I saw this post going viral, “Why You Should Almost Never Use AI to Write Anything Substantive.” I’ve defended the use of AI in writing here and here. My reasoning is that writing competently is hard for the vast majority of people, and although AI-generated writing is still obvious or inferior, (despite rapid progress at AI-produced math), if given a choice between no writing and clunky AI-generated writing, I think the latter, despite its flaws, is a worthwhile trade-off.
AI writing is thus analogous to false teeth. They have less bite force and can come loose, but are preferable to having no teeth. It’s a trade-off between not writing at all and producing an inferior imitation that, in certain situations, gets the job done. When I watch an Instagram reel or other short-form content, even if I know the text summary underneath is AI-generated, it can still provide useful context. Bland AI prose is preferable to no context, so it works for this purpose. Most viewers do not care either.
I can relate; I too find myself consuming AI-generated videos, ironically, because they have less filler compared to human content, with the annoying endless digressions and product placement commonly found in “creator” YouTube content. AI doesn’t digress unless told to, and has no concept of product placement. It isn’t trying to sell me a VPN (at least not yet). In this sense, the AI content is actually superior.
These jeremiads against AI-generated writing misdiagnose the problem. To emphasize, in bold, the problem is not a lack of creativity or some deficiency in the idea-generation process. People have much less difficulty generating ideas than these authors seem to assume. They imagine someone staring at the keyboard with a blank gaze and a blank mind, then having AI generate the ideas and, along with them, the essay. But coming up with ideas is not hard. The huge popularity of Twitter and other short-form text demonstrates that people have little difficulty coming up with ideas, albeit in bite-sized form.
My thesis is, AI is not being used as a substitute for generating ideas, according to how people actually use it. As I argue in the post, “AI-assisted writing is common because writing is hard,” before beginning to write, people already have some framework or outline of what they want to say. For example, before I sit down to write a post, such as this one, I have some ideas in my mind of the points I want to cover. I think this is true of most people: they know to some extent what they want to say before they begin to write.
So where does AI usage arise? To reiterate, people have little difficulty coming up with ideas. The hard part–as I have said time and time again, and these authors always fail to ever pick up on–is the organization, grammar, and structure. It’s everything else that goes into writing an essay or other long-form writing aside from the ideas themselves.
The author writes, “I think so because (1) the writing process is an essential part of the thinking process…,” and “The point of doing any kind of research is to form accurate beliefs about important questions, which you can then communicate to an audience. One of the best ways of doing that is in my opinion by writing.” But again, the problem is not writing or coming up with ideas per se; rather, it’s putting the writings or ideas into essay form, which many people really struggle with. The author yet again makes the critical misdiagnosis in conflating poor writing ability with an inability to come up with ideas: these are distinct.
This is why people have little difficulty composing “tweet storms,” but get writer’s block when confronted by the blank page staring back at them. They cannot articulate the ideas already present on Twitter or in their mind into essay-form, with the necessary transitions, organization, grammar, vocabulary, etc. All of this takes a lot of time. Even for this essay I found myself asking, “Is it spelled ‘trade-off’ or ‘tradeoff’?” With Twitter, this is not important, but readers have much higher expectations for prose compared to tweets, and small details matter. Even for a short essay there are dozens of choices that go into editing. It soon becomes tiring or tedious, and outsourcing it to an AI becomes an attractive preposition.
About how AI writing is vague or otherwise bad, as if human writing doesn’t have those same problems? Just read message-board posts from before 2022, before AI became popular, to see what passes for human writing. On medical forums, people commonly confuse “bowels” with “bowls,” for example, yet we accuse AI of producing “slop” because it does not always meet the high standards placed on it. Despite K–12 education, writing ability remains quite unequally distributed. Most people will never write well, just as most will never be competitive at sports despite enjoying watching them. Teachers loathe grading such poorly written essays. The feeling is mutual: the students would rather be doing anything else than writing them. AI makes the process more pleasant for both parties: Ideas are communicated, and information is consumed.
Finally, he says, “Writing with AI (and Not Labeling It as Such) Is Rude and Misleading.” I think this is only true if there is some agreement, e.g., a contract, not to use AI. But otherwise, I say let the reader decide if his or her time is wasted. The “back button” exists for this reason. Readers. But fear not, there will still be plenty of non-AI generated writing from people who are good at writing and enjoy doing it.