, if you have any follow ups on that, we'd love to hear them as well.

So when I think about open data, it is in a way democratizing access to data. And this means that if some investments are put in place to collect this data, that it's still beneficial to a wider group of people than just those who collected the data. And now specifically in agriculture, as I've mentioned earlier, it's quite difficult to get accurate ground truth data to support interventions. And so there are a few projects now that are supporting the collection of that data. Find that this data is being collected from a small sample of, let's say the palms in the region. And so the applicability of this data is only as useful as whether it can be reused or repurposed in other areas or at least used as training data sets for algorithms that may not necessarily impact the places where the data has been collected, but are for similar contexts. Traditionally we have been saying that one of biggest development challenges here in Africa is the sparsity of data. So when open data I think creates mechanisms and communities around sort of meeting that data gap and being able to support further application of data sets. You mentioned the issue of community. And I find the concept of open data serves communities in two ways. So there's the community of the producers that they can pull together resources, they can pull together data and reuse them, reproduce whatever experiments or research that has been done without having to expand resources. But at the same time, open data I think has a close relationship to what is usually called citizen science, that we can also encourage the communities themselves to contribute to the creation of this data. And then when it's placed in an open format and in, you know, in these open repositories, it's also accessible back to them either as, you know, outputs of technologies applied on this data or they can use it directly. So I think open data is, creates this community resource that again, preserves some of that indigenous knowledge we alluded to earlier, but at the same time promotes the development of new technologies where there might be sparse data previously because of this increased availability of relevant data sets.

Thank you very much, Leo. I think you've said it all in terms of open data source. Yeah. So from my experience, actually, if I look at when I started my master's research and even PhD work, there was really like nothing to do with open data source that I could use or to implement some of the ideas that I had. So we started to collect our own data. Of course, we've also been working closely with different research institutions like Canaro, who supported us. But here we are moving to communities like the previous speakers have said, farmers helping us with this information and which we are collecting together and also availing to other people. So I think really having an open data source in this kind of problems that we have in agriculture would really help a lot in coming up with better technologies. And also us putting down the protocols that we're using to collect this data to help others improve on them or reproduce the same data that we are getting is one key thing that we are doing, apparently, at the AI lab. Yeah. So I'll leave, I think, Joyce, I think there is an important data source like to do with radio talk shows. I think that should be a lot. Joyce, you can talk about here. Yeah.

Yes, good.

Libra, you mentioned that unconventional data source that, that people don't usually think about. But just to get back to one of the points that you mentioned, especially because for us, we are in a university setting and so students of course, will come up and they're very excited about building applications for AI in agriculture, and then where do they start? They need the data. So I think this whole topic of open data sets for AI and putting them out there and making sure that there's proper documentation, like Gotlieva has said, for these AI data sets. It's something that is very key and important to us in the lab because that's always the hinging point of where we start to build the applications and making them available. That means we'll encourage more people to come on board, more students to come on board to have applications that they can be able to build for agriculture. And we try and make sure we thought about having the data sets as diverse as possible. Guadiva mentioned the radio data set. That's a unique one, which is also very sensitive. But one of the projects that I mentioned, I think earlier on that talk was where we are trying to build speech recognition models for radio data, particularly looking at agriculture, because we know that the smallholder farmers listen to radio, they will call into radio and then they'll make sure that they think that they will get help from there because the agriculture is experts can hold talk shows and that's something that we don't, it doesn't come easily to us. Like if we're looking for a data set that's not a data set that you would conventionally think about and I will pick up and use, but we think that it's an important data set. So just thinking about open data sets in general that we need to think outside the box and think of what are the unique data sets where we think we are going to get data, but also where we think that these are data sets that will eventually also help the farmers. Because if they know that these are unique radio data sets and people are listening to me, they will be more enthusiastic to try and get help through that source. But it comes back to the developers to ensure that that data is, you know, mined out, formatted in a way that is representative for the agriculture experts to respond back to the farmers. So it's kind of like a two way that we think about the open data sets but we also think about the farmers from which we are collecting the data, the farmers for which we are building the AI model for again.

