In this episode of Tech Aways, I sat down with Professor Vukosi Marivate, a computer scientist, AI researcher and one of the key figures behind Africa’s growing artificial intelligence research community. The conversation begins with the origins of the Deep Learning Indaba, which Marivate co-founded in 2017 at a time when AI was not yet receiving the attention it does today. He explains why the rise of deep learning and increasingly capable models convinced a group of African researchers that the continent needed to build its own capacity rather than simply wait for technologies developed elsewhere to arrive. Almost a decade later, the Indaba has grown significantly, with more than 1,000 people attending recent editions and thousands applying. Marivate discusses what that growth says about the appetite for advanced AI research on the continent and, more importantly, the emergence of a new generation of African researchers. The challenge now, he says, is creating enough opportunities for that talent to remain connected to and contribute to the continent. The conversation then turns to one of the biggest structural challenges facing African AI: research and development. Marivate argues that Africa’s ability to shape future technologies is constrained by relatively low R&D investment. Using South Africa as an example, he points to the gap between the country’s own R&D ambitions and actual spending, while arguing that the private sector will also need to play a much larger role in funding research. From there, we explore the importance of African languages in AI. Marivate explains why the underrepresentation of African languages in datasets is not simply a problem of AI systems producing poor translations or responses. It affects whether these systems properly understand the people and environments in which they are being used. He estimates that African data accounts for less than 1% of the datasets currently being used, despite Africa being home to roughly 2,000 of the world’s 7,000 languages. The discussion looks at the work of the Masakhane ecosystem and how researchers and communities are working to improve the representation of African languages. Marivate explains the different roles played by the Masakhane Research Foundation and the Masakhane African Language Hub, as well as the importance of enabling smaller language communities to develop their own capacity rather than relying entirely on large technology companies. I also ask about the gap between African AI research and commercialisation. Marivate argues that the problem is not simply a lack of good research. Public and private institutions are often not structured to procure and test locally developed technologies, meaning promising products can remain stuck in pilot programmes instead of reaching customers. He also discusses the enormous cost of compute for deep-tech startups and draws on his experience as CTO of Lelapa AI to explain what it takes to turn AI research into a commercial technology company. The conversation then moves to talent and the challenge of retaining African AI researchers. Marivate argues that the continent should not be focused on producing a handful of exceptional individuals, but on building a much larger ecosystem around them. He discusses the importance of communities such as Deep Learning Indaba in raising the quality of African AI research and introduces the idea of “brain circulation” rather than simply viewing migration as brain drain. We also unpack the increasingly important question of AI sovereignty. For Marivate, sovereignty is ultimately about having the agency and freedom to choose which technologies Africa builds itself and which it adopts from elsewhere. That requires greater transparency around AI models, including understanding the data and design choices behind supposedly “open” systems. The conversation then expands into the geopolitics of AI and the role of African governments. Marivate discusses data sovereignty, privacy, explainability and the challenge of creating a coordinated African regulatory position. He points to the Malabo Convention as an example of the difficulties of achieving continent-wide alignment on digital policy. The episode closes with a look at the next five to ten years of AI in Africa. Marivate says what excites him most is the growing number of young African researchers moving into senior positions and producing increasingly sophisticated work. He argues that the continent now needs to give these people the opportunity to experiment, build and solve problems locally. For Marivate, Africa’s AI future will ultimately depend on a willingness to believe in what can be built on the continent, support local builders and create the conditions for them to take risks. The conversation covers the evolution of African AI research, Deep Learning Indaba, R&D investment, African-language AI, Masakhane, AI startups, compute, talent, brain circulation, AI sovereignty, regulation and the opportunities that AI could create across sectors including healthcare, agriculture and government. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit techaways.substack.com
