Deep-learning methods have proved successful recently for solving problems in image analysis and natural language processing. One of these methods, convolutional neural networks (CNNs), is revolutionizing the field of image analysis and pushing the state of the art. CNNs consist of layers of convolutions with trainable filters. The input to the network is the raw image or seismic amplitudes, removing the need for feature/attribute engineering. During the training phase, the filter coefficients are found by iterative optimization. The network thereby learns how to compute good attributes to solve the given classification task. However, CNNs require large amounts of training data and must be carefully designed and trained to perform well. We look into the intuition behind this method and discuss considerations that must be made in order to make the method reliable. In particular, we discuss how deep learning can be used for automated seismic interpretation. As an example, we show how a CNN can be used for automatic interpretation of salt bodies.