Sound Waves Give Neuromorphic Chips a Brain-Simulating Edge

2026-06-18 · Source: IEEE Spectrum · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Expert, short

Summary

A new study reveals that neuromorphic devices utilizing sound waves can more effectively emulate biological neurons, achieving faster operation and superior energy efficiency compared to electronic AI chips. Researchers, led by Xiaodong Yan at the University of Arizona, developed an acoustic synapse incorporating "phi-bits" – phase bits that encode multiple values and support quantum-like parallel computing. This novel device, constructed from three aluminum rods with ultrasonic transmitters and sensors, enables multiple simultaneous computations with significantly lower power requirements. It demonstrated 96.7 percent accuracy in classifying 150 iris flowers, using only 39 parameters and reaching peak accuracy 20 percent faster than a conventional multilayer perceptron. The acoustic synapse also consumes at most one-tenth the power of existing electronic neuromorphic hardware and can mimic complex neuromodulatory processes by simply adding an extra rod, allowing a single circuit to adapt to various functions. These findings were published on June 12 in Science Advances.

Key takeaway

For AI Hardware Engineers designing next-generation neuromorphic systems, this research suggests a significant shift towards acoustic components. You should investigate integrating sound wave-based acoustic synapses, as they offer dramatically lower power consumption and enhanced adaptability compared to conventional electronic designs. Consider exploring phi-bit technology to achieve higher connectivity and neuromodulatory capabilities within more compact hardware footprints.

Key insights

Acoustic synapses leveraging sound waves and phi-bits enable highly efficient, compact, and adaptable neuromorphic computing.

Principles

• Sound waves can encode multiple values for parallel processing.
• Synaptic plasticity is mimicked by modulating phi-bit phases.
• Neuromodulation can be achieved with simple hardware additions.

Method

An acoustic synapse was built using three aluminum rods, epoxy, and ultrasonic transducers to encode and interact sound waves, modulating phi-bit phases.

In practice

• Design neuromorphic systems with acoustic wave dynamics.
• Implement phi-bit phase modulation for adaptive learning.
• Integrate additional acoustic elements for neuromodulatory control.

Topics

• Neuromorphic Computing
• Acoustic Synapses
• Phi-bits
• Parallel Computing
• Synaptic Plasticity
• AI Hardware

Best for: AI Scientist, AI Hardware Engineer, Research Scientist

Related on AIssential

• Real-time processing of analog signals on accelerated neuromorphic hardware — Direct analog signal injection into neuromorphic hardware enables efficient, real-time, near-sensor processing.
• NeuroPlastic: A Plasticity-Modulated Optimizer for Biologically Inspired Learning Dynamics — Multi-signal plasticity modulation can enhance gradient-based deep learning optimization, especially with limited or noisy data.
• Otters++: A Time-to-first-spike Based Energy Efficient Optical Spiking Transformer — Otters++ leverages optoelectronic device decay for energy-efficient time-to-first-spike computation in spiking Transformers.
• Princeton builds bio-hybrid computer using living neurons — A 3D bio-hybrid computer integrates living neurons with electronics to create energy-efficient, pattern-recognizing neural networks.
• Amortized Inference of Neuron Parameters on Analog Neuromorphic Hardware — Amortized SBI effectively parameterizes analog neuron circuits on neuromorphic hardware, with summary networks outperforming handcrafted statistics.
• Closing the Theory-Practice Gap in Spiking Transformers via Effective Dimension — Spiking self-attention universally approximates continuous functions, enabling energy-efficient neuromorphic transformers with theoretical design guidance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by IEEE Spectrum.