At a Glance
- Researchers developed a new artificial neuron using monolayer molybdenum disulfide to advance energy-efficient, brain-inspired computing for artificial intelligence applications.
- The device combines a memory capacitor and an inverter to precisely emulate a biological neuron’s membrane potential and its ability to fire electrical spikes.
- This hardware uniquely mimics intrinsic plasticity, a complex learning process where individual neurons adapt their own firing properties, which is crucial for advanced learning.
- A 3-by-3 array of neurons successfully demonstrated visual adaptation by dynamically adjusting its light sensitivity, similar to the human eye in bright and dim conditions.
- The system was utilized to simulate a neural network for image recognition, demonstrating its potential for practical and low-power computer vision and edge intelligence applications.
Researchers at Fudan University have developed a novel artificial neuron that more closely mimics the complex adaptability of the human brain, a significant step forward for brain-inspired computing. Published in Nature Electronics, the new hardware utilizes an ultrathin semiconductor material called monolayer molybdenum disulfide, or MoS₂, to create a device that could enhance the speed and energy efficiency of artificial intelligence systems. This advancement aims to meet the growing demand for hardware that can power AI applications directly on devices, a concept known as edge intelligence, without relying on the cloud.
The device functions as an “integrate-and-fire” neuron by combining two key electronic components. The first is a type of memory known as dynamic random-access memory (DRAM), which uses a tiny capacitor to store an electrical charge. This charge acts like a biological neuron’s membrane potential, building up until it reaches a tipping point. The second component, an inverter, then triggers an electrical spike, or “fire,” mimicking how a real brain cell communicates. This unique design enables the artificial neuron to emulate not only how connections between neurons change, but also how individual neurons adapt their own firing behavior—a sophisticated process known as intrinsic plasticity.
To test their creation, the scientists built a 3-by-3 array of artificial neurons and demonstrated its ability to perform tasks inspired by human biology. The array successfully replicated the way human eyes adapt to varying levels of brightness, a process involving both photopic (daylight) and scotopic (low-light) vision. “The module can also emulate the photopic and scotopic adaptation of the human visual system by dynamically adjusting its light sensitivity,” the authors wrote. This capability makes the technology particularly promising for advanced computer vision applications.
By successfully using the neuron module to run a bioinspired neural network for image recognition, the researchers confirmed its potential for practical, real-world use. This brain-like hardware could pave the way for more powerful and efficient AI systems that learn and adapt in a manner remarkably similar to the human brain. “Mimicking the full spectrum of learning and memory processes requires the interplay of multiple plasticity mechanisms, including intrinsic plasticity,” the researchers noted, highlighting their work’s contribution to this complex field.
References
- Fadelli, I. & Phys.org. (2025, August 30). Artificial neuron merges DRAM with MoS₂ circuits to better emulate brain-like adaptability. Phys.Org; Phys.org. https://techxplore.com/news/2025-08-artificial-neuron-merges-dram-mos.html
- Wang, Y., Gou, S., Dong, X., Chen, X., Wang, X., Sun, Q., Xia, Y., Zhu, Y., Zhang, Z., Wang, D., Zhang, J., Guo, X., Tong, L., Ma, J., Xu, Z., Xie, Y., Ma, S., Zhou, P., Chai, Y., & Bao, W. (2025). A biologically inspired artificial neuron with intrinsic plasticity based on monolayer molybdenum disulfide. Nature Electronics, 8(8), 680–688. https://doi.org/10.1038/s41928-025-01433-y
