Face the truth.
A recent study has revealed that AI-generated faces can be perceived as more realistic than actual faces, particularly among white individuals. The research, featured in the latest edition of Psychological Science, demonstrates how AI has the ability to consistently deceive individuals into believing that computer-generated faces are more authentic than real ones.
The study highlights a concerning trend where individuals are misled into thinking that AI-generated faces are genuine, leading to a blurred line between reality and artificiality. This phenomenon, termed “hyper-realism,” challenges people’s perceptions of truth and falsehood.
Elizabeth Miller, a co-author of the research, noted that individuals who mistakenly identify AI-generated faces as real are ironically more confident in their judgments. This suggests a lack of awareness among those who fall for the illusion of AI authenticity.
Interestingly, the study found that this effect was only prominent in images of white faces and did not extend to images of people of color. Amy Dawel, the lead author of the study, pointed out that AI algorithms predominantly trained on white faces could inadvertently perpetuate online racial biases.
The implications of this technology are significant, particularly for communities of color, as AI systems tend to alter the features of individuals of color to align with white standards. This distortion of physical attributes blurs the distinction between real and artificial faces, leading to a misconception of what constitutes authenticity.
Dawel emphasized that the advancements in AI technology are rapidly closing the gap between real and AI-rendered faces, making it challenging to differentiate between the two. The prevalence of AI-generated faces that appear “average” further complicates the identification of real faces, posing a potential threat to online interactions and perceptions.
As the study suggests, the normalization of white AI faces as the epitome of realism may have far-reaching consequences, perpetuating biases and undermining the authenticity of diverse representations in digital spaces.
