Latent space is a mathematical representation of a multidimensional space of possibilities in which neural networks encode data. In image-based generative AI, it's where the network stores learned patterns - each coordinate corresponding to one of millions of representations of a subject. By moving the cursor across the coordinates of latent space, you can explore the depth and capacity of a neural network.