|Neuromorphic Computing (Cognitive Computing). New science research, 2020.|
Consumes the same energy as biological neurons. A chip 100 times smaller than the diameter of a human hair is capable of replicating the activity of a biological brain with the same neurological voltage.
Thirty years ago, researchers at the University of Massachusetts at Amherst discovered a proteobacteria called the Geobacter in the Potomac River that could produce nanowires of electrically conductive proteins.
Last year, those researchers developed a device that uses a natural protein produced by a microbe belonging to the Geobacter family, to generate electricity from moisture in the air.
In the third phase of this innovative process, another group of researchers from the same university have used those same protein cables to obtain a memory transistor or neuromorphic memory device.
The device works extremely efficiently with very little power, as brains do, to transport signals between neurons, the researchers explain in an article published in Nature Communications.
|Bioinspired bio-voltage memristors in Neuromorphic Computing, 2020.|
Neuromorphic ComputingThis result is especially encouraging the so-called neuromorphic computing, which aims to create chips that work as the human brain does.
Neuromorphic systems replicate at the hardware level the way neurons organize, communicate and learn.
The aim is to achieve a programmable computing model that enables intelligent electronic devices.
One of the biggest obstacles to neuromorphic computing is that most conventional computers consume more than 1 volt in each of their operations.
The human brain, however, is much more efficient with less power consumption: it sends signals between neurons consuming only 80 millivolts, much less than current neuromorphic devices.
The first stepThis feat has been made possible by the use of protein nanowires developed at the university from the Geobacter bacterium, the researchers said in a statement.
They add that it is the first time that a device with these features can work at the same voltage level as the brain.
They also note that Geobacter electrically conductive protein nanowires offer many advantages over expensive silicon nanowires (SiNWs), which require toxic chemicals and high-energy processes to achieve results.
Protein nanowires are also more stable in water or body fluids, an important feature for possible future biomedical applications.
ContextIt should be noted that achieving a device that mimics neural connections requires replicating the same feat that neurons perform with the synapses.
The electrical impulse that allows neurons to communicate and activate the nervous system runs through the neural wiring, which is not continuous.
Each wire at a certain point will be cut leaving a space between neurons that the electrical impulse must overcome by jumping into the void and reaching the other end of the neuron.
This leap into the void, known as synapse, has an energy consumption threshold of 80 millivolts, which is the barrier achieved with the new device: it has replicated the neurological voltage to achieve synchronization.
MethodologyResearchers have had the ability to cut the nanowires in bacteria to use only the electrically conductive protein.
This is one of the explanations for the low energy consumption of the device.
The researchers have also used a metal thread that serves as food for the bacteria in the protein nanowire.
Bacterial nanowires chemically break down metals to obtain their energy in the same way that we breathe oxygen, the researchers say.
And in addition it learnsIt is not however the only advantage of the device, 100 times smaller than the diameter of a human hair: it is also capable of learning.
As electrical pulses create changes in the metal filaments, new branches and connections are created in the small device.
There is then an effect similar to learning (new connections) that occurs in a real brain.
Researchers can modulate the conductivity, or plasticity, of the nanowire synapse, so it can emulate biological components and achieve brain-inspired computing.
Compared to a conventional computer, this device has non-software-based learning capabilities, they conclude.
- The work is far from finished.
- The researchers aim to fully explore the chemistry, biology, and electronics of protein nanowires in these devices.
- It is possible that the applications of these devices could include a heart rate monitoring device.
- The ultimate purpose is that one day this device will be able to communicate with real neurons in biological systems and solve health problems.
That is, to get computers as efficient as the biological brain.
So far, they have considerably blurred the line between computers and biological systems.
Bioinspired bio-voltage memories. Tianda Fu et al. Nature Communications, volume 11, Article number: 1861 (2020). DOI:https://doi.org/10.1038/s41467-020-15759-y.