Neuromorphic Computing With Sound Waves Cuts Power Use

Neuromorphic Computing With Sound Waves Cuts Power Use

https://spectrum.ieee.org/neuromorphic-computing-acoustic-chips

Publish Date: 2026-06-18 09:00:01

Source Domain: spectrum.ieee.org

Neuromorphic computing, which mimics the brain’s operations, holds the promise of dramatically reducing energy consumption compared to conventional electronic AI chips. Conventional neuromorphic devices, however, still fall short of the complexity of human neuron connections, limiting their efficiency and capabilities. A new study by Xiaodong Yan at the University of Arizona proposes a solution by using sound waves to enhance neuromorphic devices. Yan and his research team have developed an acoustic synapse using acoustic properties to perform quantum-like logic gates that facilitate parallel computing. This approach achieves greater energy efficiency than electronic neuromorphic hardware and allows for the mimicking of synaptic plasticity and neuromodulation. In tests, the acoustic synapse exhibited 96.7 percent accuracy at classifying iris flowers and performed faster and with fewer parameters than an electronic counterpart. These findings suggest future neuromorphic systems that combine wave dynamics with conventional computing could lead to more efficient information processing.

Key Points:
– Neuromorphic computing aims to emulate the brain’s computing efficiency, using exponentially less energy than conventional electronics.
– The new acoustic synapse uses sound waves to achieve quantum-like processing, enhancing parallel computing and efficiency.
– Experiments showed that the acoustic system performed significantly faster and with a smaller parameter set compared to electronic neuromorphic devices, consuming up to ten times less power.
– The acoustic synapse can mimic synaptic plasticity and neuromodulatory processes, providing adaptability crucial for contextual learning and attention modulation.