Cornell Tech Develops Optical Memory Link to Enhance AI Chip Efficiency
Researchers at Cornell Tech have developed an innovative optical receiver that can update AI model parameters using light, potentially reducing the energy demands of AI systems. This new technology involves projecting light directly onto a chip, which alters its memory through photocurrents generated by the light. The optical receiver, presented at the IEEE/JSAP Symposium on VLSI Technology & Circuits, aims to address the increasing memory demands of AI systems by reducing the reliance on energy-intensive electrical links. By using optical links, data can be transmitted with higher bandwidth and less energy loss compared to traditional metal wires. This advancement could significantly impact data centers, self-driving cars, and AI-powered robots by improving efficiency and reducing costs. 
