Comment by Millon Madhur Das (millon-madhur-das) on A short introduction to machine learning · 2023-03-15T15:44:22.189Z · LW · GW

Just like language models are trained using masked language modelling and next token prediction, Dall-E was trained for image inpainting(predicting cropped-out parts of an image).  This doesn't require explicit labels; hence it's self-supervised learning. Note this is only a part of the training procedure, which is self-supervised and not the whole training process.