Motivation and challenges for applying photonic neuromorphic computing technologies

Abstract

Artificial intelligence is playing an increasingly important role. The recent announcement of ChatGPT [1] and the following discussion on its economic impact is clearly demonstrating this. ChatGPT is based on the GPT-3 neural network; an architecture with over 100 billion parameters. The training and execution of networks with so many parameters are taking a large amount of energy [2], it is hence of utmost importance to explore new concepts and hardware that can make the important operations in neural networks more power efficient.