Being Foiled

Being Foiled (2020) is a series of artworks exploring how machine learning systems (GANs) perceive images as being fake. Algorithms for producing deepfakes have in recent years become remarkable good at producing images that look imperceptibly realistic, they do this by learning the difference between what is real and what is fake, and in turn, producing ever more realistic images. In this work we invert this objective, instead optimising the system to start generating what is sees as being fake, and in turn amplifying the uncanny nature of these machine hallucinations.

You can read more about the project in my blog post or my paper about the work.