Social media platforms have been inundated over the past several days with shocking images that purport to show former President Donald Trump being arrested by New York City police officers.
Except, Trump has not been placed under arrest by NYPD officers -- at least, not yet -- and the viral images were all created with an advanced artificial intelligence image generator, the Associated Press reported.
The "highly detailed, sensational images" have sparked legitimate concern among some about how relatively easy it has become for people to make use of increasingly sophisticated AI technology to spread misinformation and fake pictures and videos online that, at least at a glance, can appear convincingly real.
Eliot Higgins, the founder of the European-based investigative journalist collective known as Bellingcat, has admitted that he made use of the latest version of an AI text-to-image generator known as Midjourney to create dozens of images of former President Trump that were inspired by the news reports of his possible impending arrest as well as a famous movie about prison.
In a lengthy thread of tweets first posted on Monday, the pictures created by Higgins -- who claims he had no "ill intent" -- form a narrative that begins with Trump being arrested, the reactions of his family, sitting in a courtroom for his trial, life in prison in an orange jumpsuit, and eventually making his escape ala the classic film "The Shawshank Redemption," and ultimately concluded with him ending up at a McDonald's restaurant.
— Eliot Higgins (@EliotHiggins) March 20, 2023
"The Trump arrest image was really just casually showing both how good and bad Midjourney was at rendering real scenes," Higgins told the AP. "The images started to form a sort of narrative as I plugged in prompts to Midjourney, so I strung them along into a narrative, and decided to finish off the story."
Buzzfeed News reported Wednesday that Midjourney, in response to the viral images created with its technology, banned Higgins from any further use of the platform. He told Buzzfeed that he'd "been playing around with various prompts to see what's possible and how complex you can make it."
He found some of the images generated to be "amusing," so he "put it out there. I didn't intend to do any clever criticism or anything like that. But then it kind of took on a life of its own."
As for being banned by Midjourney, Higgins said, "I suspect it was pushing my luck when I did the thread," and added, "let alone when it went viral."
Wired reported that while the AI images of Trump being arrested and imprisoned can appear to be legitimate at a glance, it actually isn't too difficult to tell that they are fake with just a moderate amount of scrutiny outside of the main subject of the image.
That is because, for all of its rapid advancements, AI image-generation technology still struggles with accurately rendering things like facial expressions, fingers and hands, and natural human body movements and proportions. Other telltale signs of fake imagery can be blurred or blended details in the background as well as nonsensical lettering where actual words should be.
So what can be done about it? Higgins and others have suggested that, in addition to strengthening already existing policies against spreading false images and misinformation, social media platforms should work at creating technology that can instantly detect and make note of AI images before they go viral.
Whatever the ultimate solution turns out to be, it will need to be addressed rather soon, as "deepfakes" such as the Trump arrest images will only become even more realistic in appearance as the kinks are worked out of the AI image-generation technology, which could spell disaster in certain circumstances in the future if used for nefarious purposes.
