How worried should we be about fake AI photos?

These AI developments could have deepfake-like consequences.

How worried should we be about fake AI photos?
An AI-generated image TMS created using Midjourney with the prompt "Salt Bae running for US President" Source: Midjourney/TMS

Earlier this week, we shared viral photos of Pope Francis wearing a drippy puffer jacket. The photo has spread all over the internet and become a meme. When it became clear that AI actually created this photo, many were totally shocked. It actually does look … pretty real. But we should’ve known that the pope wouldn’t be that fashion-forward.

This pope meme shows that AI is getting better at making images that are hard to distinguish from real photographs. Usually, the uncanny valley of AI-generated images is enough for us to see that something is not quite right. Looking at fingers and hands, seeing if someone’s smile has too many teeth, and zooming in to check for image defects all definitely gave the game away. Well, not this time.

And these AI developments could have deepfake-like consequences. The latest version of Midjourney’s image production tool can produce pictures that are hard to spot as fakes. Its text-to-image model is now better at producing pictures that look like news agency photos.

When news came out that Donald Trump may get arrested, Trump himself actually shared AI-generated photos of him praying on his social media platform Truth Social. It’s clear the photo isn’t real. But, on Twitter, users were posting their own AI-generated photos of Trump running away from/getting arrested by police. With all these pictures circulating, spotting misinformation becomes a lot harder.

“It does add noise during crisis events. It also increases the cynicism level,” said Jevin West, a professor at the University of Washington who focuses on the spread of misinformation. “You start to lose trust in the system and the information that you are getting.”

Fake AI "photos" of French President Macron among protestors
Source: Twitter/@nocontextfre

On Tuesday, the popular Twitter user “No Context French” posted three AI images of French President Emmanuel Macron running in the middle of a Paris protest, surrounded by police, protesters and smoke. And some people believed the photos were real. While more savvy social media users may know what to look for to see if images are fake, casual viewers and people who aren’t online as much may have more trouble. So, don’t believe everything you see on the internet.