Hong Kong University of Science and Technology (HKUST) engineers have developed a novel computing system that brings AI closer to quantum computing by operating at extremely low temperatures. Led by Prof.
SHAO Qiming, the team's innovation significantly reduces latency between AI agents and quantum processors while boosting energy efficiency. This was achieved using magnetic topological insulator Hall-bar devices.
The system allows AI accelerators to operate just tens of centimeters from quantum processors, enhancing speed and efficiency. Tests with chromium-doped bismuth-antimony-telluride (Cr-BST) achieved high accuracy in classification tasks, with simulations indicating a performance of 724 tera-operations per second per Watt (TOPS/W) for image recognition and quantum state preparation at 2 Kelvin.
The research, published in Nature Materials, opens new avenues for topological quantum-physics-based computing schemes.