French telecom giant Orange has announced a collaboration with OpenAI and Meta to create artificial intelligence (AI) models that cater specifically to regional African languages. This initiative aims to address the gap in AI systems’ understanding of lesser-represented languages, starting with Wolof and Pulaar, spoken by millions across West Africa.
Orange plans to leverage OpenAI’s Whisper and Meta’s Llama open-source models, fine-tuning them to recognize and process these languages. According to Steve Jarrett, Orange’s chief AI officer, global AI systems often fail to understand African text and voice communications due to the lack of culturally and linguistically relevant training data.
“By fine-tuning open models, we introduce additional information, enabling AI systems to understand West African regional languages that are currently ignored by mainstream AI,” Jarrett explained.
Rollout and Focus on West Africa
The first phase of this initiative will integrate Wolof and Pulaar into Orange’s AI models by early 2025. Wolof is predominantly spoken in Senegal, the Gambia, and southern Mauritania, while Pulaar is common in Senegal. The project will expand to cover all 18 West African countries where Orange operates.
Orange will provide these customized AI models under a free license for non-commercial applications like public health and education. This approach aligns with the company’s commitment to enhancing accessibility and addressing unmet needs in the region.
Sovereign AI and Localization
The project reflects a growing global trend toward “sovereign AI,” emphasizing localized data processing and infrastructure. By hosting OpenAI’s systems in European data centers, Orange seeks to ensure secure data handling while accessing advanced AI capabilities. The company’s initiative also promotes cultural and linguistic representation, helping to build systems that resonate with local contexts.
Jarrett stressed the importance of using AI responsibly, balancing innovation with environmental concerns. The models will focus on targeted applications, avoiding the extensive energy demands of massive language models when unnecessary.
Broader Applications
Beyond language processing, Orange is integrating AI into its core business operations. The company uses AI to optimize its mobile networks, identifying and predicting failures before customers are affected. Additionally, AI-driven voice systems are being developed to improve customer service across its contact centers.
This partnership represents a step toward bridging the technological gap for underrepresented regions, fostering greater inclusivity in the global AI landscape.