Researchers made another concerning discovery about AI

Artificial Intelligence (AI) can generate faces that look more real to humans than images of real people's faces according to new research, but that wasn’t the most worrying thing a recently released study revealed about AI. Will you be able to tell the difference between real and fake? 

Published in the journal Psychological Science, researchers revealed that AI can create human faces that look hyper-realistic to the point that they’re nearly indistinguishable from those of real humans, but there’s a big catch. 

Photo by Christopher Campbell on Unsplash

Artificial intelligence can only achieve what the study’s authors defined as hyperrealism, the ability to generate images that humans perceive as more real than actual pictures of real human faces when those faces are white. 

Photo by Mitchell Griest on Unsplash

Live Science covered the new research and pointed out that one of its authors revealed how AI’s hyperrealism could affect people of color while online. But how did researchers make this discovery and how else could it matter?

Photo by Desola Lanre-Ologun on Unsplash

Australian National University (ANU) psychologist and the study’s senior author, Dr Amy Dawel, got inspired to look into the complicated world of AI after similar research published in 2022 discovered images produced by AI were found by many to be more real than reality.

Photo by Kaleidico on Unsplash

Researchers Sophie Nightingale and Hans Farid discovered that human participants in a study couldn’t tell the difference between real faces and those generated by AI, but Dawel wanted to delve into AI’s racial element.

Photo by Raj Rana on Unsplash

The researchers wrote in their study that algorithms are disproportionately developed and trained on producing white faces and they theorized that might be having an impact on the realism of white faces being produced by AI. 

Photo by Aleksandar Popovski on Unsplash

In order to see if their theory about AI was correct, the researchers recruited 610 people and showed them a series of 100 images of human and AI-generated faces using an AI image-generation tool. known as the StyleGAN2. 

Photo by KOBU Agency on Unsplash

Participants were asked to decide if they thought an image was real or generated by AI and then were told to score their confidence in their choice on a scale of 0 to 100. What was revealed was nothing short of remarkable. 

White faces were indeed passing more consistently as realer than actually human faces but the researcher revealed another extremely concerning finding, people didn’t seem to know when they were benign filled by AI. 

Photo by Aure's Notes on Unsplash

"Concerningly, people who thought that the AI faces were real most often were paradoxically the most confident their judgments were correct," explained study co-author and Ph.D. candidate at ANU Elizabeth Miller in a statement. 

Photo Credit: Screenshot Youtube @ANUchannel

"This means people who are mistaking AI imposters for real people don't know they are being tricked." However, the researchers found that AI images were entirely bulletproof and some people still found minor mistakes in them. 

Photo by Usman Yousaf on Unsplash

Dawel pointed out that there are still physical differences in AI-generated images and it is these differences that help some identify fake pictures and trick others into believing that the images are more real than reality. 

Photo by Pakk Camacho on Unsplash

"It turns out that there are still physical differences between AI and human faces, but people tend to misinterpret them. For example, White AI faces tend to be more in-proportion and people mistake this as a sign of humanness," Dawel said. 

Photo by Tamara Bellis on Unsplash

Unfortunately, Dawel also explained that if white faces are consistently perceived to be more real, than it could ultimately reinforce racial stereotypes online, a problem that the researcher believes is already a significant issue.  

Photo by Nathan Bingle on Unsplash

"This problem is already apparent in current AI technologies that are being used to create professional-looking headshots. When used for people of color, the AI is altering their skin and eye color to those of White people."

Photo by Andrey Zvyagintsev on Unsplash

"Given that humans can no longer detect AI faces, society needs tools that can accurately identify AI imposters," Dawel said. "Educating people about the perceived realism of AI faces could help make the public appropriately skeptical about the images they're seeing online."

Photo by Foto Sushi on Unsplash