The installation uses generative adversarial neural networks to reconstruct the faces of its observers from their biometric face markers that it detects. At the same time it attempts to improve its knowledge about how human faces look like by making every spectator part of its training data.