This will come as no surprise to anyone, but 4chan uses Bing Chat and DALL-E to create and distribute racist images

When you mix the political color of some 4chan users and the use ofImage Generative AIthe result is frankly not pretty to see.

Created in 2003 by Christopher Poole, the 4chan forum played a crucial role in the spread of meme culture on the Web. Very famous memes come from the activity of its members like Pepe the Frog or Rickroll. But 4chan is also a sulphurous story tinged with various scandals. Very provocative content offered thanks to the anonymity of members, raids organized against other websites or catalyst for conspiracy theories or protest movements. Even though the site has seen its glory days pass, it still attracts millions of visits per month. With the rise of image-generative AI like Midjourney, Bing Chat, or DALL-E, some of its users are using these models to generate ever more racist content.

A perverse use of powerful tools

The members concerned are organizing themselves and are not doing this randomly. By publishing posts that encourage people to share racist content automatically generated by AI, they push other users to follow suit. These tools are logically moderated and include security systems to avoid this type of use, but it must be believed that a few well-established techniques are enough to circumvent them.

A refined method

The guidelines that are thus formulated on the site are quite clear: create the content in question, modify it and share it as much as possible. There’s even a visual guide to help create this banned content, hosted on the popular Imgur site (see screenshot above). This gives the detailed procedure to allow anyone who wants to generate racist images. And the matter does not stop there, since the goal is for this content to go beyond the borders of 4chan and to be shared on other larger networks: Instagram, X.com or Telegram for example. For the vast majority, 4chan members who try this activity use Bing Chat which is based on the DALL-E 3 model.

Certainly, neither 4chan nor even the entire Web will have waited for AI to generate this type of content. On the other hand, its diverted use of its basic functions makes it possible to make image editing faster and to simplify the task for those who wish to spread hate messages. Which is very, very worrying. One wonders if there will really one day be a generative AI that is completely safe to use.

Source : 404 Media