Police and border control sniffer dogs could soon be out of a job. The familiar sight of a friendly looking spaniel straining at the leash while it investigates a bag or individual for the tell-tale aroma of dangerous chemicals, illegal drugs or other banned substances so faint only it can pick up on it, could well become a thing of the past on the news scientists have designed a computer chip that can ‘smell’.
The animals could be given early retirement a few years from now, replaced by electronic noses powered by the new chips. And their application could even extend beyond the traditional job description of the sniffer dog. There has already been research conducted into whether highly trained sniffer dogs are also able to smell diseases and preliminary results have shown some promise. It is believed that sniffer chips could well be turned to that job too.
A combined team of researchers from Cornell University and U.S. chip design giant Intel have been working with a new kind of “neuromorphic” chip named Loihi. The chip is designed to enable computers to ‘think’ and process information in a way that is more like how a biological brain works than traditional computer processing.
The scientists analysed how the brains of humans and mammals process the information picked up by the receptors in our noses to identify different scents. They then tried to replicate that process in the chip’s architecture.
When take a sniff of something, like a piece of fruit, the molecules it gives off are picked up by the receptors in in our nose. That information is then relayed to a part of the brain called the olfactory bulb. Within the olfactory bulb, interconnected neurons transmit the electrical pulses sent by the nose’s receptors, before translating them into information on what a particular smell is.
By working with olfactory neurophysiologists, scientists who study the brain activity of animals while the are smelling, the researchers were able to design electrical circuit architecture based on the neural circuits in our brains that process smells. That circuit was then recreated on a silicon chip.
An algorithm able to correctly identify different smells by replicating the patterns of electric signals that pulse through the brain’s olfactory bulb was also designed. The algorithm was then trained on the chip with ten strong, unpleasant smells that included ammonia, methane and acetone. It was able to tell them apart and correctly identify which it was smelling at any given time.
After enough training, the Loihi chip was even able to pick out the smells it had been trained to recognise from among other strong ‘background’ smells.
The technology differs from the now relatively old technology used in smoke and carbon monoxide detectors. They have sensors which detect particular harmful molecules, but do not ‘smell’ them or are able to distinguish between different odours.
The Loihi chip also proved to be a fast learner, able to pick out different odours after just a few samples. Other deep learning artificial intelligence techniques can require over 3000 training samples before a new algorithm arrives at a comparable level of accuracy.
Electronic nose technology is also being worked on by other major tech groups. Alphabet researchers last year created a machine learning algo that learned to distinguish smells on a molecular level, analysing their atoms and chemical bonds – more of an advancement on the kind techniques used in smoke and carbon monoxide detectors than the approach of the Intel and Cornell researchers, who tried to replicate the neurological structures of a biological brain.
This article is for information purposes only.
Please remember that financial investments may rise or fall and past performance does not guarantee future performance in respect of income or capital growth; you may not get back the amount you invested.
There is no obligation to purchase anything but, if you decide to do so, you are strongly advised to consult a professional adviser before making any investment decisions.