The proposed facilities, called "pico centres", would use frugal engineering methods to reduce electricity consumption and resource requirements. The study will examine their potential for AI inference applications in agriculture, space, manufacturing, healthcare and law. While large data centres are required to train large language models, smaller facilities could support AI applications closer to end users.

AI Tutors for School Students

Bodhan AI, a Centre of Excellence for AI in Education supported by the Union Ministry of Education and incubated at IIT Madras, is studying the feasibility of resource-efficient AI data centres. One potential application is personalised AI tutoring for students across India's school system, which has around 25 crore students.

The initiative aims to explore how AI-powered learning tools could provide more personalised educational support while using fewer computing and energy resources.

Promoting Responsible AI

The institute's Centre for Responsible AI (CeRAI), in partnership with The Better India, will produce 500 videos showcasing practical examples of responsible and effective AI use. The initiative aims to improve public understanding of how AI can be applied to real-world challenges.

IIT Madras also announced plans to develop Trusted AI Commons to strengthen accountability and responsibility in AI systems.

New AI Evaluation Platform

Another initiative unveiled at the AI Measurement and Governance industry conclave was a software platform for evaluating AI tools using more than 50 metrics and 100 sub-metrics across multiple languages. The platform is intended to help assess the performance and reliability of AI systems.

The institute also released a white paper examining employment trends and the potential impact of AI on jobs.

Through these initiatives, IIT Madras is focusing on cost-effective AI infrastructure, responsible technology development and evaluation methods suited to India's needs.