Real or Fake : Now that is the question ..
Have you ever watched a video and wondered if the actors in the video are really who they are or is this just a crafted video ? Once in a while organisations or individuals are hit by this particular type of attack popularly known as deep fakes. You might think this happens only in sci-fi movies but it can be a very real threat in the political domain. To give a few examples
The crux of deep fake technology lies in utilising tools to modify real video clips in a manner that is advantageous to an adversary and spread the same through social media channels. To a naive user of such channels it is very difficult to differentiate between real and fake videos. Further it is possible that the level of the attack is so sophisticated that automated tools can also fail to detect visual differences in the images that form the video.
Deepfakes are crated using facial mapping technologies to create an accurate facial symmetry data set. They use AI to swap the face of a person on to the face of another person. Apart from this, voice matching technology is used to accurately copy the user’s voice.??The illustration below shows how this works in a simplified manner:
What makes face swapping technology possible is finding a way to force both latent faces to be encoded on the same features. Deepfakes solved this by having both networks sharing the same encoder, yet using two different decoders.
领英推荐
Deep fakes are sometime used to create audience engagement on varied channels and are even considered as a good use of technology especially within arts and culture marketing, community engagement, and nonprofit development. Take for example, the Mona Lisa. Researchers from Moscow's Samsung AI Centre and the Skolkovo Institute of Science have taken portraits such as da Vinci’s “Mona Lisa” and manipulated them to talk and move as though they are real people. This is truly far more advanced than simple animation and the technology involves identifying “facial landmarks in the painting” with an advanced algorithm. More around these kind of use-cases can be found in this article.?
Given that deep fake technology is like a double-edged sword one may often wonder about the ethics behind using the same. Take for example creation of pornographic material - it may be easy to just swap faces to create more variety in these sort of videos. While it seems like an innocent prank, it can actually create a lot of mental stress for the actor whose images are involved.The article here explains the ethics behind such usage of technology and indicates how consent from the given actor is very important before any modification of images / videos. Quoting from that article
If used correctly, Deep Fakes can reach an unprecedented level of?photorealism.?Distinguishing a real video from a synthetic one will become virtually impossible in the near future. Consequently, the ability to trace and verify the origin of content is a major challenge that the Internet will have to face very soon. We live in the post-truth era, where fake news is designed and created as a tool to gain political power.?The fact that most Deep Fake videos are obviously fake does not make them less dangerous.
Food for thought: It is seen that AI & Machine Learning are tools can be used for good or evil. What is your take around Deep Fake technology - in particular where does it lie in the spectrum between good use / bad use of tech ?