Facebook's latest AI doesn't just detect deep fakes, it knows where they came from. The Facebook AI system is designed to use deep learning to track down the origins of these fake voices and faces. It actually recognizes their location, that's how powerful it is.
We're not just talking from a "this video was filmed in New York" level of detail, but the AI actually knows where the original image of that human face came from.
The AI could someday help Facebook identify victims in abusive pornography, or even identify people posting revenge porn.
A team at Facebook recently published research detailing how its AI can detect fake images and videos posted on its platform.
And the team applied a similar technique to recognize whether videos had been digitally manipulated. By recognizing individual frames and comparing them with other frames, the system could detect if anyone moved out of place. This would allow Facebook to do things like automatically blur out faces when someone takes a screenshot and posts it online, for example.
the said: “We begin with image attribution and then work on discovering properties of the model that was used to generate the image,” the team continued. “By generalizing image attribution to open-set recognition, we can infer more information about the generative model used to create a deepfake that goes beyond recognizing that it has not been seen before.”
Facebook’s AI does this through deep learning - meaning it can “learn” where people are posting from, even if a location isn't explicitly stated in the post they're looking at.