Africa’s AI Path to Health Impact
AI is opening the door to health systems that can learn, adapt and act. Can Africa harness it to leap ahead?
Editor’s note: This essay is available in our print issue, An African Manual for Debugging Empire. Buy the issue here.
A robot, Osadebe, came out to greet me. Osadebe would be performing my procedure and had already been synced to all my records in The Database. Osadebe knew my favourite musicians, my favourite books, and even made jokes that matched my sense of humour. I couldn’t believe it. Osadebe would be something of a nurse and physical therapist to me after performing my procedure. Osadebe had already created customized versions of recommended lifestyle changes for me, accounting for my quirks—like what time I liked to go to sleep and how flavourful I liked my food. Osadebe would also be calling me to make sure I was fine before and after the procedure and providing me some counselling if needed.
—‘The Winged Woman’ by Ebele Mogo
In 2021, in response to a call for speculative fiction on alternative economic futures and technologies by Edgeryders, I wrote the short story ‘The Winged Woman’. It envisioned a future where an AI agent, Osadebe, trained on biomedical data, supported a woman preparing for surgery, offering personalized diagnostics and emotional support. At the time, it seemed like science fiction from a distant future. Yet, only a few years later, that future no longer seems so distant. By 2024, the United States Food and Drug Administration had authorized over 880 AI-powered medical devices, ranging from diagnostic tools and radiology image interpreters to systems that predict deterioration or guide treatment pathways. Algorithmic agents are already helping people plan surgeries, support diagnoses, and translate complex health information, and the promise of surgical robots is very real.
In 2020, I led a team of volunteers in a distributed effort to rapidly translate the World Health Organization’s COVID-19 guidelines into multiple African languages. We onboarded translators, built a workflow, created quality assurance protocols for translation into 18 languages, and disseminated the content in under three weeks. It was a human-powered and agile response to linguistic gaps in the middle of a global emergency. In a few years, the launch of ChatGPT would accelerate the global adoption of Large Language Models. Today, a fair pass at those translations could be generated in minutes, at least for the widely represented languages.
We are at the threshold of agentic AI, models capable of goal-directed behaviour, autonomous initiative, and reasoning in constrained environments. The pace of advancement is disorienting, such that benchmark evaluations that once seemed reliable now lag behind models that evolve week to week. This gap creates challenges in how we measure real-world performance...
This essay features in our print issue, ‘An African Manual for Debugging Empire’. To continue reading this article, Subscribe.
Already a subscriber? Log in.
