As artificial intelligence drives demand for faster and more energy-efficient hardware, ferroelectric memristors are attracting growing attention. A new review outlines how these devices combine memory, computation, and learning capabilities in a single platform, offering new opportunities for intelligent computing beyond the traditional von Neumann architecture.

A research team at Peking University proposed an optoelectronic distributed computing system based on an on-chip all-optical supernode. By integrating high-speed silicon photonic transceiver chips, low-loss optical circuit switching (OCS) chips and multiple computing nodes, the system enables intermediate data between different layers of a deep neural network to be directly transmitted and switched through an on-chip high-speed all-optical supernode. This approach reduces latency overheads associated with conventional data movement, packet scheduling and buffering.

As artificial intelligence expands from cloud computing into robots, autonomous vehicles and other real-world systems, conventional electronic architectures face increasing challenges in latency, bandwidth and energy consumption. Researchers from the University of Shanghai for Science and Technology propose Brain-like Optical Computing (BOC), a new photonic framework integrating sensing, memory and computation for future edge and embodied intelligence.