Optical and electrical characteristics of the MBN. Credit: Nature Electronics (2026). DOI: 10.1038/s41928-025-01529-5

Over the past decades, computer scientists have developed increasingly advanced artificial intelligence (AI) systems that perform well on various tasks, including the analysis or generation of images, videos, audio recordings and texts. These systems power various highly performing software, including automated transcription apps, large language model (LLM)-powered conversational agents like ChatGPT, and various other platforms.

Despite their remarkable performance, most AI systems are computationally intensive and can rapidly drain the energy of existing devices. Electronics engineers have thus been trying to develop brain-inspired hardware systems that are better suited for running these systems.

Researchers at the Hong Kong University of Science and Technology, ETH Zurich and Université de Bourgogne Europe recently developed new artificial neurons based on memristors, nanoscale electronic components that change their electrical resistance based on the current that previously flowed through them. These brain cell-inspired hardware components, presented in a paper published in Nature Electronics, could be used to create dense and three-dimensional neural networks that are better suited for the implementation of AI systems.

"The continuing development of artificial intelligence requires more powerful computing architectures," Yue Zhou, Yuetong Fang and their colleagues wrote in their paper. "However, the large footprint of complementary metal–oxide–semiconductor-based neurons and constraints on electric routing hinder the scaling of conventional artificial neurons and their synaptic connectivity. We show that memristive blinking neurons can be used to build scalable photonically linked three-dimensional neural networks."

Developing more scalable AI hardware

The artificial neurons created by Zhou, Fang and their colleagues essentially consist of memristors made of silver and an insulating polymer. When they receive enough incoming electrical signals, the neurons emit light pulses that can be used to communicate with other artificial neurons.

As light can travel without wires, the team’s neurons eliminate the need for bulky electronic circuits or wiring. This means that they could be used to create compact 3D architectures in which they are densely packed, not connected by wires, and yet can still effectively communicate with each other.

"Our artificial neuron is based on a silver/poly(methyl methacrylate)/silver metal–insulator–metal memristive switching in-plane junction," wrote the authors. "Its resistive switching relies on atomic-scale filamentary dynamics and the device emits photon pulses on integrating a critical number of incoming electrical spikes, which eliminates the need for bulky peripheral circuit read-out and electrical wiring for transmitting signals."

Initial results and future possibilities

To assess the potential of their artificial neurons, the researchers used them to create two three-dimensional neural networks. They then tested these networks’ performance on two different tasks that are commonly tackled by AI algorithms, namely speech classification and the recognition of handwritten digits (i.e., numbers 0 to 9).

"We use the memristive blinking neuron, which has a footprint of 170 nm × 240 nm, to build a photonically linked three-dimensional spiking neural network," wrote Zhou, Fang and their colleagues. "We show that the network can perform a four-class classification task within the Google Speech dataset with an accuracy of 91.51%. We also create a high-density artificial neuron array with a pitch of 1 μm and show that it can perform an MNIST classification task with an accuracy of 92.27%."

The results of the first tests performed by Zhou, Fang and their colleagues highlight the potential of their artificial neurons for the development of compact and highly performing AI hardware. In the future, the team’s neurons could be improved, used to create other artificial neural networks and assessed on a broader range of tasks.

Written for you by our author Ingrid Fadelli, edited by Lisa Lock, and fact-checked and reviewed by Robert Egan—this article is the result of careful human work. We rely on readers like you to keep independent science journalism alive. If this reporting matters to you, please consider a donation (especially monthly). You’ll get an ad-free account as a thank-you.

More information: Yue Zhou et al, Photonically linked three-dimensional neural networks based on memristive blinking neurons, Nature Electronics (2026). DOI: 10.1038/s41928-025-01529-5

© 2026 Science X Network

Citation: New light-emitting artificial neurons could run AI systems more reliably (2026, January 29) retrieved 29 January 2026 from https://techxplore.com/news/2026-01-emitting-artificial-neurons-ai-reliably.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.