Smart speakers and other devices that make up the Internet of Things (IoT) of the future could harness the power of a neural network to do more with less, researchers say.

A new system called MCUNet makes it possible to design tiny neural networks on IoT devices, even with limited memory and processing power. According to a paper by MIT scientists published on the preprint server Arxiv, the technology could open up new capabilities for smart devices while saving energy and improving data security.

The research "is one of those brilliant ideas that seems obvious when you hear it," John Suit, consulting CTO at robotics company KODA, said in an email interview. "It's an elegant approach to the problem. This research is so important because it will ultimately enable real-time optimization of neural networks for any device whose resources can be known to the algorithm."

IoT devices typically run on computer chips without an operating system, making it difficult to perform pattern recognition tasks such as deep learning. For more intensive analysis, IoT-collected data is often processed in the cloud, although it is vulnerable to hacking.