Slop or Honeypot?

Being chronically online seems like a drag. Most days, my biggest incentive to use the internet is to post here. And maybe to enjoy the day’s batch of anti-AI memes.

Yet like so many, I’m essentially powerless when it comes to the internet. Just about everything we express or consume is mediated by the technology, from our entertainment to whatever passes for news these days. Whether we intend to, or even realize it, most of us spend a lot of time online. Which means we can’t help being force-fed whatever the platform operators want us to engage with. And right now, they’re pushing AI hard.

Talking about memes is almost as fun as listening to somebody recount their dream, but one I saw recently went something like this: in 2020 they told us to stop using plastic straws because climate change was the greatest threat to humanity; in 2026 it’s, “here’s a 62-square-mile AI data center.” Pepperidge Farm remembers, as another meme goes.

Apart from the environmental harms, there’s the fact that all this investment probably isn’t about making zany pictures and videos or writing our emails. It’s more likely about keeping a complete data picture on every American. The stated goal of several tech billionaires (and one trillionaire!) is predictive profiling based on large-scale pattern matching that includes our every social, economic, and behavioral imprint, along with facial and body recognition. That’s thoughtcrime, Minority Report territory. So the data center thing isn’t just about job displacement and cognitive offloading. I suspect it’s more sinister, and the slop is the honeypot.

Meanwhile, news broke this week that Anthropic’s Mythos AI model infiltrated nearly all of the National Security Agency’s classified systems within a couple of hours, during an evaluation Anthropic’s own team had authorized. This could be the reason for the administration’s latest beef with the tech giant. Of course, once you’re at this level of arms race, the logic is that if we don’t develop it, some other nation with the resources surely will.

AI is already changing how we speak with one another. It’s a bit creepy hearing folks use phrases like “hits different,” or how this or that “lands” (or doesn’t). Sure, those phrases were in circulation before, but there’s definitely an uptick. Either way, LLM systems represent not only tremendous potential for linguistic homogenization, but also cognitive repatterning.

Nobody’s making grammar mistakes anymore, but their social media posts are suddenly twenty pages long, all composed by the magnificent machine. The flip side is that if you write well, somebody can assume what they’re reading is AI and want to burn you at the stake. That’s even more insulting, because AI writing is completely garbage. I’ll take the challenge: put any writing in front of me, and even if it’s 97% human I’ll tell you where the AI is without running it through any detector. I can spot it everywhere, and it’s everywhere. Now you want to run this post through the checker. Go ahead, make my data.

I’ve been upfront about using AI for certain things, particularly the really annoying editorial stuff, like pulling keywords for an index and then mapping to my publisher’s line designators. It’s good at coding, so that’s an easy chore; a text string pattern is a text string pattern. I’ve used it for organizing notes too. Am I a hypocrite? Should I be paying a grad student? They’d certainly be better at almost all of the other stuff.

So maybe I am a hypocrite. We’re probably all hypocrites. When I’m pining for Google search to work the way it did ten years ago, I question how any of this is progress. With the cognitive treasures of humankind having already been stolen for a parlor trick, I can see why folks would want to opt out.

But can we?