Abstract
In the last year, high-quality face-swapped videos have started circulating on the Internet, showing faked situations of politicians doing odd speeches and celebrities performing in pornography. The videos and the surrounding phenomenon have been dubbed Deepfakes, alluding to the fact that they are fakes produced by deep learning—a form of machine learning. Commentators have warned that misuse of face-swapped videos in the style of fake news could have serious consequences. The technology is part of an idea of a world in which false videos are wreaking havoc. This is one of the complex ways in which the Deepfakes technology is associated with socio-technical practices. This thesis sheds light on these associations through an ethnography of the Internet and autoethnographical accounts of engaging with Deepfakes and its tools. The Deepfakes technology was born publicly and explicitly turned into a Free Software phenomenon. It is being used, by the public, to forge pornography, create memes, do satire and perform technology demonstrations. A taxonomy of Deepfakes suggested herein also encompasses the Deceptive Deepfake: An imagined video so powerfully deceptive as to have serious effects on society. There are no records of such a video being made, and at this point it is certainly not easy. However, it is not entirely implausible. Nevertheless, the idea of the Deceptive Deepfake is already associated with change, no matter its actual existence. The notion of a social imaginary has been particularly helpful in understanding this association. A social imaginary consists of world views and social practices that reinforce each other. In the social imaginary examined herein, anyone can create or modify truths by producing Deceptive Deepfakes. This understanding of Deepfakes is essential to the publicly expressed connections between Deepfakes and the notion of post-truth. This association allows the posttruth condition to be reinforced by the idea of the Deceptive Deepfake, and the Deepfakes phenomenon to be reinforced by the notion of post-truth. The post-truth condition is the source of great interest and heated debate within the field of science and technology studies (STS). Its ideas have allowed us to see any practice capable of calling forth truths to be seen as tools of knowledge production worthy of analysis in their own right. The Deepfakes phenomenon could thus be understood as the recursive public of the Free Software movement effectively fighting to shift the balance of power over knowledge production, while institutions of power are working to uphold the status quo.