Big Tech Pursues Deepfake Detection Tools

Silicon Valley is racing to develop tech that can blunt the devious distribution of deepfakes, or AI-generated audiovisual content. As both the NYT and the WSJ reported yesterday, this often feels like a cat-and-mouse game.

Let's start with the cat: Silicon Valley's top tech companies are aiming to build systems that can automatically detect deepfakes and trace the origins of synthetic photos/videos.

• Facebook is sponsoring a deepfake detection challenge. Google recently published a large dataset to help researchers ID deepfakes. Twitter is working on crafting a deepfake policy.
• Some startups, such as Deeptrace or Dessa, are also focused on detecting deepfakes.

Then there's the mouse. Sophisticated machine learning algorithms that power deepfakes have been commoditized. As costs drop and the tech spreads across the internet, it's become easier for anyone to create their own deepfake and fool the cat.

Big picture: Both sides of deepfake technology, creation and detection, are improving. It's too soon to know which will win out.