IIT Guwahati Develops Brain-Inspired AI Model That Uses Far Less Energy

Researchers at the Indian Institute of Technology, or IIT, Guwahati are developing a brain-inspired artificial intelligence model designed to process long sequences of data efficiently while consuming significantly less energy than many conventional AI approaches. The team, from the Mehta Family School of Data Science and Artificial Intelligence, presented its findings at the International Conference on Machine Learning, or ICML, 2026 in Seoul, South Korea, a top-tier international AI conference.

Modern AI systems increasingly analyze long streams of sequential data, such as health signals from wearable devices, environmental sensor readings, industrial monitoring data, and weather or traffic forecasts, said Ayon Borthakur, an assistant professor at the Mehta Family School. Widely used AI architectures often become computationally expensive as the length of that data increases, making them less suitable for battery-powered and resource-constrained devices, he said.

To address this, the team built SH2RFSSM, a brain-inspired AI architecture that mimics the event-driven communication of biological neurons. "Unlike conventional neural networks that continuously process information, spiking neural networks activate only when meaningful events occur, enabling sparse and energy-efficient computation," said Kartikay Agrawal, a PhD research scholar on the team. "We combined this principle with advanced state space modelling, allowing the system to learn long-range patterns without the heavy computational cost associated with traditional sequence models."

A distinctive feature of the model is neuronal heterogeneity, where individual artificial neurons are allowed to have different characteristics rather than behaving identically, which the team says improves its ability to capture complex temporal patterns in real-world data.

The researchers evaluated the model across 17 benchmark datasets covering long-range sequence classification, regression, human activity recognition and long-term forecasting. It delivered performance comparable to state-of-the-art sequence models while showing substantially lower estimated energy consumption, which the team says makes it promising for AI that runs directly on devices such as wearables, IoT sensors and autonomous systems, without relying heavily on cloud computing.