Real or fake? Every third American fooled by AI images

An experiment has shown that 35% of Americans could not tell photographs of famous landmarks from images generated by artificial intelligence (AI).

AI-generated images are causing waves lately. Just last week, Reddit users were fooled into thinking that an earthquake in Oregon really happened in 2001 – in fact, a series of images of the supposed aftermath were created using Midjourney, an AI tool.

There was also the image of Donald Trump, former US president, getting arrested, and Pope Francis allegedly decided to dress up in a puffer jacket.

Indeed, AI has made such significant advances that many experts worry people will soon not be able to distinguish between what’s real and what’s fake – misinformation might then enter its golden age.

In fact, a new poll by Concured, an AI-powered content recommendation system, justifies such concerns. Research shared with Cybernews shows more than a third of Americans could not tell which established landmarks shown to them were actually photographed, and which were created by an AI tool.

Of the thousand people polled, one third could not recognize the actual Eiffel Tower, half couldn’t tell which Taj Mahal was real, and a like proportion were fooled by a fake Grand Canyon image.

At least the Moon did well – 64% of the polled people successfully identified the authentic image compared to the AI-generated one.

“AI will transform content and marketing. It is already automating complex tasks, personalizing experiences, and providing data-driven insights, leading to increased efficiency, effectiveness, and profitability,” said Tom Salvat, founder and CEO of Concured.

“With 35% of Americans already being unable to differentiate between authentic and AI images, it already underlines the power of AI and the difficulty to detect AI-created content.”

One of the most effective ways for amateurs to find out if an image was created with AI is to look at the hands and teeth. But that’s helpful only if people are in the images – as the Concured poll has shown, when it comes to pictures of static landmarks it is much harder to differentiate the fake from the real.

However, AI-generated images still have issues and differ from real images. Artificial images could sometimes have less detail and appear a little bit pixelated or distorted than photographs.

Stock photo companies have already been banning AI-generated images from their platforms, Getty Images being the largest among them. The firm has sued AI image generators over unauthorized use of millions of images to train machines.