French telecommunications giant Orange announced on Tuesday, November 26, that it is collaborating with OpenAI and Meta to develop custom artificial intelligence models designed to better understand regional African languages.
Orange is working with the mentioned players of the global technology sector on the elaborating artificial intelligence models built on Whisper and Llama AI models, which are openly available systems that can be adapted to meet specific needs. The collaboration, which was publicly announced on Tuesday, focuses on developing machine intelligence models that can understand West African languages that are inaccessible to most conversational systems.
It is worth noting that currently most of the data based on which large companies operating in the artificial intelligence industry train their algorithms on originates in the United States. This means that the mentioned data may not contain important contexts such as language and culture. The relevant issue is particularly sensitive and relevant for such regions as Africa, the Middle East, and Europe.
Steve Jarrett, Orange’s chief AI officer, said during a conversation with media representatives that artificial intelligence models may find it difficult to understand text and voice messages in less widely represented languages. He noted that by using an open machine intelligence model, the user can do fine-tuning while introducing additional information that was not included when the model was first trained.
Orange plans to roll out two artificial intelligence models that incorporate two West African regional languages, Wolof and Pulaar, which are spoken by roughly 16 million people and six million people, respectively, early next year.
As we have reported earlier, KPMG Launches AI Partnership With Google Cloud.
