Thursday, March 7, 2024

The Challenge of Spotting AI-Generated Faces

A study from the University of Waterloo in Canada has shown that many people can't tell the difference between real human faces and those created by artificial intelligence (AI). Almost 40% of people get fooled by AI-generated images. This is worrying because AI images are getting better and could be used wrongly, like in spreading false information.

In this study, 260 people looked at 20 pictures without knowing where they came from. Half of these pictures were of real people found on Google, and the other half were made by AI programs called Stable Diffusion and DALL-E. These programs are very good at creating images that look like real human faces. The people in the study had to guess which pictures were real and which were made by AI. Only 61% got it right, which was much less than the 85% the researchers thought would.

The study's lead researcher, Andreea Pocol, said that people are not very good at telling real and AI-made images apart. People tried to guess based on small details like how fingers, teeth, and eyes looked. But these clues were not as helpful as they thought. What makes it harder is that most people don't look closely at pictures on the internet. They scroll fast, so it's even easier to get tricked by fake images.
AI is getting better very fast, and it's hard for research and laws to keep up. AI-made images can look very real, raising fears about their use to make fake images of public figures or to spread lies. Pocol pointed out that spreading misinformation isn't new, but the tools to do it are changing. There might come a time when it's very hard for anyone to spot fake images, no matter how much they know. This leads to a need for new ways to find and fight against fake AI images.

This study shows a big issue with how we see pictures online and trust them. It's making people think again about how much we can believe what we see in the digital world.


Image: DIW-AIgen

Read next: Latest Google Spam Explainer Update Targets Link Signals: What You Need to Know
by Mahrukh Shahid via Digital Information World

No comments:

Post a Comment