Deborah Daley 00:06
Hello, and welcome to Changemakers, a podcast series brought to you by Nature, where we shine a light on fascinating and extraordinary scientists.
I’m Deborah Daley, global chair of Springer Nature’s Black Employee Network, and I’m very proud to present this series where we explore exclusion in science and how some researchers are doing something about it.
In episode four, we meet a South African data scientist who was determined to expand the use and benefits of artificial intelligence on the African continent.
Vukosi Marivate 00:52
I’m Vukosi Marivate. I am the chair of data science and a professor of computer science at the University of Pretoria in South Africa, and also a co-founder of an AI startup called Lelapa AI. And I also wear many hats, especially looking at community-building in AI across the African continent.
My journey in science, I think, goes back to probably preschool. For me, first time interacting with a computer or as something.
It just started this fascination of, like, “What’s what’s going on there? How do you get to do this, and how can you use this?”
I think by the time I was in high school, this then became more of a fascination of, “Okay, if you are putting data into a computer, how then can you, like, you know, manipulate the data to build things that are interesting?”
So these are, like, the building blocks and thoughts about thinking about machine learning, right? So in machine learning, a machine learns patterns from data or inputs, right?
And this is just one subset of what we currently know as artificial intelligence, where there you’re trying to get a machine to be able to reason or engage with the world, right?
It’s how it does it. One of the things could be machine learning, that it first gets these inputs and outputs, and tries to learn patterns about that.
And then finalize, then moving to Rutgers University in New Jersey in the US to do a PhD in computer science, mostly on reinforcement learning.
But during that time, which is like the 2009-2010s, you started hearing this term called data science, in terms of a lot of spaces within universities, within research circles of like, yeah, we want data scientists.
And in some ways, it actually encapsulated me very perfectly that I was interested in AI or machine learning, not just for the theoretical, “Hey, let’s push just the theoretical understanding of these things.” But also, “How then can we actually look at problems within society and actually solve them?”
Yeah, I came back to South Africa in 2015 and really spent time building up a career in research in these spaces up until we already joined the academy, or academia, back into a university in 2018 and have been working, really, on machine learning, data science, especially when it comes to the intersections with African languages.
So we tend to then think about this concept called natural language processing. So how do you now get a machine to process language, right?
And then one of the ways to do that is to build machine learning tools or AI tools to be able to get that done.
But then what you find is that for languages like those in the African continent, more than 2000 of them, there aren’t really good tools, and because there’s not much data, not much interest. So, yeah, you have to then explore that space.
The AI systems that end up being developed and deployed are directly interacting with wider society, and this is where now the questions of, “Are they fit for use? Do they really do what they say on the label? Are they risky to people. And what happens with the biases that they have internally?”
And this is where, again, that more societal-driven inquiry becomes very essential. Because if we’re saying we’re going through this great AI moment at the moment, and it’s going to change the way economies work, you know, in the future, it really has to represent all of us, because it will cost us if it doesn’t.
Right, they do. People just tend to think like, “Oh no, no. For me, the thing works great for me, that’s fine.”
But if a majority of the world is really going to be affected negatively, that just puts all of us at at a disadvantage. And we also stand to also lose more, because the more our systems are also given diverse inputs and challenges, the better, actually. It also goes back to the theoretical basis, because you cover more of the landscape of possibility.
Indaba is a isiZulu word that means gathering. And as such, we created the deep learning Indaba as the deep learning gathering on the African continent.
So we wanted to have an African name. And typically, indabas are also meetings or gatherings that have some importance.
In this case, it’s become the gathering place of the African AI community on an annual basis.
The deep learning Indaba was an idea really borne out of looking around the continent about 2016 or so the African continent, in this case, where a number of friends and acquaintances who had studied in South Africa, on the continent and beyond, in AI-related fields, came together, looked at the current like, you know, the landscape at the time, and looked at what had been happening in terms of the acceleration of AI or machine learning that was happening, and thought that we didn’t have enough local activity capability and capacity to really be in this coming moment.
So the idea was: “Can we have a gathering that allows, then for us to really bring together the African community to strengthen machine learning so that we can contribute, shape, and ultimately be our own owners of these coming technologies at the time, right? Things are very different now in 2025.
So we started it now as the first meeting, akin to a summer school, in 2017, where it was like teaching to each other, poster sessions for people to talk about their research, debates, all those types of things.
But fast forward to 2025 where we will be having our Indaba, in Kigali, in Rwanda, in August, it’s a big research gathering for AI, for the continent.
It brings together multitudes of other communities within this to come together, really feel the pulse, still learn, still teach each other, and really challenge each other in terms of the views that we have. It’s not just computing people.
You will find linguists, you will find lawyers, you will find policymakers in this space.
