Deepfakes is a portmanteau of “deep learning” and “fake” where deep learning techniques—auto-encoders and generative adversarial networks (GANs) are used to transpose someone else’s image on an existing image or a video like a mask. Deepfakes use large volumes of data and artificial intelligence to learn how the face would look like when viewed from different angles. The GANs are dueling neural networks where one creates an image and the other rates it as real or fake. This enables the networks to work in tandem to create images with increasing precision until the time the grading network cannot tell between a real and fake image.
Anyone can create deepfakes
At present, images can be superimposed only below the forehead, but Stanford University has taken a leap forward to replace the image with a complete 3D head while the Heidelberg University has tested a technique that allows for superimposition of the entire body! Although the entertainment industry is not new to superimposing images, it required tremendous amounts of effort and use of complex techniques. But now, with the easy availability of software tools such as DeepFaceLab and FakeApp, it can be done in no time. And, with technology evolving at such a rapid pace, it is likely that anyone with a computer and access to the internet can create deepfakes.
A cause of concern
That’s why the potential abuse of the technology has raised fears that deepfakes can spell global disaster if left unchecked. This is because currently the most common targets of deepfakes are actors and politicians and their images may be superimposed on videos that are inciting or outright provocative. For instance, there’s a video where Donald Trump is seen advising Jared Kushner on the basics of money laundering. Of course, the video is fake as it’s a parody on Breaking Bad series with Donald Trump’s image superimposed on that of James McGill and Jared Kushner takes over as Jesse Pinkman.
With deepfakes becoming more sophisticated they can be misused to influence social and political discourses—even stock markets—with misinformation. Experts believe that deepfakes will open up unique legal and policy-related challenges that will require concerted efforts from governments, industry and private citizens. Large organizations like Facebook and Microsoft are already taking measures to remove deepfakes from their networks. Institutions including the Carnegie Mellon, Stanford University, the University of Washington, and Max Planck Institute for Informatics are experimenting with GANs to learn how the abuse of the technology can be contained.
There’s already a lot of misinformation floating on the internet, deepfakes will likely only add to the mayhem. As a responsible citizen, do not take any video at its face value. If it raises suspicion, it probably is a deepfake. Make an effort to find out the original source of the video and research for the actual context the information was disseminated. If it’s a malicious deepfake, report it so the concerned platform can take appropriate remedial action.