The Gates Foundation plans to spend at least $1 billion over the next two years to widen access to artificial intelligence in healthcare, education and agriculture. Bill Gates argues that decisions made during the next 12 to 18 months will determine whether the technology narrows or deepens inequality.
Language coverage is a central target. Gates says more than 90% of the data used to train early large language models came from English sources. He contrasts speech-recognition error rates below 6% in English with failure rates above 60% for Yoruba, illustrating how uneven performance can become a practical safety issue.
The foundation points to existing pilots: an assistant at Kenya’s Penda Health network was associated with a 16-percentage-point rise in diagnostic accuracy; students using Gemini Guided Learning in Sierra Leone recorded gains of up to 1.7 years in an eight-week pilot; and India’s MahaVISTAAR farming service reaches more than 740,000 people at a reported government cost below 18 cents per person.
Those examples are project reports rather than proof that every deployment will transfer successfully. The funding will also support the inclusion of non-English sources in datasets. The commitment forms part of the foundation’s broader plan to spend roughly $200 billion before closing in 2045.