Date: 30 January 2026
Venue: IIT Madras

This seminar addressed the energy inefficiency of modern large language models, which require significant energy to train and deploy, and contrasted them with the human brain's ability to operate on roughly twenty watts of power. It explored how spiking neural networks and neuromorphic hardware can emulate the brain's energy-efficient mechanisms for next-generation AI systems.

Talks by researchers from IIT Madras, IISc Bangalore, and the Peter Grünberg Institute at Forschungszentrum Jülich highlighted advances in simulating spiking neurons, training artificial spiking neural networks in software, and implementing them on specialised hardware. Speakers included Sridharan Devarajan, Gopalakrishnan Srinivasan, Emre Neftci, and Patanjali SLPSK, with the seminar curated and moderated by Anil Ananthaswamy.