Intel's Neuromorphic Loihi Chip Drives Electronic Nose Capable of Smart Sniffing and Categorization

Intel Labs, in partnership with Cornell University, has released details of an "electronic nose" which mimics the mammalian scent organ — using 72 chemical sensors linked to the company's prototype Loihi neuromorphic processor.

Neuromorphic computer, in which systems are designed to mimic the design and operation of the neurons of the brain rather than the on-off switches of binary hardware, is being heralded by many as a solution to some of the biggest challenges in computing: Earlier this year researchers Suhas Kumar and Jack Kendall claimed that "the future of computing will not be about cramming more components on a chip but in rethinking processor architecture from the ground up to emulate how a brain efficiently processes information."

Now, Intel is demonstrating the capabilities of neuromorphic computing with an electronic nose constructed of just 72 chemical sensors linked to its still-prototype Loihi neuromorphic processor — and capable of mimicking the mammalian sense of smell.

"We are developing neural algorithms on Loihi that mimic what happens in your brain when you smell something," Intel's Nabil Imam explains. "This work is a prime example of contemporary research at the crossroads of neuroscience and artificial intelligence and demonstrates Loihi’s potential to provide important sensing capabilities that could benefit various industries."

"My next step is to generalize this approach to a wider range of problems — from sensory scene analysis (understanding the relationships between objects you observe) to abstract problems like planning and decision-making. Understanding how the brain’s neural circuits solve these complex computational problems will provide important clues for designing efficient and robust machine intelligence."

That generalization marks a key challenge: A human may recognize a smell as being from a fruit, but not be able to specify which one — or may not recognize two different fruits of the same family as being related. "These are challenges in olfactory signal recognition that we’re working on and that we hope to solve in the next couple of years before this becomes a product that can solve real-world problems beyond the experimental ones we have demonstrated in the lab," Imam explains.

At present, the Loihi-powered system has successfully identified 10 gaseous substances present in a wind tunnel — including in the presence of interfering background odors — from acetone and ammonia to methane. It did so, the company claims, by creating "neural representations" of the smells — a very different approach to a traditional binary computer model.

More information on the work can be found in the team's paper, published in the journal Nature Machine Intelligence under open-access terms; the company has not yet, however, committed to a public release of the Loihi hardware.