Facebook didn't remove the deepfake video featuring Nancy Pelosi
12 June, 2019, 21:26
"We've used deepfake technology as way of narrative storytelling to engage audiences in some of the tensions that exist, like a cautionary tale of technology and democracy", Posters said.
That the creators are uploading these videos to Facebook subsidiary Instagram is notable because the company recently refused to take down altered clips of Democratic Speaker of the House Nancy Pelosi created to make her seem drunk or suffering the symptoms of dementia.
Posters and Howe are not trying to trick internet users, as all the videos are labelled with the hashtag #deepfake.
A sinister deepfake of Zuckerberg was uploaded to Instagram four days ago, in part to test how quickly the company can weed out fake clips.
Asked whether a doctored video of Zuckerberg would get the same treatment as the manipulated Pelosi video, Neil Potts, Facebook's director of public policy, told a parliamentary hearing in Canada last month, "If it was the same video, inserting Mr. Zuckerberg for Speaker Pelosi, it would get the same treatment". "Spectre showed me that whoever controls the data, controls the future".
"We will treat this content the same way we treat all misinformation on Instagram", an Instagram spokesperson tells Motherboard.
"If third-party fact-checkers mark it as false, we will filter it from Instagram's recommendation surfaces".
Facebook told The Verge that the video will not be removed, but will instead be stopped from being recommended on the app. Nor did Canny AI, an Israeli advertising startup that helped create the video.
Zuckerberg and other Facebook officials have said that artificial intelligence technology will ultimately allow the company to control the spread of false information, but the doctored video of Pelosi showed disagreement among big social media companies on what is worthy of removal.
The words that appeared to be coming out of his mouth were actually spoken by a voice actor reading from a script provided by Posters and another artist, Daniel Howe.
Founders of Canny, Omer Ben-Ami and Jonathan Heimann, told FXGuide that their algorithm is inspired by the works of the University of Washington and Stanford's Face2Face algorithm. As far as deepfake videos go, it's not terribly convincing, but it's realistic enough to throw off a viewer that's not paying too much attention.
In this case, Ben-Ami said, Canny AI chose a scene of less than a minute in length of Zuckerberg speaking, and used a computer to meld it with an actor's voice and appropriate facial movements. Versions of that video, though, are still on Facebook and YouTube.