The research, carried out by a team of scientists from Australia, the United Kingdom and the Netherlands, made a surprising revelation: images of white faces produced by artificial intelligence algorithms can fool people into believing that they are human – even more so than real human faces.
“Remarkably, white AI faces can convincingly pass as more real than human faces – and people do not realize they are being fooled,” the study authors reported.
This could have serious real-world implications, including identity theft via fake, hyper-realistic profile photos created by AI. People can interact with digital impostors posing as real humans in online spaces.
However, this phenomenon was limited to white faces only. The benefit of realism did not extend to AI-generated images of people of color. The researchers believe the AI system was primarily trained on white faces.
Dr. Zak Witkower, co-author of the study from the University of Amsterdam, noted that this racial disparity in AI realism could negatively impact areas such as online therapy, robots social media and many more, which rely on convincing simulated faces. “This is going to produce more realistic situations for white faces than for other racial faces,” he said.
By conflating perceptions of race and humanity, AI face generators risk exacerbating social biases, particularly in missing child alerts that rely on widely distributed AI-generated photos.
In an experiment conducted as part of the study, when shown a mix of 100 real white faces and 100 AI-generated white faces, participants were more likely to view the AI faces as real humans rather than real photos. This effect persisted even when participants were not informed that certain faces were generated by the AI.
Researchers have identified factors such as excellent facial symmetry, familiarity and memorability as the main reasons why AI faces fooled humans. Ironically, a machine learning system developed by the team could accurately identify real faces from fake faces 94% of the time, much better than humans.
Co-author Dr Clare Sutherland from the University of Aberdeen highlighted the need to tackle racial bias in AI systems. “As the world changes extremely rapidly with the introduction of AI, it is essential that we ensure no one is left behind or disadvantaged. any of them situation – whether because of ethnicity, gender, age or any other protected characteristic,” she said.
photo by It’s engineering.