Silicon photonic chip accelerates AI and reduces energy consumption
Scientists at the University of Florida have developed a photonic chip that performs key artificial intelligence operations using light, reducing energy consumption and speeding up data processing. The prototype has already demonstrated high accuracy and the ability to handle multiple streams of information simultaneously.
Ingenium
Researchers at the University of Florida have developed a silicon photonic chip that performs convolution operations using light instead of traditional electronic computations.
The device utilizes laser light and microscopic Fresnel lenses to process machine learning data. The data is converted into laser light, passes through the lenses—which carry out mathematical transformations—and is then converted back into a digital signal.
During testing, the chip was able to recognize handwritten digits with 98% accuracy. The system can simultaneously process multiple data streams by using lasers of different colors.
The Fresnel lenses are manufactured using standard semiconductor technologies and are applied directly onto the chip. The research findings have been published in the journal Advanced Photonics.
The development involved the Florida Semiconductor Institute, UCLA, and George Washington University.