Broadening the benefits of AI, starting with voice

Today, the OpenAI Foundation joined the Gates Foundation and partners around the world in a shared commitment⁠ to ensure AI works across the world’s languages.

Accessing the benefits of AI should not be restricted by which language a person speaks. Yet for billions of people, particularly those whose languages have long been underserved by technology, today’s AI models remain far less capable and useful.

To help ensure that the benefits of AI are broadly shared, we plan to make enhancing model performance in low-resource languages a significant focus of our work. We will begin with data and infrastructure for voice AI, where gaps in performance and available data are especially large, focusing on languages that are not likely to attract sufficient commercial investment.

For many people, speech is the most natural and accessible way to use technology. Capable voice models could help our grantees and many other organizations serve far more communities. A patient calling a health service could describe their symptoms, answer follow-up questions in their own language, and be connected to the right care. A farmer asking a question aloud in her own language could receive planting advice tailored to her plot and crop.

Asif Saleh⁠, the CEO of BRAC⁠, the world’s largest NGO, says that “when [BRAC] set out to build an AI scribe for our health workers in Bangladesh, the limits of the models in Bangla stopped us. For the communities we serve, voice is the only practical way into this technology, so getting it right in languages like ours will decide whether AI is genuinely useful to the 145 million people we reach, or makes very little difference to them.”

Ensuring AGI benefits all of humanity requires making meaningful progress on this challenge. We are proud to be part of the coalition of organizations, countries, and leaders committed to closing this gap and look forward to sharing more as we develop this work with our partners.