Ubuntech: A New Framework for Ethical AI

In her compelling talk, tech policy visionary Nanjira Sambuli challenges the current trajectory of artificial intelligence development by drawing on the ancient African philosophy of Ubuntu. By contrasting the destructive nature of "metaphorical elephants"—representing global powers and corporations—with the resilience of the "grass"—representing the people and ecologies affected by AI—Sambuli offers a path toward a more inclusive and ethical technological future.

The Philosophy of Ubuntu and the Metaphor of the Savannah

Sambuli uses the Yoruba proverb, "When elephants fight, it's the grass that suffers," to illustrate the current geopolitical struggle for dominance in the AI landscape. As global powers compete for resources to fuel the "intelligent age," the perspectives of marginalized communities are often "drowned out by the din of the elephants."

To counter this, she introduces Ubuntu (or Utu), which emphasizes the concept: "I am because you are." This philosophy serves as the foundation for her vision, reminding us that technology should be developed for the "collective well-being of society" rather than solely for corporate dominance. She argues that we must shift our focus to the "grass," recognizing that we are all interconnected.

Data Justice and Ubuntech

Sambuli critiques the prevailing narrative that "data is the new oil," which treats human experiences as mere raw materials for extraction. Through the lens of Ubuntu, she proposes Ubuntech, a framework that conceptualizes data as representative of "lives, cultures, and communities."

Key aspects of this approach include:

• Data Justice: Ensuring that historically marginalized groups, such as rural African women, are visible and protected within AI systems.
• Meaningful Participation: Moving beyond tokenism to ensure informed consent, self-determination, and community ownership of data sets.

Innovation Through Efficiency: Small Language Models

Challenging the conventional wisdom that only massive language models are viable, Sambuli highlights how African practitioners are creating "small-language models" that are "small, but mighty."

She cites the example of Lelapa AI’s Incuba model, which was inspired by the dung beetle. Despite being trained on significantly fewer parameters, it outperforms larger models in sentiment analysis. This approach proves that "low-resourceness of languages is not a data problem but a societal one," which can be solved through collaborative, grassroots efforts like Masahane. This community-driven initiative uses a non-traditional authorship model to include all contributors, successfully expanding NLP research to over 38 African languages.

Ancestral Intelligence and Future Coexistence

Sambuli concludes by repositioning the role of the "elephants." She suggests that if global powers shift from a mindset of dominance to one of "relationality" and "coexistence," they can transition from being destroyers to becoming "ecosystem engineers." Just as elephants in the savanna disperse seeds and create space for smaller species, powerful entities can use their influence to sustain a healthy environment for all.

By integrating the "indigenous wisdom of our ancestors" into modern technology, we can create AI that is powered by "ancestral intelligence." Sambuli’s closing plea is a powerful call to action: we must "take heed of the grass beneath our feet," as our collective future depends on our ability to build AI that is developed by, in, and for the benefit of all humanity.