"It seems like science fiction, but it's not," says a researcher, describing a groundbreaking development from the National Taiwan University. Scientists have engineered a device that mimics the brain's information storage, using quantum spin, a fundamental property of subatomic particles. This innovation promises to revolutionize artificial intelligence by creating more energy-efficient and brain-like computing systems.
For decades, computer architecture has separated processing and memory, unlike the human brain, where neurons perform both tasks simultaneously. The Taiwanese team's device bridges this gap with a physical memory that emulates synapses, the tiny connections between neurons. The device boasts 11 stable states controlled solely by electric current.
The most promising prototype utilizes tilted magnetic anisotropy, allowing precise control over electron spin. This enables the device to simulate synaptic potentiation and depression, mirroring how neural connections strengthen or weaken. The device achieved 81.51% accuracy in image classification, close to the original unquantized model. This breakthrough paves the way for neuromorphic hardware that closely replicates the brain's architecture and function.