This web page was created programmatically, to learn the article in its unique location you possibly can go to the hyperlink bellow:
https://www.sciencedaily.com/releases/2025/09/250908175458.htm
and if you wish to take away this text from our web site please contact us
Artificial intelligence (AI) methods are more and more central to know-how, powering all the pieces from facial recognition to language translation. But as AI fashions develop extra complicated, they devour huge quantities of electrical energy — posing challenges for vitality effectivity and sustainability. A brand new chip developed by researchers on the University of Florida might assist tackle this challenge through the use of gentle, moderately than simply electrical energy, to carry out one in every of AI’s most power-hungry duties. Their analysis is reported in Advanced Photonics.
The chip is designed to hold out convolution operations, a core perform in machine studying that permits AI methods to detect patterns in photographs, video, and textual content. These operations sometimes require vital computing energy. By integrating optical elements instantly onto a silicon chip, the researchers have created a system that performs convolutions utilizing laser gentle and microscopic lenses — dramatically lowering vitality consumption and rushing up processing.
“Performing a key machine learning computation at near zero energy is a leap forward for future AI systems,” mentioned examine chief Volker J. Sorger, the Rhines Endowed Professor in Semiconductor Photonics on the University of Florida. “This is critical to keep scaling up AI capabilities in years to come.”
In assessments, the prototype chip labeled handwritten digits with about 98 p.c accuracy, similar to conventional digital chips. The system makes use of two units of miniature Fresnel lenses — flat, ultrathin variations of the lenses present in lighthouses — fabricated utilizing commonplace semiconductor manufacturing strategies. These lenses are narrower than a human hair and are etched instantly onto the chip.
To carry out a convolution, machine studying knowledge is first transformed into laser gentle on the chip. The gentle passes by way of the Fresnel lenses, which perform the mathematical transformation. The result’s then transformed again right into a digital sign to finish the AI activity.
“This is the first time anyone has put this type of optical computation on a chip and applied it to an AI neural network,” mentioned Hangbo Yang, a analysis affiliate professor in Sorger’s group at UF and co-author of the examine.
The workforce additionally demonstrated that the chip might course of a number of knowledge streams concurrently through the use of lasers of various colours — a method often called wavelength multiplexing. “We can have multiple wavelengths, or colors, of light passing through the lens at the same time,” Yang mentioned. “That’s a key advantage of photonics.”
The analysis was carried out in collaboration with the Florida Semiconductor Institute, UCLA, and George Washington University. Sorger famous that chip producers resembling NVIDIA already use optical parts in some elements of their AI methods, which might make it simpler to combine this new know-how.
“In the near future, chip-based optics will become a key part of every AI chip we use daily,” Sorger mentioned. “And optical AI computing is next.”
This web page was created programmatically, to learn the article in its unique location you possibly can go to the hyperlink bellow:
https://www.sciencedaily.com/releases/2025/09/250908175458.htm
and if you wish to take away this text from our web site please contact us
