Creative, Developer, Technologist

Art & Artificial Intelligence

I utilized a GAN (Generative Adversarial Network) to generate realistic human faces and then implemented the GPT-2 Transformer-based AI text generation model to write poetry about being human.

Each of the hyper realistic faces were generated via StyleGAN (Dec 2018) using one of the pre-trained models (Trained on the CelebA-HQ and FFHQ datasets for one week using 8 Tesla V100 GPUs). None of these faces are real faces.

The text was generated with HuggingFace’s PyTorch implementation of OpenAI’s GPT-2 model. I utilized OpenAI’s pre-trained medium model (345M) for the text synthesis and conditioned the input with the following text: “What does it mean to be human?”

The text in the image was generated on the fourth attempt.