Unsettling study claims that humans can no longer differentiate between AI-generated and real faces with certainty
According to a study published by Sophie Nightingale of the University of Lancaster and Hany Farid of the University of California, Berkely, humans can no longer tell the difference between AI-generated images and real pictures with absolute certainty. A sample size of 800 photos, 400 real and 400 AI-created, of South Asian, East Asian, Black, and White faces was used to test the hypothesis.
Upon testing, researchers found that people were able to correctly tell the fake photos apart only 48.2% of the time. In other words, people confused the AI-generated images for real ones more than half of the time. However, after the participants were given some training on how to spot the fakes, they correctly pointed out the fabricated photos 59% of the instances.
Researchers also tested the trustworthiness of the AI images compared to real ones. Disturbingly, fake pictures were rated as more trustworthy. In an attempt to explain the result, researchers wrote,” A smiling face is more likely to be rated as trustworthy, but 65.5 percent of our real faces and 58.8 percent of synthetic faces are smiling, so facial expression alone cannot explain why synthetic faces are rated as more trustworthy”.
After seeing the findings of the study, authors of the study expressed their concerns about such fabricated images being used for ulterior motives. They urged the developers to weigh the benefits of these AI techniques against the risks and develop countermeasures for protecting against the abuse of the technologies.