Shaping Africa’s AI Generation - A Conversation with Kavi Ramburn and Stefan Coetzee

Welcome to the Future Learning Design Podcast.

Africa has this huge young population that's rising up, and so we need to start creating those opportunities. And so we need to be on the, on the edge of this technology and looking for opportunities to share with them and to build in Africa.

Hi everyone, thanks for joining me for this last episode in this 4-part series exploring the ways in which AI tools are impacting the lives of teachers and young people around the world. In this episode, I talked with Kavi Rambirne and Stefan Koetzi from ALX Africa about their amazing work bringing professional foundational competencies programs to young people across Africa through in-person hubs and online course offerings. They have recently announced a big partnership with Anthropic and the government of Rwanda, so I was keen to talk with them about how this will boost their mission in introducing Chidi, an AI-powered learning companion built to scaffold critical thinking and problem-solving about and with AI to learners and educators across Africa and beyond. Kavi is the Vice President of Learning at ALX Africa and has an extensive background in learning research and sustainable development economics and advocacy for social impact across many sectors. Stéphane is the AI Innovation Lead at ALX Africa, spearheading AI product research and early lifecycle product development. He has a huge depth of knowledge as a data scientist, content developer, and educator. Amazing. Thank you for taking the time, Kaby, Stéphane, to have the conversation with me. As I said to you, as part of a broader series that we're doing with colleagues from all over the world who are engaging with Anthropic, really to make a difference to young people in education systems all over the world. So, no, really happy to be able to chat with you about it today. So thank you so much. So if we dive in first, I mean, if it's okay with you, I think it would be helpful for listeners just to understand a little bit more broadly about ALX and kind of the broad sweep of things that you are already engaging in, amazing work that you're doing. And then we'll get into more of the specifics around the AI piece with Anthropic. But yeah, Kavi, do you want to open it up perhaps?

Sounds good. Yeah, happy to talk about that. So Alex is a pan-African institution providing world-class learning experiences for hundreds of thousands of learners across the continent. So we've enrolled over a million learners over the past several years and graduated over 400,000. We offer programs in the tech space, creative space, and in entrepreneurship. A key feature of our offering is that all of our programs are underpinned by a core soft skills training program, and that's proven to be one of the key differentiators of our programming in terms of how our graduates are able to set themselves apart from other tech graduates or other creative graduates and from other learning institutions. We have a very strong community-based approach to our learning model where our learners get to engage quite heavily with their peers, with our technical mentors, and the broader ALX community, which is an integral part of that learning experience. We have a physical presence in 8 different African countries currently, but we also, through our hybrid model, so our online learning model, we've reached historically learners in 42 of the 54 African countries with a goal of actually reaching all 54 within the next 5 years.

Amazing. Yeah, very cool. And do you mind if you just say a little bit more about that capabilities, that soft skills underpinning? Because I'm quite intrigued by what you said there because there's definitely, for me, a very human aspect to the engagement with technology. So I'd be keen to go a bit deeper on that if you don't mind.

Absolutely. So ALX is actually part of a wider group of institutions. So the African Leadership Group, leadership being the key word here. We have the African Leadership Academy, which is a high school institution in South Africa, African Leadership University based in Rwanda and a campus in Mauritius. And ALX was the third institution as part of the group. So leadership being, like I said, a key part of our curriculum, which actually runs through the entire organization. So a key part of the, our learners' journey is when actually they start their journey is to go through a series of leadership training, everything from self-leadership where they get to learn about some, some key skills to, to better manage themselves, interact with others, develop some key habits and principles for, for their own learning. But we also teach them one of the most key skills that we probably impart with our learners is learning how to learn. That's also part of that core leadership curriculum. They also learn a series of other very practical skills that cut across any domain. So project management skills, communication skills, core data skills as well, all of which are now obviously data, AI skills that they can use in a whole range of different fields and verticals so that they're well equipped as the job market evolves and changes, they can quickly adapt., like I said, learn how to learn those new skills and keep evolving even after they've long left ALA. Yeah, fantastic.

And it's good for listeners as well because I've had some amazing young people from African Leadership Academy on the podcast before, just at the end of last year. And I've also interviewed Hatim El Tayeb quite a few years ago now, actually, about the amazing stuff they're doing at ALA. So yeah, good to make that connection as well. Perhaps, yeah, maybe then thinking about, obviously, you know, the significant shifts that have been happening over the last few years in terms of artificial intelligence. Stefan, perhaps, yeah, could you say a bit maybe about, because clearly ALX has been going for a lot longer than that, how have you engaged with that AI question in terms of, you know, AI literacy or course, different courses or integrating it across different offerings that you've already got?

