Microsoft’s light-powered computer could run AI 100x faster

Microsoft Research has developed a prototype analog optical computer (AOC) that uses beams of light instead of electrons to perform computations. This novel approach enables the AOC to solve complex optimization problems and potentially run artificial intelligence (AI) workloads with up to 100 times greater speed and energy efficiency compared to traditional silicon-based digital processors. The prototype is constructed from commercially available components such as micro-LEDs, optical lenses, and smartphone camera sensors, which could facilitate cost-effective mass production. Additionally, the team created a “digital twin” software model that simulates the hardware’s behavior, allowing researchers to test and scale AI and optimization tasks and collaborate externally. Early demonstrations of the AOC’s capabilities include optimizing financial transactions in partnership with Barclays Bank, where it efficiently handled thousands of delivery-versus-payment securities transactions among multiple parties, showcasing its potential impact on large-scale financial systems. In healthcare, the digital twin was used to reconstruct MRI scans with promising accuracy, suggesting the technology could reduce scan times from 30