MTN Pledges Support for Research into African Languages for AI
MTN Group has committed to partnering in the collection and development of datasets for African languages, in response to Nigeria’s national call to ensure the continent is not excluded from the global AI revolution. This work is anchored in building inclusive large language models (LLMs) that reflect Africa’s linguistic diversity.
Quick Insight: Nigeria’s “N-ATLAS” initiative is already underway — an open-source multilingual LLM designed to catalogue and generate Nigeria’s many languages. MTN’s backing could accelerate efforts to expand dataset reach and include many under-represented tongues across Africa.
1. What MTN Agreed to Do
• MTN accepted the challenge to mobilise resources for funding academic research into African language datasets.
• The company supports Nigeria’s Nigerian Atlas for Languages & AI at Scale (N-ATLAS), a project launched to understand, digitise, and preserve Nigeria’s over 500 languages via open-source frameworks.
• MTN Group President & CEO Ralph Mupita committed during “The Y’ello Chair Vodcast” that MTN would take up this initiative and help avoid Africa becoming a “digital underclass.”
• The initiative is intended not only for Nigeria, but designed as a framework that other African countries can adopt.
2. Why This Matters
• There are 1.5 billion people in Africa, speaking over 2,000 languages; most are very poorly represented in today’s AI models. Without datasets and research, many could be excluded from AI services in education, health, commerce, governance.
• Data-driven AI requires representative data — speech, text, usage patterns — for each language. Gaps in data mean models could misinterpret, misrepresent, or neglect particular languages.
• Projects like N-ATLAS help preserve endangered languages, provide inclusion, and ensure local relevance of AI systems.
• Bridging this gap can enable new opportunities: local content creation, improved user experiences in vernaculars, and better civic engagement via technology.
3. Challenges & What to Watch For
• Collecting high-quality data in many African languages is difficult: some languages are undocumented, or spoken only in oral form, with few written resources.
• Ensuring ethical data handling (consent, privacy) across diverse linguistic communities is essential.
• Funding and resourcing: academic institutions, governments, and private sector need well-structured incentives and backing.
• Technical issues: model bias, dialectal variation, transcription challenges; representation of tone and idiom.
• Building infrastructure: computational resources, hosting, storage and computing power for training multilingual/multimodal models.
What to Expect Next
• MTN will likely collaborate with academic institutions, researchers, tech companies to build out datasets.
• Expansion of N-ATLAS framework beyond Nigeria to neighboring African countries.
• Potentially, new AI tools or products from MTN that support local languages — in voice, messaging, content generation.
• Government policy or regulation to standardize data collection, open access, and ensure fair participation from communities.
• Monitoring for how ML/LLM models trained on these data perform, and whether inclusion improves real-user outcomes (education, health, local services).
