DeepMind AI teaches itself about the world by watching videos

He showed the image recognition network stills taken from short videos while the audio recognition network was trained on 1-second audio clips taken from the same point in each video. The algorithm learned to recognise audio and visual concepts, including crowds, tap dancing and water, without ever seeing a specific label for a single concept. But until we reach that point, self-supervised learning might be a good way of training image and audio recognition algorithms without input from vast amounts of human-labelled data. "Most of the data in the world is unlabelled and therefore it makes sense to develop systems that can learn from unlabelled data," Agrawal says.