Gates Foundation Launches Coalition to Broaden AI Language Representation

The Bill & Melinda Gates Foundation has established a coalition of 60 organizations, including major AI developers like Anthropic, Google, and the OpenAI Foundation, to enhance the representation of underrepresented languages in artificial intelligence tools. The initiative aims to reach more than 3 billion people over the next five years by coordinating existing efforts to combat global inequality through more inclusive AI.

Mark Suzman, CEO of the Gates Foundation, emphasized the urgency of this work, stating that it must continue "full speed ahead" even amidst calls from some AI companies to slow advanced model development. He argued that governments should focus on regulating AI's impacts on cybersecurity and child development while simultaneously expanding its humanitarian applications to underserved communities.

"Even if AI was frozen right now — which I don’t expect and I am not calling for — we would want to be building these language sets and making them usable with the tools that we have available right now," Suzman told The Associated Press. He highlighted that a lack of representative language data can lead to significant mistranslations, citing an example from the foundation's recent Goalkeepers report where a pregnant woman's statement about her water breaking could be misinterpreted in English.

This new coalition follows the foundation's recent commitment of $1 billion toward AI-focused initiatives aimed at improving health outcomes, educational tools, and agricultural practices globally. The report also pointed out that many AI tools are trained on data scraped from the internet, which is not a representative sample of global languages and cultures.

E.M. Lewis-Jong, CEO of the Mozilla Data Collective, stated that the internet's inherent lack of representation means AI systems trained on it will also lack cultural diversity. Her organization works to help communities upload their own linguistic and cultural data on their terms, rather than having it collected from the web without explicit consent.

Details regarding the coalition's governance are still being finalized, according to Suzman. A secretariat will be established to monitor each signatory's progress, and the Gates Foundation may intervene to help fill identified gaps. Google, a coalition member, is already involved in Project Vaani, an effort to collect over 150,000 hours of audio across districts in India to capture speech data from various dialects. Anthropic, while acknowledging its products "lag in many African languages in particular," is working with the foundation to improve AI's performance in areas like vaccine development and local crop information, recognizing that language inclusivity is critical for realizing AI's global benefits.