Researchers at the Massachusetts Institute of Technology have developed a "brain-on-a-chip" that includes tens of thousands of artificial brain synapses.
Called memristors, the artificial synapses are made of silver and copper alloys, rendering them more efficient than their earlier non-alloyed counterparts.
They published their work on Monday in the journal Nature Nanotechnology.
In the future, you just might be able to carry around a tiny machine in your pocket that works like the human brain. It could do anything that the most advanced artificial neural network could—but with no internet, and no software download required. That is, if new brain-on-a-chip research published Monday in the journal Nature Nanotechnology is any indication of where miniature computers are headed.
At the Massachusetts Institute of Technology, scientists have been developing that tech, squeezing tens of thousands of artificial brain synapses, called memristors (short for memory transistors), onto a confetti-sized slab of silicon and metal alloys. The goal is to produce tiny devices that can pack the punch of artificial intelligence, locally, without having to connect to the cloud, or rely on a supercomputer.
"Imagine connecting a neuromorphic device to a camera on your car, and having it recognize lights and objects and make a decision immediately, without having to connect to the internet," Jeehwan Kim, associate professor of mechanical engineering at MIT who led the work, said in a prepared statement. "We hope to use energy-efficient memristors to do those tasks on-site, in real-time.”
Brain-on-a-chip research is a step toward neuromorphic devices, or electronics that contain a new kind of circuitry meant to mimic the brain's neural architecture. These devices are meant to be energy efficient and fluent in cognitive tasks like object recognition, association, adaption, and learning. One day, researchers hope to make them as capable as today's supercomputers—and MIT is a step toward that reality.
To fabricate the memristors, the team used silver and copper alloys, along with silicon. The resulting chip is only about one millimeter square, and can "remember" stored images and reproduce them over and over again. Those replications were more exact than in past efforts, which relied on unalloyed metals in the chip.
In one test, researchers recreated a grayscale image of the Captain America shield. Each pixel corresponded to one memristor in the chip. So, the team altered the conductivity of each memristor such that it was relative in strength to the color in the pixel it would illustrate. They were able to reproduce the image many times over.
Then, in an image processing task, the team programmed the memristors to change the appearance of an image depicting MIT's Killian Court, a manicured lawn before a classical domed building on campus. They found that they could better sharpen and blur the original image with these memristors than the prior state-of-the-art circuits.
All of this adds up to a move away from large-scale computing. Because researchers could complete these tasks on the tiny memristors, rather than through software on a computer, it's a step toward small devices that can use artificial intelligence without reliance on a massive data center. That could mean better onboard cameras in smartphones, or more intelligent computers in our cars.
"We would like to develop this technology further to have larger-scale arrays to do image recognition tasks" Kim said. "And some day, you might be able to carry around artificial brains to do these kinds of tasks, without connecting to supercomputers, the internet, or the cloud."
You Might Also Like