Accepting a problem is the first step towards its resolution. Deepfakes may have been developed with non-malicious intent, but the human mind can put any tool to work in two modes: good and bad. The concerns about the misuse of deepfake technology are definitely not dumbfound as was demonstrated by the efforts of actor Jordan Peele, BuzzFeed, and Monkeypaw Productions who made a deepfake of former US President Barack Obama in which he was seen calling Donald Trump names. In reality it was just Peele, whose voice and face were switched with Obama’s. This video made using Adobe After Effects and the AI face-swapping tool, FakeApp, was intended to serve as a Public Service Announcement on the malicious use of deepfake technology.
While this was a PSA, there can be real-world implications of deepfake tech and we are not far from a time when deepfake detection technology would be used to identify and tell the synthetic videos from the real ones as a necessary step before disseminating news or taking any actions on the basis of any video content.
The Bad| Fabricated News and Deepfake Detection | Fighting disinformation
Consider this hypothetical scene: A war-crimes journalist receives videos of a political leader unleashing cruelty on prisoners of war from an anonymous person. The news is published, it goes viral, and the leader in question is condemned worldwide. A few days later, the video is declared a ‘deepfake’ and the journalists loses his or her job, and the organization he worked with, is sued for libel. To avoid such and more weaponized propagandas from being unleashed, AI with deep learning jumps to the rescue by helping distinguish the deepfake from the real.
Deepfake detection is done through visible parameters such as deepfakes in which the impersonator does not blink. But when detection becomes easier to both the human eye as well as detecting technology, deepfakes become smarter. It is a cat and mouse game in which one will always trump the other. Over time, deepfakes have become evolved and are closer mimicking the real videos making it harder to be detected. AI can now be fed to record patterns of Elton John or any celebrity’s many videos available online to detect how he speaks, how he pauses, what his facial and hand gestures are, etc. to detect what’s fake and what is real. Besides, AI teams too, help fish out other inconsistencies from one video frame to another. That said, a lot needs to be done in the field of ‘deepfake detection’ to work on low quality videos available in enormous numbers online.
In such scenarios, AI needs human intelligence support to win the fight against disinformation which will be a growing challenge in the times to come. Facebook, which has an audience worldwide at its disposal, and could be used in a malicious way to spread disinformation, set precedent as it took a huge positive step by announcing earlier in 2020 that it will ban AI modified videos from its platform. Other platforms such as Twitter and TikTok have similar policies in place too.
The Good | Deepfakes can be creative and fun too
Video editing and film-making
If you have ever been in the business of video editing, especially when a famous celebrity was involved, and no matter how much you try to salvage the video, you cannot; deepfakes will help with that clip and your video will be made uninterrupted. Earlier, movie directors would spend millions on creating the perfect location for their movies. This is now achievable with a far lesser sum using deepfake and AI. Deepfakes can be real game-changers in film-making. Imagine new movies starring Charlie Chaplin opening to full houses. This, ofcourse, after all the copyright issues are taken into consideration..
Make the dead talk | Recreational purposes
Imagine Leonardo da Vinci talking about his own artworks from a younger age at the famed Hall 15 of the Uffizi Gallery in Florence. This is the kind of positive impact that deepfake tech can create when used wisely. In fact, this has been already done when St. Petersburg’s Dalí Museum implemented deepfake technology for its exhibition Dalí Lives, where a life-sized deepfake of the artist was created from his old interviews. It delivered numerous quotes which are attributed to Dalí.
Make a newsroom come alive
Deepfakes have entered the newsroom in a positive manner too. Reuters collaborated with an AI startup in making world’s first presenter-led news reports, using deepfake technology. New videos were created from old pre-recorded clips of a news presenter for doing so.
How we use deepfakes can indeed be a blessing or a curse. We decide.