Nanowire Network Mimics Brain, Learns Handwriting with 93.4% Accuracy

A cutting-edge computing system, modeled on the biological brain, achieved 93.4% accuracy in identifying handwritten numbers through an innovative training algorithm providing real-time feedback.

In development, the nanowire network, from the California NanoSystems Institute at UCLA, promises lower power consumption than silicon-based AI. Over 15 years, researchers created a brain-inspired system with nanoscale wires dynamically reconfiguring in response to stimuli, resembling biological synapses.

The collaboration with the University of Sydney produced a streamlined algorithm for efficient input and output interpretation. The nanowire system, made of silver and selenium, self-organizes and, using brief electrical pulses, demonstrated the potential for significantly reduced power requirements compared to traditional AI.

Beyond energy efficiency, the nanowire network excels in tasks challenging for current AI, making it suitable for edge computing. This includes applications in robotics, autonomous navigation, IoT devices, and health monitoring, offering continuous adaptation and learning in physical systems.

Author: Neurologica