The Rise of Anti-AI AI Slop - The Atlantic

Americans are wary of AI in general, and they are especially suspicious of the AI data centers that are popping up across the country like enormous mushrooms. A majority do not want a new data center built in their town. Across the country, community groups have organized to protest individual projects, and activists have successfully lobbied local and state politicians to place moratoriums on the facilities’ construction. But online, the movement has been mutated by some of the same forces it’s protesting. Defenders of the AI industry have claimed that the social-media conversation about the dangers of AI is inauthentic—that, in fact, it’s AI-generated—and to some extent, they’re right. There is a lot of anti-AI AI slop. Much of it is very strange.

Read: Inside the dirty, dystopian world of AI data centers<br>Last week, I perused dozens of local anti-data-center groups on Facebook, and in almost every one, I found people sharing AI-generated materials. Even in these groups, users posted screenshots of AI-generated summaries as backup for their arguments. In the comments under a post about data centers in Texas, a woman shared her concern about the fact that data centers use human stem cells. When someone called her a 🤡, she replied with a screenshot of a Google AI summary for the search Do data centers use stem cells. One Australian start-up is experimenting with the idea, but the AI summary made the practice sound widespread: “Yes, pioneering facilities are starting to utilize living human neurons grown from stem cells as biological processors,” it said. The same week, a town supervisor on Long Island had to debunk a rumor about a new data-center project after an inaccurate AI-generated search summary attracted so much attention that residents planned a protest (which they promoted with a flyer that itself appeared to be AI-generated).

A weirder, more disturbing type of AI-generated anti-AI content started proliferating on Facebook in March. The memes, which show broadly nostalgic images of the American countryside, are shared on state-themed pages with names such as “Life in Michigan” and “North Carolina Life.” In one repeating format, someone has mowed a spiky message into their grass or crops: “NOT WORTH GIVING UP AN INCH OF THIS TO A DATA CENTER,” for instance. (Sometimes they also mow a middle finger.) Another meme shows a boxy new industrial building—presumably a data center—right next door to a beautiful old farmhouse.

An accompanying caption will generally call out the unique qualities that make the state in question so worth fighting for: “quiet roads stretching beside cornfields and barns 🌽,” “Friday night football and county fairs 🎡,” “dark skies over peaceful countryside ✨.” Which state is that? Almost any of them. They’re all the same, but they’re all very special. AI data centers must not infringe on Indiana’s “quiet country roads, golden cornfields, old barns, peaceful sunsets, and the feeling of home that comes with them.” Nor should they be allowed to tarnish Kentucky’s “quiet country roads, golden fields, old barns, peaceful sunsets, and the feeling of home that comes with them.”

By far the most common template pairs an aerial image of pristine farmland with a copy-pasted story about a proud farmer making headlines after turning down a data-center developer’s offer of millions of dollars for his or her land. Although many commenters recognize that the stories are fake, many others offer apparently credulous responses: “Thank you”s and “God bless you”s and “#Respect.” One commenter gently fact-checked a post about an Alabama farmer, based on similar content that he’d come across in other places: “It was actually a Pennsylvania farmer that rejected the $15 million offer,” he wrote, “but there is supposedly a farmer in my home state of Kentucky that rejected a $33 million offer for his 650 acres.” (Actually, one farmer in Kentucky did reportedly turn down a huge offer from an unnamed company in March, but it was for $26 million, and the farmer was a woman.)

That many of these posts are AI-generated is not in question. They are not typically photorealistic. Some images include a deformed (or upside-down) state outline. Others name a state in the image that doesn’t match the one named in the caption. I found one in which the poster seemed to have forgotten to cut out some extra AI-generated text before sharing: “Here’s a Michigan version in the same style,” it says at the top. I also saw a depiction of Pennsylvania with a New York flag flying over the landscape. And in a picture of Texas residents coming together to protest a new data center on the Gulf Coast, one activist holds a sign that says, nonsensically, PRESERVE BEFORE CLOUDS.

Who is making this stuff, and to what end? Maybe foreign actors are to blame. (Kevin O’Leary, the entrepreneur and Shark Tank star, has suggested that opposition to a 40,000-acre data-center project he is...