Stop Sending Me Slop
in which I take offense in defense of writing
My daughter’s college sends me a regular email with campus news: who’s received what prizes, what building was reopened after refurbishing, what wisdom was dispensed by this year’s commencement speaker.
I don’t usually read these emails. There’s something about them that repels me, and I realized just recently what it was.
Slop.
Intermingled with actual campus news are filler articles like “16 Life Skills Your Student Should Know” and “5 Ways Parents and Families Can Stay Supportive Through the College Years.” These articles are not written by college staff but manufactured by a company called CampusESP, which provides “AI-powered parent engagement” for institutions of higher education.1 Engaging with parents apparently means sending them regular helpings of content by an AI Content Agent. As the website explains to potential clients, “the CampusESP Content Network creates, curates, and prepares your family content for you.”
Did you just read the word “content” so many times you’re no longer sure what it is? Here, I’ll help. Something is called “content” when it was created not for its own sake (like a letter, a poem, or an angry essay) but to fill some kind of container. A business website. A social media feed. Emails addressed to me from a small, elite, liberal arts college.
To me!
While college signifies a move towards adulthood, your student still cherishes having you as their anchor. ... Embrace this journey alongside them, celebrating their strides towards independence, while always standing ready to guide and support them along the way.
I think people have forgotten that email is a kind of mail and that an email address is a kind of address, which is attached to a human being whose time and attention they are claiming when they click “send.”
“Most teams are stuck creating family content from scratch,” the CampusESP platform pitches. “Every email, every deadline, every reminder.” Oof. That does sound tedious. It sounds like a perfect entry-level job for, say, a recent graduate from a small, elite, liberal arts college.
I’m getting angrier as I write. I meant for this essay to provide a calm and rational elucidation of the practical and moral limits we should set on using generative AI.2 Each of us needs to decide our limits — and quickly. ChatGPT was released to the public in November 2022. In less than four years, we went from “Hey look, I made this machine write a limerick” to “Am I talking to a human?” and “Is my job at risk?”
In the old days — say, three years ago — saying “no” to generative AI was easy. It was an opt-in system, so all you had to do was ... not open ChatGPT.
Now it offers to finish sentences in my emails, jumps in and answers first if I ask Google a question, and pretends to be an employee of the online company where I shop. It’s intrusive and omnipresent; it fawns over me (“great question!”) while stonewalling me, spying on me, and giving me unreliable information.3
What it does not do is write. Let’s be clear about that.
Generative AI does not write because it does not think. When ChatGPT or Claude strings words together, it isn’t choosing words. It’s using a large language model to predict the most likely next word, based on the vast looted treasure of human-made text on which it was trained. If that still sounds like thinking, it might help to remember that it doesn’t understand words. It recognizes units of data called tokens, each of which is assigned a numerical ID.4 It’s predicting numbers.
So large is the corpus on which it is trained, so quickly does it work, so cleverly is it designed to adjust its output, and so generously do we feed it more data with each “free” use, that it is a truly incredible illusion, the greatest fake imaginable. Bravo, everyone!
But it is fake. Text extruded by LLMs is only posing as thoughts communicated by a human being. If you use it in your writing to skip the hard work of finding examples or structuring your argument, you are short-circuiting the connection-making, serendipity-producing, bulb-flashing process that makes writing “the most astonishing of all human technologies,” as the poet Dan Chiasson puts it.5 If you use it as your writing you are lying.
Oh, wait, does it save you time? Do you have a container to fill?
I’m sorry, I’m not mad at you.
I’m going to assume that if you subscribe to Scratch — or if you’ve made it this far into this essay — you do not worship the capitalist gods of convenience and cost savings, no matter how relentlessly you are proselytized. That CampusESP’s boast about how little time its clients spend on “content management”6 makes you wince. But I worry that even those of us who value what is painstakingly made by human hands and minds have succumbed to a kind of fatalism, accepting as inevitable AI’s takeover of modern life.
If we could go back in time and put the brakes on plastics, would we? The plastic revolution began with a booming optimism: the notion that, between the laboratory and the factory, we could preserve scarce natural resources used in material goods (silk in hosiery, tortoiseshell in combs) while expanding the buying power of consumers.
