A new generation of computer chip powered by light rather than electricity could revolutionise artificial intelligence, researchers say.
A team of engineers in the United States has developed a prototype device capable of making AI calculations between 10 and 100 times more energy-efficient than today’s best chips.
The breakthrough centres on one of the most power-hungry operations in machine learning: convolution. This process allows AI systems to recognise patterns in photos, videos and even written text, but it consumes vast amounts of electricity on conventional processors.
Read More: From bandages with brains to reborn teeth — and Google’s $425M headache
The new design instead uses lasers and microscopic lenses etched directly onto circuit boards. In laboratory tests, the chip matched the accuracy of electronic chips—achieving 98% success when identifying handwritten digits—while using only a fraction of the energy.
“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 at the University of Florida and co-author of the study.
Lead researcher Volker J. Sorger, from the University of Florida’s Florida Semiconductor Institute, described the advance as “a leap forward for future AI systems”. He added:
“Performing a key machine learning computation at near-zero energy is critical to keep scaling up AI capabilities in years to come.”
The study, published on 8 September in Advanced Photonics, involved collaboration between the University of Florida, UCLA and George Washington University.
Also Read: AI avatars bring comfort but can’t replace grieving
How the chip works
The prototype integrates two sets of ultra-thin Fresnel lenses—miniature versions of those found in lighthouses but only a fraction of a human hair wide. Machine learning data is first converted into laser light on-chip, passed through the lenses, and then transformed back into digital signals to complete the task.
The use of light has another advantage: different coloured lasers can be used simultaneously to process multiple data streams in parallel, a feature known as wavelength multiplexing.
“We can have multiple wavelengths, or colours, of light passing through the lens at the same time. That’s a key advantage of photonics,” Yang explained.
Industry implications
Chipmakers, including market leader NVIDIA, already use optical components in some parts of their AI hardware. Sorger believes convolution lenses could soon be integrated into mainstream products.
“In the near future, chip-based optics will become a key part of every AI chip we use daily,” he said. “And optical AI computing is next.”
