In an era filled with fake news, many believe that video is more credible because it provides a clearer picture of reality. Alas, this is an erroneous opinion, according to Google engineer Suparsan Suvajanakorn, who has developed a tool that, based on the analysis of the original original video information, can create a realistic fake video image that simulates the features of any person's speech, ensuring perfect lip-sync with the given fake text.
Like any technology, the novelty has tremendous potential, both in positive and negative directions. So Suwajanakorn is also partnering with the AI Foundation, an artificial intelligence company, to create a Reality Defender application that can automatically embed itself in Internet browsers to detect and tag fake photos and videos.
“I fed my computer 14 hours of Barack Obama material and then synthesized his speech,” said Suwajanakorn during a TED (Technology, Entertainment, Design) conference in Vancouver, where he showed impressively convincing results.
According to the author, this technology can be used to create virtual versions of those people who have already left our world: grandparents who can be asked for advice; actors, whose fans are ready to do anything for another meeting with their idol; great teachers who will give lessons, as well as authors who read their own works.
In particular, he spoke about the New Dimensions in Evidence project, which allows museum visitors to talk to holographic images of Holocaust survivors.
“The results were intriguing, but alarming at the same time. I am seriously concerned about the potential for abuse,”said the inventor. “So I am also working on countermeasures to identify fake photos and videos.”
He is concerned, for example, that a fake video depicting one of the world leaders announcing a nuclear strike against the enemy could provoke a world war and a catastrophe for humanity.
Reality Defender, which automatically scans images for signs of tampering, also allows users to report tampering. In this way, the "power of the crowd" is connected to the good cause of defending the truth.
Promotional video:
"Video manipulation can be used for harmful purposes if the necessary countermeasures are not taken," the author said in an interview with Agence France-Presse. “We need to make the creation of fake video images extremely risky and inefficient in terms of costs.
Writing fake news may be easy and cheap, Suwajarakorn said, but manipulating video images without any trace is a much more difficult task. Videos are streams of thousands of images, each of which must be flawlessly modified to create a good fake.
“There is still a long way to go before we can effectively simulate human video,” said Suwajanakorn, whose work in the field began several years ago when he was a student at the University of Washington.
“We must be extremely careful that this tool does not fall into the wrong hands,” he told Agence France-Presse.
Igor Abramov