Milestone towards neuromorphic computing: Quantum materials exhibit brain-like ‘non-local’ behavior

According to the latest issue of Nano Letters, the Quantum Materials for Energy-Efficient Neuromorphic Computing (Q-MEEN-C) project led by the University of California, San Diego, USA, reported the latest research results: they discovered that the electricity transferred between adjacent electrodes Stimulation also affects non-adjacent electrodes, which is called nonlocality. This achievement is an important milestone towards the development of neuromorphic computing devices that mimic the functions of the brain.

Electrical stimulation delivered between adjacent electrodes also affects non-adjacent electrodes, which is called nonlocality.
Photo credit: Mario Rojas/UC San Diego

Computers are often thought to be more efficient than humans and can complete a complex mathematical equation in an instant. However, the human brain can process complex information quickly and accurately, such as identifying who is who by looking at a face only once, or instantly knowing the difference between a mountain and an ocean, and with almost no energy input. These simple human actions require a lot of processing and energy input for computers, and accuracy is not guaranteed.

Creating brain-inspired computers with minimal energy requirements will revolutionize every aspect of modern life. Previously, in the first phase of research, the Q-MEEN-C team successfully found ways to create or simulate the properties of single brain elements (such as neurons or synapses) in quantum materials.

This time, the research team has taken another important step forward in understanding and simulating brain functions. They performed computations on arrays containing multiple devices to simulate multiple neurons and synapses in the brain. While conducting these tests, they discovered that non-locality was theoretically possible and further refined the idea by translating simulations into real devices in the lab.

Traditionally, creating a transmission network to power devices such as laptops required complex circuits with continuous contacts, which was inefficient and expensive. The design concept of Q-MEEN-C is much simpler because the non-local behavior in the experiment means that all wires in the circuit do not have to be connected to each other.

Until now, pattern recognition tasks that the human brain can perform brilliantly can only be simulated by computer software. Artificial intelligence programs like ChatGPT and Bard use complex algorithms to simulate brain-based activities such as thinking and writing, but without corresponding advanced hardware support, the software will reach its limit at some point.

The research team has demonstrated that it is possible to replicate non-local behavior in a synthetic material. Next, they will find ways to improve the hardware and create more efficient learning machines, which will bring a new paradigm to the field of artificial intelligence.