Moore's law predicts the rapid growth of computing power, but the demands of AI are outpacing it.
Light-based 'optical computers' using photons instead of electrons offer advantages like higher bandwidth and efficiency.
Researchers have used light for AI since the 1980s, with recent focus on matrix multiplication for neural networks.
In 2017, a breakthrough showed how light could be used for matrix multiplication in optical neural networks.
Recent advancements like the HITOP optical network aim to improve computation throughput and energy efficiency in optical computing.
[link] [comments]