Deepfake and Its Impact

0
Machine Learning

With the help of modern technology and really smart software, we can make anything possible. Recently, the world’s been introduced to an advanced new technology called the Deepfake.

This technology has the potential to create audio and video files by altering the original video with a new superimposed image.

In simple words, this technology has the potential to create fake videos and can make them look so real that it becomes impossible to tell the difference. It can make counterfeit jeopardizing videos that can seriously affect real human lives and their fame. 

What is a Deepfake

The word Deepfake comes from the combination of “deep learning” and “fake.”  It is an Artificial Intelligence (AI) based technology that produces or alters video so that it presents something that didn’t occur.

In other words, deepfake is taking an existing video or image and replacing it with an alternative image of someone or something.

This is possible through the AI which uses a deep-learning system that can produce a persuasive counterfeit by studying photographs and videos of a target person from multiple angles, and then mimicking its behaviour and speech patterns.

Deepfakes and the technology behind it – Source BBC

Creating deepfakes

Creating deepfakes is not that hard. This process requires a lot of pictures of a person’s face and facial expressions.

For instance, if we look back at some movies that came out during this decade, we notice that this technology has been used more than we thought it did.

Digital de-aging is when the software makes an old person look young. This was particularly used on Samuel L. Jackson and Clark Gregg who were digitally de-aged by approximately 25 years for their roles in Captain Marvel.

Though it is not so popular yet digital resurrection is also used to show late actors and actresses. This was seen first time in the 2017 movie Star Wars installment Rogue One, where Peter Cushing, made an appearance on screen despite having passed away in 1994.

Origins of Deepfakes

Deepfake first surfaced in December 2017, when a Reddit user, also known as deepfakes, published a pornographic video of an actress. This was fake but very difficult to spot.

He used a strong machine-learning algorithm to studied faces, body language, and speech of a person and then would make a false video based on the original photos, videos, and audio files.

This technology clearly has the potential to ruin a lot of reputation and can be easily be used for blackmailing and revenge.

How to Identify a Deepfake

Deepfakes are well-crafted videos that were made to improve with time, after learning all the intricate details about the original content.

Advanced deepfakes are nearly impossible to spot and can only be declared fake by machines and software that are specifically created to do so.

John Villasenor at Brookings’ Villasenor wrote in February 2019, while AI are used to make deepfakes, it can also be used to detect them.

The problem is that the sophisticated technology needed to detect deepfakes is not readily available and by the time it is detected the misinformation has already been spread.

It’s Getting Harder to Spot a Deep Fake Video – Source Bloomberg

What are Generative Adversarial Networks

Generative Adversarial Networks or GAN was a neural network developed by Ian Goodfellow and his colleagues in 2014, that is responsible for learning details from old photographs and making forged ones that are too realistic.

GAN is a method that makes a deepfake more believable. The method detects flaws and improves it to make the video or image look more real.

To understand deepfakes, you need to understand GAN. The video below fully explains how GAN works.

Can deepfakes become a threat

Deepfakes are being used to defame and misrepresent prominent politicians in videos. One such instance was when the face of the Argentine President Mauricio Macri was deepfaked on the face of Adolf Hitler.

Large corporations like Microsoft and Facebook have taken countermeasures against deep fake videos and most of the social media platforms have banned these kinds of videos as well.

But it is getting harder and harder to detect the “tells” of these deepfake videos. Soon we won’t be able to trust any video or audio files we come across.

To spread awareness of this threatening technology, Jordan Peele and Buzzfeed collaborated and created a deepfake of Barack Obama with Peele’s imitation voice of the president. It served as a public service announcement to increase awareness of deepfakes.

Source: BuzzFeedVideo

The Affect of Deepfakes on the World?

One of the biggest threats the world faces from deepfakes is that of blackmail. Because this technology was first used for making fake pornographic content of celebrity, this can further involve people using it to blackmail politicians, businessmen, and even ex-girlfriends or boyfriends.

The spread of misinformation using deepfakes

Deepfakes can also be used to spread misinformation during campaigns for politicians and can heavily influence elections. For instance, an MIT conference showed how they could use deepfake technology to create a fake interview with Russian President Vladimir Putin.

As a result of this, the state of Texas passed a law that criminalizes the publicity and distribution of such deepfake videos that have the potential to harm any person or political candidate.

Following this California passed a similar law in October that made it illegal for anyone to share or distribute deep fakes that could deceive voters and harm any candidate’s reputation.

Disclaimer