It’s a sunny winter’s day in Joburg when I meet with Vukosi Marivate—a multi-hyphenate academic, entrepreneur and African Artificial Intelligence (AI) advocate. When he finally enters the pre-lunch emptiness of the steak restaurant he chose for this face-to-face (I have a flight to catch, so every minute counts, and he’s running a little late) his infectious energy fills the room and all is immediately forgiven. 

“It’s a microcontroller,” says the 38-year-old Fulbright Scholar and ABSA UP Chair of Data Science at the University of Pretoria (where he also leads the Data Science for Social Impact research group). He’s holding aloft the device whose purchase was the reason for his delay. “I could get one of those fancy Dyson ones,” he says, referring to a type of air quality monitor—the one piece of automation missing from his otherwise fully automated home, “but it’s more fun building my own. Besides, those are way overpriced,” smiles the CTO and co-founder of Lelapa AI, a startup “focused on AI for Africans by Africans”.

In a February 2023 article about Lelapa by tech publication Wired, Marivate was more emphatic: “We don’t want to be left behind,” he’s quoted as saying. “In technological revolutions, those left behind pay a big price as a society.” The burgeoning field of AI in which Marivate works is one of many from which the African continent’s 1.3 billion people desperately need to ensure they are not entirely waylaid. 

Marivate—who trained as an electrical engineer before winning a Fulbright Scholarship to Rutgers University in New Jersey, which led to his PhD in Machine Learning—is not unfamiliar with the sense of being on the outside looking in. 

Born in South Africa during apartheid, Marivate’s family history is all too common to Black people in the country. His maternal and paternal grandparents were born and raised in Ga-Rankuwa, a farming area outside of Tshwane (formerly Pretoria) before their families were forcibly removed from their farms to make way for white farmers. Both families landed in a township in Tshwane, where Marivate was born and raised. 

After graduating in 2012, he interned at Google and did stints at a few other tech companies in the US before returning in 2015 to stay in his family’s bustling multi-generational household. Working at a government lab to develop a biometric system for use across government services, it was during this time that his interest in technology and its place in society took root.

“When I think about technology, I think of how I can work to ensure that data science, or AI, can work for people in that area and others like it, too. Otherwise, I’m investing and contributing to stuff that benefits other people’s neighbourhoods and not my own.” 

The nearly empty restaurant of just over an hour ago is now bursting. The tables are all busy, the temperature a little higher, and the voices a little louder. As our food arrives—two medium-rare steaks with mustard, truffle chips and a salad to share—I hear the words “ChatGPT” from a neighbouring table. It’s inescapable.

“Is it all hype?” I ask Marivate of the ChatGPT/AI craze that began consuming the world’s attention in 2023. “Or are we witnessing something truly extraordinary?”

“Yes, there is hype… but the current state of AI is too significant to ignore,” Marivate says sharply. “Some giant leaps that cannot be ignored have been made,” he adds, “and whoever controls this will control so much of the international economy in the future. The question is, is the African continent ready? We will not have any say in it, let alone control, if we don’t build up a critical mass of people who can build their own version of these technologies, for themselves, on the continent.”

Believing strongly that Africa must build and develop its own ecosystems of highly skilled innovators committed to building custom-made technologies relevant to very particular African needs, Marivate is also one of the founders of the Deep Learning Indaba, a gathering of researchers in the fields of machine learning (ML), AI, and natural language processing (NLP). Founded in 2017, the conference now attracts participants from over 40 African countries. During the pandemic, he led a team of experts and volunteers to create the Covid-19 Data Repository for South Africa, which became the go-to source for researchers, policymakers and others seeking data on all aspects of the pandemic. 

“I can work with people in the health sector, for example: try to understand their problems, see how data flows, and then come in, and challenge them to really look at the challenges of my own city and the neighbourhood I grew up in, and the one I live in now—which is much wealthier—with the same level of interest. What starts as a conversation about health soon becomes [one] about how people access electricity; how they get their water, and the quality of that water; how they get to and from work, and how they live. I’m interested in how technology can be informed by those stories.”

“You are an urbanist,” I chime in. 

“I’m interested in everything we do as a collective to keep the city going. The city and life in it are very connected to my work,” he continues. “I’m inspired by the city and how the work I do feeds back into the city and into people’s lives. Otherwise, what does it matter?” he says. “We have now built a kind of tradition in our lab at the university, which can sometimes feel like being inside a fortress, that says ‘let’s get out into the city and walk it’. So, if we have a meeting outside of campus—at a government department or any of our local partners or collaborators’ offices a few blocks away—we walk there or sometimes take the bus.” 

This is an unusual act in South African cities, which are famously hostile to pedestrians and very private car-friendly, especially for wealthier residents and among professionals like Marivate. 

“What is an African city?” Marivate asks me unexpectedly. 

I’m a little taken aback. I start going through my now-standard semantic contortionist spiel about how I prefer to talk about “cities in Africa” rather than “African cities” and quickly catch myself. “Perhaps it’s a useful phrase in some cases, but it’s too broad a generalisation,” I say instead.

This plays straight into the point Marivate wants to make. 

“When I’m in Nairobi, my experiences there are very different from my experience of Kigali. It’s the same as when one is in Lagos versus Abuja. Durban is not like Joburg, Pretoria or Cape Town. In some ways, these [cities] could be in different countries. As we are designing things, I’m trying to think about that. So, rather than talk about AI in Africa, we are instead trying to focus on the context of where you are on the continent rather than a definitive catch-all that ignores the massive differences in place and experience. It’s a much closer reading of the ground rather than a flattening. We are engaging with that complexity,” he says. 

