Researchers at the University of Florida have developed a silicon chip that utilizes laser light and microscopic Fresnel lenses to perform convolution operations, a core function in artificial intelligence (AI), with significantly reduced energy consumption.
The chip integrates optical components directly onto silicon, enabling it to perform convolutions using light instead of solely relying on electricity. This approach drastically cuts down on energy consumption while accelerating processing speeds. Volker J. Sorger, the Rhines Endowed Professor in Semiconductor Photonics at the University of Florida and the study’s leader, emphasized the importance of this breakthrough, stating, “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.”
In prototype tests, the chip demonstrated its ability to classify handwritten digits with approximately 98% accuracy, a performance level comparable to that of conventional electronic chips. The system employs two sets of miniaturized Fresnel lenses, flat and ultrathin versions of traditional lenses, fabricated using standard semiconductor manufacturing techniques. These lenses, narrower than a human hair, are etched directly onto the chip’s surface.
The convolution process involves converting machine-learning data into laser light on the chip. This light then passes through the Fresnel lenses, which execute the mathematical transformation required for convolution. Finally, the result is converted back into a digital signal to complete the AI task.
Hangbo Yang, a research associate professor in Sorger’s group at UF and co‑author of the study, noted the novelty of this approach, stating, “This is the first time anyone has put this type of optical computation on a chip and applied it to an AI neural network.”
The team also demonstrated the chip’s capability to process multiple data streams concurrently using lasers of different colors, a technique known as wavelength multiplexing. Yang explained, “We can have multiple wavelengths, or colors, of light passing through the lens at the same time. That’s a key advantage of photonics.”
The research was a collaborative effort involving the Florida Semiconductor Institute, UCLA, and George Washington University. Sorger pointed out that companies like NVIDIA already incorporate optical elements in certain AI systems, which could facilitate the integration of this new technology.
Sorger envisions a future where chip-based optics are integral to everyday AI chips, stating, “In the near future, chip‑based optics will become a key part of every AI chip we use daily. And optical AI computing is next.”