Yeah, I'm happy to. I think, you know, AI to me brings a fundamental shift in how we teach, how we learn and what we need to know in the future. So, The moment that ChatGPT came out, I started using it inside of our team. So, you know, creating content, creating story-driven content, coding up examples for learners. And so it is really embedded in my workflow as an SME and as a teacher at ALX. And part of that means we also need to spread that awareness about AI literacy. And so I think it was in 2023 or 2024, we created an AI literacy course that just unpacks the whole mysticism around generative AI, what it is just at a basic level, how that changes your personal life and your professional life, and what are some of the things you can do to start getting onto that curve and getting ready for this, this new wave. So the first approach was to use it ourselves. The second approach was to start using this technology internally to build products so that our learners can benefit from that technology as well. And now we are in the phase where we are using that technology to now scale up our efforts. You know, the success of these coding agents are so— well, the performance of these agents are so incredible right now that we can just create a product and launch it in very quick time and turnaround and then do an experiment, see what the learners are doing with these tools, and then Either I'll promote that or scrap it if it doesn't work. So it's really a core part of my DNA and I think it's a core part of what ALX is and what we are heading towards is understanding that AI brings this fundamental shift. Yeah.

Amazing. And I mean, for people not on the continent, what, how do you see AI and Africa? You know, there's obviously there's multiple dimensions to that, like with, you know, the way that exploited labor in Africa has been used as part of the training process, you know, so ethical, really deep ethical questions about it. But then there's also kind of access questions and device access and network access. I mean, does, how much does that factor in? I mean, in a way that's my ignorance asking the question, but how, how much does that factor into your thinking about doing the work? Or is, or is it not so much a factor and you just keep going as you would in a, in a different context?

No, it is core of the problem, right? Africa has this huge young population that's rising up. And so we need to start creating those opportunities. And so we need to be on that, on the edge of this technology and looking for opportunities to share with them and to build in Africa. You know, from an ethical point of view, there are many scenarios that, that either make it very difficult and give us a lot of opportunity. So, no, there's a lot of forecasting that A lot of jobs are going to change, and I suspect, you know, for Africa, if that change doesn't happen quick enough, it'll kind of worsen that gap that we've, we've had classically. But there's also this opportunity where it provides us with a toolset that was classically available to privileged communities, to systems and places where we had the luxury to develop learning institutes for a long time and nurture teaching and learning at those institutes. And now we have the opportunity to take that and apply that to a mass scale where instead of a learner sitting in a class with 80 other learners and all of the challenges that come with that, you know, power, access, attention, teachers perhaps not trained in the subject that they are supposed to teach. And so AI brings a very great opportunity to start implementing really impactful tools that augment this process of learning. So instead of a teacher focusing on trying to get everyone on to work and to engage, they, their attention turns more towards the human elements of teaching. And so at this stage, everything is early game. It's, it's very difficult to predict how things are going to work, what is going to work, what are the longer-term effects of teaching with AI and interacting with AI, and what the effect would be on human relationships. So we, we are trying to keep all of this in mind. A core part of Aleksa's DNA is also the aspect of community, where we understand that that is a core part of how we teach, how we learn, is that there needs to be this human element. So if you just, you know, blanket apply AI into your organization, what you'll find is you'll just isolate learners where, you know, GDE or this tool just gives them all the answers. It scaffolds so well that they don't ever need to speak to anyone else. What'll happen is they go into the workplace and they, they don't know how to work with other people. And so you have to simulate as closely as possible what the real world would be. And so, yeah, we definitely try and think about this very deeply before we put pen to paper, so to speak. Hmm.

Yeah. Interesting. And Kavi, maybe could I ask you just to, I mean, Stephan mentioned Chidi there, like perhaps you could, yeah, just talk about some of the specifics of the way in which this is working in community or like co-creating with rural communities or, you know, getting this into young people who, as Stefan mentions, you know, maybe are in very large classes. Like, what does some of that actually, that work actually look like on the ground?

Yeah, absolutely. So a couple of things that we're seeking to do to make this technology a lot more accessible to people, especially in rural areas, is first to ensure that our learning experiences are primarily mobile-based. So ensuring that our learning curriculum, the content, our learning management system is adapted for mobile-based learning. And side by side with that, the technology that we're using to enhance the experience, such as Cheetah, our AI Socratic tutor, is also embedded in a way that allows learners in different parts of the continent, whether that's in rural or urban areas, to do so through mobile access. As we know, only a small proportion of them have access to either desktop or laptop, but the spread of mobile phones is, is quite wide. So that's why that's one of the objectives that we have is really to sort of re-engineer our offering to make sure that it's primarily mobile-based. The second one, our engineering team is actually grappling with this right now. We're looking at a few different hardware solutions. That would allow us to make our learning management system accessible in areas that have very low connectivity or no connectivity. Still in the early conceptual stage, but our engineering team is working on actually creating those devices that would allow learners to access our learning management system and the tools associated to that completely in an offline environment, which would then allow us to enter spaces such as refugee camps, for example, and other hard-to-access areas in a much easier fashion.

