Everything you see, it may not always be true. From KOL’s videos to Hollywood movies, we will dismantle the filters layer by layer. As long as you know about them, you can be anyone or anything you want in videos. Despite the magical facade, editing apps, Deepfake and computer graphics (CG) are all about algorithms. 

Before we get into our focus of the day, Deepfake, let’s watch the following videos:

In the video, Donald Trump gave the speech in his usual tone. He criticized Belgium on the issue of climate change, ‘As you know, I had the balls to withdraw from the Paris climate agreement and so should you.’ This is followed by random insults, ‘You only blablabla bingbangboom.’ and ‘Shame. Total shame.’ Later it even escalated to Trump’s narcissism, ‘People love me because I am a fair person. I am the fairest person on Earth.’ 

If you are aware of American politics, perhaps the video isn’t able to fool you. Obviously, Trump’s voice is dubbed with an exaggeration in Trump’s attitude, not to mention the subtle details, including the unnaturalness in the corners of his eyes and mouth. In fact, this video is uploaded by a Belgium political party, namely the Socialistische Partij Anders, on its official social media on Facebook and Twitter sp.a. The spokesperson explained the unsatisfactory quality was to remind netizens not to believe in it, and more importantly, to raise awareness towards the petition on climate issues.

BuzzFeedVideo uploaded this video of Barack Obama on Youtube. Obama’s facial expression and speaking style are less artificial comparing to Trump’s in the last video. Obama’s speech eventually became fishier and fishier. He started with the risks on Internet, emphasizing on the advancement of technology. He added with the technology, we were granted the ability to make anyone say anything, with Obama himself attacking Trump, ‘President Trump is a total and complete dips**t.’, even if he would never do so in public.

Well, not Obama, but someone else might. This person is Jordan Peele. From 00:36 onwards, Obama on the left and Peele on the right were in sync, warning citizens to be careful and critical. Now honestly, did you realize anything wrong before that? For us, all we found was Obama’s voice was a little too deep. 

If you examine Deepfake videos closely, you can actually see clues. However, these clues are not so easy to find admittedly. 

So, what exactly is Deepfake? Deepfake originated from the words ‘Deep Learning’ and ‘fake’. It is a synthetic technology using artificial intelligence. Overlapping images and videos, it can create realistic illusions. Deepfake might involve face switching and voice synchronization etc. 

The system is developed by Ph.D. student and Apple employee, Ian Goodfellow, with Generative Adversarial Network (GAN), an unsupervised machine learning, as basis. On one hand, a generator generates graphics. On the other hand, a discriminator differentiates actual graphics and synthesised ones. With a generator and a discriminator, the system achieves deep learning automatically.

v2 d5a3bfce3cb0b692e70fef236e435e8d r

Given the authenticity, you may find Deepfake complicated. Actually, not quite. The flow of Deepfake consists of 3 steps: Step 1, extract figures from images. Step 2, train computers within hours to understand rights and wrongs. Lastly, step 3, switch characters in images and videos. And, viola! There you have it. 

It really costs you nothing to use Deepfake, especially when there is an app, FakeApp, for it. AI research professor David Doermann from the State University of New York described Deepfake to be so easy to use that a highschool student with a mediocre computer can do it.

Given the user-friendliness, FaceApp was often inappropriately utilized to create parodies and pornographies. Starting from 2017, celebrities including Taylor Swift, Scarlett Johansson and Gal Godot were victims of Deepfake and have had appearances as porn stars.

According to Reddit’s guidelines, any forms of involuntary pornography are prohibited. As a matter of fact, Twitter, Reddit and adult site Pornhub all banned videos of Deepfake. Virginia, a state in the United States, even declared transmission of Deepfake videos to be illegal. Violation of the law would lead to a fine of USD $2500 and an imprisonment of 12 months. Apparently, our society considers Deepfake as a threat. 

With risks come opportunities. The business of video authenticity analysis is thriving. Lecturer, Shiva Kintali, at Princeton University and started a business in Silicon Valley. Uploading any files, users can check the modification record of videos and audios. Truepic is also a platform for verification. You get the codes from the organic information by uploading the files. After encryption, you can print the block chain for comparison with the original codes. SRI International tries to untangle the problem with the algorithms. Readers as smart as you can recall, this method is basically the same as the deep learning mentioned above. At the same time, researchers are still exploring at the moment. 

Technology helps fake our eyes out. Deepfake is just the tip of an iceberg. Coming up next in following few articles are video-editing apps and computer graphics.