AI is getting better at generating faces – but you can train to spot fakes

Images of faces generated by artificial intelligence (AI) are so realistic that even “super-recognizers” – an elite group with exceptionally strong face processing skills – are no better than chance at detecting fake faces.

People with typical recognition skills are worse than chance: most of the time they think that AI-generated faces are real.

This emerges from a study published on November 12th in the journal Royal Society Open Science. However, the study also found that just five minutes of training on common AI rendering errors significantly improves an individual’s ability to spot fakes.

“I think it was encouraging that our fairly short training procedure significantly increased performance in both groups,” said the study’s lead author Katie Greyan associate professor of psychology at the University of Reading in the United Kingdom, told Live Science.

Surprisingly, the training increased the accuracy of super recognizers and typical recognizers by similar amounts…