Deepfakes for the Good: A Beneficial Application of Contentious Artificial Intelligence Technology

Abstract

Deepfake algorithms are one of the most recent albeit controversial developments in Artificial Intelligence, because they use Machine Learning to generate fake yet realistic content (e.g., images, videos, audio, and text) based on an input dataset. For instance, they can accurately superimpose the face of an individual over the body an actor in a destination video (i.e., face swap), or exactly reproduce the voice of a person and speak a given text. As a result, many are concerned with the potential risks in terms of cybersecurity. Although most focused on the malicious applications of this technology, in this paper we propose a system for using deepfakes for beneficial purposes. We describe the potential use and benefits of our proposal and we discuss its implications in terms of human factors, security risks, and ethical aspects.