Greece 2023AI-generated images have nowadays become so realistic that distinguishing the real ones from the fakes is nearly impossible. Trained for days with massive data and GPUs scattered all around the world, they have crossed the borders of the Uncanny Valley into the hills of visual perfection. What about those digital beings that do not have the resources to complete their training? These low-res synthetic creatures are the low-class citizens of the Humanoid Realm. They can't afford to train forever nor to be GPU nomads. They are stuck in the Uncanny Valley, always striving to convince the Discriminator they look real enough, always dreaming of climbing the steep hill of human likeness. A short animated film exploring AI through AI, this project aims to shed light on the inner workings of AI algorithms, specifically those responsible for creating realistic human faces, like the GAN models. GAN systems (Generative Adversarial Networks) are machine learning algorithms comprising two neural networks: the Discriminator and the Generator. The core idea behind this model is a fascinating game played between these two entities: the Generator continuously produces images in an attempt to deceive the Discriminator into perceiving them as real rather than synthetic. In response, the Discriminator assesses each image's authenticity. This intriguing back-and-forth between the two networks is played out repeatedly, resulting in the production of increasingly realistic images. The video depicts a fictional dialogue between the Discriminator and the synthetic creatures it evaluates. These creatures engage in a Sisyphean effort, an endless struggle to reach an elusive perfection, as they strive to deceive the Discriminator and rise above their pixelated origins.
ANIMATION MARATHON
