Boon or Bane: Deepfakes
Deepfakes have grabbed the minds of many enthusiasts, however, is there a risk?
In a world where technological advancements have breakneck speeds, we need to consider the potential risks many of them pose to be. Countless technologies were created for constructive purposes only to be misused to spread chaos and fear in the lives of people. One such area that is being exploited is deep fake technology.
Deepfakes use machine learning algorithms to create fake human images and replicate them in pictures and videos. Using this technology, it’s possible to synthesize a person’s face and superimpose it upon another video or photo.
The film industry is one of the largest markets for this technology and its applications are extensive. You could have a stunt performer perform a risky scene and then use deepfakes to replace their face with the actor’s visage. If a movie scene needs to be reshot and the actors aren’t available, you could have random cast play the part and replace the faces.
One example was the portrayal of Princess Leia in Rogue One: A Star Wars Story. Carrie Fisher’s (Princess Leia) face was superimposed on the face of Danish actress Ingvild Deila’s to play a younger version of Princess Leia. As one can see, the applications are mind-blowing.
In 2017, a video was released in which Obama was talking about deepfakes and their potential misuse. Towards the end of his talk, he revealed that the video was an example of a deep fake and he never actually spoke the words.
It was a slap in the face for many people and showed potential disasters deepfakes could create. It revealed how one could spread false news and make controversial statements from people in power.
Deepfakes could be used for propaganda and defamation of political leaders, especially before elections. Another place where it has garnered interest is pornography. You could fit any celebrity face to a porn video and publicize it across social media.
Revenge porn is another terrible way to manipulate this technology and retaliate. The list is huge on how one could exploit this technological breakthrough.
Deepfakes, as the name suggests is a morph of deep learning and fakes. It is based on a branch of machine learning called Generative Adversarial Networks or GANs. GANs are made up of two components — a generator and a discriminator.
The generator creates fake videos or images that are sent to the discriminator to determine its authenticity. If identified as a fake, it’s sent back to the generator along with inputs on how to make it more convincing.
Over time, the generator becomes better at creating fake content and the discriminator at identifying them. The two components help each other out and collectively make a perfect video. The technology is readily available and with a few lines of code and any amateur can create a deep fake.
With such powerful technology available at the click of a button, isn’t it time to pause and think about the harm that can accrue on society?
Computer Science Undergraduate at NIT Trichy, Web Developer and a Machine Learning Enthusiast