There is an image that destroys ChatGPT.

Send images as directions to ChatGPT This is still a fairly new feature, but in my own tests it works fine most of the time. However, someone just found an image that ChatGPT can’t handle, and it’s definitely not what you’re expecting.

The image, discovered by user Brandon_xyzw on X (formerly Twitter), contains some digital noise. Nothing special really, just a black background with vertical lines all around. But if you try to show it to ChatGPT, the chatbot will definitely break the image every time.

I’ve tried passing the image to ChatGPT with and without additional text prompts, as part of a conversation or at the start of a new chat. The only response I get is a ChatGPT error message: “Hmm… Looks like something went wrong.” Trying to create a new answer doesn’t help.

Whatever you do, don’t show ChatGPT this image. pic.twitter.com/DwSkmz0xP6

— Brandon (@brandon_xyzw) January 10, 2024

Interestingly, ChatGPT responded when I sent them a screenshot of a tweet containing the image and described it very well. However, when in the same conversation I tried to show him the real picture, he broke down again.

What is it about this image that ChatGPT hates it so much? Hard to tell. I searched the internet for similar images and found that ChatGPT breaks when encountering some of them, but not all of them. Most likely this is just a bug, given that it processed the screenshot well, as well as similar images with digital noise.

I tested the image with some other popular chatbots and they had no problem telling me more about the image. While ChatGPT struggled to answer my queries, Bing Chat analyzed the image and described it to me in detail. While Bing Chat opted for technical analysis, Google Bard began interpreting the image, stating the following: “Red and blue are often used together to represent opposing or complementary forces. “In this case, the red and blue lines can be seen as symbols of positive and negative energy, or order and chaos.”

Ultimately, while ChatGPT works quite well most of the time, it is not without its flaws. Sometimes he forgets what he can and cannot do, refusing to perform simple tasks that he had no problem with just two messages earlier. Errors when forming responses also happen quite often, but it is this image that constantly breaks the chatbot. We’ll have to try again in a couple of days and see if OpenAI has fixed the problem.