Community-centered innovation
Masakhane, Lelapa, and Jacaranda Health’s language models
In line with our 2024-25 organizational strategy, we looked for examples of AI that strengthen social connections and preserve local knowledge. While there are limited AI initiatives that meet these criteria, we have come across some projects that stand out for their efforts to support African languages and facilitate local knowledge sharing.
One example is Lelapa, an African AI research lab, and another example is Masakhane, a community-led project. Both these examples exemplify community-centered innovation through the various various language models they have developed for a variety of African languages.
Lelapa’s InkubaLM (which translates to dung beetle language model), is a small language model designed for low-resource African languages. The language model supports languages including IsiZulu, Yoruba, and Swahili, underscoring the importance of language justice in AI.
Masakhane’s work on creating machine translation tools for over 30 African languages, demonstrates a more grassroots-based approach to language preservation, whereby African developers and researchers lead efforts to provide language tools for their own communities. Both of these initiatives demonstrate how AI tools can facilitate community connections within diverse language ecosystems.
A third example we came across is UlizaLlama, a language model developed by Jacaranda Health that was built to respond to pregnancy-related questions in Swahili. Tailored specifically to Kenyan mothers, UlizaLlama’s design shows how AI tools can adapt to the health service provision needs of local communities.
Last updated