What is a diffusion model?

The idea of the diffusion model is not that old. In the 2015 paper called “Deep Unsupervised Learning using Nonequilibrium Thermodynamics”[1], the Authors described it like this:

The essential idea, inspired by non-equilibrium statistical physics, is to systematically and slowly destroy structure in a data distribution through an iterative forward diffusion process. We then learn a reverse diffusion process that restores structure in data, yielding a highly flexible and tractable generative model of the data.