1 h 2 min
Seminar, Hosted by International Telecommunication Union · AI for Good
Environmentally Sustainable AI: Spike-Based Machine Intelligence
University of Southern California
Recording
Abstract
Priya Panda examines how spiking neural networks and hardware–algorithm co-design can reduce the energy and latency costs of AI. She explains how event-driven computation encodes information in the timing of spikes and discusses how sparsity and memory management can improve the execution of these models on general-purpose hardware. The session connects brain-inspired representations with the practical design choices that determine their efficiency.