Mathematics Computations Using Light Waves
University of Pennsylvania engineers have created a chip that makes use of light waves, rather than electricity, to perform math needed to train AI. This new Silicon Photonic chip will dramatically accelerate the processing speed of computers while cutting down on their energy requirements.
The silicon-photonic (SiPh) chip's design is the first to use pioneering research in manipulating materials at the nanoscale to perform mathematical computations using light with the SiPh technology. Benjamin Franklin Medal Laureate and H. Nedwill Ramsey Professor Nader Engheta teamed up to develop this new chip. Light is the fastest possible known means of communication, providing possibilities for faster computing speeds.
Overcoming Limitations with Traditional Chips
Light interacting with solid matter is one possible way to develop computers that break through the limits of today's chips, which are still the same in principle as chips from 1960s.
"We decided to join forces," says Engheta, leveraging Aflatouni's research group's new nanoscale silicon devices.
They set out to develop a device for performing vector-matrix multiplication, a key math operation in the development and function of neural networks.
How the Silicon Photonic Chip Works
The silicon wafer used is thinner in some specific regions. This allows variations in height without adding other material layers. This provides a means of controlling light travel through the chip by causing light to scatter in specified patterns, enabling math calculations at light speed.
This design is already ready for commercial applications, and could potentially be altered to create new graphics processing units (GPUs). The demand for GPUs has recently massively increased. The Silicon Photonics chip could be used as an add-on to hardware used for AI systems.
Less Hackable
Additionally, Engheta and Aflatouni's chip has enhanced privacy: Because many computations can happen simultaneously, there is no need to store sensitive information in a computer's working memory, which will make a future computer using the SiPh based system essentially unhackable.
This is a remarkable breakthrough which could create a new paradigm in computing and AI systems.
Source: Phys.org
