Uganda’s Homegrown AI

When it comes to machine learning and artificial intelligence, one size does not fit all.

Photo by Getty Images/mtcurado.

No single label (except multitalented, perhaps) can fully describe Daniel Mutembesa. He’s a poet, musician, DJ, semi-professional dancer, national Karate champion, and, of course, an engineer—all despite being only twenty-seven years old. When I met him front of the AI lab at Makerere we became fast friends and immediately fell into a spirited conversation about the cultures, traditions, and politics of Uganda. I was fascinated to meet someone in the lab who saw his work in computing as perfectly aligned with his interests in supporting his countrymen.

Mutembesa took me off campus to different and interesting parts of the city in his tiny Korean car. He taught me the indigenous names for Uganda’s national parks. By using these names he hoped to revive interest in the histories and traditions of his country. Part of the incredible diversity of Uganda comes from its more than forty indigenous languages. They are spoken in local communities without a national language to lean on for universal communication, except for the English used primarily in metropolitan areas.

Like nearly everyone in Uganda, Mutembesa comes from its tribal regions: the Ankole and Tooro parts of the southwest of the country (the latter borders the Democratic Republic of Congo). Learning about his nation opened my eyes to a sad trend of historic and cultural erasure. Mainstream entertainment produced for and accessed by Ugandan audiences often adheres to the Western brand, as when DJs fake American accents on the radio and African TV programs mimic American Idol and Survivor. Western webpages dominate the internet. Mutembesa sees this as an invasion of national and cultural consciousness by Western norms and products.9 I asked him how the AI lab sees itself relative to these cultural and global forces. Is computing simply seen as a vehicle of “being modern” (aka imitating the West) or can it revive and support the identities and values of the culture’s richly diverse traditions and communities?

Mutembesa passionately believes in the latter. He is designing AI systems that learn from Ugandan farmers’ decisions about harvesting crops. He has built a machine-learning model that considers the way farmers make choices about using treatments that agricultural scientists suggest. Potential problems arise, he tells me, when applying a model developed from afar to the Ugandan context. But an alternate approach may come from understanding that the strength of game theory lies in the relatively few assumptions upon which it is built.

Game theory considers decisions made to either cooperate or compete with participants in social settings, and the factors that shape these choices. The lab’s model allows the Makerere team to learn the unexpected. For example, they found that the National Agricultural Institute, a partner of the AI lab in Mutembesa’s project, has great sway with farmers—far more, in fact, then the lab would have in paying them to participate. Praise matters more than money; knowing this important cultural value, the team collaborated with farmers and applied technology to maximize yields and harvests.

Facebook Instagram Twitter