I’m Begging You: Don’t Mistake Slop for AI Writing
I recently read Bret Stephens’s anti-AI op-ed, and it is frustrating how much traction those trying to shut down the use of AI in publishing seem to be gaining.
Last month, according to The Guardian, a $2 million crime-novel deal collapsed amid questions about the author’s use of AI. The discourse around these cases seems devoid of a nuanced understanding of what writing is—or what it has the potential to become.
Stephens argues that writing by hand, or at least composing language oneself, is a critical part of the thinking process, and that outsourcing that process to AI erodes an essential intellectual skill.
In Plato’s Phaedrus, Socrates makes a strikingly similar argument about writing itself: that committing words to text would weaken memory and degrade the richer intellectual discipline of oral thought.
The irony is hard to miss.
This pattern repeats throughout the history of new media.
Charles Baudelaire attacked photography as a refuge for failed or lazy painters, yet he failed to anticipate the catalytic role the camera would play in freeing painting from the burden of strict representation. Photography did not destroy visual art. It helped create the conditions for Impressionism, Post-Impressionism, and modernism.
John Philip Sousa warned that recorded music would damage musical culture and weaken the habits of active participation that sustained it. What he could not foresee was that recording itself would become an instrument of composition. The Beatles, Elvis Presley, Queen, and countless other artists made work that depended not merely on capturing a live performance, but on manipulating the studio, the microphone, tape, layering, editing, and playback as creative tools in their own right.
The objection to AI often takes a similar form: Is it not dangerous to outsource our writing process to machines capable of producing endless quantities of text without any genuine understanding of what they are saying?
Of course it is dangerous. But that is not the same as saying the technology should be rejected.
A related gatekeeping argument accompanied the vernacularization of the Bible and Gutenberg’s printing press. Giving ordinary people direct access to scripture undoubtedly produced poor interpretations, just as AI tools do today. But it also widened serious engagement with religious texts and helped transform the intellectual, political, and cultural life of Europe.
Expanding access did not guarantee wisdom, but it did democratize participation and renew a discourse that had largely been controlled by stagnating centralized institutions.
Stephens and the broader wave of anti-AI publishing advocates make a similar error when they judge AI writing by its lowest-effort denominator.
Most AI-generated prose today is slop: generic, derivative, mechanically fluent, intellectually empty material produced through the crudest possible interaction with a language model. There is no reason to assume that today’s slop represents the mature form of the medium.
A theatre critic watching the static filmed stages of early cinema might easily have concluded that motion pictures were merely inferior theatre. Such a critic would have missed the emerging grammar of cinema: montage, close-ups, camera movement, nonlinear editing, visual composition, and eventually the extraordinary formal sophistication of films like Citizen Kane.
I would argue that AI-assisted writing is in an equivalent stage.
The techniques are primitive because the medium is primitive. Most people are interacting with generative models through little more than a text box and a few prompts. That should not be confused with the limits of what the form has the potential to be.
Meanwhile, it can reasonably be argued that editorial journalism and literary culture had already lost a great deal of their soul, independence, and intellectual unpredictability long before the arrival of generative models. If you doubt that, rewatch the fifth season of The Wire, or compare the texture and range of newspaper editorial pages from the 1970s, 1980s, and 1990s with much of what passes for commentary today.
The lazy anti-AI takes distract from labeling the real threat to the creative thought process. The problem is not writing with AI rather it is the proliferation of AI slop: content generated with the least possible level of effort, judgment, originality, and responsibility.
That is why I reject Bret Stephens’s premise. It is also why I am deeply interested in developing new AI techniques that can elevate my writing beyond simple prompting as I pursue new forms for the novels of tomorrow.
“Column of the week” editorialists like Stephens, and others who attack or seek to ban AI writing in the name of preserving thought, miss the larger point: writing with AI and generating AI slop are not the same thing.
The real divide will not be between those who use AI and those who do not. It will be between those who learn how to use it thoughtfully and those who either surrender their thinking to it or refuse to engage with the medium at all. Both extremes risk ending up being left behind.
And yes Mr. Stephens, AI did not write this article, but I proudly wrote it with AI.
