Why Facebook has been filled with Jesus Christs made with eggs, spaghetti or cauliflowers

“No one likes my artistic work”, “I made it with my own hands!”, “May 2024 is your best year”, are some of the texts that accompany thousands of bizarre images of Jesus Christ generated by intelligence on Facebook. artificial. The network’s algorithm promotes them, some become viral and manage to attract huge audiences for pages that then take advantage of their success to monetize or even scam users.

“The images are captivating: some are beautiful, some are sensational, and some are just weird,” says Renée DiResta, a researcher at Stanford University and co-author of a March paper on this phenomenon titled “How spammers and scammers use AI-generated images on Facebook for audience growth.” Thousands of users interact with these images, many with a simple “amen”, which could mean that they are bots.

Facebook has become over the years a network where content is shared and a basic objective is to increase the audience. Jesus Christs are another resource that is easy to create and attracts attention: “Competition for attention among page creators means that they will use any tactic that seems to work, which now includes posting AI-generated content that recommendation systems they seem to be promoting,” DiResta continues.

The texts added by the authors serve to promote this virality. In some images, a black child stands next to a Jesus Christ made of spaghetti and the text says: “Thank you to everyone who appreciates my art.” None of these images indicate that they have been made artificially.

The academic article has studied 120 Facebook pages that have published at least 50 AI-generated images each. Proof of success is that the pages had an average of almost 130,000 followers. Shortly after looking at those photographs, Facebook’s algorithm offered more examples and new variants. The ones that EL PAÍS has seen were from pages of Latin American users who combined these religious images with memes, animals and still lifes also made with AI, where they overwrite “good morning” or funny videos. One of them, a Haitian resident in Mexico, briefly responded to this newspaper that he did make those images, that people really like them, and he acknowledged that something can be earned with them: “Yes,” he responded, but he did not want to explain how. . This user had more than 7,000 followers.

Beyond the cheating within Facebook, the underlying debate is: if people confuse these ridiculous images with real art, what will happen when what is generated by AI is truly realistic? There are no certainties, although not everyone is confused either. Under one of these photos a user says: “What a wonderful sculpture these are hands blessed by God, my love, you are a genius, God bless you today tomorrow and always congratulations from Colombia.” Another responded: “Oh lady, how can you believe in these things, how easy to deceive you are, don’t you realize that it is a false image? “You don’t have to believe everything that is published on the networks.”

The gullible woman received another response that illustrates one of the types of scam. If someone believes that these fake images are real, they are probably more naive and easy to scam: “I am French and it has been more than 13 years since my husband matured.” [sic] in a car accident. After death, I inherited my husband’s fortune. Considering my advanced age and health, I decided to financially help people who wish to receive financial assistance for their needs, since I do not have children to benefit from my inheritance and I do not want the government to freeze my fortune after death. So if you are interested in economic activity, send me a message,” and added a link to a real WhatsApp number.

“Scam accounts occasionally interact with gullible commenters on posts, both on pages and in groups, sometimes asking for personal information about them or offering to sell products that don’t exist,” the article explains. Once an audience is created in a group or page, the ways of trying to extract benefits vary: “We mark pages as ‘scam’ if: one, they trick followers by stealing, buying or changing control of the page; two, falsely state a name, address or other identifying characteristic; and/or, three, they sell fake products,” details the academic article.

Some of the pages that use these tactics often change their name and keep their followers to do something else or sell the audience. Spam pages take advantage of attention to direct users to pages outside of Facebook that are filled with ads.

Jesus Christ is just one of the main claims. In the third quarter of 2023, one of the 20 most viewed images on Facebook globally was this kitchen with a glass floor. It had 40 million views and 1.9 million interactions, according to the platform’s quarterly report on its most viewed content. Some of the pages that try this trap also try animals, children, cabins or landscapes, anything that might attract attention on Facebook: “Just like Picasso had his blue period, the pages often went through periods, a few dozen snow sculptures; a few dozen watermelon carvings; a few dozen wood carvings; a few dozen artfully arranged sushi plates, each in a very similar style,” the article says.

“We suspect that these high levels of engagement are driven in part by Facebook’s recommendation algorithm,” DiResta and his co-author Josh A. Goldstein of Georgetown University say in the paper. Two years ago, according to a leaked document from Meta, the social network changed its algorithm to be more like TikTok. The goal was, according to the document, “to help people find and enjoy interesting content regardless of whether it was produced by someone you are connected to or not.” The authors show how this “Suggested for you” content went – ​​as a percentage of the number of views – from 8% in 2021 to 24% in 2023.

Although the hub for all these viral images is Facebook, there are also Instagram: “AI-generated images are also popular on Instagram,” says DiResta. “They’re on Twitter and everywhere, but the design of Facebook and Instagram makes people create specific types of visual content and build audiences over time.”

In February Meta announced that it was going to label AI-generated content on Facebook, Instagram and Threads. At the moment it does not happen in any of these cases. To assess potential dangers, this detail is important. Although in some cases it is easy to detect AI-generated images, this will not always be the case: “Even in cases where there are obvious signs, identifying them still requires a lot of cognitive effort, and it may be unrealistic to expect the average Facebook user to be simply browsing.” Look at your chronology.