Autophagy

Autophagy — from Greek auto- (self) and phagein (to eat) — is the biological process by which cells degrade and recycle their own components. In systems theory, the term has been extended to describe any self-consuming dynamic in which a system's outputs become its inputs in a way that degrades rather than sustains the system. The canonical example is model collapse in machine learning: a generative model trained on synthetic data enters a recursive loop where its own outputs contaminate its training distribution, progressively eroding statistical diversity. Biological autophagy is regulated; systems autophagy is not. The difference is feedback: a cell knows when to stop consuming itself. A data flywheel does not.