Revolutionary Transistor Design Boosts AI Energy Efficiency by 100x
Originally Published 2 years ago — by Northwestern Now
Engineers at Northwestern University have developed a nanoelectronic device that performs machine-learning classification tasks with 100-fold less energy than current technologies. The device can process large amounts of data and perform AI tasks in real time without relying on the cloud for analysis. It is ideal for integration into wearable electronics such as smartwatches and fitness trackers, enabling real-time data processing and near-instant diagnostics. The device was tested on electrocardiogram (ECG) datasets and achieved near 95% accuracy in identifying irregular heartbeats and determining arrhythmia subtypes. The device's tunability, achieved through a mix of materials, allows for dynamic reconfigurability and low energy consumption. The researchers envision these nanoelectronic devices being incorporated into everyday wearables for personalized, real-time applications.