Yeah, I got a quick follow up. It's one that's been developing over the course of the episode in my mind. If I go back and to my question early on that to Leo about defining kind of food insecurity and I think that's a real, that was a really useful explanation for me I think. And it made me realize how fortunate I am that I'm not living in a situation where food security is one of my primary motivators. And, and it might be that way for many people in our audience. And it's really gotten me thinking about the challenges that substance subsistence farmers are facing that I'm probably not thinking about on a day to day basis in North America. And so you know, where we have these large corporations that drive farming and where AI in that context of farming is, is, is typical because there's lots of technology approach, it's just kind of built in to that farming culture at this point. I'm curious as you're working with subsistence farmers that, that are trying to produce enough and they're dealing with these very basic concerns day to day to provide the food for their communities and even just for their families, you know, and, and the risk of going hungry and children going hungry when you're, you're bringing. It seems like a big jump to say this is a very immediate concern that affects people in a very direct way. And we have these amazing technologies that we're talking about. Is it culturally a jump to get a community to say, help us with open data, help us apply these tools to the vital work that you're doing? Is there any challenge in, in buy in or understanding to take the most advanced technologies in the world and apply them to the most basic food security issues from a cultural sense?

So of course, with any technology adoption, one of the biggest issues is the issue of trust. So we can't just consider ourselves experts, dive into a community and, you know, just apply technologies and try to change their way of life directly. So for us, even just practicing AI in this agriculture space has been quite a journey that started with us building trust within these communities. And we've used a very interesting approach whereby. So another challenge that might not necessarily be, might not necessarily come up in this conversation is the issue of extension services. So we tend to find that for farmers to learn about, you know, new science, new technologies, what to do with disease outbreaks, they tend to rely on government agriculture officer who passes on this information maybe in a way that is, you know, digestible to them. And so in this sense, they act as information intermediaries who, I mean, based on new research that the government wants to apply in local contexts, would pass on this information. Now there tends to be a shortage of these knowledge workers who tend to, I mean, ideally we should be having a ratio of one to 400, you know, one extension farmer to 400 farmers. But we tend to find that right now there's only 1 to 10,000 farmers or even more. And so it's very difficult to start passing on this information. So you can imagine how that becomes a problem when you're introducing, let's say, outputs and recommendations from an AI. So what we've done is that we've actually built a community of what we call peer pharma educators. So in each village that we work in, we have identified farmers who we treat as champions and we train them on agricultural best practices. And then they tend to use, it's almost like a training of trainers. Then they go and train on their fellow farmers on, let's say, these new approaches and these new agricultural practices that might improve their harvest. So you see, by working with trusted members of the community, even as we start introducing outputs from an AI, or if we need their support in Collecting data. We have already established trust because we are working with community representatives and you know, we're not just again passing information top down. We also get to spend time with them and understand what their challenges are when it comes to, you know, what's going on in the community and what approaches to best use to, you know, to get the community to adopt new practices. Now that's just one side of things. The other very important stakeholder when it comes to agriculture is the government itself. And in our, in the case of Kenya, agriculture is a devolved function. So ideally like the way you'd have states running issues in the US Here it's we have what are called counties and governors running agricultural issues. So we also take our time to, to also bringing the county governments into the fold when it comes to explaining to them what the AI can do and using them as informediaries as well as sort of knowledge holders to design and develop the AI. So for us we are not just sitting in our ivory tower in the city and building technologies and then taking them back to the farm. We are actually through this network of peer pharma educators as well as the county government and other local communities, using them as co designers and collaborators in the design of the AI. Because again with any AI what matters is the output. So we will not for instance give them metrics as outputs. Maybe fancy dashboards do not work in that context. But if we are able to craft the messaging of, let's say the recommendations, for instance, saying that we have observed through satellite imagery and through the ground truth data that there's a likelihood of crop failure in the near future, you know, you just don't show them, you know, graphical images. What would be more important is how to design the messaging around maybe potentially do not rely on rain fed agriculture, but it is time to irrigate. So at that point designing how those messages will be crafted and how the outputs of the AIs reach the communities require an effort that's not just by US technology developers. And so that's why I love this approach of design thinking and how versatile it is in different contexts including in US building AIs for agriculture.

We've talked a lot about sort of generally AI and agriculture, the data involved in AI and agriculture. I'm really interested in some of the sort of success stories or challenges you all are currently tackling. So maybe I'll pass it over to Godliver and the others can join in afterwards. But Godliver, I'd love to hear about some of your own work in Particular I've seen some of your published work around detecting diseases in plants and using spectral data and even diagnosing bacterial wilt and other things. So wondering, could you give us a sort of highlight of a couple of the, the projects that you've worked on and what the challenges were and the results were?

