New Chips Open the Door to Speed of Light AI Computing

Engineers at the University of Pennsylvania have developed a new type of chip that uses light instead of electricity to perform complex mathematical operations necessary for training artificial intelligence (AI). This chip has the potential to fundamentally accelerate the processing speed of computers while also reducing energy consumption. The relevant research was published in the latest issue of Nature Photonics.

This chip combines for the first time the pioneering research of Benjamin Franklin Medal winner Nader Engeta on manipulating materials at the nanoscale with the Silicon Photon (SiPh) platform. The former involves using light for mathematical calculations; The latter uses silicon, a cheap and abundant element used for large-scale production of computer chips.
The interaction between light waves and matter represents a possible way to develop computers, which is not limited by the limitations of current chips. The principle of the new chip is essentially the same as that of chips in the early stages of the computing revolution in the 1960s.
The researchers described the development process of this chip in the paper. Their goal is to develop a platform for performing vector matrix multiplication. Vector matrix multiplication is a core mathematical operation in the development and functionality of neural networks, and neural networks are the computer architecture that supports AI tools today.
Enheta explained that they can make silicon wafers thinner, such as 150 nanometers, and use highly uneven silicon wafers, but this is limited to specific areas. Without the need to add any other materials, these height changes provide a way to control the propagation of light in the chip, as height changes can cause light to scatter in specific modes, allowing the chip to perform mathematical calculations at the speed of light.
In addition to faster speed and less energy consumption, the new chip also has privacy advantages. Due to the fact that many calculations can be performed simultaneously, there is no need to store sensitive information in the computer’s working memory, making future computers using such technologies almost inviolable.