New Semiconductor Enables AI Processing at Light Speed

New Semiconductor Enables AI Processing at Light Speed

Engineers at the University of Pennsylvania have created a novel chip that employs light waves instead of electricity to execute the difficult calculations required for AI training. The chip has the ability to significantly increase the processing speed of computers while also lowering energy usage.

The silicon-photonic (SiPh) chip design is the first to combine Benjamin Franklin Medal Laureate and H. Nedwill Ramsey Professor Nader Engheta’s pioneering research in manipulating materials at the nanoscale to perform mathematical computations using light—the fastest possible mode of communication—with the SiPh platform, which employs silicon, the low-cost, abundant element used to mass-produce computers.

The interaction of light waves with matter is one potential option for constructing computers that outperform the constraints of today’s chips, which are largely built on the same concepts as chips from the beginning of the computing revolution in the 1960s.

New Semiconductor Enables AI Processing at Light Speed

In a report published in Nature Photonics, Engheta’s group, along with Firooz Aflatouni, Associate Professor in Electrical and Systems Engineering, describe the novel chip’s design.

“We decided to join forces,” adds Engheta, citing the fact that Aflatouni’s research group pioneered nanoscale silicon devices.

Their goal was to provide a platform for doing vector-matrix multiplication, a fundamental mathematical operation in the design and operation of neural networks, the computer architecture that underpins today’s AI products.

Instead of utilizing a uniform-height silicon wafer, Engheta explains, “you make the silicon thinner, say 150 nanometers,” but only in specified areas. Without the addition of any other materials, the variations in height provide a means of controlling the propagation of light through the chip, as the variations in height can be distributed to cause light to scatter in specific patterns, allowing the chip to perform mathematical calculations at the speed of light.

Because of the constraints imposed by the commercial foundry that produced the chips, Aflatouni claims that this design is already ready for commercial applications and could potentially be adapted for use in graphics processing units (GPUs), which are in high demand due to widespread interest in developing new AI systems.

“They can adopt the Silicon Photonics platform as an add-on,” Aflatouni explains, “and then you could speed up training and classification.”

In addition to quicker speeds and lower energy consumption, Engheta and Aflatouni’s chip features privacy benefits: Because several computations can occur concurrently, there will be no need to store sensitive information in a computer’s working memory, making a future computer powered by such technology virtually unhackable.

“No one can hack into a non-existing memory to access your information,” Aflatouni said.

Co-authors from Penn Engineering include Vahid Nikkhah, Ali Pirmoradi, Farshid Ashtiani, and Brian Edwards.