Scientists have created a new brain-inspired electronic device that can quickly detect unusual events while using far less energy than today’s artificial intelligence (AI) systems.
The breakthrough could help power future wearable health devices, self-driving cars, robots and cybersecurity systems that need to react instantly without draining their batteries.
The research, published in the journal Nature Communications, was led by engineers at Northwestern University.
Instead of copying the part of the brain responsible for thinking, the team looked to another important area called the cerebellum.
The cerebellum helps the brain work efficiently. Rather than carefully analyzing every piece of information, it mostly ignores normal, expected events.
It only reacts when something unexpected happens.
This allows the brain to save energy while still responding very quickly to danger or important changes.
The researchers designed a new electronic device that works in a similar way. Instead of constantly processing every bit of data, it focuses on detecting unusual events. This greatly reduces the amount of computing needed.
In laboratory tests, the device analyzed electrocardiogram (ECG) recordings, which measure the heart’s electrical activity.
It ignored normal heartbeats but quickly detected abnormal heart rhythms, known as arrhythmias. Impressively, it recognized an irregular heartbeat within about one-fifth of a heartbeat, achieving more than 98% accuracy.
The new system also required about 10,000 times fewer computer operations than traditional AI methods. Because it performs much less work, it uses far less energy while still producing fast and accurate results.
According to the researchers, this technology could improve many everyday devices. A wearable heart monitor could warn people about dangerous heart rhythm problems much sooner. A self-driving car could detect a person suddenly stepping onto the road more quickly. Robots could react faster to unexpected obstacles, while cybersecurity systems could spot suspicious computer activity before it becomes a major attack.
The new device builds on earlier work by the same research team. Traditional computers store information in one place and process it somewhere else. Constantly moving data between memory and processors uses a lot of electricity. To solve this problem, the researchers developed a special component called a memtransistor, which combines memory and computing into a single device, making it much more energy efficient.
The team also designed the memtransistor to imitate two types of signals found in the cerebellum. Under normal conditions, these signals balance each other. When something unusual happens, the balance changes, allowing the brain to notice the event almost instantly. The electronic device copies this behavior, helping it recognize important changes without wasting energy on routine information.
The researchers built the device using an extremely thin semiconductor material called molybdenum disulfide. By changing the direction of the electrical voltage, the same device can switch between two different operating modes, similar to how brain cells respond to changing situations.
The researchers now hope to make the technology even smarter by giving it the ability to learn. Just as the human brain gradually stops treating familiar events as surprising, future versions of the device may be able to adapt over time, making brain-inspired AI systems even more efficient and useful.
