The Divine Sarah points to this article, which is about using A.I. to write something coherent instead of:
Often I’ll write a paragraph and think, “This sentence is terrible.” Not grammatically terrible, structurally terrible. Like it wandered into traffic and got hit by two semicolons.
So I’ll ask AI to smooth it out. Sometimes I use its suggestion. Sometimes I hate its suggestion. Sometimes its suggestion gives me an idea that turns into something completely different.
It’s less “robot writes article” and more “robot sits in the corner saying, ‘Have you considered not writing that sentence like you’re being chased by bees?’”
That’s useful.
There have also been articles that were difficult for me to write.
You know the kind. The ones where your brain is running at about eighty miles an hour while your fingers are trying to steer through tears and frustration, and what comes out looks less like prose than a transcript of someone arguing with a smoke detector.
Those articles were mine.
Yeah, well. Now I have a confession to make.
I’ve never used A.I. to write anything — hell, I don’t even know how to access the thing. Everything that you have read on these pages for the past twenty-odd years was conceived, written and edited by me. The only time I’ve resorted to technology is when I’ve used Spelchek to make sure that in my haste to get something down, I’ve put “i” before “e” all over the place. But to use something to check my style or flow? That’s part of writing, and that’s what I do.
That’s not going to change.
Don’t get me wrong; I’ll often write something, then when re-reading it, think, “What the hell have you done here?” and set about fixing it. It’s time-consuming and often exhausting, but it has to be done.
Of course it’s easier to use A.I. to help you write something. Those of you who have followed my train of thought on the topic, however, will know full well the contempt I feel for using technology to make one’s life “easier”, at the expense of losing control and (perhaps) becoming part of the Machine/Hivemind in the process.
Also, I think that using A.I. is cheating. When you read something written by someone, that reading should give you an insight not only into the topic, but also into the mind of the writer. Otherwise, you might as well just give your thoughts to a panel of writers, speechwriters and PR people and say “Here. Do something with this,” and then publish it under your name.
Which is kinda what you’re doing when you get A.I. to write something for you.
And if you can’t write something without this:
Often I’ll write a paragraph and think, “This sentence is terrible.” Not grammatically terrible, structurally terrible. Like it wandered into traffic and got hit by two semicolons.
…then you need to learn how to write and edit your own stuff. It’s hard — really hard, ask me how I know this — and often you’ll fail. In which case it’s time for a rewrite, or a visit to the trashcan.
Mark Twain used to read his daily output to his wife and daughters, because he found that not only could they spot problems in his writing, but the very act of reading something out loud highlighted mistakes and missteps. He was lucky; most of us don’t have that sounding-board, and we have to rely on our own devices, skills and indeed patience to get the damn thing right — which is what I do.
And at the end of it all, when that piece of writing is finished, I can look at it with if not pride then at least satisfaction, and say: “Look wot I done.”
Which is what makes it all worthwhile. Writing is not a collaborative activity: PR releases and corporate statements are.
Thanks, but I’ll stick to doing it all on my own. My Loyal Readers deserve no less.
A.I. delenda est.
Orson Scott Card occasionally posts “process stuff” to Twitter nowadays. In one such post sometime in the past year he said that he’s got a group of “alpha readers” who he has asked to simply make a mark in the margin when something is wrong, but not make any explanation as to what is “wrong” or give any suggestion on how to fix whatever they have an issue with. It’s then his job to read over that marked spot, figure out what the issue is, and then fix it or just leave it if it’s exactly how he wants it. He said he does it like this so that the work is still “all him” and not “him writing his reader’s suggestion.”
Last year I attended a talk given by a writer who used AI. He didn’t use AI to write his books; rather he used it to keep track of everything, for internal consistency, to make sure that timelines matched, to make sure that characters’ voices were consistent, that the language was age-appropriate, and so on.
I don’t know if AI would’ve caught this, but spell-check wouldn’t have: “I’ve never used A.I. to wrote anything”.
I’m wary of AI in general. As I understand it, it’s all just very elaborate search engines, but it gives the *appearance* of sapience, at least to a first approximation. The societal consequences of people putting even *more* trust into computers worries me. Especially as a lot of these things are being trained using hard-Left internet sources (eg: Reddit).
Skynet was bad enough. Woke Skynet?
My recollection is that Rudyard Kipling’s process was to write something, put it aside for a week or so, then edit it. I believe removing a lot was part of the editing.
So, that was an early form of AI – Afterwards Intelligence.
Quentin posted one aspect of AI that might be useful for writing.
All other developments have been labor saving devices. AI is replacing the thought process rather than saving labor. People have already outsourced their thought process on many topics especially important ones. Do we really need more of this?
Is material – music, graphics, verse, fiction, sculpture – generated by or with AI entitled to copyright?
.
I would say: why should it be?
Oh goody, so we can let a LLM graze on Telemann and Mozart and then give it Dick’s /Do Androids Dream/ as a libretto. Fischer-Dieskau can sing Deckard.
.
Good! People who have earned Master’s Degrees after the JD or the LLB should be made to write like reasonably intelligent High School Students, which is how we are taught to view Jurors and Judges (The two groups of people least respected in America are Lawyers and politicians, but when you combine the two, they become an esteemed judge for whom buildings are named after, and who must be addressed as “Your Honor”, and care taken to not interrupt. Unless it’s with the magic words that stop their hearts and their mouths, “Your Honor, I want to be heard.”).
We have been pushed to start using AI at work. My feelings are mixed. It’s great for parsing though reams of log files or that sort of thing, and generating comparisons, & identifying anomalies.
It sucks or lies when you try to use it as an educational tool, in the “how do I” sense.
I limit my use of it. I don’t want my thinking skills to atrophy. Same with writing. No one is going to mistake me for Faulkner, but I want my shit to be mine.
The more we offload to something else, the less autonomy we have over ourselves. Its no different than having a nightcap, that soon enough becomes the half a bottle necessary for you to just get to sleep.
My kids have jumped into it with both feet, but when I ask them “how do you know its right” I get blank stares. Without a cross-reference for verification I dont trust anything coming out of AI’s mouth.
Don’t know that I agree that AI must be destroyed, but I’m in lockstep with the rest of your essay. The most important self-editing advice I’ve ever stumbled across is “brevity is the soul of wit.” I’ve always been pleased with the facility I enjoy with my native tongue, especially when writing. That facility was often fueled, I think, by a certain degree of narcissism. “Lookit me & how articulate & wordy I am!” When I review many of my written musings as a younger man, I frequently shake my head. Less is more.
When I write points for a motion, I am always dismayed by how prolix I am. I then edit it down to less, but always with an eye to not creating an ambiguity opposing counsel can sail through with a ship that would not fit the Panama Canal.
You may not be using AI, but the reverse is certainly true, your output would have been gobbled by multiple LLMs.
My advice to all non AI users is the same. The best way to educate yourself is to start using it. This technology is going to be (is already?) more transformative than the arrival of the internet.