A fake video of Mark Zuckerberg shows him giving a disturbing speech about Facebook’s power
A video has emerged on Instagram over the weekend of Facebook founder Mark Zuckerberg giving a sinister, provocative speech about the social network’s dark omnipotence – only it isn’t Zuckerberg, it’s a deepfake avatar. “Imagine this for a second – one man, with control of billions of people’s stolen data, all their secrets, their lives, their futures,” the deepfake says. “I owe it all to Spectre. Spectre showed me that whoever controls the data, controls the future.”
The video is formatted like a news segment, with a ‘CBSN’ strapline ‘We’re increasing transparency on ads’ along the bottom. It features a deepfake made by artists Bill Posters and Daniel Howe, in collaboration with the ad company Canny AI. The original video is a speech Zuckerberg gave about Russian election hacking on Facebook back in September 2017.
An edited, manipulated, and now viral video of US house speaker Nancy Pelosi had sparked concerns around Facebook’s video policy and community guidelines, and this new deepfake puts its policies to the test. The video of Pelosi was slowed down to make it seem like she was slurring her words, and it was circulated widely on Facebook, while YouTube swiftly removed the video for violating its policies.
Facebook, which acquired Instagram in 2012, maintains that the video did not breach community guidelines, and stated that the platform “does not have a policy that stipulates information people publish to Facebook must be true”. Instead of deleting then, Facebook de-prioritised the video on site, pushing it to the bottom of the interaction pile. Facebook’s director of policy claimed that a fake video of Zuckerberg would stay up when previously asked, but this new video with its explicit message puts that to the test. The question now is what happens to deepfake Zuck?
The deepfake uses VDR (video dialogue replacement) technology to manipulate a video. The growth in deepfake technology raises valid concerns over fake news, revenge porn, and as Pelosi told KQED, political propaganda. “I think they have proven – by not taking down something they know is false – that they were willing enablers of the Russian interference in our election.”
The video is part of an upcoming art exhibition by Francis and Howe called Spectre, which “examines and subverts the power of the digital influence industry”. The two commissioned Canny AI to manipulate a series of clips of pop culture icons, Silicon Valley heavyweights, and Zuck, all in an effort to generate awareness of computational propaganda, and the dangers that this presents. The clip itself is pretty disturbing – the deepfake blinks and hesitates on words, and uses narrative markers like a real person. The experts at Canny trained the algorithm on the original, real 2017 clip and voice actors, and then reconstructed the original video using Zuckerberg’s real movements and an actor’s. Canny engineers told FXGuide they were inspired by the work of the creators of the Obama deepfake, and the Stanford Face2Face program.
An updated statement from Facebook, following the Pelosi incident, announced partnerships with the universities of Maryland and California to create further research into the new wave of so-called “manipulated media”.