Diffusion
The technique behind most AI image generators: start with noise, remove it step by step until a picture emerges.
A diffusion model is trained by taking real images and adding noise until they are unrecognizable, then learning to reverse the process. Given a text description, it starts from pure noise and removes the noise in stages, guided toward images that match the words. DALL-E 2, Stable Diffusion, and Midjourney all work this way, and the run of entries on AI art from 2022 onward is largely the story of this one idea.
Where it mattered in the record
-
Stability AI publishes the Stable Diffusion weights
Stability AI, working with LMU Munich researchers, RunwayML, EleutherAI and LAION, published the full weights of Stable Diffusion on Hugging Face on 22 August 2022 under the CreativeML OpenRAIL-M license. The model ran on a single consumer graphics card.