What does a neuromorphic chip that can "smell" 10 harmful substances look like?

According to Leifeng.com, Intel and Cornell University published a joint paper today, demonstrating the new capabilities of Intel's neuromorphic chip Loihi: it can learn and identify 10 harmful substances from smells - even in the presence of "significant" data noise and occlusion.

The coauthors say it shows how neuromorphic computing could be used to detect odor precursors of explosives, narcotics, polymers and more.

The study was published this week in the journal Nature Machine Intelligence. Researchers from Intel and Cornell University used a data set of 72 chemical sensors responding to different smells to describe how to "teach" Loihi to "smell" by configuring the circuit diagram of biological olfaction.

They say their technique does not destroy the chip’s odor memory; it has “superior” recognition accuracy compared to previous state-of-the-art traditional methods, including a machine learning solution that required 3,000 times more training samples per class to achieve the same level of classification accuracy.

Nabil Imam, a senior research scientist in Intel's Neuromorphic Computing Lab, believes the research will pave the way for neuromorphic systems that can diagnose disease, detect weapons and explosives, spot drugs, and detect signs of smoke and carbon monoxide.

"We are developing neural algorithms on Loihi that can simulate what happens in the brain when a person smells something," he said in a statement. "This work is a great example of the crossroads of contemporary neuroscience and artificial intelligence research, and demonstrates the potential of Loihi to provide important perception capabilities that could benefit a wide range of industries."

Neuromorphic engineering, also known as neuromorphic computing, describes the use of circuits that mimic the neurobiological structure of the nervous system. Researchers at Intel, IBM, Hewlett-Packard, MIT, Purdue, Stanford and other institutions hope to use it to develop supercomputers that are expected to be perhaps a thousand times more powerful than anything available today.

Intel's 14nm Loihi chip has a 60mm die size and contains more than 2 billion transistors, 130,000 artificial neurons and 130 million synapses, with three manageable Lakemont cores for coordination.

What makes Loihi unique is that it has a programmable microcode engine for on-chip training of asynchronous spiking neural networks (SNNs), or AI models, which incorporate time into their operating models so that components of the model do not process input data at the same time. Intel claims this will be used to "efficiently" implement adaptive self-modifying, event-driven, and fine-grained parallel computing.

According to Intel, Loihi processes information 1,000 times faster and 10,000 times more efficiently than traditional processors, and can solve certain types of optimization problems with more than three orders of magnitude greater speed and energy efficiency.

Furthermore, Loihi maintains real-time performance results, using only 30% of the power when scaling 50 times (while traditional hardware consumes 500% more power), and consumes approximately 100 times less power than widely used CPUs running simultaneous localization and mapping methods.

In addition to the field of neuromorphic computing, researchers at Google, the Canadian Institute for Advanced Research, the Vector Institute for Artificial Intelligence, the University of Toronto, Arizona State University, and other institutions have studied the use of artificial intelligence methods to solve molecular recognition and odor prediction problems. Google recently demonstrated a model that outperformed the state-of-the-art methods and the best performing models from the Dream Odor Prediction Challenge (a competition to characterize the chemical properties of odors).

Separately, IBM has developed Hypertaste, an “artificial tongue” that can identify drinks and other liquids that are “not ideal for ingestion.”

Intel trains neuromorphic chip to detect 10 different odors