What exactly are Deepfakes and how do they work?
Deepfakes have been ruling the internet and doing the rounds for quite some time. But how serious are they?
These days apps like ‘Beauty Plus’ and ‘Snapchat’ have allowed users to elongate legs, remove pimples, and even give an opportunity to add animal ears and faces. Some of these can even create false videos that look absolutely real. These days anybody who has a computer and internet access can technically produce a ‘deepfake’ video. Industry experts have opined thus.
Imagine a person as strong and powerful as the President of the United States falling prey before this. Even the Speaker of the House of Representatives in the USA. Deepfakes can entirely shake a nation!
Even several cybersecurity experts have said that technology can be used to make people believe that is something is real, when actually it is not.
What Exactly Are Deep Fakes?
Deepfakes are commonly used for referring to manipulated videos, or any other digital representations, that are produced by sophisticated artificial intelligence. Most of the time, this yields fabricated images and sounds that appear real.
The term ‘deepfake’ is a combination of the terms ‘deep learning’ and ‘fake’, this is a form of artificial intelligence. Simply put, deepfakes are falsified videos that are made through deep learning. Deep learning is a part of AI and refers to the arrangements of algorithms that are intelligent enough to make decisions on its own.
How Do Deepfakes Work?
Normally by studying photographs and videos of a particular person, a deep learning system can produce a persuasive counterpart. Thought leader Barrett explained that once a preliminary fake has been produced, a method called GANs (generative adversarial networks) makes it more believable. This process tries to detect flaws in the forgery, thus leading to improvements addressing the flaws.
As per an MIT technology report, a device that is capable of enabling deepfakes might even be a ‘perfect weapon for purveyors of fake news who want to influence everything from stock prices to elections. Experts were of the opinion that though GANs didn’t create the problem, yet they made it worse.
Some Major Examples
- Donald Trump Joins Breaking Bad
Most deepfakes strive to fool the viewer. Consider Better Call Trump: Money Laundering 101, which is a straight up parody. They take a scene from the famous Netflix web series, ‘Breaking Bad’ and introduces Donald Trump as James McGill, later on, who used the pseudonym Saul Goodman in the series ‘Better Call Saul’.
In this case, James McGill explains to Walter White’s business partner Jesse Pinkman about the basics of money laundering. It’s seen here that Trump’s son-in-law Jared Kushner takes over from Jesse Pinkman in the scene. The parody gets almost a heart to heart feeling.
- Barack Obama’s Public Service Announcement
Most of the deepfakes have been produced by using impersonators that are capable of mimicking the source’s voice and gestures. This was first seen in a warning produced by BuzzFeed and comedian Jordan Peele using After Effects CC and FakeApp.
They had pasted Peele’s mouth over Obama’s. The former President’s jawline was replaced with the one that followed Peele’s mouth movement, then they used FakeApp to refine the footage for more than 50 hours of automatic processing.
- Mark Zuckerberg Too Isn’t Spared
Recently, two artists from an ad agency created a deepfake of Facebook founder Mark Zuckerberg. It mentioned several things that Mark never said, and it was uploaded on Instagram.
How To Avoid And Control Deepfakes?
At a time when AI can be used to create deepfakes, it can also be used to detect them. As technology becomes more and more advanced, more and more researchers are focusing on deepfake detection and also looking forward to a way to regulate it.
MNCs like Facebook and Microsoft have taken great initiatives in order to detect and remove deepfake videos. The tech giants announced earlier that they would collaborate with top universities in the US, and create a large database of fake videos for research, as per Reuters.
It seems that even as technology continues to evolve, the detection techniques often lag behind the most advanced creation methods. The question now is – ‘Would people repose their faith in this form of fabricated technology?
Video sources: Youtube
I do not know do you know that Zemana has created a program to detect deepfake content. According to the Deepfake Detection Challenge worldwide they are really close to making the best solution for this problem: deepware .ai
I m making research about this problem and I would love to read about them more on your site…
Greetings,