Companies are trying to train AI to do all sorts of things, but a team at Intel has been attempting to train AI to something unorthorodox for a computer: smell. Researchers from Intel Labs and Cornell University have announced that Intel’s neuromorphic research chip, dubbed Loihi, is able to “learn and recognize hazardous chemicals,” through algorithms that mimic the brain’s olfactory system.
Obviously, lacking a nose, a computer chip can’t actually smell, but in a paper titled “Rapid online learning and robust recall in a neuromorphic olfactory circuit,” published to Nature Machine Intelligence, Intel’s Nabil Imam and Cornell’s Thomas A. Cleland explain that Loihi was able to learn to identify 10 different odors.
“Imam and team took a dataset consisting of the activity of 72 chemical sensors in response to 10 gaseous substances (odors) circulating within a wind tunnel,” Intel wrote on its website. “The sensors’ responses to the individual scents were transmitted to Loihi where silicon circuits mimicked the circuitry of the brain underlying the sense of smell.”
The company says that Loihi was able to learn each odor after only a single sample, with Intel comparing that to a similar “deep learning solution that required 3,000 times more training samples per class to reach the same level of classification accuracy.” Loihi even managed to identify those odors through “significant noise,” so it seems that not only was Loihi a quick learner, but its identification skills held up in the presence of other odors as well.
Now that Imam and his team have figured out a way to make Loihi detect odors, he says that his next task is “to generalize this approach to a wider range of problems – from sensory scene analysis (understanding the relationships between the objects you observe) to abstract problems like planning and decision-making.” In the end, the work that Intel and Cornell have done here could help lead to more robust machine learning AI, so hopefully we’ll hear more from the two about the Loihi project soon.