Generative AI promises a similar magic leap, putting previously resource-intensive creative intellectual goods within easy reach of everyone. Just issue a prompt and voilà: an academic paper, an article, or a children’s book pops out. No training, no collaborating, no learning, and certainly no hiring required. We are giddy with our new powers, like 1950s housewives drunk on disposable dishware.
Just as irresistibly cheap single-use plastics now clog our oceans (and microplastics lurk in our bloodstreams), so too will simulacra of human creativity clog our inboxes, our internet, and our brains. Amazon already teems with AI-generated “books” and well over two-thirds of social media posts are AI-generated.
And as cheaply as we value material goods in the age of plastic, cultural goods will fare worse. A plastic comb may not be made of tortoiseshell, but it still combs. A plastic English essay, on the other hand, though it includes sentences, paragraphs, and a title, no longer essays. It’s no longer a place where a student wrestles thoughts, evidence, logic, and syntax into ideas expressed to a reader. It’s just content.7
This hollowing out of words, this emptying writing of effort, may feel like freedom. But we will pay a hefty price for it. Since these fakes do not in fact do what art and writing do — do not genuinely engage a human mind at either end of the communication — the more of it we consume, the more we will doubt the value of the thing it is pretending to be. We won’t just want to pay less for it; we won’t even understand what it’s for.
I want to be smart about this. I want to recognize that all AI isn’t generative AI, that machine learning can be a force for good — just as plastics are essential in modern health care — and even that some consumer uses of AI can free me of tedious, noncreative tasks without cheating or misleading anyone else.8 But when it comes to generative AI, I instinctively recoil from a product designed to trick me, devalue my work, and clog my culture with trash.
Recoiling is not enough, though. It’s not enough to boo commencement speakers who tout the wonders of AI to graduating seniors whose education was AI-cheapened and whose futures are AI-threatened.9
We should be asking, at every opportunity, who this faster-cheaper-more is for. We should be imposing moratoria on data centers, passing regulations, filing lawsuits. And at the very least we should take offense when an institution supposedly committed to high cultural standards10 sends us “content.”
Rather than using generic copy extruded by a machine, I wrote to my daughter’s college, might I suggest they hire a person to write articles? “Maybe an English major who hopes to enter a world where specific humans do the thinking and writing?”
Parent, engaged.
p.s. I meant for this essay to provide a calm and rational elucidation of the practical and moral limits of ... etc. etc. That’s what I sat down to write, and that’s what you would have gotten if I had “prompted” ChatGPT to produce it. Writing is funny that way.
I know this because they are clearly labeled as such. No one was trying to trick me.
There are many kinds of what we commonly call “artificial intelligence.” Generative AI produces things that seem human-made (text, sound, images). Large language models is one subset of generative AI; an LLM is what you chat with on ChatGPT, Gemini, or Claude.
As I read over that paragraph, the solution is obvious. Just dispense with the portals through which AI reaches me: my phone and my computer. It may come to that. But these little bundles of plastic and precious metals is how I reach you, too; I’m not ready to abandon them.
Chaisson’s “Think for Yourself” in the New York Review of Books is terrific, as is, in a completely different key, Garrett Buck’s “You Will Never Win at AI” in The White Pages.
“Penn State reaches over 100,000 parents each week with about one hour of staff time.”
I’m well aware that writing-as-content predates the invention of LLMs. I’ve been asked to produce my share of “boilerplate” and I know exactly what John Warner means when he says that most students are taught “to write Potemkin essays, fakes designed to pass surface-level muster that are revealed as hollow facades when inspected more closely” (Why They Can’t Write). Humans can make slop too.
Then there’s the matter of energy and water use, the enriching of already rich men, and the consolidating of capital and power in the hands of a few corporations. Every honest moral reckoning may well conclude that making my life slightly easier makes something else worse.
Just to be clear: I’m pro-booing. University of Arizona commencement, Game 3 of the NBA Finals, what have you.
I wonder how many of the colleges sending out CampusESP content also have strict policies around their students’ use of AI. Honor code violation for thee, smart business practice for me!
I love the analogy to plastics. The AI gurus seem to *want* us to swim in the Great Garbage Patch.
The temptation to use generative AI is powerful, especially to us who find writing a struggle. But trying to write helps me think. And thinking is so necessary. As you say, “…Writing is funny that way.” Yes! Thank you for a great essay.