Google releases thousands of deepfakes to combat deepfakes

Deepfake videos, synthesized with a deep neural network, show the face of one person (the target) in another one's video (the source).

The clips are created with so realism that one can easily make a person, say a celebrity or politician, do or say something that they wouldn't normally.

Now, this exactly makes up the new avatar of misinformation that threatens the web community.

Currently, you can detect some blips in deepfakes to tell that a video is fabricated, but as this technology continues to evolve, it would be impossible to figure out the differences.

Deepfake researchers/engineers, around the world, realize this problem and are working on AI-powered detection tools to flag and get such videos taken down from the internet.

Now, with 2020 US Presidential elections on the way, Google is offering help in expediting the development of deepfake detection tools.

The search giant has created and released as many as 3,000 deepfakes of actors (with their permission of course), giving engineers working on detection methods a huge dataset to train their algorithms and make them as accurate as possible.

Google published the videos after filming the actors in a variety of scenes and employing public tools to turn those videos to deepfakes.

However, it notes that it will continue this process and add more videos to keep with up the evolving deepfake detection tech in the future.

As the tools will evolve, the video quality will improve and so will its detection methods.