While artificial intelligence (AI) continues to inflame the West, pitting doomsayers against promoters of "techno-solutionism," Africa is not lagging behind. But the tone is different. "The great threat is colonization by AI," warned Senegalese professor Seydina Ndiaye at the end of November. Ndiaye is one of 38 experts selected by the UN to form a think tank on artificial intelligence.
Interviewed by the UN Info website, Ndiaye was concerned about the implications of a profoundly unbalanced situation. The bulk of African data today is feeding into foreign multinationals, at the risk of leaving "no room for the creation of local solutions." Put simply, cutting-edge technologies could perpetuate the extractive approach prevailing even before the colonial era – namely, exploitation of the African continent's raw resources for the benefit of other markets.
This feeling of dispossession is reinforced by the idea that AI development is taking place on the backs of an ignored workforce, living in the countries of the global South. From their computers in Kenya, Nigeria or Madagascar, these vulnerable workers are being tasked with moderating content and "training" machines through repetitive tasks, on behalf of the big tech companies.
"Kenyans and many other Africans have helped make ChatGPT the phenomenon it is today. (...) In fact, they're the ones making Sam Altman [CEO of OpenAI, creator of ChatGPT] and Mark Zuckerberg [CEO of Meta] rich, because without these Africans, their platforms would be unusable. But I bet the people of Africa don't even cross the minds of Altman and his colleagues," said Kenyan Mozilla Foundation researcher Odanga Madung in an op-ed published by The Guardian.
Clearly, the development of Africa-made AI is facing serious handicaps, from a lack of skills – most African specialists are located outside the continent – to a lack of infrastructure, in a sector where computing power and connectivity play a decisive role. All this exists in an Africa-wide context of chronic underfunding for science. No country is meeting the 2006 commitment, under the aegis of the African Union, to devote 1% of its gross domestic product to research.
Improving agricultural yields
Some are also upset that most machine learning algorithms are trained on datasets collected outside the continent. One example is genetics: 95% of existing data comes from European genomes. This uniquely limits the usefulness of AI for disease detection in Africa.
The priority, however, is developing an artificial intelligence aligned with the continent's challenges. A so-called "axiological" [integrating an understanding of values and their role in research] AI, according to one of the leading African figures in the sector, Senegalese researcher and former Google employee Moustapha Cissé. It's all the more a priority given that needs in Africa are legion. It's not so much about driving autonomous vehicles as improving agricultural yields, better managing water resources or compensating for the shortcomings of healthcare systems.
You have 19.5% of this article left to read. The rest is for subscribers only.