New chip opens door to AI computing at the speed of light

Engineers at the University of Pennsylvania have developed a new chip that uses light waves rather than electricity to perform complex calculations essential to training AI. The chip has the potential to radically accelerate computer processing speed while reducing energy consumption.

The design of the silicon photonic (SiPh) chip integrates for the first time Benjamin Franklin Medal recipient and H. Nedwill Ramsey Professor Nader Enheta’s pioneering work in material manipulation at the nanoscale, allowing for the fastest possible means of communication. Perform mathematical calculations using some light. The SiPh platform uses silicon, a cheap and abundant element used in the mass production of computer chips.

The interaction of light waves and matter represents one possible avenue for computer development to overcome the limitations of today’s chips. This chip is based on essentially the same principles as chips from the early days of the computing revolution in the 1960s.

In a paper published in natural photonicsEngheta’s group, along with the group of Firooz Aflatouni, associate professor of electrical and systems engineering, describes the development of the new chip.

“We decided to join forces,” Engheta said, taking advantage of the fact that Aflatuni’s research group has developed nanoscale silicon devices.

Their goal was to develop a platform for performing what is known as vector-matrix multiplication, a central mathematical operation in the development and functioning of neural networks, the computer architectures that power today’s AI tools.

Instead of using a silicon wafer of uniform height, “we thin the silicon to, say, 150 nanometers” in only certain areas, Engheta explains. Without adding any other materials, these height variations provide a means to control the propagation of light through the chip. This is because it distributes the height changes and scatters the light in a specific pattern, allowing the chip to perform mathematical calculations. At the speed of light.

Due to constraints imposed by the commercial foundry that manufactured the chip, the design is already ready for commercial use and could be adapted for use in graphics processing units (GPUs), according to Aflatuni. It is said that there is a sex. He is interested in developing new AI systems.

“We can adopt the silicon photonics platform as an add-on, which will speed up training and classification,” Aflatuni says.

In addition to faster speeds and lower energy consumption, Enheta and Aflatuni’s chip also has privacy benefits. The ability to perform many calculations simultaneously eliminates the need to store sensitive information in the computer’s working memory, making future computers equipped with such technology virtually unhackable. .

“No one can access your information by hacking into a memory that doesn’t exist,” Aflatuni says.

Other co-authors include Vahid Nikkhah, Ali Pirmoradi, Farshid Ashtiani, and Brian Edwards of Penn Engineering.