(Bloomberg) -- French telecommunications firm Orange SA has partnered with OpenAI and Meta Platforms Inc. to begin training artificial intelligence programs on African languages, addressing a shortage of models for the continent’s thousands of dialects.
The project will start during the first half of next year, the companies said in a statement on Tuesday. Initially, the companies will focus on two West African languages, Wolof and Pulaar, which are spoken by 22 million people in the region. Orange will work with Meta and OpenAI to fine-tune their models and test them with native speakers.
The French phone carrier ultimately wants to expand the project to more AI technology companies to build large language models that can help it communicate with customers across its 18 markets in the Middle East and Africa. Africa is home to about a third of the world’s languages. That scope, together with limited funding and patchy infrastructure, means few of those dialects have been included in AI training models.
Orange said it will use public cloud capacity in Europe and Africa to help train the models, as well as its own data centers. It will train the models using internal, anonymized data from customer interactions with its contact centers and a public dataset of languages from Meta, the company added.
Orange will use the local language model to engage with its customers, and plans to make it available for free public health and education services, as well as local companies and business people, Steve Jarrett, Orange’s chief AI officer, said. The project will target training on other sub-Saharan languages including Lingala, Swahili and Bambara next year.
“We see the initiative as a blueprint for how AI can be used to benefit those currently excluded,” Jarrett said in an email. “Orange’s vision is to make AI and other related advances accessible to all, including illiterate populations, who are currently unable to benefit from the potential of artificial intelligence.”
OpenAI will also give Orange early access to its AI models so the carrier can develop use cases, such as AI-based voice interactions with its customers, as well as provide additional data-processing and hosting capacity in European data centers.
(Updates with context on partnership and details about training data)
©2024 Bloomberg L.P.