Scientists from Aristotle University of Thessaloniki’s (AUTH) Department of Informatics have developed the fastest artificial intelligence (AI) processor in the world that runs on light rather than electricity.
Based on a unique photonic neuron design, the processor performs algebraic operations at high rates while simultaneously leveraging neural network training methodologies. The result is an experimental demonstration of a photonic processor that runs at speeds of up to 50GHz, which is 25 times faster than existing GPU processors while requiring 10 times less power.
Photonic technology has advantages in information processing with neural networks because it uses data in the form of light and photonic integration technologies to create deep neural networks. In the world of computers, these networks try to “mimic” the functioning of the human brain and neurons, resulting in quicker and more efficient information processing.
Miltiadis Moralis-Peios, a postdoctoral researcher at the AUTH IT department, notes that a global effort is ongoing to build alternative computing systems since microelectronics technologies have hit physical limits in terms of the size of electronic transistors. Light, which is quicker and less disruptive to integrated optical circuits than electricity, is described as a promising technology.
The sources for this piece include an article in GreekReporter.