AI network simulations become unstable following continuous periods of unsupervised learning. When the networks were exposed to states that are analogous to human brain waves during sleep, the stability was restored.