[ad_1]

Penn engineers have developed a new chip that uses light waves instead of electricity to perform the complex math needed to train AI. This chip is capable of drastically speeding up the processing speed of computers while also reducing their energy consumption.

The design of the first silicon-photonic (SiPh) chip brings together pioneering research in manipulating materials at the nanoscale to perform mathematical computations using light by Benjamin Franklin Medal winner and H. Nedwill Ramsey Professor Nader Nghita. Is. – with the SiPh platform, which uses silicon, a cheap, abundant element used to mass-produce computer chips.

The interaction of light waves with matter represents a potential avenue for the development of computers that overcome the limitations of today’s chips, which are based on essentially the same principles as the early days of the computing revolution in the 1960s. were on the chips.

I in a paper Nature PhotonicsEnghita’s group, along with Feroz Platoni, associate professor of electrical and systems engineering, describes the development of the new chip. “We decided to join forces,” says Anghita, taking advantage of the fact that Platon’s research group has pioneered nanoscale silicon devices.

Their goal was to develop a platform to perform what is known as vector-matrix multiplication, a fundamental mathematical process in the development and operation of neural networks, the computer architecture that powers today’s AI tools. .

Instead of using a silicon wafer of uniform height, Nghetta explains, “you make the silicon thinner, say 150 nanometers,” but only in certain areas. These variations in height — without the addition of any other material — provide a means of controlling the propagation of light through the chip, as variations in height can be distributed to cause the light to scatter in specific patterns. , which allows the chip to perform Mathematical calculations on the speed of light.

Because of the constraints imposed by commercial foundries producing chips, the design is already ready for commercial applications, Platoni says, and could potentially be adapted for use in graphics processing units (GPUs). can go, for which the demand has increased massively. Interest in developing new AI systems. “They can adopt the silicon photonics platform as an add-on, and then you can accelerate training and classification,” Platoni says.

In addition to high speed and low power consumption, Engheta and Aflatouni’s chip has privacy advantages: because many calculations can be done simultaneously, there is no need to store sensitive information in the computer’s working memory, which In the future, computers powered by such technology may be rendered virtually unusable. . “No one can hack into non-existent memory to access your information,” Platoni says.

This study was conducted at the University of Pennsylvania School of Engineering and Applied Science and was supported in part by the US Air Force Office of Scientific Research (AFOSR) Multidisciplinary University Research Initiative (MURI) to Enghetta (FA9550-21). -1-) was supported by a grant to 0312) and a grant from the US Office of Naval Research (ONR) to Afaltouni (N00014-19-1-2248).

Other co-authors include Wahid Nikah, Ali Permradi, Farshid Ashtiani and Brian Edwards of Penn Engineering.

[ad_2]