Orange will work with African languages using OpenAI’s most recent models
Orange (ORAN.PA), a French mobile provider, announced on Tuesday that it will employ the most recent AI models from OpenAI with African languages.
Researchers at Cornell University in the United States in a report published in the journal Nature claim that the advantages of AI models have mainly eschewed African languages, which number over 2,000, because of issues like a lack of data and restricted computer resources.
In order to gain access to OpenAI’s pre-release AI models and improve its large language models for translating regional African languages, Orange, a telecom provider in 18 African nations, inked a partnership with the company last year.
It claimed that while OpenAI’s Whisper voice model was used to begin working with African languages this year, the new models can expand this work to much more intricate applications.
The trained parameters, or weights, of OpenAI’s initial open-weight models are publicly available and allow developers like Orange to modify the models for certain tasks without needing the original training data.
Orange intends to use its gathered samples of regional African languages to refine the models before implementing them locally.
Steve Jarrett, Orange’s Chief AI Officer, told Reuters, “We intend to give the improved models to local governments and public authorities for free.”
“We see this initiative as a blueprint for how AI can help bridge the digital divide: by collaborating with local startups and communities, Orange and OpenAI hope to catalyze an ecosystem where African languages are first-class citizens in the AI realm,” Jarrett stated.
