A fake video of Mark Zuckerberg giving a speech about Facebook's power and control of the wider public has gone viral online.
The deepfake, spoken by a voice actor, moves seamlessly and blinks and gestures like Zuckerberg would, making it very realistic, except for his voice. The clip was framed with a news strap and CNBC logo, making it look like something from a TV update.
The art installation Spectre uses deepfake artworks to explore the power of the digital influence industry, technology and their impacts on privacy and democracy. Canny engineers clipped a 21-second segment from the original seven minute video then trained the algorithm as well as videos of the voice actor speaking, and then reconstructed the frames in Zuckerberg's video to match the facial movements of the voice actor to realistically sync.
Despite their intent to use AI media for art, satire and political debate, there still remains concerns around truth, identity and consent, including citing an unauthorised trademark of a news station, which potentially stirred fake news.
Who created the deepfake?
Was the content disclosed as a deepfake?
Was the deepfake consensual?
How was the deepfake created?
Inspired by Face2Face real-time facial reenactment
A proprietary AI algorithm trained on 20 to 45 second scenes of the target face for between 12-24 hours