When Reality Looks Like AI, and AI Looks Real

There has been a lot of debate about a recent photo circulating in the news related to the Louvre burglary. Is it real or simply another example of fake AI attempting to grab our attention?

The image shows French police officers in tactical gear standing at the Louvre Museum, the scene of a recent and perhaps one of the boldest jewelry robberies.  The thieves stole priceless artifacts, gems, and a golden crown in broad daylight, escaping on scooters, whisking through the streets in Paris as if they were in an action thriller movie.

Among the police officers stands a well-dressed man wearing a long coat, a striking yellow vest, and a fedora. He looks like he stepped out of a 1940s era detective novel and into a modern police crime scene. The contrast of imagery is what captivated people. Social media was convinced it was AI, The New York Times journalist referred to him as the “Dapper Man” and began to look in to it. AI experts questioned its authenticity.

Too Good to be True

It is real. I admit at first, I thought it was AI. This makes this story more interesting because we all seem to be at a point where we struggle with identifying AI. We have heard enough of the AI hype of its capabilities, that many things seem plausible.

Generative AI tools have vastly improved over their initial versions in the past year, so it is reasonable to question the images we see. A little more than a year ago, it was easier to spot AI images with their subjects’ off-balance stares, awkward gestures, or other ‘uncanny valley’ appearances. Sometimes things just do not look right. We might not have been able to describe why (we are not all experts in subtle lighting differences and skin tone), but our natural instincts made us wary.

The man in the picture appears as a modern Jules Maigret or Inspector Clouseau. Like these fictional characters, he is just a bit out of place with the scene around him. We know this is often how social media works. It tries to grab our attention. Our suspicions are appropriately raised, which is good.

Experts in AI mimicry might point to our skepticism as the first clue. Why would a real person be in this group of police officers? The ‘Dapper Man’ is solving the crime. Is this how things work in France? Maybe, maybe not. Many on social media concluded he was AI, others had doubts. The scene is plausible, another building block of deception, yet still a little off. As one AI expert noted, he looks too good to be real. And when things are too good to be true…

Trust but Verify

So how do we know it is real? The New York Times did what good journalists do and went directly to the source. They spoke with the A.P. photographer Thibault Camus, who confirmed the photo and shared many other similar photos he took that day of people standing around the police officers.

If you do not have the good fortune to work at the New York Times, what can you do to verify a suspicious online photo?  Here are some ideas:

• Go to the source. Search for the photographer if credited. Is the photographer verifiable, and did they have similar photos in other news stories?
• Context and Content. Does the photo seem plausible? Could there be some motivation for someone to fake the image?
• Technical ‘tells’. This is increasingly difficult as AI gets better. Look for details that are off or blurred for no reason.

The Bigger Picture

We have crossed a threshold from trusting reality to distrusting reality. This only happens when AI leads us to distraction, when in the hands of those with nefarious intentions seek to send us in another direction, when we begin to question what is real and what is not.

We are saturated with AI daily, numbing our ability to see things for what they are.

The initial picture might not have been surprising to us in another time or context. We are saturated with AI daily, numbing our ability to see things for what they are. A big crime committed in a public space is interesting. It naturally draws attention from onlookers, and it’s Paris after all, where being fashionable is, well, fashionable. One picture of a well-dressed man strolling the streets could be expected.

However, this will get more complicated for us all to distinguish the truth. To demonstrate how much further this could go down the path of AI deception, I added the animated scene by simply using the original photo. One picture is all it takes, and the results are incredibly good. AI videos are not like those grainy apparitions of the past. In this example, the video is sharp, the movement natural, with only one ‘tell’.

Did you see it? The floating umbrella? AI is not perfect…yet.

For now, the AI photo mystery at the Louvre is solved, but what about the next one?

---

Thanks for reading. Please consider adding your thoughts in the comments below.

More AI Videos

Take 1

Another Take – Magic Hat Flip

AI Adds a Strange Looking Car