In this SBIR Phase I program, we propose to evaluate the use of our novel optical neural network technology to enable efficient, low latency, edge-enabled artificial intelligence (AI) applications within the Air Force. As intelligence systems grow, their SWaP grows to an unmanageable point, due inherent inefficiencies in using conventional Complementary metaloxidesemiconductor (CMOS) based general purpose processors or AI accelerators like GPUs. Although, photonic AI accelerators have been attempted, the current approaches are lacking and are unable to operate at light speed due to numerous conversions back and forth between electrical and photonic domain, to enable non-linear activation. GenXComm technology enables photonic AI accelerators to which are low power, high throughput, and operate purely in the optical domain, enabling operation at light speed, with a SWaP enabling deployment on small Airborne platforms.optical neural networks,Silicon photonics,Neural networks,artificial intelligence,machine learning,deep neural networks
