For the first time, researchers have designed an artificial synaptic network in hardware, mimicking the information-transmitting synapses in the human brain.
Smaller than a piece of confetti, the chip is made from tens of thousands of artificial brain synapses. These silicon-based memory transistors – or memristors – could pave the way for small and portable brain-inspired circuits that don’t require data connections.
“So far, artificial synapse networks exist as software. We’re trying to build real neural network hardware for portable artificial intelligence systems,” said Jeehwan Kim, Associate Professor of Mechanical Engineering at the Massachusetts Institute of Technology (MIT).
“Imagine connecting a neuromorphic device to a camera on your car, and having it recognise lights and objects and make a decision immediately, without having to connect to the internet. We hope to use energy-efficient memristors to do those tasks on-site, in real-time.”
How does it work?
While a transistor in a conventional circuit transmits information by switching between one of only two values, 0 and 1, memristors mimic brain synapses by working along a gradient. This enables a memristor to carry out a much wider range of operations than binary transistors, allowing it to ‘remember’ the value associated with a given current strength.
While memristors were first devised in the early 1970s, such designs have been somewhat limited in performance. Kim and his colleagues used metallurgy to improve upon the previous memristor design, incorporating silver and copper alloys along with silicon.
“Traditionally, metallurgists try to add different atoms into a bulk matrix to strengthen materials, and we thought, why not tweak the atomic interactions in our memristor, and add some alloying element to control the movement of ions in our medium,” Kim said.
And the adjustments proved successful. When the researchers ran the new chip through visual tasks including equating each pixel in an image of the Captain America shield to a corresponding memristor, the chip was able to ‘remember’ stored images and reproduce them many times over, more reliably than existing memristor designs.
Published in Nature Nanotechnology, the research into this new design could advance the development of small, portable Artificial Intelligence (AI) devices.
“We would like to develop this technology further to have larger-scale arrays to do image recognition tasks,” Kim said.
“Someday, you might be able to carry around artificial brains to do these kinds of tasks, without connecting to supercomputers, the internet, or the cloud.”