Hilke Schellmann reports at the Wall Street Journal:
Seeing isn’t believing anymore. Deep-learning computer applications can now generate fake video and audio recordings that look strikingly real.
In a recent video published by researchers to show how the technology works, an actor sits in front of a camera moving his face. The computer then generates the same expressions in real time on an existing video of Barack Obama. When the actor shakes his head, the former president shakes his head as well. When he speaks, Mr. Obama speaks as well.
“This is a big deal,” Hany Farid, computer science professor at Dartmouth College, told The Wall Street Journal. “You can literally put into a person’s mouth anything you want.”
Prof. Christian Theobalt, part of a team working on the technology at the Max-Planck-Institute for Informatics in Germany, said he is motivated by the creative possibilities that it holds for the future.
He said researchers have developed forensic methods to detect fakes.
But Prof. Farid says researchers who push computer-generated technology need to think about the consequences these computer-generated fakes could have for society. He believes forensic experts are being outpaced by the development of fakes and that there is no method yet that can detect them all.
“How are we going to believe anything anymore that we see? And so to me that’s a real threat to our democracy,” Mr. Farid said.
In the video above, WSJ’s Jason Bellini explores this world of realistic video fakes. He gets deepfaked himself, and thanks to a deep-learning application, he can now dance like Bruno Mars. He also learns of the dark side of this technology, through one victim whose life has been deeply affected by deepfakes, and why others believe they could even lead to war.
Read more: Wall Street Journal