Predictive coding—the theory that the brain predicts sensory input and adjusts predictions based on prediction errors—is a plausible computational model for how biological systems learn, and self-supervised learning in AI mirrors this fundamental learning mechanism.