Yeah, interesting. And I should have asked this earlier, but what's the age of the learners that we're talking about?

Yeah, so generally speaking, the vast majority of our learners are usually in the age group of 22 to 28. That's where the bulk of our enrollments come from. But our programs are open to learners of any age, so we do have learners who are much older as well join us and also younger learners who are actually juggling many things in their lives. So other studies at other institutions, working in a lot of cases, or family responsibilities that they're in charge of. So we see our learners as, in that sense, jugglers, and we do call them, quote unquote, jugglers. And we try to design and craft our offerings catering to the fact that they do have a lot happening in their lives. So we do want to provide them with the flexibility and support that allows them to continue their journey while they also have thousands of other things that they're doing in their day-to-day.

And how accessible is it in terms of cost? What's the model there?

Yes, in terms of our cost, it's actually very accessible, thankfully, because of the support of many institutions such as the Mastercard Foundation, which is one of our biggest funders historically. So we offer all of our learning catalog for $5 a month to all of our learners. So for a subscription of $5, they can access the entire Lx Learning catalog. And can take any number of learning courses that they want at the same time. This may change, of course, in the future as we decide to adapt our pricing model to our new offerings, but as it stands, that's what we're offering.

No, great. Thank you. I appreciate it. And is that just, again, excuse my ignorance, Coco and Caribou also are kind of other elements that you're, or products or course offerings?

Absolutely. I think, yeah, Stéphane's actually the brain behind all three. So I'll let him speak to those three tools. Yeah.

Yeah, I think it's important to note that AI tools on their own or in isolation are not going to get us to a place of very usable products or very effective products. So you have this chatbot that might have an interaction with Alena, but there's a lot of context you have that missing kind of, just imagining if you are a teacher and you meet a student for the first time. A lot about teaching one-on-one happens in the interplay between you, like getting to know how they think and setting some traps for them so you can expose some misconceptions. So there's a huge part of AI that's kind of incompatible in, in that sense where, you know, it's the stateless machine. You put something in, you get something out. And so there's no real growth or understanding of that learner. And so that's something we have to artificially build. And so how we are thinking about AI is, well, you have these interconnected products. Each one of them has a role. To play in the learner journey from, you know, introduction all the way to post-graduation. And all of them help us understand the learner better and in different aspects. And all of them help the learner in turn with different aspects. So for example, CoCo is an AI grading application and it has pedagogy baked into it, into its feedback system. And so we get to understand how they think about a problem, how they unpack that problem in an essay or in a, an assignment and then use that in order to make our chatbot experience better. And so you have this interconnected web of data flowing between these things that really power and empower these products to work. So yeah, Karibu is right at the front. You land on our website and it helps you to understand what we do and what type of courses we offer. Koko, Chidi facilitate your learning. And later on, there are some other products that we would like to do for more postgraduate.

So yeah, okay, thank you. That's useful. And, and then, so making the, the slight left turn towards Anthropic, I mean, you mentioned other AI providers, Stefan, earlier. I mean, what was it about Anthropic? Because obviously, you know, you've just announced some partnership work with them, and also in collaboration with the government of Rwanda and the Ministry of Education there. Yeah, what was it about Anthropic and, and that possibility that it opened up for ALX?

