Communications of the ACM
Fake Faces Created by AI Look More Trustworthy than Real People
Generative adversarial networks can learn to create fake images that are less and less distinguishable from real images by pitting two neural networks against each other.
Credit: Anatolii Babii/Alamy
Researchers at the U.K.'s Lancaster University and the University of California, Berkeley found that people have a hard time distinguishing images of human faces created by artificial intelligence from images of real faces.
The researchers asked a group of 315 people to distinguish a selection of 400 fake photos from 400 photos of real people; they were accurate less than half (48.2%) of the time (a second group trained to recognize computer-generated faces did slightly better, with an accuracy rate of 59%).
The researchers found that white faces were hardest for participants to distinguish, possibly due to the software being trained on disproportionally more white faces.
From New Scientist
View Full Article
Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA
No entries found