Deep fakes: A scoping study of the evolution and use of the technology over time
There is a growing public debate and concern among key stakeholders in governments, civil liberty organizations and industry about the rapid proliferation of lifelike audio and video clips online of real people doing and saying things they never did or said. The so-called ‘deep fakes’ combine image manipulation with AI and machine learning algorithms to produce content that is as realistic as possible. Over the past 18 months, deep fakes have become associated with a range of social harms from revenge porn to the manufacture of political statements and the duping of CEOs to sign off large sums of money.
However, current debates are problematic in two ways. They are ahistorical and singular in the presentation of the technology as entirely new therefore divorced from its precursors in the form of image manipulation that originate with the advent of photography or its evolution in film production. The singularity is manifest in the neglect of the pioneering work done in the film to track and clone facial performances including the use of computer manipulations to bring dead actors back to life on the screen. The consequence is that current debates risk be skewed.
Our premise is that the technology is Janus-faced and both the sinister and entertaining need to be considered and historically located if governments looking to regulate are to moot appropriate responses.
Secondly, the academic debates are problematic. Most studies of deep fakes and their variants have been in computing and digital forensics and concerned with detection. Most social science/humanities studies on disinformation have focused on fake news as words rather than visuals and most film studies research on CGI has looked at animation rather than the cultural and social significance of the technology.
In mapping the evolving uses of the technology over time, we aim to contribute to current public debates on it and to set a foundation for more solid academic theorisation. We propose to set the basis for the work through a scoping study of existing source materials that support humanities and social science perspectives.