INDIA'S DEEPSEEK FOR ITS OWN AI

FROM LARGE LANGUAGE MODELS CATERING TO ITS LINGUISTIC DIVERSITY TO SOLUTION-DRIVEN AI AGENTS THAT CAN BE DEPLOYED ACROSS SECTORS, THE NATION LOOKS POISED TO RIDE THE NEXT-GEN INNOVATION WAVE BACKED BY THE GOVERNMENT’S RS 10,300 CRORE ‘INDIA AI MISSION’

- Ajay Sukumaran

In the past five months, both Jensen Huang, CEO of the American computer chipmaker Nvidia, and Sam Altman, who heads OpenAI, have visited India, emphasising the country’s growing significance in the tech world, including in the field of Artificial Intelligence (AI). Nvidia reports a tenfold increase in the deployment of its graphics processing units (GPUs)—specialised chips used to train AI models—in India over 18 months, while for OpenAI, India has emerged as the second-largest market, with users tripling in the past year.

During his February 5 visit, Altman said that India should be “one of the leaders of the AI revolution”, an endorsement in stark contrast to his 2023 stance that the country “should not even try” to build foundational AI models. What has changed in between, of course, is the end-January release of industry-disruptive DeepSeek-R1, a large language model (LLM) developed by a Chinese startup based in Hangzhou.

Foundational LLMs are deep learning systems trained on gigantic amounts of data to perform tasks such as natural language processing, question-answering and image classification. They form the foundation on which conversational agents and other Generative AI (or GenAI) products are built. If prohibitive costs were what prompted some experts to advise caution in committing billions of dollars to the development of LLMs such as OpenAI’s GPT-4, the arrival of DeepSeek has been a game-changer because the Chinese startup claims to have developed its eponymous model at a fraction of the cost incurred by its western forerunners.