Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore how to make artificial intelligence more sustainable through event-based neural networks and neuromorphic computing in this 38-minute keynote presentation. Discover the fundamentals of spiking neural networks (SNNs) and learn how they differ from traditional artificial neural networks while addressing AI's growing energy crisis as systems like ChatGPT scale and their energy demands skyrocket. Master the adjoint method for calculating exact gradients in SNNs and understand how this approach improves training efficiency and scales with longer input sequences. Examine the deployment of trained networks on Intel's Loihi 2 neuromorphic chips and analyze real-world results demonstrating over 1000x energy savings in speech recognition tasks. Gain insights into how neuromorphic computing, combined with new training algorithms like EventProp, enables spiking models to achieve state-of-the-art performance while consuming a fraction of the energy. Learn about the future applications of low-power AI in edge computing, speech recognition, and other domains, and understand how biologically inspired neural models that mimic the efficiency of the human brain offer compelling solutions for decarbonizing artificial intelligence and unlocking new paths for sustainable AI development.
