As we know Nandan Nilkeni is one of the founders of Infosys and is now the chairman of the company. He is also an architect of Aadhar. Recently at Build with AI Summit October 23, at Bangalore, he asked India not to be a party to building LLMS as they do in the (Silicon) Valley but focus on small language models which are cost effective, and which could be trained easily by using appropriate data. India should build infrastructure, according to Nilkeni, for collecting appropriate data. India could be made the use case capital of AI.

He projected that India would have high quality, accurate inference at very low cost in a year or year and a half. Nilkeni was all praise for Facebook’s Open source LLM which has become a game changer for India. Llama 3.2 launched in September 2024 has multimodal capabilities. It understands text and images at the same time. There are vision versions and separate text-only versions which are light weight.

Open source contributes to Digital Public Infrastructure (DPI). India too is building DPI and would accelerate our march to exploit AI. Indian language models could be made digital public goods. India would not like to stifle innovation by heavy regulation and should adopt thoughtful guardrails.

The right approach, according to Nilkeni, is to use open-source foundation model and fine tune it to deploy it for specific applications across various domains.