Caucasian faces created by artificial intelligence (AI) now look almost even more “realistic” than real human faces, according to Australian National University (ANU) researchers.
Although participants in this new study found white human faces created by artificial intelligence more realistic than real faces, the same was not true for images of black people. The senior author of the article, Dr. According to Amy Dawel, this is because AI algorithms are trained to a much greater extent on white faces.
“If white faces created by AI are consistently perceived as more realistic, this technology could ultimately have serious consequences for people of color by reinforcing racial biases online,” Dawel said in a statement. It is already evident in the current AI technologies used. “When artificial intelligence is used for people of different races, it transforms their skin and eye color to that of white people.”
We fall for artificial intelligence very quickly
Rapid advances in the ability of artificial intelligence are now starting to make it difficult for us to appreciate how powerful this technology truly is. As this research shows, people don’t always realize they’re being fooled by AI’s “hyper-realistic visuals.” “Worryingly, people who thought the AI faces were real were, paradoxically, the most confident that their judgment was correct,” said study co-author and PhD candidate Elizabeth Miller at ANU. It means that people who confuse people do not know that they are being deceived.”
Interestingly, the team believes they have a reason why people can be fooled so easily. There still appear to be physical differences between AI and real human faces, but people are misinterpreting them. Dawel explained that, for example, white faces created by AI are often more proportional, but viewers see this as a sign of their “humanity”: “But we can’t rely on these physical cues for long. “AI technology is advancing so quickly that the differences between AI and human faces will likely disappear soon.”
The danger of spreading false information on the Internet
It can easily be said that such developments may make it easier for misinformation to spread on the internet. The team argues that future action is needed to limit the spread of misleading information and the potential identity theft that comes with AI images.
Dr. “AI technology cannot be compartmentalized, so only technology companies can know what is going on behind the scenes,” Dawel said. “More transparency around AI is needed so that researchers and civil society can detect issues before they become major problems.”
The team emphasizes that it is important for the public to become more aware of the potential misuse of AI technologies to reduce risks. Since individuals can no longer accurately distinguish between real and AI-generated faces, it can be seen that society needs tools that can accurately identify an AI impostor.
The study was published in the journal Psychological Science.
40927
