The virus generated by AI The images of Donald Trump being arrested that you may see on social media are definitely fake. But some of these photorealistic creations are quite convincing. Others are more like stills from a video game or a lucid dream. A Twitter feed by Eliot Higgins, founder of Bellingcatwhich shows that Trump is invaded by synthetic copsrun everywhere on fireAnd pick a prison jumpsuit has been viewed over 3 million times on the social media platform.
What does Higgins think viewers can do to tell the difference between fake AI footage, like the ones in his article, and real photographs that could be from the former president’s potential arrest?
“Having created a lot of images for the thread, it’s obvious that he’s often focus on the first object described — in this case, the various members of the Trump family — with everything around it often having more flaws,” Higgins said via email. Look outside the focal point of the image. Does the rest of the image seem like an afterthought?
Even though the latest versions of AI image tools, like Midjourney (version 5 of which was used for the aforementioned thread) and Stable Diffusion, are making great strides, errors in small details are still a common sign of fake images. As the art of AI grows in popularity, many the artists point out that algorithms still struggle to reproduce the human body in a consistent and natural way.
Looking at the AI images of Trump from the Twitter feed, the face looks pretty convincing in many posts, as do the hands, but his body proportions may appear twisted or fade into a nearby police officer. Even though it’s obvious now, it’s possible that the algorithm could avoid odd-looking body parts with more practice and refinement.
Need another say? Look for strange writing on walls, clothing, or other visible objects. Higgins points to messy text as a means of differentiating fake images from real photos. For example, police are wearing badges, hats and other materials that appear to have letters, on the face of it, in the fake images of officers arresting Trump. On closer inspection, the words make no sense.
Another way to tell that an image is AI-generated is sometimes to notice exaggerated facial expressions. “I’ve also noticed that if you ask for expressions, Midjourney tends to render them in an exaggerated way, with the skin folds from things like smiling being very pronounced,” Higgins said. The pained expression on Melania Trump’s face looks more like a recreation of Edvard Munch The Scream or a still image from an unreleased A24 horror film than a snapshot from a human photographer.
Keep in mind that world leaders, celebrities, social media influencers, and anyone with large amounts of photos circulating online may look more convincing in doctored photos than AI-generated images of people with a less visible Internet presence. “Clearly, the more famous a person is, the more images the AI had to learn,” Higgins said. “Very very the famous characters are extremely well rendered, while less famous people are usually a little wonky. For added peace of mind about the algorithm’s ability to recreate your face, it might be worth thinking twice about posting a photo of selfies after a fun night out with friends. (Although it is likely that the AI generators already retrieved your image data from the web.)
As the upcoming US presidential election approaches, what is Twitter’s policy regarding AI-generated images? The social media platform current policy reads, in part, “You may not share synthetic, manipulated, or out-of-context media that may mislead or confuse people and cause harm (“deceptive media”).” Twitter provides multiple exceptions for memes, comments, and posts that were not created with the intent to mislead viewers.
Just a few years ago it was almost unfathomable that the average person would soon be able to make photorealistic deepfakes of world leaders at home. As AI images become harder to differentiate from reality, social media platforms may need to reevaluate their approach to synthetic content and try to find ways to guide users through the complex and often confusing world of AI. generative AI.