Yeah, I think, I mean, at the start we were thinking about it, you know, primarily from a cost perspective, and Anthropic hasn't, you know, isn't the cheapest model provider. There are cheaper models out there, but as soon as you start applying this as a chatbot or a teacher, you know, you enter some ethical and some safety gray areas where you want to be careful, right? And one of the key decision drivers was performance and adherence, where, you know, we, we use mostly Haiku to drive this because it's, it's really cost effective and its performance and its adherence to what we instructed to do is very high. But at the time that Haiku was published, it was the safest model available. And so that is one of the most important parts to this whole thing. So, you know, we applied it into an AI literacy course for the first kind of proof of concept run. And we applied it in a sort of a reflective way where, you know, in this ACE course, the course really poses a question to you, like, AI is here and what are you going to do about it? And it helps you start to understand the problem and then scaffold your new way of thinking around this. And so part of that journey is that, hey, here's this, how AI affects this part of your professional life. And there's this question that used to be just a rhetorical question, like, you know, how does this apply to you specifically? Those questions are now chats with GPT, where those conversations are going pretty deep with learners, where people are ideating business ideas. People are coming with common frustrations that they have in an old world kind of management perspective and how AI might help out and change that. So there are some deep and impactful conversations happening and to make sure that those conversations don't go off rail. Having someone like Anthropic on our side to have these models available, but secondly, to give us some advice when our guardrails might not be as strong as they need to be. So the collaborative aspect of that is really important, but, you know, Anthropic just, in my perspective, kind of lead the pack in terms of model quality at all three tiers, Opus, Sunet, and Haiku. So it's a very simple choice for us to make, just safety first. And performance second, and then a lot of collaboration and understanding from them.

Ah, that's really interesting. Thank you. And Kavi, maybe you could speak to the aspirations, I suppose, like the rationale for partnering with Anthropix. It sounds really clear. And then what would be the hopes about what you're actually wanting to do there? I mean, why does that support you in your mission? What, you know, where is the alignments then with Rwanda, the government in Rwanda, et cetera? So yeah, how is that opening up?

Absolutely. Yeah. So the partnership itself and what we seek to do there is sort of core. I mean, this question sits at the heart of what ALX stands for. So historically we've tried to reach as many people as we could across the continent and provide them with access to world-class experiences from a learning standpoint. And before I get sort of deeper into your question, I actually want to reference a study that's a core part to who we are as ALX. Back in the 1980s, there was this educational psychologist, Benjamin Bloom, who conducted this study around the impact of one-on-one tutoring on learners. Through that study, he found that learners who had access to one-on-one tutoring performed at 2 standard deviations higher than learners who did not have access to, to one-on-one tutoring. And this became known as the 2 Sigma problem. Why a problem? Because obviously it's quite costly to provide at-scale one-on-one tutoring at that time, or through most of, of human history. It was usually people with a, with a privilege and financial access to higher-class institutions that allowed them to get this type of tailored support and one-on-one support. With the coming of age of AI, we now have the, the possibility of addressing this historical problem, right? So through Chidi, through our AI Socratic Tutor that we're developing, in partnership with Anthropic, we are actually now able to provide this type of tailored support to learners 24/7, right? So a learner that sits in a rural part of Ghana, for example, has direct access to a tutor that helps them think critically through the learning material in the same way that historically someone would've needed to, to be in an affluent institution to have that type of personalized tailored support. So that's a big reason of why we decided to get into this partnership with Anthropic, to be able to provide that large-scale access to technology that greatly enhances the learning experience of all of our students. In terms of the government of Rwanda, one thing that we're obviously very grateful for is for their forward-thinking leadership, right? As a government, they've shown tremendous boldness in their national vision for, for technology and human development, right? So the, the Ministry of Education isn't only treating AI as a distant or, or future concern, they're really leaning into it in their day-to-day. And through our partnership that we announced with them at the end of last year, we're seeking to, to launch in a first stage, a pilot that would train their public servants. So in the, in the education space and the healthcare space, with critical AI skills that will allow them to greatly enhance their productivity in terms of what they do on, on their day-to-day. So that's also very exciting for us. Not only are we partnering with sort of world-class leading AI institutions such as Anthropic, but also have the partnership with the government of Rwanda where we're able to test and implement this technology and this new approach to learning very concretely. In the context of Rwanda, and of course, use the learnings and everything that happens there to then expand our impact and our reach more broadly across the continent and potentially beyond.

Yeah, interesting. I mean, it's such a compelling story, isn't it? That the two sigma problem and the AI capabilities to address that two sigma problem is really compelling. And at the same time, I worry personally that there is so much more going on in a responsive relational human interaction between a teacher and a young person that it's, it's, it's just a really interesting challenge for me in how to hold this. Cuz I, I hear you 100% on like access, you know, historic exclusion from many aspects of capability building, knowledge building, et cetera. Because it was clearly the ultimately kind of elitist, exclusionary, aristocratic tutoring system originally, but way back, right? And then this increasing access, but how to hold that with also this like ecological piece of humans in a room together are doing so much more than just the, the exchange and tutoring on the, the thing that is the seemingly the explicit object of the learning, right? There's so much more like learning going on in this kind of incredible collidoscope of all this other stuff. And that's— it's a fascinating challenge, I think. I just wanted to say that because I think there's something— I don't know quite— I have no answer to that, but just wanted to raise it, I suppose, in relation to rolling out one-to-one tutors as fast and as widely as we can.

