Penn Engineers have developed a brand new chip that uses light waves, reasonably than electricity, to perform the complex math essential to training AI. The chip has the potential to radically speed up the processing speed of computers while also reducing their energy consumption.
The silicon-photonic (SiPh) chip’s design is the primary to bring together Benjamin Franklin Medal Laureate and H. Nedwill Ramsey Professor Nader Engheta’s pioneering research in manipulating materials on the nanoscale to perform mathematical computations using light — the fastest possible technique of communication — with the SiPh platform, which uses silicon, the low-cost, abundant element used to mass-produce computer chips.
The interaction of sunshine waves with matter represents one possible avenue for developing computers that supersede the restrictions of today’s chips, that are essentially based on the identical principles as chips from the earliest days of the computing revolution within the Nineteen Sixties.
In a paper in Nature Photonics, Engheta’s group, along with that of Firooz Aflatouni, Associate Professor in Electrical and Systems Engineering, describes the event of the brand new chip. “We decided to hitch forces,” says Engheta, leveraging the proven fact that Aflatouni’s research group has pioneered nanoscale silicon devices.
Their goal was to develop a platform for performing what’s generally known as vector-matrix multiplication, a core mathematical operation in the event and performance of neural networks, the pc architecture that powers today’s AI tools.
As a substitute of using a silicon wafer of uniform height, explains Engheta, “you make the silicon thinner, say 150 nanometers,” but only in specific regions. Those variations in height — without the addition of another materials — provide a way of controlling the propagation of sunshine through the chip, because the variations in height may be distributed to cause light to scatter in specific patterns, allowing the chip to perform mathematical calculations on the speed of sunshine.
Resulting from the constraints imposed by the industrial foundry that produced the chips, Aflatouni says, this design is already ready for industrial applications, and will potentially be adapted to be used in graphics processing units (GPUs), the demand for which has skyrocketed with the widespread interest in developing latest AI systems. “They will adopt the Silicon Photonics platform as an add-on,” says Aflatouni, “after which you possibly can speed up training and classification.”
Along with faster speed and fewer energy consumption, Engheta and Aflatouni’s chip has privacy benefits: because many computations can occur concurrently, there might be no have to store sensitive information in a pc’s working memory, rendering a future computer powered by such technology virtually unhackable. “Nobody can hack right into a non-existing memory to access your information,” says Aflatouni.
This study was conducted on the University of Pennsylvania School of Engineering and Applied science and supported partly by a grant from the U.S. Air Force Office of Scientific Research’s (AFOSR) Multidisciplinary University Research Initiative (MURI)to Engheta (FA9550-21-1-0312)and a grant from the U.S. Office of Naval Research (ONR) to Afaltouni (N00014-19-1-2248).
Other co-authors include Vahid Nikkhah, Ali Pirmoradi, Farshid Ashtiani and Brian Edwards of Penn Engineering.
