Deepfakes in the web TV show

Deepfake Mark Zuckerberg

I’m not sure how many people are aware of this technology now, but it kind of followed Rule 34. Apart from academic research, it gained traction with porn clips, where celebrities faces were superimposed on porn actors bodies – at least that’s what my friend told me… Deeptrace estimated that 96% of all deepfakes were pornographic.
Deepfake technology was made possible by developments in machine learning and its subset – image/face recognition technology. It was treated as a challenge by AI enthusiasts and with fair dose of fearmongering from the media. One of the first prominent examples was Buzzfeed’s deepfake of Obama:

2020 comedic use of deepfake technology by South Park creators Troy Parker and Matt Stone showed how good (and still creepy) the tech is:

Watching the Zuckerberg ad remained me of the scene from Back to the Future II, where Ronald Reagan and Michael Jackson imitations are helping customers in 80’s style cafe:

Deepfakes use in the media production is going to rise as the tech develops and prices drop. If you can create virtual models, there is no need for TV anchors or models in fashion and other industries. I can see one interesting use of this technology in a manner where it was used in “Marjorie Prime” movie, where a character is interacting with her holographic late husband.
The deepfakes have been used to misrepresent politicians like Trump, Angela Merkel and Belgian Prime Minister Sophie Wilmès. Potential for scam (vishing) is also quite real and already reported.
Simultaneously there are efforts to create software to detect doctored footage (one of them funded by DARPA). What’s surprising, one of the most viral “fakes”, the Nanci Pelosi’s speech wasn’t in fact a deepfake, just a slowed down video. Social media platforms are taking efforts to detect synthetic media and label them as misleading where appropriate, but viral content equals more revenue, so they aren’t really inclined to do so.|
The fully synthesized body isn’t far off. Malicious deepfakes are another weapon in the arsenal of disinformation and a business with potential to grow. It will further undermine trust in news media and play into confirmation bias of a polarized society. Companies like Deeptrace Labs or Sherlock AI try to create detection tools, but even the short exposure (before takedown) can do a lot of damage. The race is on.


To see some interesting movie/actors mashups (SFW) see this Reddit sub .

Share

Leave a Reply

Your email address will not be published. Required fields are marked *

Post comment