The nature of neural networks makes them leverage the patterns and features found in the training data to try and fool the viewer, to make them believe what they see is real. The visuals created with Generative Adversarial Networks appeal to us through the ambiguous imagery that evokes nostalgia and emotions. However, this visual generative process is not able to encapsulate the complex logic beneath the familiar concepts, such as language. With this work, we explore how the machine is pretending to speak but is struggling to make sense. The almost recognizable signage is stripped of meaning and takes a form and life of its own.
Generative AI work created by training a GAN model on a custom collection of 3000 night photographs.
2022
