An artificial intelligence designed to work as a human could need periods of rest similar to those needed by human brains.
Researchers at Los Alamos National Laboratory discovered that neural networks experienced high benefits that were "similar to a good night's rest" when exposed to an artificial simulation of sleep.
"We were fascinated by the prospect of training a neuromorphic processor in a manner analogous to how humans and other biological systems learn from their environment during childhood development," said Yijing Watkins, a computer scientist at Los Alamos.
The discovery, made by the team of researchers, happened while working on a form of artificial intelligence designed to mimic how humans learn to see.
They realized that the AI became unstable during long periods of automated learning, as it attempted to classify objects using their dictionary definitions, without ever having any prior examples to compare them to.
When the AI was exposed to a state mimicking what a human brain experiences during sleep, the neural network's stability was restored.
The study detailing the research was presented at the Women in Computer Vision Workshop in Seattle on 14 June.
"The issue of how to keep learning systems from becoming unstable really only arises when attempting to utilize biologically realistic, spiking neuromorphic processors or when trying to understand biology itself," said Garrett Kenyon, a Los Alamos computer scientist and co-author of the study.
"The vast majority of machine learning, deep learning, and AI researchers never encounter this issue because in the very artificial systems they study they have the luxury of performing global mathematical operations that have the effect of regulating the overall dynamical gain of the system."