Cornell Tech Develops Optical Memory Link to Enhance AI Efficiency in Robotics
Cornell Tech researchers have developed a new optical receiver that could significantly enhance AI efficiency in robotics by reducing energy demands. The receiver uses light to directly alter its memory, bypassing traditional energy-intensive electrical links. This innovation was presented at the IEEE/JSAP Symposium on VLSI Technology & Circuits. The technology aims to address the growing memory demands of AI systems, particularly in data centers, self-driving cars, and edge applications like AI-powered robots. By using optical links, data can be moved at high bandwidth with less energy loss compared to metal wires. The new receiver design allows chips to adjust AI model parameters without relying on power-hungry analog circuits, enabling fully digital optical communication. 
