6 Signs To Help Spot An AI Deepfake Image

Images created by artificial intelligence (AI) have long been associated with certain “tells”, including anatomically impossible hands and a “glossy”, soft-focus finish.

Still, enough of us are fooled by them that multiple fake pictures have gone viral. These include a made-up mugshot of US President Donald Trump, a Prince William and Harry reunion that never happened, and AI-generated images of Dua Lipa and Callum Turner’s wedding that multiple publications shared.

Tom Holland even claimed he had to debunk deepfakes of his nuptials to his own family.

Worse, AI detection tools, which are designed to tell real text and images from those made by machines, are seen by experts as ineffective and inaccurate. So how are we meant to spot the real from the fake?

Well, a study published in the Proceedings of the National Academy of Sciences says us humans can use six tools – none of which are machine-based.

Researchers from the Australian National University (ANU) Emotions and Faces Lab trained people to spot AI-generated images of faces from real ones.

“Training on visual artifacts, like looking for a sixth finger or odd earrings, has had limited success, partly because the AI is getting too good, and fraudsters may avoid using pictures with obvious flaws anyway,” lead researcher Associate Professor Amy Dawel said.

“Our training directs people’s attention to global qualities that differ between AI and human faces. AI faces tend to be more symmetrical, proportional and attractive, but without training we often think these are markers of being human.”

By the end of the training, all people were better at spotting AI faces. And those who were already good at it became “near perfect”.

The scientists trained participants by asking them to focus on six factors:

1. distinctiveness,
2. memorability,
3. proportionality,
4. symmetry,
5. attractiveness, and
6. expressiveness.

Speaking to HuffPost UK, Associate Professor Dawel explained: “AI faces tend to be more symmetrical, proportional and attractive but less distinctive, memorable and expressive than human faces”.

Student Tanya George, who helped to train the participants, said: “We found that even relatively short training sessions helped participants improve their accuracy in detecting AI-generated faces, highlighting the potential for practical education tools in this area.”

She added, “AI image-generation technology is improving extremely quickly, and many people underestimate how convincing these faces can be. Research like this can help people navigate increasingly complex online environments.”

By the way, if the topic takes your interest, the Australian National University has said their Emotions and Faces Lab are interested in hearing from people who might want to undertake training or participate in related studies.

You can register to participate here.