Engineers at Penn have developed the first programmable chip capable of training nonlinear neural networks using light—a major breakthrough that could significantly accelerate AI training, lower energy consumption, and potentially lead to fully light-powered computing systems.
Unlike conventional AI chips that rely on electricity, this new chip is photonic, meaning it performs calculations using beams of light. Published in Nature Photonics, the research demonstrates how the chip manipulates light to execute the complex nonlinear operations essential for modern artificial intelligence.
“Nonlinear functions are critical for training deep neural networks,” explains Liang Feng, Professor of Materials Science and Engineering and Electrical and Systems Engineering, and senior author of the study. “Our aim was to make this happen in photonics for the first time.”\
To read more, click here.