Sarvam, an Indian startup, has launched a pioneering 105-billion-parameter foundational large language model (LLM), aimed at enhancing India’s AI landscape. Co-founder Vivek Raghavan highlighted the model’s unique architecture, which focuses on Indian knowledge and context, differentiating it from larger global competitors like ChatGPT and Gemini. While acknowledging that their model isn’t the largest, Sarvam asserts its efficiency in real-world applications, especially in Indic languages. The company prioritizes Indian languages over global trends, boasting superior performance in speech recognition and natural language processing. Their AI models were trained entirely on domestic infrastructure, emphasizing affordability and accessibility. Sarvam aims to integrate AI into smart devices like glasses and feature phones, promoting inclusion and natural language interaction. Addressing challenges like hallucination and cultural bias, they focus on developing robust grounding mechanisms. As they navigate a competitive landscape dominated by free global services, Sarvam seeks to innovate within the specific needs of the Indian market.
Source link
Share
Read more