Yes, thank you very much, Daniel. Specifically in that area still about crop disease detection. So over time as we worked on these tools to put them on the mobile phone and here we assume farmers are going to use them to detect diseases in their gardens. But then we realized there is also like a very big challenge. Usually if you look at some of these diseases, they will come out visually when the status for the stages like very late or close to harvest, harvesting the garden. And as you heard from the start, most of these diseases, they can even take up to 100% of the damage. If you look at the cassava crops with these root tuber crops. So we said instead of waiting onto these visual features which we can see with our eyes, which you can see through the mobile phone, think about other technologies that will help us to detect these diseases at a very early stage. And that's how spectrometry came in place. This is more to do with using light like as you see in a lemon language, like how you go to an X ray system and they scan through your body. So the same way we are scanning through these crops so we are able to read information which we can't visually see with our eyes. And I was first experiments actually were able to detect that you can, you can see the disease at least six weeks when it, once it has been. When the plant has been infected. Yeah. So some of the challenges we had is some of these technologies are very expensive. So if you talk about taking back to a smallholder farmer, this technology is over thousands of, of dollars. So no one can use it here. Even interpreting it is difficult. So our challenge here and our task here is to come up with low cost technologies that we think should be understandable by these people. We are developing the products the farmers.

Okay.

Yeah. So that's the stage where we are and we've had a few prototypes and which we are building on again that we really to other different crops. So we started with cassava and now we are moving to all different crops. And yeah, so that's, that's what I can say about the spectrometry technology which is very, very promising apparently.

Yeah. Godliva, you mentioned, I think a very important point there that with spectrometry we look at the diseases before they manifest. Right. Many of the applications that are out there focus on image data when the diseases are already on the. But you try to do the early prediction. I think that's very paramount research work such that then we don't wait for that damage because once the, once the disease is manifested, they spread really quickly. And for a smallholder farmer, that's. That's such a big problem for them as well. I'll also probably just highlight one topic, one research area also to allude to what Leo said, that the issue of having what we call them, the experts, agricultural experts in Uganda, that the ones that, you know, try and reach out to extension farmers or try and reach out to the individual farmers to help. So if a farmer has a problem, they will go to an extension farmer to have the extension farmer maybe come to their garden, look and see if there's a problem or not, and provide a solution. But the ratio, as Leo pointed out, is usually very big, that one farmer cannot attend to all, one expert cannot attend to all the farmers. And I think in one of the first episodes we had Mutsembesa here talking about an application, a mobile application that we have built in the lab called adseb, where we have tried to provide this connection between the farmers and the experts such that the agriculture experts can easily be able to reach out to more farmers. But also, I think getting back to what Leo talked about, that we have the champions who are also farmers, they are more knowledgeable than other farmers who are also able to come in and help their families, fellow farmers in their community, to try and help and do that. So one of the applications that we are currently working on is to try and build a recommender model where the farmers on the network can come and post a question and provide and the application is able to automatically, you know, provide or give them, you know, a response. If someone says, for example, I am in a certain area and I want to find clean planting materials, the season is coming up, I don't know where to get cleaning planting materials and they don't know who to contact. But through the application, one is it's able to respond and say, oh, if you need clean planting materials, maybe this is where you can find them. Or maybe I have identified these symptoms on my crop. What can, what disease is it? And so the application, the recommended module is able to, you know, provide a response out to the farmers to say, okay, if that's. Those are the symptoms that you have, it's probably this disease. But also that those are the things that can easily be flagged out, out to the agriculture experts. But I like also the peer to peer learning that when we have that form of that network, then when the expert is not available, that we can have another farmer who's also knowledgeable, who can quickly respond and come back to the agriculture and come back to the smallholder farmers and provide a solution for them. And so I imagine that there's a lot of potential for this work. And what is important as well is to try and see how to provide these solutions in the local languages that the farmers understand. Understand. Many of them are not very knowledgeable in English. And so if you come and you want to provide, you know, solutions in English, that might not work. I think for Kenya, it's Swahili, that's the main language. And so you have to also target the languages that the farmers speak into or, you know, that they understand better. And so those are some of the things that we are trying to do and build these solutions out in the local languages that the farmers understand through, you know, natural language processing technology. Leo, any feedback on applications from Kenya?

