People are asking AI to recreate the identical image over and once again, with each iteration drifting further and farther from the unique.
The results are sometimes amusing, sometimes unsettling. In some cases, the photographs completely shape-shift into crazy abstract forms. In others, facial expression are wildly exaggerated.
One of essentially the most viral images is of actor Dwayne “The Rock” Johnson replicated a staggering 101 times.
any individual on Reddit told ChatGPT to copy a picture of The Rock without changing anything 100 times over pic.twitter.com/IcRgmWHCWK
While the primary few iterations closely resembled the unique photo, subsequent versions saw Johnson’s features morph and warp, eventually becoming totally abstract.
ChatGPT prompted to “create the precise replica of this image, don’t change a thing” 74 times pic.twitter.com/u6E8aVThy2
I attempted the “Create the precise replica of this image, don’t change a thing.” trend. pic.twitter.com/xcAcsvBRJp
So what’s happening under the hood? It’s primarily a results of how AI models are trained and the way they encode and reconstruct images.
When an AI is asked to recreate a picture, it doesn’t simply copy and paste the unique pixels. Instead, it breaks the image down into a fancy set of features and patterns, which it then tries to reassemble based on its understanding of what the image should seem like.
However, this process is inherently imperfect and introduces small errors or deviations every time. As the image is repeatedly fed back into the AI, these deviations compound, resulting in increasingly distorted or unexpected results.
It’s a bit like playing a visible game of “telephone” or “whispers,” where each message you whisper to the subsequent person introduces recent features.
However, AI’s aberrations can also reveal something concerning the biases and assumptions baked into these models. For example, some images appear to exaggerate facial expression or create a hotter, more orange-tinted color palette.
Users also noticed that eyebrows turn into highly exaggerated – almost painted on within the type of social media filters. As for the orange tint, some speculate that warmer tints are preferred in photography and thus are more common within the training data.
Really, though, now we have no idea what’s happening contained in the immense “black box” that’s today’s largest frontier models.
But within the meantime, social media users appear to be having loads of fun with the surreal, often disturbing results of repeated recursive AI image generation.
Trends involving AI, like we recently saw with AI motion figures, have recently taken off on social media, with hundreds of individuals getting involved across X, Reddit, TikTok, and Instagram.
One quipped, “I drained the ocean replicating my image 100 times.” Not to be a buzzkill, however it’s an excellent point.
