Intel is inching one step closer to bringing human capabilities to a computer. The chipmaker has published new research that details an algorithm capable of smelling. The idea here is not to smell but recognize what it is smelling using algorithms. The company is introducing neuromorphic chips that sniff out hazardous chemicals. The company detailed the new chip in a joint research paper published with researchers from Cornell University.
The neuromorphic computing group at Intel Labs built a mathematical algorithm that works similar to olfactory cells in our nose. The algorithm has been implemented on Intel’s Loihi neuromorphic computing chip. The chip fabricated using a 14nm process node, has a design similar to that of the brain. Intel is calling the chip neuromorphic since the terms refers to making computers think and function like brains.
Nabil Imam, senior research scientist in Intel’s Neuromorphic Computing Lab, said Loihi will learn and recognize the scents of 10 hazardous chemicals. “We are developing neural algorithms on Loihi that mimic what happens in your brain when you smell something. 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,” he said in a statement.
The researchers at Intel and Cornell used a dataset consisting of the activity of 72 chemical sensors in response to these smells. They also configured the circuit diagram of biological olfaction on Loihi. Imam sees the potential of robots being equipped with neuromorphic chips for environmental monitoring and hazardous materials detection. Intel has already scaled neuromorphic research systems to 100 million neurons.
Intel first detailed Loihi as a new self-learning chip in 2017. Now, the company is looking to add more senses to the chip in the future. Imam says the next steps would be to generalize the approach to a wider range of problems. “Understanding how the brain’s neural circuits solve these complex computational problems will provide important clues for designing efficient and robust machine intelligence,” he said in a statement.