Intel’s Loihi neuromorphic chip can identify hazardous chemicals from their smell

Photo: Intel Labs

Researchers from Intel Labs and Cornell University have demonstrated the unique ability of Intel Corp.’s neuromorphic research chip Loihi to identify a range of hazardous chemicals based on their odor alone.

Loihi could identify each chemical based on its odor from just a single test sample, and without disrupting its memory of previously learned scents, the researchers said. Not only that, the chip demonstrated superior accuracy compared to any conventional recognition system, including a deep learning system that requires around 3,000 times more training samples to reach the same level of accuracy.

“We are developing neural algorithms on Loihi that mimic what happens in your brain when you smell something,” said Intel Labs Senior Research Scientist Nabil Imam (pictured). “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.”

Intel’s Loihi chip is a bit of hardware that aims to mimic how the human brain processes and solves problems. The chip was first announced in September 2017, when Intel said it was capable of “incredible” learning speeds. The chip is unique in that it can leverage knowledge it already possesses to make inferences about new data, helping to speed up its learning process exponentially over time.

The chip is based on a “neuromorphic computing” architecture that’s inspired by scientists’ current understanding of the human brain and how it solves problems.

The latest research, published in the journal Nature Machine Intelligence today, describes how Intel Labs and Cornell’s research teams built a neural algorithm from scratch based on the architecture and dynamics of the human brain’s olfactory circuits. The chip could then learn and recognize the scent of 10 different hazardous chemicals.

To understand how Loihi did this, it helps to know how the human brain perceives different odors. When a person picks up a grapefruit for example and smells it, the fruit’s molecules stimulate olfactory cells in the nose that sends signals to the brain. Then, electrical pulses within an interconnected group of neurons generate a unique sensation of that smell.

“Whether you’re smelling a grapefruit, a rose or a noxious gas, networks of neurons in your brain create sensations specific to the object,” Intel’s researchers explained. “Similarly, your senses of sight and sound, your recall of memory, your emotions, your decision-making each have individual neural networks that compute in particular ways.”

In the latest research, Intel’s team used a dataset that consists of the activity of 72 known chemical sensors in the brain and how they respond to the smell of each chemical. That data was used to configure what the team calls “a circuit diagram of biological olfaction” on Loihi. With this, Loihi was then able to recognize the neural representation of each smell and identify each one, even with significant occlusion.

Analyst Patrick Moorhead of Moor Insights & Strategy told SiliconANGLE that Intel’s research shows that determining the smells of various hazardous chemicals is a great use case for neuromorphic computing.

“Neuromorhic computing trains and infers just like the human brain, so it’s nice to see Intel blazing a trail in this research space,” Moorhead said.

Imam said in a blog post that Loihi’s olfactory capabilities could be put to use on new “electronic nose systems” that help doctors to diagnose diseases. Other uses include developing more effective smoke and carbon monoxide detectors, or systems for detecting weapons and explosives at airports.

“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,” Imam said. “Understanding how the brain’s neural circuits solve these complex computational problems will provide important clues for designing efficient and robust machine intelligence.”

Form：