Yeah, absolutely. So, I mean, really critical point that you raised. One way that we're also seeing this play out is in terms of how we're enhancing the capabilities of the teachers themselves, right? You're absolutely right. The type of interactions that the learners need to have with their teachers should not go away. Stephan touched on that earlier in terms of that critical importance of those human interactions. The great thing with this technology is that it also greatly enhances the productivity of the teachers and potentially even takes away some of the mundane tasks that would have been quite time-consuming for them, which allows them then to really focus on the key parts around sort of providing inspiration to their learners, one-on-one guidance when needed. So the fact that we're enhancing their productivity, providing them with the ability to sort of set aside a lot more time to really focus on that inspirational part that they're spending with their learners, I do think that could also, yeah, counterbalance a little bit the—

for sure—

some of the worries that, that we have around that.

Maybe to add on to that, there are different models that we are operating in and that we are speaking about. You know, if you have a fully online platform, an AI is probably your best bet because it's better than static content or videos that take a hell of a lot of data. It's a very efficient way to teach. But absolutely, in a teacher classroom space, I don't think we've figured that out. You won't have a simple chatbot solving that problem and, and augmenting that teacher in that space. I think there's a lot of of thinking that you'll need to do and work that you need to do. You know, it's, it's in that same regard, you know, you have augmentation of a human being. So just a quick example, instead of a teacher explaining a concept on a whiteboard, they can now ask an AI to build a little snippet of code that they put up on the screen. They press play and you see a live simulation for exactly the thing that you wanted to explain to your learners. And so there's a lot of those opportunities that come along, but I think for us to build an effective in-class tutoring or augmentation system, I think that we're quite a far way away. There's a lot of other aspects, as you say, there's— we don't want to interfere with the human part. And it's, you know, if you have superfluous tools assessing or helping out with learning planning, I don't see them as being highly effective, I think. You'll need an integrated kind of solution that, yeah, that really understands the problem well and then tries to understand where are the pain points or the problems in that classroom and try and solve them.

Yeah, sure.

No, I appreciate that. I mean, and you know, I just raise it because I think, you know, there is a, sometimes the capabilities of the technology leads and just remembering the human aspect, I think is, and I've heard that a lot from you. And also just to reflect on what you were saying, Kavi, about the juggling. I mean, I think. Clearly that's also a really important piece, right? Is that people are holding a lot in their lives and if there's a moment of access where they can access something where being in a classroom with a bunch of humans learning that thing isn't an option. So actually there's a really important role for meeting people where they are with the capabilities to support. And I think that's, you know, that sounds a really powerful aspect of what you are doing as well. Amazing. Thank you so much. I don't know if there's anything else, just, you know, in terms of if anybody finding out more or if they're listening to this and, and, you know, obviously I'll share all links, et cetera, in the show notes, but just, you know, if there was anything coming up that's exciting or anything that you just haven't said that you'd like to say.

Yeah.

So an interesting new area that we're adventuring ourselves into is through the creation of an institution that we're now calling the Frontier Institute of Technology. It's the latest institution within the broader group of institutions that I referred to earlier. It is a US-based accredited tertiary education institution, fully powered by African-developed technology and know-how, right? So the people who have gone out and set up the Frontier Institute of Technology have come directly from, from ALX and the platforms that Frontier will be using, the technology that they'll be using from our AI chatbots. And so on have all been developed in Africa by ALX. So quite an interesting story in terms of how we're bringing that technology to a US and global audience. The other really interesting aspect of what we're offering with Word Frontier is that given that it is a US accredited institution, it will be able to offer accredited degrees to all enrolled learners. So we're starting off there with a bachelor's degree in software engineering and one on the creative front as well. The really interesting aspect for our learners based in Africa is that once they complete their initial learning with ALX, they will have the ability to directly enroll in Frontier, in that US institution, and earn a US accredited degree, which obviously carries quite a bit of weight in our local markets. So that's, yeah, just another really interesting aspect of the latest, the newcomer in the group.

Yeah, fantastic. Thank you, Gaby. That's, I mean, it's Really inspiring work. And now I just really appreciate everything you've shared today and everything you're all doing. As we said, you know, it's not straightforward, and I love that you're really emphasizing the community aspect and the relational aspect and the human aspect as well as you're building it out. So now, thank you, Stefan. Thank you, Kavi. Really nice to meet you. Enchanté.

Our pleasure. Thank you so much, Tim.

We hope you've enjoyed this episode. Please feel free to continue the dialogues with our guests, with us at Good Impact Labs, on our blog or on social media, or within your own networks.