Making AI More Energy Efficient

Northwestern University engineers have created a new tiny electronic gadget. It’s super efficient at doing smart computer tasks, using way less power than what we have now. This little champ can handle big loads of data and do AI stuff right away, without needing to send everything to the internet for help.

Because it’s small, doesn’t need much energy, and doesn’t slow down, it’s perfect for putting in wearable gadgets like smartwatches and fitness trackers. This means these devices can process data quickly and give you instant health information.

Put to the test

To check how well it works, the engineers tested it with heart data. The gadget not only spotted irregular heartbeats accurately but also figured out the exact problem type in nearly 95% of cases.

“Today, most sensors collect data and then send it to the cloud, where the analysis occurs on energy-hungry servers before the results are finally sent back to the user,” the researchers explain. “This approach is incredibly expensive, consumes significant energy and adds a time delay. Our device is so energy efficient that it can be deployed directly in wearable electronics for real-time detection and data processing, enabling more rapid intervention for health emergencies.”

Machine learning tools have a big task before they can work their magic on new data. They must learn to correctly sort training data into different categories. For instance, if we want a tool to group photos by color, it needs to know which ones are red, yellow, or blue. Humans do this easily, but it’s a tough job for a machine and eats up a lot of energy.

More efficient

Our current technology, which relies on silicon, needs over 100 transistors, each using its own energy, to sort data from big sets like electrocardiograms (ECGs). But Northwestern University’s tiny electronic device achieves the same result with just two components. This means it uses much less power and can fit nicely into everyday wearable devices.

The special thing about this device is its ability to adapt, thanks to its unique materials. Instead of using traditional silicon, the researchers built these tiny parts from two-dimensional molybdenum disulfide and one-dimensional carbon nanotubes. This choice allows the transistors to switch between different data-processing tasks, so we don’t need loads of separate silicon transistors for each step.

“The integration of two disparate materials into one device allows us to strongly modulate the current flow with applied voltages, enabling dynamic reconfigurability,” the researchers explain. “Having a high degree of tunability in a single device allows us to perform sophisticated classification algorithms with a small footprint and low energy consumption.”

In practice

To check how well the device works, the researchers used publicly available medical data. They started by teaching the device to understand information from heart rate readings (ECGs), a task that usually requires a lot of time from trained healthcare professionals. Then, they asked the device to tell apart six different types of heartbeats: normal, atrial premature beat, premature ventricular contraction, paced beat, left bundle branch block beat, and right bundle branch block beat.

The tiny electronic device did a great job, accurately recognizing each type of heart problem in a group of 10,000 ECG samples. What’s cool is that it does all of this without sending your data to the internet, which not only saves time for patients but also keeps their information private.

“Every time data are passed around, it increases the likelihood of the data being stolen,” the authors conclude. “If personal health data is processed locally — such as on your wrist in your watch — that presents a much lower security risk. In this manner, our device improves privacy and reduces the risk of a breach.”

Facebooktwitterredditpinterestlinkedinmail