Deepfakes
Synthetic media produced by AI technologies that superimpose existing images or videos onto source images or videos to create realistic likenesses.
Deepfakes are created through AI-based techniques, particularly ML (Machine Learning) methods such as Deep Learning. The resulting media is so persuasively realistic that it can be hard to distinguish from authentic materials. The term 'deepfake' combines 'deep learning' and 'fake', reflecting the underlying technology and the faked output. Deepfakes have compelling applications in entertainment and creative arts, such as dubbing films in different languages using the original actors' voices and faces. However, they also pose significant ethical and societal implications, particularly manipulating public opinion, impersonating individuals, or creating non-consensual explicit content.
The concept of Deepfakes came into existence around late 2017, born in online communities (like Reddit) where labs began creating and distributing software that utilized Deep Learning techniques to replace faces in video content. By 2018, the phenomenon had reached mainstream awareness due to increasing quality and disseminated instances of misuse.
Hany Farid, a digital forensics expert who specializes in analyzing the credibility of photos and videos, has become a notable figure in the fight against deepfakes. Other major contributors include software developers and AI researchers from companies and institutions such as Google, Facebook, and DARPA, focusing both on creating and detecting deepfake technology.