Brain-Inspired Chip Mimics the Cerebellum to Spot Anomalies 10,000× More Efficiently

Northwestern University engineers have built a memtransistor chip that emulates the brain's reflex center, detecting heart arrhythmias within a fraction of a heartbeat using 10,000 times fewer operations than conventional AI.

Engineers at Northwestern University have unveiled a new kind of AI chip that doesn't think like a brain — it reacts like one. Instead of mimicking the cerebrum, the brain's conscious thought center, the device emulates the cerebellum, the ancient reflex gatekeeper that ignores the mundane and fires only when something unexpected happens.

The breakthrough, published in Nature Communications, uses a memtransistor built from atomically thin molybdenum disulfide (MoS₂). Unlike traditional computer chips that constantly shuttle data between separate processing and memory units — the infamous Von Neumann bottleneck — the memtransistor performs both memory and logic operations in the same physical space, slashing energy demands.

The chip's secret is its asymmetric design. By reversing the voltage direction, the device toggles between two biological modes: excitatory, where it builds and strengthens its response as a signal persists, and inhibitory, where it fires at maximum intensity then rapidly fades to zero. This mirrors the cerebellum's natural balance of competing neural signals.

In proof-of-concept trials, the chip detected abnormal heart rhythms within one-fifth of a heartbeat — and did it with more than 98% accuracy. Critically, it achieved this while performing 10,000 times fewer computer operations than conventional AI systems, pointing toward a future where wearable health monitors, self-driving vehicles, and industrial robots can react instantly without draining their batteries (Knowridge Science).

Because the cerebellum processes reflex-like sensorimotor data rather than abstract reasoning, the research opens a distinct path in neuromorphic computing — one focused not on making machines smarter, but on making them faster, leaner, and more reactive to the unpredictable (Digital Journal).