Deep Fakes – The ultimate fake.


Deep fakes are the latest innovation to give everyone sleepless nights. Now its impossible to tell if the person is actually saying or doing something in a video as it can  appear what its not. Take a look at the video below.

This is an example of AI trained video manipulation in practice, in this case its called Neural Voice Puppetry ( . More info on this technique and how its done at  I say AI  because its vastly more complex and time consuming than standard video manipulation.  AI technology, trained neural networks, can switch anyone anywhere into a video or photo they were never in and say things they never said. Deepfakes leverage powerful techniques from machine learning and artificial intelligence to manipulate or generate visual and audio content with a high potential to deceive.

The main machine learning methods used to create deepfakes are based on deep learning and involve training generative neural network architectures, such as autoencoders which learns to represent some input information very efficiently, and subsequently how to reconstruct the input from it’s compressed form or generative adversarial networks (GANs) which use an adversarial feedback loop to learn how to generate some information that “seems real” (i.e. looks the same/sounds the same/is otherwise indistinguishable from some real data, although a lot of deepfake videos these days are apparently generated by algorithms in which GANs don’t play a major role.

What this really means is that in the future, once deepfakes get better, you will not be able to tell the difference between reality and fake. The only hope is foundations such as Reality Defender 2020 a non-partisan and non-commercial effort to help reporters and campaigns uphold truth and ethical standards ( This is an initiative by the AI Foundation non-profit and is worth support.  Reality Defender, a platform under construction by the company AI Foundation, similarly hopes to tag and bag manipulated images and video before they can do any damage. Another tools is the defence of truth is Deeptrace , which works on an API that will act like a hybrid antivirus/spam filter, prescreening incoming media and diverting obvious manipulations to a quarantine zone, much like how Gmail automatically diverts spam before it reaches your inbox.