# AI Gets a Cerebellum > Source: > Published: 2026-07-10 19:16:46+00:00 #### The Problem Conventional AI continuously analyzes incoming data even when nothing has changed, wasting energy on unnecessary computation. Conventional AI continuously analyzes incoming data even when nothing has changed, wasting energy on unnecessary computation. Researchers developed a cerebellum-inspired memtransistor that ignores expected inputs and rapidly detects unexpected events while using far less energy than conventional AI. This approach could enable fast, always-on AI for applications like wearable health monitors, autonomous vehicles, robotics, and cybersecurity without the high energy demands of today's systems. Professor Mark Hersam, Research associate professor Vinod K. Sangwan The brain’s cerebellum doesn’t waste energy analyzing every moment. Instead, it constantly monitors the world for the unexpected—and springs into action only when something suddenly changes. Inspired by this remarkably efficient strategy, Northwestern researchers developed a new brain-like electronic device that consumes very little energy and detects novelties almost instantly. In proof-of-concept experiments, the device identified abnormal heart rhythms within one-fifth of a heartbeat and with more than 98 percent accuracy. The device also required roughly 10,000 times fewer computer operations than conventional AI approaches—paving the way for more energy-efficient AI. The breakthrough could enable a new generation of low-power, always-on AI systems for wearable health monitors, self-driving automobiles, autonomous robots, and cybersecurity systems that need to instantaneously recognize and react to unusual events without relying on massive, energy-hungry data centers. The [study](https://www.nature.com/articles/s41467-026-75212-4) was published July 10 in the journal *Nature Communications*. “In the world of brain-like computing, researchers typically try to mimic the cerebrum, which is often viewed as the brain’s ‘thought center,’” said Northwestern Engineering’s [Mark C. Hersam](https://www.mccormick.northwestern.edu/research-faculty/directory/profiles/hersam-mark.html), who co-led the study. “In our work, we developed a device that mimics the cerebellum, which controls reflex reactions seemingly without even thinking. The cerebellum is excellent at ignoring the expected and reserving its resources for reacting to the unexpected. That approach ultimately translates into lower energy consumption, and that is where we achieve orders of magnitude improvement.” An expert in brain-like computing, Hersam is the Walter P. Murphy Professor of Materials Science and Engineering, professor of medicine, and professor of chemistry at Northwestern, where he has appointments in the McCormick School of Engineering, [Northwestern University Feinberg School of Medicine](https://www.feinberg.northwestern.edu/index.html), and [Weinberg College of Arts and Sciences](https://weinberg.northwestern.edu/). He also is the chair of the department of materials science and engineering, director of the [Materials Research Science and Engineering Center](https://mrsec.northwestern.edu/), and member of the [International Institute for Nanotechnology](https://www.iinano.org/). Hersam co-led the study with [Vinod K. Sangwan](https://vinodksangwan.com/), a research associate professor at McCormick; [Indira M. Raman](https://neurobiology.northwestern.edu/people/core-faculty/raman-indira-m.html), the Bill and Gayle Cook Professor of Neurobiology at Weinberg; and [Amit Trivedi](https://ece.uic.edu/profiles/amit-ranjan-trivedi-phd/), an associate professor of electrical and computer engineering at the University of Illinois Chicago. **Mark C. Hersam****Walter P. Murphy Professor and Chair of Materials Science and Engineering** The new device represents the latest advance in [Hersam’s lab’s](https://www.hersam-group.northwestern.edu/) broader effort to rethink AI hardware from the ground up. Conventional computers constantly shuttle data back and forth between physically separate memory and processors—a process that consumes a significant amount of energy. Hersam’s group instead collapses memory and computation into a single device called a memtransistor. In a [2023 study published in Nature Electronics](https://www.mccormick.northwestern.edu/news/articles/2023/10/ai-just-got-100-fold-more-energy-efficient/), the team demonstrated that just two memtransistors could perform AI classification tasks that otherwise required more than 100 conventional transistors. That approach reduced energy consumption by roughly 100-fold. The new study pushes that concept beyond low-energy classification. Rather than simply making AI hardware more efficient, the Northwestern team redesigned the device to mimic a specific circuit in the cerebellum, which excels at detecting novelties and making split-second decisions. The approach allows AI to ignore routine information while immediately flagging unexpected events. For wearable heart monitors, that might mean detecting the first signs of an irregular heartbeat. For robots, it could mean recognizing when a person suddenly steps into their path. And for cybersecurity systems, it could mean spotting suspicious network activity before it escalates into a full-scale attack. “Today’s AI is remarkably good at recognizing patterns, but it often spends enormous amounts of computing power to continuously analyze streams of data—even when nothing has changed,” Hersam said. “Therefore, it burns energy on unnecessary analysis.” In the cerebellum, neural circuits contain two competing signals—one excitatory and one inhibitory—that constantly balance one another. During normal activity, the signals remain in equilibrium. But when something surprising occurs, that balance briefly shifts, alerting the brain that it needs to react. The Northwestern team recreated this same dynamic in its hardware. The engineers developed the device to perform two distinct roles. In one mode, it behaves like an excitatory synapse, gradually strengthening its response as signals continue. In the other mode, it acts like an inhibitory synapse, responding strongly at first before quickly fading away. Together, these complementary behaviors enable the device to distinguish ordinary occurrences from genuinely novel events—just as the cerebellum does. To build the device, the researchers used molybdenum disulfide, an atomically thin semiconductor known for its electrical properties. Then, they engineered an asymmetric transistor architecture in which one electrode partially overlapped the semiconductor through a thin insulating layer. That seemingly small design change fundamentally altered how electricity flows through the device. Simply reversing the direction of the applied voltage switches the memtransistor between excitatory and inhibitory modes. To test the system, the researchers gave the device a series of electrocardiogram (ECG) recordings, which contained both normal heart rhythms and arrhythmias. Instead of wasting energy by fully analyzing each heartbeat, the device successfully ignored normal heartbeats. But then it rapidly identified an abnormal heartbeat within mere milliseconds. “Our cerebellum-inspired memtransistor detected an irregular heartbeat within a fraction of a second, before the heartbeat even ended,” Hersam said. “That is more than twice as fast as conventional AI.” Next, Hersam plans to explore ways to mimic the cerebellum’s ability to learn and adapt over time. If a once-unexpected event occurs repeatedly, for example, the brain gradually learns and stops treating the repeated event as a novelty. “We have demonstrated one part of the cerebellum neural circuit, but there is more that we have not yet emulated,” Hersam said. “We intend to continue going down this path to mimic more and more of this complicated system.”