Greek Scientists Create Fastest Ever AI Processor Harnessing Light

One of the fastest Artificial Intelligence (AI) processors in the world, which consists of photonic neurons and works with light instead of electricity, was created by the scientific team of the Department of Informatics of the Aristotle University of Thessaloniki (AUTH).

According to an announcement by AUTH, the processor is based on an original photonic neuron architecture, which allows the implementation of algebraic operations at very high speeds, simultaneously utilizing original neural network training techniques.

This combination led to the experimental demonstration of a photonic processor operating at speeds up to 50GHz (25 times faster than the speed of modern GPU processors) with the simultaneous possibility of 10 times lower power consumption.

Postdoctoral researcher in the IT department of AUTH Miltiadis Moralis – Peios explained that a global effort is underway to build computing systems in alternative ways because microelectronics technologies have reached their physical limits related to the size of electronic transistors.

One emerging technology uses light, which, as an electromagnetic wave, is faster and less disruptive to integrated optical circuits than electricity. The benefits of using light are speed and energy consumption.

At present, Moralis says, the effort is at a research level, however, moves are being made to put processors that work with light into production.

Fastest-ever AI processor at speeds higher than GPU and TPU

The research of the Department of Informatics of AUTH highlights that the use of light in the implementation of the operations of multiplication and addition allows the processors to operate at speeds many times higher than today’s corresponding computing units GPU of NVIDIA and TPU of Google.

The advantages of photonic technology in the area of information processing using neural networks is that it utilizes data in the form of light and photonic integration technologies to implement deep neural networks, which try to “mimic” the functioning of the human brain and of human neurons in the field of computers, with the aim of faster and more efficient processing of information.

Last March, some of the results of the University of Thessaloniki research team were presented at the world’s largest Optical Fiber Communication Conference (OFC), in San Diego, USA.

The three main researchers announcing results were the PhD candidates of the Department of Informatics of the AUTH Apostolos Tsakiridis, Georgios Giamougiannis and Christos Pappas, supervised by the Professor of the Department of Informatics of the AUTH Nikolaos Pleros.

The three postdoctoral fellows were selected by the International Committee of the Conference, out of 350 students from all over the world, to be among the seven finalists for the prestigious international OFC Corning Student Award.