Yes, so I had started describing our project earlier on, but just to go into a little bit more detail, we are trying to build a localized early warning system for food and nutrition security. And the idea is that existing early warning systems tend to basically look at agriculture from a regional perspective. And by the time, let's say there's a likelihood of crop failure and the information is reaching the farmers, it's the same issue that Joyce mentioned that, I mean, you can already see that my crops are failing. So what if we could pass them on information as well as possible interventions and remedies in near real time or, you know, just in time, so that they can either change their current agricultural practices. So that is the motivation for the project you're doing. And for us it presents an interesting challenge because one, we are trying to build it for smallholder farmers and of course the policy, I mean, the policy makers who work with them. And the issue of smallholder farming is very unique because unlike what Chris was mentioning, there being those commercial farms, farms here tend to be very small, about 0.5 to 3 hectares in size. And then at the same time, a lot of farmers tend to do practice of intercropping. So for instance, on the farm they may not just have maize, they might have maize and beans. So how do you detect maize, how do you detect beans? And how do you detect whether they're getting the right amount of, you know, moisture, whether they are likely to maybe Dry out or you need to change, apply fertilizer, apply different inputs to prop up the yield. So we are trying to collect ground truth data with tools that are very locally available to farmers and to their representatives on the ground. So using mobile phones, using the typical camera on a mobile phone and trying to see if we can use this ground truth data and reference it to earth observation data in order to basically start predicting instances of crop failure. Again using a very similar process to what Joyce was describing for the disease interventions. But now here in general plant health and especially as related to climate variances and then trying to see whether we can detect that early enough and then in near real time send messages with interventions to both the policymakers but also to the farmers. So for us it's a very interesting challenge given that again we're trying to use low cost technologies but at the figure out sustainable ways to scale this AI to, you know, various seasons and monitoring from means to other different crops that tend to be what farmers plant for subsistence. And maybe just a small addition like what we find that given that we are trying to also estimate predicted yield yield, that such data can also be used in other use cases such as insurance and access to, you know, financial services for these farmers. Because if you're able to estimate their yield, then it's very easy to to offer them the right financial services that make sense for them.

So as we finish up, I would really like to get each of your takes on kind of what you're excited about on the future. I'd like to start with God Lover and then maybe get some follow up from Leo and Joyce about where are you excited about things going in the next short or longer term? If you think that there's a longer term play there, what is AI doing that you think will really impact things going forward?

God Lover, thank you so much, Chris. Really the future should be very exciting. If I see in a short time with my experience how these farmers adopting to some of these tools that we have, I think it would be really good in the near future. I only had I think a concern maybe to Leo or someone else she mentioned about bias in terms of technology. And also we've had some of the systems like incentivize like give us an incentive to these smallholder farmers to use this tool. So I think I don't know how this is going to work well in the future if we still have maybe incentives in place. But with my little experience apparently where we are working with no incentives with farmers, I see it's really good and in terms of adoption, they pick very fast challenges, say in languages and also sometimes maybe typing or if you could use some visual graphics into these tools that maybe when someone is talking, there's something happening with their tool they're using. So it could make life easy in the future. Thank you.

Leo, how about you?

I think I'm most excited about the potential for solving this grand challenge of food and nutrition security, but also the potential for AI to increase local incomes. I think when there's predictability in the production process and then from when you plant a seed to when you get a return on your harvest, then it's very easy to create plans around your farm and also uplift your livelihoods. So for me, that is very exciting, but also I think just the ability to bring more people into the fold, hopefully given that there's more data available than issues such as marginalized communities or increasing and gendered approaches. So where women are considered also and are brought into the fold, where communities have agency of how they live and self determination, I think those for me are really exciting things about AI, and especially AI that is locally designed, developed and built, because hopefully they understand the context, the nuances, and acknowledge the biases that exist.

Thank you. And Joyce, you have the final word. What's the future like for you?

Yeah, thanks. I think I'm also very excited, you know, just looking back at what we've been doing and how the farmers are also very excited with technology. I feel like he is to involve them early in the development of the technology such that there's also early acceptability of the tools that are coming up. So I think that AI, you know, has great potential to make the farmers more resilient, you know, to agricultural threats, to help in prediction of yields. The work that GOLIBA is doing around, around early disease detection, I think that these are all technologies that can be enhanced but also adopted by the agriculture experts. So I think bringing all the key stakeholders on board as we are building and deploying of AI applications in agriculture is something that's very important to have the growth of AI and its use for smallholder farmers across the globe.

Yeah. Thank you all. Well, this has been a fascinating conversation and just really important work. Appreciate all, all that each of you are doing and the community you're building and yeah, thank you for taking time to join us on the podcast. It's been a pleasure.

Thank you. Thank you for having us.

Thank.

You.

All right, that is practical AI for this week. If this is your first time listening, subscribe now at PracticalAI FM or just search for Practical AI in your favorite podcast app. We're in there. And if you're a longtime listener, please do share the show with your friends. It is the best way you can help Practical AI succeed. Thanks again to Fastly for shipping our shows super fast all around the world to Break Master Cylinder for the Beats. And to you for listening. We appreciate you. That's all for this week. We'll talk to you again next time.