AI-Generated White Faces Deemed More "Human" Than Real Ones

A recent study reveals that people are more likely to perceive images of white faces generated by artificial intelligence (AI) as real than actual human photographs. The study also emphasizes that individuals are often oblivious to the fact that they are being deceived by these AI-generated faces.

Study Details

The research, conducted by scientists from Australia, the United Kingdom, and the Netherlands, underscores the profound impact this new discovery could have in areas like identity theft, given the risk of people being fooled by digital impostors.

The team outlined their study in the journal Psychological Science, describing how they carried out two experiments. In the first experiment, each participant was shown a mix of 100 AI-generated white faces and 100 real human white faces and asked to identify which were real and which were AI-generated.

The Results

The team found that 66% of AI images were rated as human compared to 51% of real images. Interestingly, the results did not hold for images of people of color. The researchers speculate this may be due to the algorithm used to generate AI faces being primarily trained on images of white people.

Ramifications and Biases

The co-author of the research, Dr. Zak Witkower from the University of Amsterdam, warned that this could have implications for areas ranging from online therapy to robots. He noted that the AI “is going to produce more realistic situations for white faces than other race faces.”

The research team also expressed concern that such a situation could confound perceptions of race with perceptions of being “human”, potentially perpetuating social biases. For instance, the search for missing children could be affected as it often relies on AI-generated faces.

Towards Tackling AI Biases

The researchers stress the importance of addressing biases in AI, as evidenced by Dr. Clare Sutherland from the University of Aberdeen. She highlights that “as the world changes extremely rapidly with the introduction of AI, it’s critical that we make sure that no one is left behind or disadvantaged in any situation.”

Despite humans struggling to differentiate real faces from AI-created ones, the research team developed a machine learning system that can do so with 94% accuracy. This highlights the potential for AI to both deceive and discern in our increasingly digitized world.