Word has recently leaked out about an interesting project that’s ongoing at Intel’s research labs. The company has built an algorithm that mirrors the brain’s activity to detect and identify smells. The company developed a computer chip for the project. Called a “neuromorphic” chip, it was designed to allow a computer to process information the same way that a biological brain does.
The project is being pursued in partnership with researchers at Cornell University. It began when the Intel research team began investigating what happens inside the brain of mammals when they smell something.
As you might expect, there’s a lot that goes on inside a mammalian brain when a scent is detected. There are more than 450 different kinds of olfactory receptors in our noses that send signals to the brain. From there, electrical impulses within a group of neurons generate the sense of particular odors.
Then, of course, there’s the fact that our brains act as biological databases. They’re capable of storing not only memories of previous scents, but can cross-reference known scents and accept new information that allows us to track and differentiate hundreds of thousands of different scents.
Intel’s current goals are significantly more modest than developing a chip that can differentiate that many different smells. They’re starting with a goal of ten. They’ve recorded the responses of a total of 72 different chemical sensors sitting in a wind tunnel as a small number of different scents (including acetone, ammonia, and methane) were circulated.
The data gleaned from the sensors was then fed into the Neuromorphic chip, which was able to draw neural representations of each smell.
Ultimately, the point behind the research is to better understand how the brain’s neural circuitry solves complex problems and use that understanding to design the next generation of machine intelligence. The future just got another step closer.
