Engineers at Northwestern University have developed an artificial intelligence device that identifies abnormal heart rhythms with over 98% accuracy within one-fifth of a heartbeat. Inspired by the cerebellum, the brain region responsible for reflex reactions, the device achieves this near-instantaneous detection while using roughly 10,000 times fewer computer operations than conventional AI. This energy efficiency stems from mimicking how the cerebellum prioritizes unexpected changes, reserving resources for novel events rather than processing constant streams of ordinary data. “In our work, we developed a device that mimics the cerebellum, which controls reflex reactions seemingly without even thinking,” explains Northwestern’s Mark C. Hersam, who co-led the study; this breakthrough enables low-power, always-on AI systems for applications ranging from wearable health monitors to autonomous robots.
Cerebellum-Inspired Memtransistor Design Enables Efficient Novelty Detection
The pursuit of energy-efficient artificial intelligence has shifted focus from the cerebrum to the cerebellum. A new device developed by Northwestern University researchers achieves near-instantaneous novelty detection with significant power savings. This performance improvement stems from a fundamental redesign of AI hardware, moving beyond mimicking the brain’s processing center to emulate the cerebellum’s specialized role in filtering information and responding to change. The team’s innovation centers on a memtransistor, a device that integrates memory and computation into a single unit, drastically reducing energy consumption; this new iteration requires approximately 10,000 times fewer computer operations than standard AI approaches. “In brain-like computing, researchers typically try to mimic the cerebrum, often viewed as the brain’s ‘thought center,’” explained Mark C. Hersam, Walter P.
Constructed from molybdenum disulfide, an atomically thin semiconductor, the memtransistor’s asymmetric design allows it to seamlessly switch between excitatory and inhibitory modes based on voltage direction. This allows the AI to ignore routine data, conserving power, and instantly react to unexpected events, such as an irregular heartbeat identified within one-fifth of a heartbeat. The researchers envision applications ranging from wearable health monitors and self-driving cars to cybersecurity systems, all benefiting from this low-power, always-on capability. Hersam’s team intends to continue refining the design to further emulate the cerebellum’s adaptive learning capabilities, building on this initial success.
Our cerebellum-inspired memtransistor detected an irregular heartbeat within a fraction of a second, before the heartbeat even ended.
This design allows the device to function as both an excitatory and inhibitory synapse, mirroring the competing signals within cerebellar circuits. The team successfully recreated the cerebellar dynamic where signals remain balanced during normal activity, but shift when something unexpected occurs. This is achieved through a memtransistor that switches between modes simply by reversing the applied voltage. Hersam, co-lead author and Walter P. Murphy Professor of Materials Science and Engineering at Northwestern, explained that by ignoring routine data and focusing only on anomalies, the device significantly reduces energy consumption. The researchers are now focused on enabling the device to learn and adapt over time, further refining its ability to distinguish between truly novel events and those that have become familiar.
We have demonstrated one part of the cerebellum neural circuit, but there is more that we have not yet emulated.
Rapid ECG Analysis Demonstrates Sub-Heartbeat Anomaly Identification
Researchers are increasingly turning to biological systems for inspiration in artificial intelligence, and a team led by Mark C. Hersam at Northwestern University has demonstrated an efficient approach by modeling AI hardware after the cerebellum. Unlike conventional brain-inspired computing, which focuses on the cerebrum, Hersam’s group developed a device that mimics the cerebellum’s ability to rapidly detect unexpected changes, achieving a significant reduction in energy consumption. This efficiency stems from a fundamental shift in design; the team collapsed memory and computation into a single component called a memtransistor, reducing the need for constant data transfer between separate processing units. A prior study in 2023, published in Nature Electronics, showed that just two memtransistors could perform AI classification tasks that otherwise required more than 100 conventional transistors, achieving a 100-fold reduction in energy use.
The current work builds on this foundation by engineering the memtransistor to mimic the cerebellar circuit responsible for novelty detection. The device distinguishes between expected and unexpected events by utilizing both excitatory and inhibitory signals, mirroring the dynamic equilibrium found in the cerebellum. To test the device, the researchers analyzed electrocardiogram recordings, finding that it successfully ignored normal heartbeats while identifying abnormal heart rhythms within one-fifth of a heartbeat. This speed highlights the potential for real-time health monitoring and other applications requiring instantaneous responses.
In the world of brain-like computing, researchers typically try to mimic the cerebrum, which is often viewed as the brain’s ‘thought center.’
