Researchers Create AI Silicon-Photonic Chip Capable of Multiple Computations Within A short Time

In Education

University of Pennsylvania engineers have developed a groundbreaking silicon-photonic (SiPh) chip that can propel artificial intelligence at unprecedented speeds. Unlike conventional chips, which depend on electrical signals, this innovative chip utilizes light waves for intricate mathematical computations, potentially transforming the efficiency and processing power of AI systems.

SiPh chip enables light manipulation for faster calculations

The chip’s development stems from groundbreaking research by Nader Engheta, a distinguished professor at Penn, renowned for his innovations in nanoscale material manipulation for light-based computations. This advancement combines Engheta’s expertise with the SiPh platform, utilizing silicon for cost-effective mass production.

The new SiPh chip is developed to overcome limitations faced by computers, leveraging the interaction between light waves and matter. Developed through collaboration between Engheta’s and Aflatouni’s teams at Penn, the chip showcases advancements in nanoscale silicon devices, promising innovation in AI’s computational backbone.

Researchers have improved vector-matrix multiplication, crucial for neural networks in AI, by manipulating silicon thickness to 150 nanometers. This manipulation enables light manipulation for faster calculations, marking a significant advancement in processing efficiency.

Aflatouni highlights the design’s suitability for commercial use, especially in GPUs crucial for AI development. The SiPh platform’s compatibility with current GPU technology allows seamless integration, promising faster speeds and lower energy consumption for AI training and classification tasks.

Silicon photonics platform allows enhanced data privacy

The Silicon Photonics platform offers accelerated training and classification capabilities, providing performance benefits. Additionally, it enhances data privacy by executing multiple computations simultaneously, eliminating the need to store sensitive data in a computer’s memory, thus reducing vulnerability to hacking attempts.

The study documented in the journal Nature Photonics, carried out at the School of Engineering and Applied Science, University of Pennsylvania, garnered backing from both the U.S. Air Force Office of Scientific Research and the U.S. Office of Naval Research. The ramifications of this study reach far beyond mere expedited and enhanced computing capabilities; it signifies a notable advancement in the quest for computers that boast heightened potency while concurrently prioritizing security and eco-friendliness.

Mobile Sliding Menu