University of Tokyo and NTT ran a physical neural net through a 200-meter optical fiber with a state-skipping trainer that barely needs intermediate measurements, so the model is computing in light instead of transistors.
pull down to refresh
pull down to refresh
University of Tokyo and NTT ran a physical neural net through a 200-meter optical fiber with a state-skipping trainer that barely needs intermediate measurements, so the model is computing in light instead of transistors.