People rely on 'outdated visual cues' to spot AI-generated faces

I like to think I'm pretty good at spotting AI-produced images — soft lighting, blurry backgrounds, and missing small details are my usual clues — but the latest test from UNSW Sydney knocked me down a peg. Their demo shows 20 faces that participants must label as either "Human" or "Computer-generated (AI)." I scored 14/20, above the average of 11/20, though that average is only slightly better than chance.

The researchers gave the test to 125 people and published the findings in the British Journal of Psychology. Among the group, 36 were identified as "super-recognizers" — folks who "excel at a range of face processing tasks including recognizing faces they have only seen briefly before and perceiving differences between very similar faces" — and they were better than average at telling AI faces from real ones.

The team noted that "AI discrimination ability was also associated with individuals' sensitivity to the ‘hyper-average’ appearance of AI faces." Lead researcher Dr. James D.

Australia, Sydney

ai faces, ai-generated, computer generated, unsw sydney, british journal, super-recognizers, face processing, hyper-average, 125 participants, face test