Skip to main content

Computing—Building a brain

With applications ranging from autonomous vehicle sensing to daytime astronomy to robotics, researchers attending the ORNL-hosted ICONS conference are interested not only in novel uses of neuromorphic computing, but also in the role it might play in building future supercomputers as traditional systems hit power and performance limits. Credit: Jason B. Smith/Oak Ridge National Laboratory, U.S. Dept. of Energy

Researchers at Oak Ridge National Laboratory are taking inspiration from neural networks to create computers that mimic the human brain—a quickly growing field known as neuromorphic computing. By replacing traditional memory and CPUs with electronic neurons and synapses, scientists aim to create systems that solve complex problems more quickly using less power. “The computing community is starting to understand that this future beyond the GPU-CPU environment is coming,” ORNL’s Catherine Schuman said. As scientists imagine supercomputers after ORNL’s Summit, the world’s fastest supercomputer, and its successor Frontier, they will look for ways to surpass power and performance limits of traditional computing. “One of those paths forward is to incorporate more novel computing architectures into the supercomputer,” she said. ORNL will host the second annual International Conference on Neuromorphic Systems in Knoxville from July 23-25, bringing together government, industry and academic professionals to collaborate on neuromorphic computing.—Abby Bower