Accelerating Artificial Intelligence (AI) Computing to the Speed of Light

-By Eniola Elizabeth Fase

Artificial Intelligence (AI) is as of now a vital piece of our regular day-to-day existence on the web. For instance, search engines, for example, Google utilize intelligent ranking calculations, and video streaming features, for example, Netflix utilizes AI to customize film proposals.

Artificial Intelligence (AI) is as of now a vital piece of our regular day-to-day existence on the web. For instance, search engines, for example, Google utilize intelligent ranking calculations, and video streaming features, for example, Netflix utilizes AI to customize film proposals.

In the present scenario, the demand for AI is increasing, there is a need to discover approaches that increase AI speed without compromising its energy utilization. University of Washington-led group has presently brought up a framework that could be of help. This is an optical computing core model that utilizes phase change material.
On Jan 4, a paper was published by this team in Nature Communications titled “Programmable phase-change metasurfaces on waveguides for multimode photonic convolutional neural network”
An Illustration by Ryan Hoover demonstrating how Neural network computing speed and precision.
Source: https://www.ece.uw.edu/spotlight/ai-computing/

This framework is quick, energy effective, and fit for accelerating the neural networks utilized in machine learning and AI. The technology is additionally adaptable and directly favorable to cloud computing.

A UW partner teacher of both electrical and PC designing and material science, Mo Li explained that the hardware they created is upgraded to run algorithms of an artificial neural network, which is actually a backbone algorithm for machine learning and AI. He explained that this research advance will bring about AI centers and cloud computing more energy effective and run a lot quicker.”

The research group is among the first on the planet to utilize phase change material in optical computing to empower picture acknowledgment by an artificial neural network. Acknowledging an image in a photograph is simple for people to do. Still, this is computationally demanding for AI.

Since recognizing images is substantial, it is viewed as a benchmark trial of a neural network’s computing speed and accuracy. The group exhibited that their optical computing core, running an artificial neural network, could undoubtedly be successful.

 A UW electrical and computer engineering graduate student, Changming Wu also explained that Optical computing was first showed up as a concept during the 1980s, yet then it blurred in the shadow of microelectronics. He explained that due to the furthest limit of Moore’s law, advances in incorporated photonics and the demands of AI computing have been revised.

Reference:

Research Paper: Wu, C., Yu, H., Lee, S. et al. Programmable phase-change metasurfaces on waveguides for multimode photonic convolutional neural network. Nat Commun 12, 96 (2021). https://doi.org/10.1038/s41467-020-20365-z

https://techxplore.com/news/2021-01-ai.html

TechThoroughFare

Share on facebook
Share on twitter
Share on linkedin
Archives

One Response

Leave a Reply

Your email address will not be published. Required fields are marked *

Science

Latest

Trending