Edited by Chance Lai
______

Artificial Intelligence (AI) is reshaping our world at an unprecedented pace, driving innovations from autonomous vehicles to medical diagnostics. However, as the demand for AI applications soars, the energy consumption associated with traditional AI computation has become a pressing concern.

A research team from the International College of Semiconductor Technology at National Yang Ming Chiao Tung University (NYCU) has achieved a key technological breakthrough, paving the way for a more efficient and energy-saving computing paradigm. The study, now published in the prestigious journal Nano Letters, represents a significant step forward in neuromorphic computing—an approach that emulates the human brain’s ability to learn and adapt.

Advancing Neuromorphic Computing through Low-Energy AI Models

The team’s innovation centers around “field-free switching” (FFS), accomplished through a heterostructure device based on magnetic materials (W/Pt/Co/NiO/Pt), utilizing spintronics technology. This design eliminates the need for an external magnetic field, enabling magnetic switching with dramatically reduced energy consumption and enhanced efficiency.

Based on this foundation, the researchers developed low-energy artificial synapses and neurons, applying them to a three-layer artificial neural network (ANN). Their neural networks demonstrated high-accuracy performance on the MNIST and Fashion MNIST datasets, showcasing a human-brain-like operational model. These advancements open new doors for rapid processing in AI applications such as autonomous driving, intelligent surveillance, and medical imaging diagnostics.