Deepfakes explained: How artificial intelligence is being misused to create fake porn videos | tech | Hindustan Times
Today in New Delhi, India
Aug 21, 2018-Tuesday
-°C
New Delhi
  • Humidity
    -
  • Wind
    -

Deepfakes explained: How artificial intelligence is being misused to create fake porn videos

Here’s everything you need to know about Deepfakes, the latest menace on the internet.

tech Updated: Feb 08, 2018 15:34 IST
Kul Bhushan
Kul Bhushan
Hindustan Times
Deepfakes,Reddit Deepfakes,Reddit
What are Deepfakes? And why it’s being banned by Reddit, Twitter and Pornhub?(Shutterstock Image)

Artificial intelligence (AI) can achieve great things, but it can be a menace too. Deepfakes falls to the latter category that has already affected various Hollywood celebrities including Gal Gadot, Arianna Grande and Taylor Swift among others. The AI uses image recognition technology to super impose face of any person on to another.

Deepfakes are increasingly being misused to create fake porn videos, forcing companies such as Twitter and Reddit to impose a ban on the use of the technology on their platforms. Even Pornhub has banned Deepfakes videos.

“Reddit prohibits any sexual or suggestive content involving minors or someone who appears to be a minor,” said the company while updating the policy.

“Depending on the context, this can in some cases include depictions of minors that are fully clothed and not engaged in overtly sexual acts. If you are unsure about a piece of content involving a minor or someone who appears to be a minor, do not post it,” it added.

What are Deepfakes? How it works?

Deepfakes uses the artificial intelligence software to create an artificial version of a person’s face that accurately matches the other person’s face in a video. Consider this as a high-end version of Photoshop, which has also been notorious for being used to create fake still photos. In case of Deepfakes, the software not only swaps the face but also matches the expression. And if the body figure of the person is very similar to the victim’s, the fake video can look very real.

While there’s no information how this software was developed, but it’s believed to have gotten attention sometime late last year when Hollywood actress Gal Gadot’s fake porn video was spotted on the web.

Shortly after a new software called FakeApp, said to be created by a Reddit user called deepfakeapp, was released. The application makes it much easier to swap face of any person on to other. The application analyses hundreds of photos and videos available on the internet of a person whose face has to be imposed. Then it imposes the photo in a video frame by frame, making the face swap completely glitch free. The process of creating a Deepfakes-based video may take a very long time, even for a small clip.

Other Face Swap apps

Already a number of applications are available on Google Play Store and Apple App Stores that let you swap your face with someone else’s. Even Snapchat has one such filter. A lot of people use these filters to create funny videos on the internet. But none of these look as real as the one created by Deepfakes.

Can Deepfakes just create fake porn?

In last couple of months a number of Deepfakes-based videos have been circulated online, but not all of them are pornographic in nature. Recently, the tool was utilised to insert Nicolas Cage into any movie. Check out the video below.

Another video doing the rounds is of Donald Trump’s face imposed onto Michael Scott from a scene in The Office.

Who have banned it?

Apart from Reddit, Twitter and Pornhub have banned the use of the technology on their platforms. Twitter told Motherboard that it has banning all the accounts that are posting AI-generated videos or are used to post such clips.

Pornhub, which is one of the largest adult content website, said such AI generated videos were in violation of its policies.

“We do not tolerate any nonconsensual content on the site and we remove all said content as soon as we are made aware of it,” a PornHub spokesperson said.

“Nonconsensual content directly violates our TOS [terms of service] and consists of content such as revenge porn, deepfakes or anything published without a person’s consent or permission.”

Other internet companies such as Google and Facebook are also expected to announce measures against the Deepfakes-based content on their platforms.

First Published: Feb 08, 2018 13:16 IST