I recently finished a personal project that uses a machine learning model called a variational autoencoder (VAE) to generate new Lego faces. Here is some example output: The results range from fairly ...