A new study published in the Proceedings of the National Academy of Sciences finds AI-synthesized faces are indistinguishable from real faces and humans actually find them slightly more trustworthy. The results highlight not only the risks of deepfakes, but also a tremendous opportunity for researchers and businesses.
The study looked at three experiments. In the first experiment, 315 participants classified 128 faces taken from a set of 800 as either real or synthesized. Their accuracy rate was 48%, no better than a 50-50 guess.
In a second experiment, 219 new participants were trained on ways to identify deepfakes and given feedback on how to classify faces. This group classified 128 faces taken from the same set of 800 faces – but despite training, the accuracy rate improved to only 59%.
Finally, a third group of 223 participants rated a selection of 128 of the images for trustworthiness on a scale of one (very untrustworthy) to seven (very trustworthy). This group rated synthetic faces a slightly higher average of 4.82, compared with 4.48 for real people.“We found that not only are synthetic faces highly realistic, they are deemed more trustworthy than real faces,” says study co-author Hany Farid, a professor at the University of California, Berkeley.
The results make clear that humans intuition and discernment alone will not be able to effectively combat deepfakes. Researchers should be proactively focused on countermeasures and other tools and techniques that can help detect deepfakes.
While the results suggest deepfakes can be highly effective when used for nefarious purposes, it also highlights the effectiveness to which marketers can use AI-synthesized faces and tools in promoting their services and products. The fact that humans find AI-synthesized to be more trustworthy than photos of actual humans suggests marketers might be able to leverage this attribute to form a stronger bond between consumers and their marketing message. Already, companies like LG are using synthetic humans to promote their products. The study results also highlights the potential role AI-synthesized humans might play in the metaverse.
