Context:
The India Artificial Intelligence (AI) Mission has signed an MoU with the Parliament to access its data for training indigenous large language models (LLMs). Other datasets from Doordarshan and All India Radio will be used.
Computational Infrastructure
- A common compute facility under the India AI Mission is made available with 14,000 GPUs.
- India aims to create domestic facilities for the manufacture and deployment of GPUs in the next 3–5 years.
- The government is assessing two competing instruction set architectures for indigenous GPU development, the decision will ultimately rest in the hands of experts.
Focus on Indigenous and Proprietary Tech
- The Minister emphasized the need for India’s own LLMs and expressed concerns that “open source today may not remain open source tomorrow,” citing examples from the world stage such as OpenAI.
- Priority will be given to the development of proprietary technology and AI models so as to realize self reliance.
Collaboration and Talent Development
- The government is planning to collaborate with industry, professors, and startups in developing indigenous LLMs.
- Following the success of setting up 100 labs for 5G training in universities, a similar approach will be taken for AI skill development.
India U.S. Tech Partnership
- Despite trade tensions and export controls, Mr. Vaishnaw highlighted the India U.S. technology partnership based on trust and mutual respect for IP rights.
- He emphasized the decade long co development and co creation efforts as a strength for both countries in the AI race.
Strategic Vision
- The overall vision should be such that India is not merely a consumer but is also a producer of cutting edge AI technology.
- Building indigenous models along with compute infrastructure will be crucial for maintaining strategic autonomy and global competitiveness.
