Facebook's New AI Will Distort People's Faces To Hide Them From Facial Recognition Systems

Facebook has stopped making use of facial recognition but now, it is offering users a way to avoid the technology altogether. The company’s researchers have developed an AI-enabled software that de-identifies people, even if they appear in live videos in real-time.

The system is made up of a trained face classifier and an adversarial auto-encoder. What it does is to slightly distort a person’s face in a way that puzzles facial recognition systems; however, the subjects still remain recognizable to human eyes. For example, you may notice that a celebrity’s face looks a little different, but you will not doubt that you are looking at the right person.

The best part is, the system does not need to be trained again when it works on another person or video. All it does is to make a bit of time distortion.

As the fear of facial recognition is on the rise, such a system will be of great help for privacy-conscious Internet users. However, do not expect the AI to be used on Facebook anytime soon. A company’s spokesperson said that Facebook had no plan to add the new feature to its social network platform.

We may not see the day the tech is widely applied but some of its benefits are obvious. The AI can be used to prevent third-party organizations from tracking users or creating deepfakes with facial recognition systems. For example, you can freely upload a video of a party without worrying that your friends’ and family’s facial data will be exploited by someone else.

Back in September, Facebook announced that it would stop using facial recognition on its site. However, this act came after the company had been under fire for some time about its harvesting data too aggressively.