Through that, it’s also inspired many young people to now go in for research careers, go on, build up their work, more developers and engineers to work within there, because now they’re seeing their own supports, like, you know, support systems and really support also all the other grblockroots communities across the continent who work in these spaces.
The Indaba is just one of them. It is just nice because we have this annual gathering. It’s a big tent where, if you come to an Indaba, you will also meet the all like, you know, a lot of the other grblockroots organizations from across the continent, and it’s really been something I’m very proud of, if I look back at this decade being back in South Africa, that we were able to do, yeah.
Lelapa AI is actually a Setswana word that means home.
And the reason we called a company Lelapa AI was we wanted to be a home for the top talent and researchers on the continent, to be able to do research that then impacts people through then the products that come out on the other side, right?
So you want to be a research-driven company.
Our first areas of work are in building natural language processing systems for African languages, specifically right now, especially targeting speech recognition for African languages, and translation systems that then can be used by other organizations.
So we’re targeting other organizations and getting them to say that they can enable, like, you know, new features to engage with their stakeholders, but then engaging with them in their own languages.
So you can think about, you know, our ambition to say, how do we make impact in education, in health, where people can speak, really talk about their ideas or ailments in their language, and not have that lost in translation feeling of trying to explain something and not being able to do that.
So the journey still feels like it’s beginning, even though it’s 10 years later, because there’s now lots of other kind of mountains we have to climb in making sure we can, like, you know, succeed, incubate more, build these tools into some things that really become infrastructure for the continent.
I think a lot of what drives me is that opportunity to invent the future. There’s a lot that you get from looking back, as I said, like right now, maybe having that opportunity to look back a decade and say, like, where were we before, and what do we have now, and what we have then? And you’re like, oh, you know, I was part of that.
I was talking to a colleague of a different university earlier in the week, and we were just talking about large language models, or language like, you know, language models, and saying, “Okay, yeah, but this thing, you know, don’t worry. They’ll have these languages and they’ll start working in about a year or two.”
And I had to try to explain to them, I’m like, “Yeah, yeah. The reason that’s going to be there is because there’s a lot of us working on that.”
Like, you know, it’s on one side, it’s like, people just expect progress, right?
And then on the other side, it’s like, “Oh no, I fit. I’m one of those gears that gets that to really be a reality.”
And it was something that, like, you know, trying to explain to them. And they were not completely getting it at the beginning. I’m like, “No, it is not just a given. It is not an eventuality. People actually have to actively be doing the research and the development for that to become the actuality.”
And it’s been very rewarding seeing that happen and also being able to do it with so many young people.
Right, Africa is a very young country. Lot of young people, and giving as many of them opportunity to be in this space and play and discover and really blockist the rest of us, meaning the whole globe, learn more about ourselves, is something we should strive for.
And if I go back and think about my childhood, and I would have never, ever thought I would be doing what I do today. And that’s the thing that sometimes, you know, people ask, like, “How did you….?” Like, I didn’t know.
It was just this part of, there is some interest, and then, you know, what? How would I get through some of those interests and meeting the mentors I was able to do over the time, where there’s my Master’s supervisors, my PhD advisors.
The like, you know, bosses in different research organizations, the colleagues have right now, co-founders that I have in all those spaces, was more really trusting that ultimately, that means that we keep marching forward bit by bit, and that will then create this place where, if I think about my children right now, who are about three and a half, they will have all these new tools they can engage with that for a lot of people, even within the African continent, would have seemed impossible even a short time ago.
And by also being very communal, you then, in some ways, de-risk you being the only path to those things being resolved.
I think it very arrogant to put it up as an individual contributor. It’s really the community in lots of different ways.
There’s probably somebody right now in the DRC thinking about some way to solve a problem in natural language processing that ultimately will intertwine with some of the stuff we do 18 months from now.
And that’s exciting, right? That we’ve created an environment that allows that to, to really grow by itself.
I think an advice to young researchers is to, yeah, speak about what you’re doing. I know we’re very self-conscious because you think you’re not there yet. Yes, you know, you may not be a Nobel prizewinner, or you don’t think. You always think, but we always like, to me, we’re always a work in progress.
So the thing is, speaking about your interest and what you’re doing allows you to really expand knowledge, because you’re likely going to find somebody else is then going to give you input that you would have not had before.
So there was a lot of times of anxiousness as a young, much younger researcher, where you just like, I don’t think I’m good enough.
And a lot I know that affects people of colour. vIt affects minorities in different spheres who are not for minorities within the science sphere.
It’s still something that we should not take for granted. I know when I go to our top conferences in AI, I’m more the oddity than the norm.
So seek, just seeking out and being comfortable to speak, and then also finding communities like like deep learning, even Indaba, is really good for for you to really have a validation that you’re you know, you can try.
Deborah Daley 15:52
In the next episode, we meet an Iranian environmental scientist who struck up an unlikely partnership with NASA and created, once in a lifetime, opportunities for her minority students in.
