An AI that mimics how mammals smell recognizes scents better than other AI – Science News

When it comes to identifying scents, a neuromorphic artificialintelligence beats other AI by more than a nose.

The new AI learns to recognize smells more efficiently and reliablythan other algorithms. And unlike other AI, this system can keep learning newaromas without forgetting others, researchers report online March 16 in NatureMachine Intelligence. The key to the programs success is its neuromorphicstructure, which resembles the neural circuitry in mammalian brains more thanother AI designs.

This kind of algorithm, which excels at detecting faint signalsamidst background noise and continually learning on thejob, could someday be used for air quality monitoring, toxic waste detection ormedical diagnoses.

The new AI is an artificialneural network, composed of many computing elements that mimic nerve cells toprocess scent information (SN: 5/2/19). The AI sniffs by taking inelectrical voltage readouts from chemical sensors in a wind tunnel that wereexposed to plumes of different scents, such as methane or ammonia. When the AIwhiffs a new smell, that triggers a cascade of electrical activity among its nervecells, or neurons, which the system remembers and can recognize in the future.

Headlines and summaries of the latest Science News articles, delivered to your inbox

Like the olfactory system in the mammal brain, some of the AIsneurons are designed to react to chemical sensor inputs by emitting differentlytimed pulses. Other neurons learn to recognize patterns in those blips thatmake up the odors electrical signature.

This brain-inspired setup primes the neuromorphic AI for learningnew smells more than a traditional artificial neural network, which starts as auniform web of identical, blank slate neurons. If a neuromorphic neural networkis like a sports team whose players have assigned positions and know the rulesof the game, an ordinary neural network is initially like a bunch of randomnewbies.

As a result, the neuromorphic system is a quicker, nimbler study.Just as a sports team may need to watch a play only once to understand thestrategy and implement it in new situations, the neuromorphic AI can sniff asingle sample of a new odor to recognize the scent in the future, even amidstother unknown smells.

In contrast, a bunch of beginners may need to watch a play manytimes to reenact the choreography and still struggle to adapt it to futuregame-play scenarios. Likewise, a standard AI has to study a single scent samplemany times, and still might not recognize it when the scent is mixed up withother odors.

Thomas Cleland of Cornell University and Nabil Imam of Intel inSan Francisco pitted their neuromorphic AI against a traditional neural networkin a smell test of 10 odors. To train, the neuromorphic system sniffed a singlesample of each odor. The traditional AI underwent hundreds of training trialsto learn each odor. During the test, each AI sniffed samples in which a learnedsmell was only 20 to 80 percent of the overall scent mimicking real-worldconditions where target smells are often intermingled with other aromas. Theneuromorphic AI identified the right smell 92 percent of the time. The standardAI achieved 52 percent accuracy.

Priyadarshini Panda, a neuromorphic engineer at Yale University,is impressed by the neuromorphic AIs keen sense of smell in muddled samples.The new AIs one-and-done learning strategy is also moreenergy-efficient than traditional AI systems, which tend to be very powerhungry, she says (SN: 9/26/18).

Another perk of the neuromorphic setup is that the AI can keeplearning new smells after its original training if new neurons are added to thenetwork, similar to the way that new cells continually form in the brain.

As new neurons are added to the AI, they can become attuned to newscents without disrupting the other neurons. Its a different story fortraditional AI, where the neural connections involved in recognizing a certain odor,or set of odors, are more broadly distributed across the network. Adding a newsmell to the mix is liable to disturb those existing connections, so a typical AIstruggles to learn new scents without forgetting others unless its retrainedfrom scratch, using both the original and new scent samples.

To demonstrate this, Cleland and Imam trained their neuromorphicAI and a standard AI to specialize in recognizing toluene, which is used tomake paints and fingernail polish. Then, the researchers tried to teach theneural networks to recognize acetone, an ingredient of nail polish remover. Theneuromorphic AI simply added acetone to its scent-recognition repertoire, butthe standard AI couldnt learn acetone without forgetting the smell of toluene.These kinds of memorylapses are a major limitation of current AI (SN: 5/14/19).

Continual learning seems to work well for the neuromorphic systemwhen there are few scents involved, Panda says. But what if you make it large-scale?In the future, researchers could test whether this neuromorphic system can learna much broader array of scents. But this is a good start, she says.

Read the original here:
An AI that mimics how mammals smell recognizes scents better than other AI - Science News