"Light-based Machine Learning System Boosts Efficiency and Power of Large Language Models"

Researchers at MIT have developed a machine-learning system that uses light-based computations instead of electrons, potentially enabling machine-learning programs that are orders of magnitude more powerful and energy-efficient than current models. The system, demonstrated using hundreds of micron-scale lasers, achieved over 100-fold improvement in energy efficiency and a 25-fold improvement in compute density compared to state-of-the-art digital computers. The technology could pave the way for large-scale optoelectronic processors capable of running advanced machine-learning tasks on small devices like cellphones. The researchers anticipate further substantial improvements and expect the system to be commercially scalable within a few years.