More than a decade ago, scientists were working on what they hoped would open a new frontier of neuromorphic computing. At the time, making a device using miniature tools that worked like real brain synapses was almost a fantasy. Although an Australian start-up (Cortical Labs) has gone one step further: It is building miniature, disembodied brains, using real biological neurons embedded in a specialized microchip.
Cortical Labs, based in Melbourne, hopes to teach these mini hybrid brains to do many of the tasks that software-based artificial intelligence formalizes, but at a fraction of the power consumption. The company is currently working to obtain its mini brains, which until now are approaching the processing power of a dragonfly brain, to play the arcade video game “Pong” by Atari. Pong was one of the first games DeepMind used to demonstrate its performance as AI. Precisely this demonstration led to Google launching to buy DeepMind the following year, in 2014.
The biggest obstacle to neuromorphic computing is that most standard computers operate at more than 1 volt to perform their operations. However, the brain sends signals between neurons consuming around 80 millivolts, which is much lower. After decades of experiments, scientists achieved a memory voltage in a range similar to conventional computers, but being below that seemed unlikely.
Now, a team at the University of Massachusetts has discovered, while trying to better understand protein nanowires, how to use these biological strands that conduct electricity to create a neuromorphic device or “memory transistor.” They used protein nanowires developed in UMass Amherst from the Geobacter bacterium by microbiologist and co-author Derek Lovely and conducted experiments in which the device reached neurological voltages.
It works efficiently with very little power, just like human brains do and it is 100 times smaller than the diameter of a human hair. And yes, it is capable of replicating the activity of a biological brain with the same neurological voltage. This is the first time that a device can operate at the same voltage level as the brain. No one would have expected that we could create a device that is as energy efficient as the biological counterparts in a brain. Still, we now have real evidence of ultra-low-power computing capabilities.
Geobacter’s electrically conductive protein nanowires offer many advantages over expensive silicon nanowires, which require toxic chemicals and high-energy processes to be manufactured. Furthermore, they are more stable in water or body fluids, an important feature for biomedical applications. It is clear that this project is far from being completed, that is, achieving computers as efficiently as the biological brain, but it is a truly outstanding step that brings electronics and biology a little closer.
The conductivity or plasticity of the nanowire-memristor synapse can be modulated so that it can emulate biological components for brain-inspired computing. Compared to a conventional computer, this device has non-software-based learning capabilities.