Eliot Higgins – the founder of the Netherlands-based investigative journalism group, Bellingcat – took to Twitter to post versions he made on Midjourney, a popular AI image generator. Some of the images are photorealistic, while some feel more like CGI from a video game. — AP
Images of former president Donald Trump in handcuffs and flanked by police officers went viral across social media Monday and Tuesday. But they aren't real.
As news consumers wait to see if Trump will be indicted, AI-generated images depicting his arrest are spreading online, raising questions and concerns over media literacy and deepfakes as the tools to create them become more accessible.
Eliot Higgins – the founder of the Netherlands-based investigative journalism group, Bellingcat – took to Twitter to post versions he made on Midjourney, a popular AI image generator. Some of the images are photorealistic, while some feel more like CGI from a video game.
Higgins disclosed that his images were fake, captioning the Twitter thread "Making pictures of Trump getting arrested while waiting for Trump's arrest." He and others made the art as a form of parody. Several versions of fake Trump arrest images began widely circulating on social media, initially by their creators.
But along the way, the art was reposted and reshared across Twitter, Instagram, and TikTok, as if they were "photos" of a major news event, not AI-generated art of a fake occurrence.
How can you tell if a photo is AI generated?
In some cases, social media platforms' content moderation services are attempting to flag the images of Trump getting arrested as fake. But without a content flag, some of the art can be convincing.
As noted by Wired, even while image generators are getting smarter, they still make mistakes in the smaller details. While the main subject of the image may look legit, the rest might appear to be an afterthought. For instance, Trump's face looks realistic in most of the fakes, but his body proportions look contorted and almost melted or surreal.
Other hints that an image is fake include odd or messy-looking text – like in police officers' uniforms and badges – and over-the-top facial expressions.
When in doubt, tools like Google's reverse image search can help pin down the source of a photo. Additionally, online tools for AI-detection are becoming more accessible. Hugging Face, a company that focuses on AI technology, has a free AI-image detector that lets users drag and drop a photo into its portal and will determine if it is human or artificial.
What does this mean for the future?
For fans of AI, the system's advancement is exciting. But critics worry about its ramifications when it comes to media literacy and, separately, the devaluation of creators' time and labour.
"I had assumed that people would realise Donald Trump has two legs, not three," Higgins from Bellingcat told the Associated Press in an email. "But that appears not to have stopped some people passing them off as genuine, which highlights that lack of critical thinking skills in our educational system."
Artificial intelligence in the art scene is viewed with both advantages and disadvantages. Artists like New Jersey's Ryan Evans value the new tools for the ability to streamline the art he's already making and open up new possibilities. But he also acknowledges the services' potential to take jobs away from artists.
In the case of the Trump art, Evans says the images' spread without context is "pretty horrifying."
"The younger generations will have no problem recognising AI imagery due to art, memes, and overexposure. But I'm concerned about older generations," he said. "It's like that special frequency of sound only teenagers can hear. I think it's time we sit our parents down for an AI literacy check." – The Philadelphia Inquirer/Tribune News Service
