The Digital Age reshaped everything. Even our conception of reality. In the XXth century, philosophers questioned the originality of an artwork (“The Work of Art in the Age of Mechanical Reproduction” by Walter Benjamin). In the XXIst century, we question if something is real at all (and if there is something like a reality). With Machine and Deep Learning’s power, entirely new possibilities emerged: we can generate something that never existed before. And ontological discussions continue in new dimensions…
The technical progress is going on at a rapid pace. For example, Generative Adversarial Networks. Here you can see the quality of GAN progress on face generation, presented in a tweet by Ian Goodfellow (creator and developer of GANs):
And it was a 2018 tweet. The following year, StyleGAN came. And then, StyleGAN2. Based on the co-play of Generator and Discriminator this framework was able to generate photorealistic imagery. Not only the AI community was fascinated by new visual freshness. A website named “ThisPersonDoesNotExist.com” became viral and acted as a bridge between Data Science and a broad audience. It showcased AI’s capability to create lifelike portraits of not existing persons… or Deepfakes (a favorite media topic if it’s time to speak about visual AI).
DeepFake as a term means already delusion, fraud. But for me, ML/DL creations are full of unusual beauty. Forget “Uncanny Valley”.
It became a cultural movement, a metaphor for AI’s creative (and dangerous) potential, a meme. The “This XY Does Not Exist” meme.
In the following, I want to present the existing representations of non-existing things, crafted with Machine and Deep Learning’s power. The list will be ongoing updated with new findings.