The intelligence we leave behind | Maja Zavrsnik | TEDxEsei School Barcelona

Key Concepts

• Homogeneous Teams: Groups lacking diversity, leading to biased or exclusionary product development.
• Monosynthetic vs. Polysynthetic Languages: A technical distinction in linguistics where monosynthetic languages (like English) require spaces between words, while polysynthetic languages (like many indigenous languages) combine multiple morphemes into single words, often causing current AI systems to fail.
• Digital Sovereignty: The concept that communities must own the servers and data that house their cultural and linguistic heritage.
• AI Gender Gap: The disparity in representation, where women hold only 22% of AI roles and 15% of executive positions.

---

The Problem: Homogeneity in AI Development

The speaker argues that current AI development is concentrated in a few global hubs (Seattle, New York, Beijing, San Francisco) and driven by homogeneous teams. This lack of diversity results in products that fail to address the needs of the global population. Currently, AI systems are built using only 100 of the world’s 7,000 languages, effectively excluding 6,900 languages from the digital revolution.

Case Studies in Inclusive AI

The speaker highlights three real-world applications where localized, diverse teams solved specific problems that Silicon Valley-centric models ignored:

1. Agricultural Empowerment in Kenya (Kericho County):

   • The Solution: The "Virtual Agronomist" app by ISDA.
   • Methodology: Farmers send crop data via WhatsApp. The AI cross-references this with a cross-continental soil map, ergonomic science, and crop history to provide a personalized fertilizer and management plan in the local language.
   • Impact: The service costs only $1.50 per season. Within months, 78% of farmers adopted it, resulting in a 1.4x to 1.9x increase in crop yields.

2. Sustainable Storage in Burkina Faso:

   • The Problem: 40% of onion crops were failing post-harvest due to high humidity and climate conditions.
   • The Solution: An AI-powered, solar-operated storage system built from eco-recyclable materials designed to withstand local climate fluctuations.

3. Preserving Indigenous Languages (North USA):

   • The Challenge: Michael Running Wolf noted that AI systems are built on "monosynthetic" logic, which fails to process "polysynthetic" indigenous languages.
   • The Methodology: Running Wolf is documenting over 200 endangered languages by empowering youth to record their ancestors' speech.
   • Key Perspective: He emphasizes Digital Sovereignty, stating: "If your grandmother's language lives on a server that you don't own, you haven't been preserved, you've been archived."

The Gender Gap and Automation

The speaker presents alarming statistics regarding gender inequality in the tech sector:

• Representation: Women hold only 22% of AI roles and 15% of executive roles.
• Vulnerability: Women’s jobs are three times more likely to be automated by AI.
• The Paradox: The demographic most likely to suffer the negative consequences of AI bias is the same demographic currently excluded from the rooms where these systems are designed.

Initiative: She AI

In response to these disparities, the speaker introduced "She AI," an organization dedicated to closing the gender gap through accessible AI education. The initiative operates globally—including in Colombia and Ukraine—with the goal of ensuring women gain access to the decision-making rooms where the future of AI is being shaped.

Conclusion

The central argument is that the history of civilization is defined by what is built and, more importantly, by who is allowed to build it. To create a future that serves everyone, we must move beyond the "small rooms" of current AI development. The speaker concludes with a call to action: we must actively expand these rooms, prioritize inclusive design, and ensure that diverse voices have the agency to build the systems that will define our future.