It's hard to believe that only a year and a half ago, artificial intelligence tools began to open up to the public, and since then they've improved from being ridiculed for drawing humans with 6 fingers in each hand to a stage where they manage to completely mislead us. According to a new study by researchers from the Australian National University, the University of Toronto, the University of Aberdeen and University College London, and published in Psychological Science, a journal of the Society for Psychological Science, white human faces created by artificial intelligence are now perceived by observers as more real than pictures of real people. We used artificial intelligence Perplexity to bring the news that AI is already completely fooling us.
The researchers conducted two experiments to test the perception of faces created by artificial intelligence. In the first experiment, 124 participants were shown images of 100 faces created by AI tools and 100 real faces, and they were asked to identify which faces were real and at what level they were confident in their choice. 66% of AI images were identified by participants as humans, compared to only 51% of real images. In the second experiment, more than 600 people rated different qualities of the images, such as attractiveness, familiarity, and image quality. There remain clear differences between AI and real images, but according to the researchers, participants misinterpreted the details or missed them altogether. They also noted that what trashed many was the "averageness" of the artificial face—that is, AI draws people who are neither too pretty nor too ugly, who look simply "ordinary."
If we were to cut the body from the picture, would you recognize that it is a fake?, Photo: Israel Hayom via Leonardo.Ai
The phenomenon of falling into the trap of faces created by artificial intelligence, dubbed "hyperrealism," has an unfortunate catch: it only applies to white humans. In images of blacks, artificial intelligence is much less good, because the algorithms are mostly trained on images of whites.
The study's findings raise concerns about the spread of misinformation and identity theft, and the researchers argue that more transparency is needed around AI technology, along with raising public awareness about the perceived realism of AI-generated faces. "Given that humans can no longer recognize AI faces, society needs tools that can accurately identify AI-generated impostors," said Dr. Amy Dowell, one of the study's authors. "Educating people about the perceived realism of AI faces can help make people skeptical about the images they see online."
Wrong? We'll fix it! If you find a mistake in the article, please share with us