The universal absence of African languages in computer systems that play a critical role in all aspects of life in cities across the continent exemplifies the complexity Marivate is interested in. 

Despite African languages representing almost a third of the just over 7,000 known spoken languages globally, that they are not factored into how we think of computers and computing is a universal truth. It’s a good indicator of the problem Marivate and his collaborators want to face head on, and he believes that AI can be a powerful tool in addressing this issue. 

One of the multiple projects Marivate has played a key role in is Masakhane, an open source NLP participatory research community focused on African indigenous languages. The project’s aim is two-fold: One, to have the web available in as many of the 2,000 African languages as possible and to democratise access to knowledge and preserve these languages, many of which are disappearing. 

The other aim is to develop and support a community of researchers in this field. Ultimately, users of any web or computer system should be able to access the knowledge and services they require on the web in their native language if they prefer, Marivate argues. This “plays a crucial role for information accessibility and communication worldwide,” according to a paper which was co-authored by Marivate and published by the Association for Computational Linguistics in 2020. It’s an admirable dream, but progress is slow. As of 2020, about 1,000 participants from 30 countries across the continent have published in 38 languages.

The reality is, given the significant challenges across the continent—health, poverty, housing, all these big, weighty, seemingly intractable problems—it’s easy to keep on saying now is not the time to focus on this kind of effort. There are many reasons not to: It’s too expensive; the resources could be better spent elsewhere on more urgent needs; it takes time and skills that might not necessarily be there. But, Marivate says, “the cost of doing it becomes more and more the longer we wait. Once you start interacting with language as more than symbols, as more than just sounds, but instead as communities, as people, as history, as culture, you realise no one is fully empowered to take complete control of their fate until these things matter too.”

I’m convinced. By now, we have been speaking for almost two hours. The table has been long cleared, and the restaurant staff are slowly starting the setup for the dinner. There are a few other occupied tables with people speaking in hushed tones—in my conspiratorial mind, they are all whispering about ChatGPT. There aren’t any subtle signals from the waiters still clearing up around us that we are in the way and need to move on soon. A dessert menu is offered, we both quickly decline and order some coffee instead. We keep talking.

In the half-hour that follows, Marivate explains how all the organisations, communities, activities, academic research, and research and development he has been involved in are part of an effort to contribute towards the building of that brigade of thinkers, coders, researchers, translators and developers who will take on the task of building an alternative AI system that not only addresses the needs of the continent and its people in all their diversity, but also acts as a buffer against an otherwise assured dominance of a Western variant that is not fit for purpose. The technical part is one thing, he says. Engaging with the political and organising are the other components.

“At this point, I feel like the scaffolding is there on which we can build,” he says. “We’ve got the research going on, we have built some quite robust communities that have developed a momentum of their own and are starting to do some remarkable things that we didn’t think possible in such a short amount of time. We have a steady, growing pipeline of people with the political and development skills we will need in the field. We have Master’s and PhD students who will take this work to another level. Now, the new challenge is to build viable enterprises on the continent, rather than leaving skilled people with moving abroad, working locally in the finance sector, or staying in academia as the only viable options.” 

There are numerous players across Africa looking at problems on the continent as a basis of AI research, and working to establish enterprises that can be financially sustainable, well-resourced, connected to economic activity, and, most of all, genuinely transformative. “We are not on our own. A larger ecosystem is emerging and it’s exciting to be part of that,” he says.

What about regulation? Is there an effort to address some of the significant gaps in policy around issues like user privacy? Don’t we need some kind of continent-wide regulatory blueprint, something akin to the EU’s general data protection regulation (GDPR), to ensure a more robust set of protections that can act as a guardrail against exploitation not only from global companies but also new enterprises like Lelapa that are beginning to emerge? Surely, they can’t be expected to self-police as they work in different parts of the continent, each country with its own regulations and varying capacity to enforce them? 

Quick to the draw, Marivate has a ready answer. 

“Have you heard of The Malabo Convention?” he asks. I had not. “It was signed in 2014. And it is exactly that: it is the African GDPR. It took almost a decade before it could be ratified at the African Union when Mauritania became the 15th country to do so as required under its articles. It’s the only binding regional treaty on data protection outside Europe.” 

I wonder aloud whether it has any relevance since only 15 countries from a union of 55 are signatories. The convention is the equivalent of a brand-new car leaving the dealership already aged, with 10-year-old technology under the hood. After all, a decade is a lifetime in tech. 

Perhaps not, but it’s a start, Marivate says. 

We realise we need to wrap up, and promise to continue the conversation. I have a flight to catch, and Marivate needs to pop into the Lelapa office, which is nearby. 

What’s next? I ask him as we both prepare to leave. Marivate quietly tells me about Vulavula, Lelapa’s “NLP-as-a-service” platform that the startup is preparing to launch “soon” as their first commercial product. If successful, Vulavula promises to test the context-first approach that Marivate professes is at the core of his sprawling ambitions that cut across academia, the non-profit/activist sector and commercial enterprises. That is, the creation of a series of networked, homegrown algorithms that could transform the technology industry’s relationship to Africa and its people.

Whether the city, national government, and industry bigwigs across the continent are listening to the call Marivate and his growing communities of collaborators are making on the urgency of undertaking such a task remains to be seen.