Artificial intelligence is becoming a backbone of modern technology, from facial recognition to translation apps. But powering AI models demands heavy electricity, raising questions about efficiency and sustainability.
Researchers at the University of Florida believe they have found a way to tackle this problem. Their new chip uses light, not just electricity, to perform one of AI’s most demanding tasks.
Light-based computing breakthrough
The chip is built to handle convolution operations, a core function in machine learning. These operations let AI detect patterns in images, video, and text. They also consume large amounts of computing power.
The team integrated optical components directly onto a silicon chip. Laser light and microscopic lenses then carry out convolutions faster and with lower energy needs.
“Performing a key machine learning computation at near zero energy is a leap forward for future AI systems,” said study leader Volker J. Sorger, the Rhines Endowed Professor in Semiconductor Photonics at the University of Florida. “This is critical to keep scaling up AI capabilities in years to come.”
Tests showed the prototype classified handwritten digits with about 98 percent accuracy, matching conventional chips.
Illustration of new silicon photonic chip. Credit – SPIE
The system relies on two sets of Fresnel lenses, flat ultrathin structures similar to those in lighthouses. Each lens is narrower than a human hair and etched onto the chip with standard semiconductor techniques.
To run a convolution, data converts into laser light on the chip.
That light passes through the Fresnel lenses, which perform the mathematical transformation. The result converts back into a digital signal for the AI model.
“This is the first time anyone has put this type of optical computation on a chip and applied it to an AI neural network,” said Hangbo Yang, a research associate professor in Sorger’s group and co-author of the study.
Multiplexing for parallel processing
The chip can also process multiple data streams at once. The team achieved this by using lasers of different colors. The approach is known as wavelength multiplexing.
“We can have multiple wavelengths, or colors, of light passing through the lens at the same time,” Yang said. “That’s a key advantage of photonics.”
The project involved the Florida Semiconductor Institute, UCLA, and George Washington University.
Sorger noted that major players such as NVIDIA already use optical components in their AI systems. This could make it easier to bring the new chip into commercial use.
“In the near future, chip-based optics will become a key part of every AI chip we use daily,” Sorger said. “And optical AI computing is next.”
By cutting energy use while keeping accuracy high, the Florida team’s chip could help scale AI to meet global demand.
If the technology advances beyond the lab, light-based chips may soon power many of the AI tools people rely on every day.
The study is published in the journal Advanced Photonics.