Researchers at the University of Florida have developed a new chip that uses light instead of electricity to perform one of the most power-hungry tasks in artificial intelligence, a breakthrough that could help address the growing energy consumption of complex AI models. The research, reported in Advanced Photonics, details a system that dramatically reduces energy use and speeds up processing.

As AI systems become more central to technology, their electricity consumption poses significant challenges to sustainability. This new chip offers a potential solution by integrating optical components directly onto a silicon chip to handle a core AI function.

How the optical chip works

The chip is specifically designed to carry out convolution operations, a fundamental process in machine learning that allows AI systems to detect patterns in images, video, and text. These operations typically require a large amount of computing power.

The new system performs these convolutions using laser light and microscopic lenses. Machine learning data is first converted into laser light on the chip. This light then passes through two sets of miniature Fresnel lenses—flat, ultrathin versions of the lenses found in lighthouses—which are etched directly onto the chip and are narrower than a human hair. These lenses perform the mathematical transformation, and the result is then converted back into a digital signal to complete the AI task.

“Performing a key machine learning computation at near zero energy is a leap forward for future AI systems. This is critical to keep scaling up AI capabilities in years to come.”

said study leader Volker J. Sorger, a professor at the University of Florida.

Performance and advantages

In tests, the prototype chip successfully classified handwritten digits with approximately 98 percent accuracy, a level of performance comparable to traditional electronic chips.

A key advantage of the photonic approach is its ability to process multiple data streams at the same time. By using lasers of different colors, a technique known as wavelength multiplexing, the chip can handle several operations simultaneously. “We can have multiple wavelengths, or colors, of light passing through the lens at the same time,” said Hangbo Yang, a research associate professor and co-author of the study. “That’s a key advantage of photonics.”

Future integration

The research was conducted in collaboration with the Florida Semiconductor Institute, UCLA, and George Washington University. The researchers noted that major chip manufacturers like NVIDIA already use optical elements in some of their AI systems, which could facilitate the integration of this new technology.

“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.”