According to a recent study, training people to pay attention to the right visual cues nearly doubled how accurately they could spot AI-generated faces. This might not sound like a big deal, but it’s a crucial step in understanding how to prevent the spread of deepfakes – AI-generated videos and images that can be used for malicious purposes.
What makes a face ‘fake’?
The researchers identified several key visual cues that are common in AI-generated faces. For example, AI faces often have a “smoothness” to them, which is because they’re generated using algorithms that try to recreate the perfect, pore-free complexion. Real faces, on the other hand, tend to have a more irregular texture. Eyes are another key area – AI faces often struggle to get the shape and positioning of the eyes just right, which can give them away.
The researchers also found that people are more likely to spot a fake face if it’s looking straight at the camera, rather than slightly averted. This is because AI faces often struggle to create the subtle, nuanced expressions that humans take for granted. And then there’s the matter of facial proportions – AI faces often have a slightly distorted or oversized quality, especially in the eyes and mouth.
Training the human eye
The researchers developed a training program that taught people to pay attention to these visual cues. They were able to improve detection rates by nearly 100%, which is a huge leap forward. The training program wasn’t just about showing people lots of examples of fake and real faces – it was about teaching them to think critically about what they were seeing.
What this means in practice is that as AI-generated faces become more sophisticated, there will be a growing need for people to be able to spot them. This might involve training law enforcement officers, journalists, and other professionals to be able to identify deepfakes. It’s a challenging task, but by understanding the visual cues that give away AI-generated faces, we can start to build a defense against them.
A cat-and-mouse game
The researchers acknowledge that this is going to be a cat-and-mouse game, with AI developers continually pushing the boundaries of what’s possible and researchers working to stay one step ahead. But by identifying the visual cues that give away AI-generated faces, we can start to build a more nuanced understanding of how to prevent the spread of deepfakes.
