Virus AI-generated The images of Donald Trump’s arrest that you see on social media are definitely fake. But some of these photorealistic creations are quite convincing. Others seem like stills from a video game or a vivid dream. A Twitter line By Eliot Higgins, the founder of Bellingcat, it shows the Trump jam Artificial policeRunning around On the cowAnd Choosing a prison jumpsuit It has been viewed over 3 million times on the social media platform.
What does Higgins think viewers can do to distinguish between fake, AI images from real photos of the former president?
“After creating a lot of images for the thread, it’s often seen. It focuses on the first thing In this case, various members of the Trump family, the surrounding material has often been described as highly flawed, Higgins said in an email. Look outside the focal point of the image, does the rest of the image look like an afterthought?
Although new AI-image tools like Midjourney (version 5 used for the mentioned thread) and stable distribution are making great strides, errors in small details remain a common sign of fake images. As the popularity of AI art increases, many artists point out that the algorithms still struggle to replicate the human body in a balanced and natural way.
Looking at Trump’s AI images from the Twitter thread, his face looks fairly convincing in many posts and his hands, but his Body size may appear distorted. Or melt into the nearest police officer. Although it’s unclear at this point, the algorithm may be able to eliminate seemingly unique body parts with further training and further refinement.
Want to tell me something else? Look for unusual writing on walls, clothes, or other visible objects. Higgins points in the direction Messy text As a way to distinguish fake images from real photos. For example, in the fake footage of police arresting Trump, at first glance the police are wearing badges, hats and other documents that appear to have letters. On closer inspection, the words are meaningless.
A third way you can sometimes tell if an image is AI-generated is by perception. Excessive facial features. “I’ve noticed that if you ask Midjourney for expressions, he exaggerates them. The pain on Melania Trump’s face looks like Edvard Munch’s entertainment The noise Or still rather than a human photographer’s snapshot from some unreleased A24 horror film.
Remember that world leaders, celebrities, social media influencers, and anyone with a photo that goes viral online can look more convincing than AI-generated images from a deep internet presence. Higgins writes, “It’s clear that the more famous a person is, the more images the AI has to learn; Celebrities are greatLess famous people get a little angry. For added peace of mind about the algorithm that creates your face, it might be worth thinking twice before posting selfies after a night out with friends. (Although AI generators may have already scraped your image data from the web.)
Ahead of the next US presidential election, what is Twitter’s policy on AI-generated images? Social media platform Current policy In part, “You may not share artificial, fraudulent, or out-of-context media that is likely to deceive or mislead people and cause harm (‘misleading media’).” Twitter carves out several exceptions for memes, comments, and posts that aren’t created with the intent to mislead viewers.
A few years ago, it would have been unimaginable that an ordinary person would soon be able to create photo-realistic portraits of world leaders at home. As AI images become harder to distinguish from the real deal, social media platforms may need to reevaluate their approach to artificial content and try to navigate users through the complex and often unsettling world of generative AI.
