The human brain is a marvelous thing. It’s made up of billions of neurons that carry out high-level computing functions in an instant. By comparison, computer chips are archaic. Nonetheless, chipmakers and innovators strive to replicate the brain in a piece of hardware.
Researchers from MIT claim that their “brain-on-a-chip” design brings us closer to that vision. A team of engineers found a way to pack tens of thousands of artificial brain synapses, called memristors, onto a single computer chip. Notably, it is smaller than a piece of confetti.
Future of Computing
There are endless uses for artificial intelligence (AI). Many experts predict that it will be one of the most impactful technologies of the decade. However, AI doesn’t always play nicely with traditional computer chips. That being said, it’s clear that there is a need for compact, AI-oriented chips to power the devices of the future.
At the core of their innovation are memristors. Essentially, these are simulated brain synapses created using silicon. The MIT team also used alloys of silver and copper. Their unique product was a chip that is able to remember and recall images with a high degree of detail. Notably, the chip did so with routine accuracy that far surpasses other types of simulated brain circuits.
Unlike a standard transistor that can only switch between zero or one, memristors can take on a range of values. That trait resembles the human brain. It also allows the memristors to remember these states and recreate them over and over.
The MIT team was able to pack tens of thousands of memristors onto one chip. Notably, the chip was only one square millimeter in size. It successfully recreated a gray-scale image of Captain America’s shield and consistently altered a photo of MIT’s Killian Court. Although those tests might not seem impressive, they are a positive sign.
The researchers involved believe that the advancements could lead to a future where portable, artificial brain computers could perform complex tasks on the same level as a supercomputer without the massive power or network connectivity requirements.
Jeehwan Kim, an associate professor of mechanical engineering at MIT, has a bright outlook on the future of so-called “brain-on-a-chip” applications. He says, “So far, artificial synapse networks exist as software. We’re trying to build real neural network hardware for portable artificial intelligence systems.”
“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,” adds Kim.
It’s worth noting that MIT isn’t the only group working on this type of chip. Worldwide tech companies like Apple, Google, Microsoft, and Nvidia are also creating machine learning hardware. Intel’s Loihi chip uses 1,024 artificial neurons to mimic the brain.
Moving forward, the MIT team has big plans for its brain-inspired chip technology. Kim says, “We would like to develop this technology further to have larger-scale arrays to do image recognition tasks. 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.”