Note on Masley’s “I think the data center backlash is mostly about data centers”

Three related points

Andy Masley has written a post on the data center backlash that I think is mostly right and worth a read. He argues that as much as everyone would like the data center backlash to be about something bigger (and related to their specific concerns about AI) it really is mostly about people’s beliefs about data centers, not AI, Big Tech, or something else.

I think the data center backlash is mostly about data centers

Still let me add my two cents. Or rather three cents, in the form of three observations about why these concerns became so dominant.

First it’s notable to me that the data center story is one that is easily templatable. Not all stories are. Any person on a beat can easily see how they can make this a local story, a state story, a national story. It has clear intersecting elements that allow you to target different segments of an audience or electorate. They bought out farms, they put it in a poor neighborhood, etc. They killed the environment, they killed rural America, they polluted urban America, this town should be getting factories but they are getting a data center, this town should be getting low-income housing but they are getting a data center. If you see how many costumes people put on this you realize that this is a perfect example of what Dave Karpf calls conflict expansion, based out of Schattschneider’s theories in Semi-Sovereign People. If you want to win a fight against something you try to make your fight about everyone else’s fight. And to do that you it helps to have an issue you can make about other people’s issues. And it turns out data centers can be about anything. They’re like an issue stem cell.

This isn't to say that data center backlash is about these specific issues (farmers, social justice, the environment, egghead industries replacing factory workers, etc) but that data centers can be broadly framed to be against your politics no matter what those politics are.

Which brings us to the second point. What if the reason data centers are so unpopular is, ironically, because they have no pre-existing political lean? That there is no existing politics about them?

Bear with me. I think this case is actually pretty clear.

Data centers are a rare issue that people are hearing about at a high degree of salience before they have been told by their political elites how to think about it. The fact that data centers seem to in some cases poll worse than coal-fired power plants shows why this matters. Political elite opinion generally puts floors on things. I know how I am supposed to feel about a coal-fired power plant and a wind farm, and other people know they are supposed to feel the opposite. Those elite signals generally reach most people before a flood of stories about those things reach most people, which prevents anything from getting too much support or opposition. Data centers hit so fast that most people were seeing dozens of negative stories as their first introduction to it, a rare case where opinions get formed before the politics gets processed.

Final point. Though people may not want to admit this, information elites (influencers, reporters, researchers, opinion writers, etc) who have an outsized impact on what we pay attention to have a lot feelings about AI. And I don’t think it’s controversial to say information elites have felt more directly impacted by this technology than the average person.¹ Many use it but they also rightly fear what professions already been hollowed out by the web’s dynamics will look like as AI advances. They know it’s much worse at writing and media production than they are, but they also have come to realize the market for good media and writing is smaller than they thought when the low-quality stuff is free.

So if you’re wondering why each one of us sees

• a gazillion stories about AI data centers

• only one news cycle of news stories about the Trump administration announcing they are bringing online the first new coal-fired plants since 2013 this past June,

• or only one news cycle about the administration preventing 17GW of coal power from going offline this year…

…there are many factors at play, but I think it’s rational to think part of it is an AI story just feels more relatable and “now” to people in information fields than one about heart attacks from coal plants.²

After all, some of the first negative externalities of LLMs were felt by university professors. For many, it has made their daily lives as teachers miserable, both becuase of the impacts on their courses and their students (who they care about a lot!).³ Prominent influencers are now competing with an never-ending stream of AI-created slop. Researchers are being asked to review endless AI-produced garbage journal articles. Reporters are watching the owners of their industry salivate at the potential union-busting possibilities of AI production. That doesn’t even count the daily aggravation of people you know saying they “wrote” something with AI and they are a writer now. (Just like you! The person who spent 20 years learning a craft!)

If you’re part of the information elite and this doesn’t impact your sense of what’s a story, I have one question: why not?

So this is one other place I differ with Andy. While I don’t think public opinion about data centers is about AI feelings more generally I think we should consider a related proposition. Namely, that one reason this dominates the reporting, research, and influencer cycles that form the public’s sense of issue salience is because AI issues are naturally salient to reporters, researchers, and influencers.

I’m not really talking about bias; just our collective intuition that “this is a story or research area people will be interested in”. ⁴ After all, coal plants don’t screw up your end of quarter term papers and tank your class’s performance, bust your reporting unions, or force you to compete on TikTok with Grok videos of sheep making friends with ravens. So perhaps when deciding what to cover, share, or discuss the things that hit closest to home for people who produce information and content are just going to seem more engaging to them in general, and take up more of the public discourse as a result.⁵

Or maybe not. But I think we should at least consider the possibility.

1. I’m not excusing myself here. I know that I am a minor information elite, or used to be at any rate. And as I note below I may not be as negative about AI but I’m absolutely talking about AI a lot more than it probably deserves relative to other social threats!

2. Some people will just pivot and say “but that is just about AI too”. This has the disadvatage that it is not true, as Trump ran on a promise of starting new coal plants and keeping coal-fired plants burning quite apart from any concerns about AI compute. If you read the articles as well you’ll understand the older plants being kept open are horribly inefficient.

3. I think there are ways to respond to the AI challenge in education that are really promising, but they are not by any stretch of the imagination simple or certain to succeed.

4. I’m here doing the same thing! I’m not writing a blog post about coal-fired plants, I am writing one about data center perceptions, and the reason I am doing that is because concerns about AI are central to my field and work. I am absolutely influenced by the salience of this issue to me and my audience as a result of my profession. How could I not be?

5. I’ll note here that it is a well-known fact that during the rock radio years when DJs chose songs (rather than central programming) that songs with the word “DJ” or “radio” in them took up a highly disproportionate share of airplay. I don’t think that was intentional. I think as they listened to a variety of songs they were just thinking “This song really speaks to me for some reason. I think this is the hit!”