The concept of data augmentation involves increasing the size of a training set while intentionally distorting the images in a way that doesn't change the image's nature. This technique has evolved in recent years and has become vital in self-supervised learning techniques to pre-train vision systems.