Neuromorphic and optical computing research advances brain-like AI

Researchers are developing new computing frameworks and hardware prototypes designed to mimic biological nervous systems to improve efficiency in artificial intelligence and machine vision.

At RMIT University, a team led by Professor Sumeet Walia has developed an early-stage neuromorphic vision prototype. This system aims to integrate sensing, memory, and information processing onto a single chip to reduce data transfer, energy use, and latency. The prototype, which utilizes atom-thin semiconductor materials like molybdenum disulfide, has demonstrated the ability to recognize patterns, shapes, and movement locally. The university has filed an international patent for this technology, which could eventually support smart bionic eyes.

In a separate development, researchers from the University of Shanghai for Science and Technology have proposed the Brain-like Optical Computing (BOC) framework. Published in the National Science Review, this framework seeks to address the bottlenecks of conventional electronic computing by integrating memory and decision-making directly at the hardware level using photonic architectures. The BOC framework combines optical neural networks, spiking neural networks, and optical memristive devices to enable real-time, event-driven sensing and learning for autonomous systems like robots and vehicles.