This is at least what can be deduced from
Posted: Tue Jan 07, 2025 4:08 am
artificial intelligence , most of us would probably boast that we can tell the difference between the two without a doubt. However, not only are we not very good at differentiating real faces from "fake" faces, but we are also more inclined to trust the latter than the former.
a study published in Proceedings of the National Academy of Sciences and written by Hany Farid , a professor at the University of California, and Sophie J. Nightingale , a professor at Lancaster University.
Farid has been researching AI-generated images (and how people distinguish them norway number data from real ones) for years . However, the fascinating world of synthetic nature images has taken a giant leap forward in recent years, so much so that real faces are now almost indistinguishable from fake faces, Farid says.
The problems arising from the proliferation of such extraordinarily realistic AI-generated faces are many and varied. “ Fraudulent online profiles are a good example . So are fraudulent photos that make their way into passports. Still images lend themselves to a number of nefarious uses. But where things are really going to get tricky is in the world of video and audio,” warns Farid.
AI-generated images can trick people
Given the rapid pace of AI-powered image generation, Farid and Nightingale wanted to explore in their research whether "fake" images can really trick people into believing they are authentic.
Their paper includes three experiments with the ultimate goal of elucidating whether people can actually discern a real face from a synthetic one created by Nvidia's StyleGAN2 generative network.
After confronting participants with more than 800 images of real and fake faces, only 48.2% of the answers were correct.
In a second experiment , the researchers then gave participants a series of cues to identify AI-generated faces and took the trouble to provide some feedback, but this did not dramatically increase their ability to distinguish real faces from fake ones. The participants' correct guess rate in this second experiment was 59%.
a study published in Proceedings of the National Academy of Sciences and written by Hany Farid , a professor at the University of California, and Sophie J. Nightingale , a professor at Lancaster University.
Farid has been researching AI-generated images (and how people distinguish them norway number data from real ones) for years . However, the fascinating world of synthetic nature images has taken a giant leap forward in recent years, so much so that real faces are now almost indistinguishable from fake faces, Farid says.
The problems arising from the proliferation of such extraordinarily realistic AI-generated faces are many and varied. “ Fraudulent online profiles are a good example . So are fraudulent photos that make their way into passports. Still images lend themselves to a number of nefarious uses. But where things are really going to get tricky is in the world of video and audio,” warns Farid.
AI-generated images can trick people
Given the rapid pace of AI-powered image generation, Farid and Nightingale wanted to explore in their research whether "fake" images can really trick people into believing they are authentic.
Their paper includes three experiments with the ultimate goal of elucidating whether people can actually discern a real face from a synthetic one created by Nvidia's StyleGAN2 generative network.
After confronting participants with more than 800 images of real and fake faces, only 48.2% of the answers were correct.
In a second experiment , the researchers then gave participants a series of cues to identify AI-generated faces and took the trouble to provide some feedback, but this did not dramatically increase their ability to distinguish real faces from fake ones. The participants' correct guess rate in this second experiment was 59%.