Africa's AI Future: Focus on Local Context Over Models

Mustafa Ehsan, who runs Convly, a research firm, has analyzed the economics of AI, revealing that the cost of running AI workloads has dramatically decreased. For instance, a workload that costs a Nigerian startup approximately $1,750 monthly on a premium American model can now be executed for about $21 using cheaper alternatives.

Ehsan argues that as model access becomes more affordable, the focus should shift to making AI useful in local contexts. Jephte Ioudom, founder of FoubsLabs, contributed to a World Bank Group study in the Republic of Benin, which tested an AI mathematics tutoring platform for pre-service mathematics teachers.

Ioudom believes that generating data from such platforms is crucial for understanding user patterns and shaping AI products. Both experts agree that Africa should prioritize owning the layers around AI rather than striving to develop frontier models, as the costs of training such models are prohibitively high and often lead to rapid obsolescence.