Deepfake

Deepfake is a term used to describe a specific kind of imaging software, one that has recently become more and more important in the last two years. The term itself is a portmaneau, in which the terms deep learning and fake were mashed together in order to describe a type of advanced form of human image synthesis that relies on the use of artificial intelligence in order to perform its intended function. The phrase was coined only in 2017, and is now known across the world as one of the potentially most invasive kinds of imaging software in common use amongst the general populace. What deepfake does that has caused such a significant controversy is to combine and superimpose specific digital images or videos onto the source images or videos. This means that by using a sophisticated kind of machine learning technique known as ‘generative adversarial network’, deepfake can allow any user to create things like fake celebrity pornography or other malicious fake news.  The research that led to the eventual creation of deepfake began as far back as 1997, with the Video Rewrite program that was designed to modify videos of people talking so that their faces matched up to a different audio track. For some reason, there seems to be very little mention of this kind of machine learning techniques being used for that purpose for almost twenty years after the release of the Video Rewrite program. Perhaps the potential for mischief making simply slipped under the radar …