The human brain has a power consumption of only 20 watts, yet can easily perform tasks that are far beyond the capabilities of today’s supercomputers. Computation happens at many levels in the brain, from subcellular biochemical processes, through individual neurons and dendrites, small local circuits, up to the most complex computations that involve interactions of large neuronal assemblies across the whole brain. Investigating these computations helps us both to understand how our own brains works and to construct bio-inspired artificial computing systems, including prostheses and robotic devices.
What computations are performed by different components and structures within the brain? What are the contributions of synapses, neurons, network connectivity? What algorithms are being used, and can we use these algorithms in artificial systems? How does the brain deal with noise and unpredictability in the environment and in its own operations?
Different teams in NeuroPSI address these questions using experimental, theoretical and computational approaches, including physics-based models, large-scale simulations, hybrid biological-artificial circuits, and brain-computer interfaces. We collaborate with physicists, computer scientists and engineers to use the knowledge gained to develop bio-mimetic, “neuromorphic” computing systems.
This work has the potential for significant impacts in technology and health: breaking through current bottlenecks in artificial intelligence, helping to create more robust, smaller, and more energy-efficient computing devices, and helping to design treatments for problems such as spinal cord injury, depression and schizophrenia.