A team of engineers at Northwestern University has developed a new nanoelectronic device that can perform machine learning tasks with 100 times less energy than current technologies. This advancement, described in a paper published today in Nature Electronics, enables artificial intelligence (AI) data analysis to occur directly within wearable devices, eliminating the need to transmit data to the cloud.
“Today, most sensors collect data and then send it to the cloud, where the analysis occurs on energy-hungry servers before the results are finally sent back to the user,” said senior author Mark C. Hersam, professor of materials science and engineering at Northwestern. “This approach is incredibly expensive, consumes significant energy and adds a time delay. Our device is so energy efficient that it can be deployed directly in wearable electronics for real-time detection and data processing, enabling more rapid intervention for health emergencies.”
The nano-sized chip can classify data and perform AI tasks using just two transistors, compared to over 100 required by standard silicon chips. This is achieved by constructing the transistors from a mix of molybdenum disulfide and carbon nanotubes, giving them unprecedented tunability and allowing a single device to serve multiple functions.
To test the chip’s capabilities, the researchers analyzed publicly available electrocardiogram (ECG) data. The device successfully identified six types of heart rhythms from 10,000 ECG samples with 95% accuracy. “Having a high degree of tunability in a single device allows us to perform sophisticated classification algorithms with a small footprint and low energy consumption,” said Hersam.
Performing analysis directly on wearable devices not only saves time compared to cloud-based AI, but also improves data privacy. “Every time data are passed around, it increases the likelihood of the data being stolen,” Hersam explained. “If personal health data is processed locally — such as on your wrist in your watch — that presents a much lower security risk.”
Hersam envisions the chip being incorporated into smart watches and fitness trackers in the future, personalized to each user’s health data. “It would enable people to make the most of the data they already collect without sapping power,” he said.
Han Wang, professor at the University of Southern California and co-leader of the study, said, “This technology significantly reduces the hardware overhead for AI model implementation, enabling AI algorithms to be run with very low energy consumption.”
Vinod Sangwan, co-author and research assistant professor at Northwestern, added, “The reconfigurable nature of our chip allows for personalized medicine by tuning the device characteristics for each patient.”
The researchers say this nanotechnology provides a path towards sustainable AI by reducing power demands. “Artificial intelligence tools are consuming an increasing fraction of the power grid. It is an unsustainable path if we continue relying on conventional computer hardware,” concluded Hersam.