Deepfakes (contraction of “deep learning” and “fake news”) are videos in which anyone can say anything, thanks to artificial intelligence. We have seen, for example, videos of Barak Obama or Donald Trump appearing to say words that in reality they never said, but it was for demonstration. For the first time, a politician, a candidate in the parliamentary elections in Delhi, India, decided to use this technology in a completely assumed way to campaign in a language he does not speak.
Manoj Tiwari made two videos. In the first real video, he speaks English, his mother tongue. In the other video, apparently identical, he seems to be speaking in the Hryanvi dialect, when he does not know a treacherous word of this language. After the holograms in the meetings, here are the deepfakes as weapons of political communication.
Certainly, if you look closely, the result is not perfect. The main part of the haryanvi video is identical to the English version, but at the level of the mouth we can see artifacts related to the manipulation of the image. It’s the artificial intelligence program that taped specific lip movements, and added a few pauses, to match the audio message, which is spoken by an actor. In the past, this kind of rigging required very powerful computers and long hours of processing. Today, this can be done more easily with software available for free on the internet.
All this obviously raises ethical questions. We feared that deepfakes would be used to make disinformation, but now they are coming in to make information or, at least, political communication. This candidate considers that it brings him closer to his voters, that it puts less distance than subtitles. However, this is a hoax, which raises questions. In any case, the video was widely shared in Whatsapp groups but that did not prevent the high-tech candidate from losing the elections.