Description
According to an estimate, by 2030, more than 20% of the world's total energy usage will be consumed by artificial intelligence-based information technology. To prevent data centers from consuming the world's electricity production, in parallel with the development of artificial intelligence algorithms, new energy-efficient hardware devices are also needed. The human brain, with its ~20W power consumption, is an excellent example for energy-efficient information processing. This is the basis for the emerging field of brain-inspired computing, or neuromorphic computing. In my talk, I review how novel nanoscale devices can be used to implement artificial synapses and neurons in electrical circuits, and how these devices can be assembled into energy-efficient information processing networks.