Eight projects selected as part of IndiaAI mission

As part of the Rs 10,738 crore IndiaAI Mission, the Governemnt has selected eight projects intending to drive ethical development of artificial intelligence in India. These projects were chosen from over 2,000 proposals submitted in response to an Expression of Interest (EoI) issued under the "Safe and Trusted AI" pillar of the mission.

This vertical of the government's AI initiative aims to address critical themes such as bias mitigation, privacy, and AI governance. Other pillars include building a high-end computing ecosystem with over 10,000 GPUs, developing cutting-edge AI models for India and so on.

In March, the government greenlit the ₹10,732-crore IndiaAI Mission to develop computing infrastructure in the country, establish AI centers for multi-modal LLMs, and acquire 10,000 graphics processing units, among other initiatives.

The Indian Institute of Technology (IIT) Jodhpur will concentrate on "machine unlearning" within generative foundation models. This technique enables models to "forget" or eliminate specific training data points without requiring a complete retraining from the beginning.

IIT Roorkee will focus on "creating synthetic data to reduce bias in datasets," while the National Institute of Technology (NIT) Raipur will formulate strategies to tackle bias in healthcare systems.

The Defence Institute of Advanced Technology (DIAT) in Pune with Hyderabad-based Mindgraph Technology, will create a framework "enabling explainable and privacy-preserving AI for security".

Meanwhile, IIT Delhi, in partnership with IIT Dharwad, IIIT Delhi, and the Telecommunication Engineering Center (TEC), will work on developing privacy-preserving machine learning models.

In addition, Indraprastha Institute of Information Technology (IIIT) Delhi, working with TEC, will create tools to assess the fairness of AI models under the project named "Nishpaksh."

Civic Data Labs has been selected to develop ParakhAI, an open-source toolkit for algorithm auditing. Amrita Vishwa Vidyapeetham, in collaboration with TEC, will work on Track-LLM, a framework for transparency and risk assessment of large language models.