The previous chapter introduced Generative Adversarial Networks (GANs), which is specific deep learning architecture learns the latent representation, or the coding, of the data. Another architecture that learns the latent representation of the data are autoencoders. Both architectures learn the latent representation of the data, but the ways that they understand and represent the data are different, and because of that the generated data is different. One of the main differences is the fact that autoencoders copy the input data, whereas GANs generate random data that is similar, but not the same. In this chapter, you will learn about autoencoders and how they can be used to generate new data.