Life Hack: How to spot a fake, AI-generated image
Brad Pitt arrives at the premiere of "Bullet Train" on Monday, Aug. 1, 2022, at the Regency Village Theatre in Los Angeles. (Photo by Jordan Strauss/Invision/AP)
In recent days, social media has been abuzz following the news that a French woman was duped out of over €800,000 after believing she was in a romantic relationship with A-list actor Brad Pitt.
While it seems more like a headline than a real one, the lady in question was convinced the online romance was legit thanks to AI-generated images and videos of ‘Brad’ professing his love for her.
This is not something that merely happens on foreign shores.
In Ireland recently, Daniel O’Donnell had to (again) warn against falling for online scammers pretending to be the singer and asking his fans for money.
In another sphere, AI has been used in Irish political campaigning, with Michael Healy-Rae a frequent creator of AI videos where he looks more like a character from a video game than a real person.
Last October he shared a fake video of ‘Taylor Swift’ endorsing him for the election. In this video, the voice sounds like Swift, though speaking too fast and with unnaturally long pauses between sentences.
Her mouth also does not sit at a natural angle on her face when talking. But to a casual online scroller unfamiliar with Tay-Tay’s mannerisms, it could just slip past their judgement.
It seems likely we will be seeing more sophisticated image and video-based scams as AI technology advances, which will put even tech-savvy people at risk of losing a lot of money.
When it comes to spotting an AI image or video for what it is, the devil is in the details.
Much like my own drawing skills, AI struggles to create convincing fine details like hands, which is why you may spot 12-fingered hands at the end of arms or unnaturally shaped limbs — think joints that bend in the wrong direction.
It’s not just the subject of the imagery that you should focus on. Take in the background and determine whether it is realistic.
The person represented in the picture might look real, but where are they located?
If it’s supposedly an Irish person, for example, why are they in a tropical location or in a fantasy world?
Another indicator is writing on an image. If there is a body of text within the scene, such as a jacket logo or information on a poster, zoom in and take a closer look.
Often these are made up of gibberish, with shapes that resemble letters at a glance but mean nothing or a mish-mash of letters creating nonsense words.
Some AI details can look pixelated, for example, the image of Kate Middleton with her children which was shared on Britain’s royal family’s social media last year had some pixelation which caused royal-watchers to take a closer look and spot some questionable details that hinted at AI interference.
On the other hand, some AI creations can be too perfect: skin airbrushed to remove all texture and detail, giving a more ‘plastic’ look to the subject.
There are many examples of AI-generated pictures of famous people that we can learn from.
Some are so out of character they can be spotted quickly, like one of Pope Francis in 2023 wearing a Balenciaga puffer jacket, but others are very convincing.
Last year pictures of Katy Perry, Rihanna, and Billie Eilish at the Met Gala went viral with people sharing them on X, formerly Twitter. However none of the three were actually in attendance.
There were varying images of Perry, in particular, in two completely different gowns — the first red flag.
In cases like this, some quick fact checking can be done by searching news articles around an event to see if they were at it or checking the celeb’s own social media (who can resist sharing pictures from inside the fashion’s most glamorous event?).
Thousands of people were tricked by the fake Perry pictures, including the star’s own mother.
The singer shared the mix-up on her Instagram account to clarify the images were fake and how easily they could fool someone.
Perry was busy in a recording studio during the event when her mom sent her one of the fake images: Perry replied: “lol mom the AI got you too, BEWARE!”
In this world of increasing AI presence, ‘beware’ is right.
