How to spot deepfakes: Reuters and Facebook release course
Deepfakes and other manipulated media are now easy to make. The tools are freely available and the results can be convincing. It is now possible to put one public figure’s words in the mouth of another by superimposing and lip-syncing the latter’s face. It is also possible to have anybody appear to say things that nobody ever said in reality except the faker. The technology is now mature enough to fool the casual viewer. And since the media are consumed almost entirely by casual viewers, deepfakes are a new and pernicious danger to the public’s perceptions of reality.
Journalists need to have their eyes on stalks when reviewing media they have received. While in most cases manipulated media can be easily debunked by comparison with other filmed versions of the same event, it is getting harder to label any one version as definitive. And when a video is the only one there is of an event, how can its provenance and veracity be determined, since there is nothing else to compare it with?
In an attempt to tackle the problem, objective journalism stalwart Reuters has teamed up with a social platform that needs to get up to speed on deepfakes to protect its own credibility: Facebook. They have produced a free online course that helps journalists spot and avoid sharing manipulated pictures, videos and audio clips. Facebook has footed the six-figure bill as part of its Facebook Journalism Project – the Zuckerbot’s attempt to mend fences with journalists and encourage everyone to read critically and fight fake news.
In English, Spanish, French and Arabic, with many more languages planned, journalists learn how media can be manipulated, taken out of context, staged, and otherwise designed to deceive or mislead. Best practices to identify manipulated media are given, along with a verification workflow to be used during breaking news scenarios that reduces the chances of sharing media that later turns out to have been faked.