Light-based Computer System Could Revolutionize Machine Learning

An MIT-led team has developed a groundbreaking system that uses light-based computations instead of traditional electronics to significantly improve energy efficiency and compute density for machine-learning programs. The team’s first experimental demonstration of the system reported in Nature Photonics showed over 100-fold enhancement in energy efficiency and a 25-fold improvement in compute density compared to state-of-the-art digital computers. This breakthrough could potentially revolutionize machine-learning programs, making them several orders of magnitude more powerful than current models and paving the way for large-scale optoelectronic processors.