Demonstration of Neuromodulation‐inspired Stashing System for Energy‐efficient Learning of Spiking Neural Network using a Self‐Rectifying Memristor Array

Brain-inspired stashing system slashes AI energy consumption

28 July 2022

Mimicking the brain’s system for storing and retrieving memories in hardware based on artificial intelligence (AI) cut its energy consumption by more than one third.

The development of computing systems that run AI is advancing rapidly, but such systems consume a lot of energy. In their efforts to reduce this power consumption, engineers are seeking inspiration from the brain, which is highly energy efficient.

Now, a team led by researchers from the Korea Advanced Institute of Science and Technology (KAIST) in South Korea has demonstrated that AI systems can cut power consumption by mimicking neuromodulation in the human brain.

Specifically, they showed that a ‘stashing system’ that emulates the way the brain changes its neural topology in real time can cut power consumption by 37%.

Supported content

References

1. Advanced Functional Materials 2022, 2200337 (2022). doi: 10.1002/adfm.202200337