"Deepfake": Facebook reflects on a specific policy for these fake videos
Mark Zuckerberg felt it would be "sensible" Facebook is arming itself against these hyper-realistic fake videos that make it possible to say what you want to a person, only from a few photos. & nbsp;
Mark Zuckerberg said it would be "sensible" for Facebook to arm themselves against these hyper-realistic fake videos that make it possible to say what you want to a person, only from a few photos.
Facebook CEO Mark Zuckerberg said on Wednesday that it would be "sensible" for the social network to adopt a specific "deepfake" video policy, with hyper-realistic tweaks telling personalities never held.
Like other content, "deepfakes" in particular are a topic for which we are evaluating what the "Facebook" policy should be, said Mark Zuckerberg, himself recently a victim of a video of this type posted on his affiliate Instagram, which had not removed it. "It would probably make sense to have a different policy and treat it differently from the way we treat" classic "misinformation, he said, pointing out that Facebook did not want to be the arbiter of the truth. He was speaking at a conference in Aspen, Colorado, which was webcast.
A manipulation of a formidable efficiency
The "deepfakes" videos are multiplying, made more and more realistic thanks to the advances and the democratization of the tools of artificial intelligence, opening the door to handling actions of a formidable efficiency. Facebook uses external "fact-checkers", including AFP, to label certain content as "false" or misleading. The social network then modifies its algorithms to limit their propagation but refuses to delete them. It is also, he justified, which explains why the group has refused to remove a recently manipulated video of the president of the US House of Representatives Nancy Pelosi, much to the chagrin of the latter.
The video, quickly turned viral, was not a "deepfake" but had been slowed down so that the girl seemed to express herself with difficulty, as if she were drunk. On the other hand, "the mistake on our part, was that it took our systems (...) more than a day to report" as manipulated users, he acknowledged, adding that the video had therefore spread more than it should have.