The human brain is extremely energy efficient. Researchers at the California NanoSystems institute at the University of California Los Angeles are developing systems that mirror the brain’s structure with the hope of matching some of the brain’s computational and energy efficiency. The device they built is much different from conventional computers. The current pilot is a mesh of silver nanowires connected by artificial synapses that is only 2 millimeters by 2 millimeters. Therefore, the device is much messier than silicon circuitry, with no “geometric precision.” The silver mesh network has 1 billion artificial synapses per square centimeter, and his is within a couple orders of magnitude of a real brain network. The networks electrical activity displays “critcality,” a term coined by Danish physicist Per Bak. Criticality is a state in between order and chaos that indicates maximum efficiency. The existence of this device suggests that one day we may be able to build devices that can compute with energy efficiency close to that of the brain.
The device also sets itself apart by the fact that instead of being designed, it organized itself through random chemical and electrical processes. Back in 2007 a group in Japan was studying stingle atomic switches when they discovered something interesting. The more often switches were turned on, the more easily they would turn on, and if left unused for a period of time the switch would slowly turn itself off. The switches also seemed to interact with one another.
Initially they wanted to engineer these properties out, but members from UCLA were reminded of synapses in the brain. They then had the idea to try and embed them in a structure similar to the cortex of the brain of a mammal. Many people doubted this device would work, but input currents kept changing the paths it followed through the device-proving activity in the network was not localized but distributed, as in the brain. The device also shows power-law behavior, a quality observed in the brain. Power-law behavior basically indicates maximum efficiency.
Preliminary experiments suggest that the device can solve computing tasks, though it is extremely different from a traditional computer. There is no software, however they believe voice of image recognition is possible. The challenge with this device is to find the right outputs and decode them so that they can figure out how to best encode information for the device to understand. Although the device is still in a preliminary stage and will not be able to do much that is useful in the next few years, the potential of this device is massive. This device does not separate processing and memory, unlike a conventional computer which needs to move information to different areas to be able to handle the two functions. If researchers are able to get the silver mesh network to solve tasks as effectively as algorithms on conventional computers, the power efficiency struggle of computers will essentially be over. And preliminary tests show that devices like this may be able to predict statistical trends and other complex processes better than traditional computers. Either way the silver mesh device is an exciting new advancement in technology and shows remarkable potential.
Bubnoff, Andreas von. “Artificial Synapses Could Lead to Brainier, Super-Efficient Computers.” Wired, Conde Nast, 3 Oct. 2017, www.wired.com/story/brain-built-on-switches/.
Article and Image Retrieved from: https://www.wired.com/story/brain-built-on-switches/
