Meta’s AI Head Outlines Efforts to Integrate Indian Languages into Llama Model for Enhanced Accessibility

Meta’s AI Head Outlines Efforts to Integrate Indian Languages into Llama Model for Enhanced Accessibility

Meta AI, which now serves over 500 million active users worldwide, is advancing efforts to support Indian languages in its AI ecosystem, with its open-source large language model (LLM), Llama, positioned at the forefront. At Meta's first Build with AI Summit in India, Manohar Paluri, Meta's Vice President of AI, shared plans to incorporate high-quality Indian language tokens into Llama, ensuring its accessibility to the millions of Indians using Meta’s platforms like WhatsApp, Facebook, and Instagram.

Paluri emphasized the importance of expanding language support beyond English, noting that while Hindi is currently available, Meta’s roadmap includes further Indian language integration. He also addressed the complexity of adding languages, explaining that it involves fine-tuning data and aligning model inputs to maintain safety and reliability across multiple languages. Through collaborations with local companies such as Sarvam, Meta has developed Indic Llama, a specialized model for Indian languages, highlighting the adaptability of Meta’s open-source model, which developers can customize independently.

Yann LeCun, Meta’s VP and Chief AI Scientist, reinforced India’s critical role in shaping the global AI landscape, underscoring the significance of open AI innovation.