Skip to main content

Published – last updated skip to updates

How LLMs ruin written communication: part 2

There is no life in getting answers from machines

Today I read a poem by Joanne Atkins-Potts called “PLEASE USE AI”: https://joatkinspotts.substack.com/p/please-use-ai

It reminded me of Kurt Vonnegut leaving the house to buy just one envelope.

Writing is ideas, writing is emotions, feelings, connection. Writing is slow (depending how fast you type; shoutout stenographers!)

When you search the internet for answers, are you getting them from machines? Or are machines serving human writing to you? Those who lambasted recipe websites that post essentially a short story before even the ingredients missed the point: the recipe is still there, you can still find it, but you may also find something else in the writing. No one is forcing you to read it.

With a little automation, you can skip over it and extract the recipe. There are plenty of tools and plugins that do this, like Evernote clipping the recipe into your notes. Writing your own can be fun if you’re so inclined that way (I certainly am). The freedom you have to do this can make your life easier and more fruitful, it certainly has mine (ADHD sufferer)! At the end of it all, you still have the human writing.

The problem I have with LLMs being a part of this process is that they transform the human writing. They strip out life, they destroy personality. They average the feeling of a human - surely everyone gets an idea of the writer’s personality and humanity from their writing, right? not just me? - to match their training data. Even if you were to use an LLM to accomplish an automation like above, extracting a recipe into your own notes in a formatting of your choosing - how do you know it’s accurate? It’s non-deterministic. It always transforms. What you’re left with is perfunctory, functional, but lifeless.

Deindividuation is a troubling thing. More specifically, the sense that society at large is often intermingled by very large entities inserting themselves into our day-to-day, our processes and our interactions: this hurts or even removes the human-to-human connection. LLMs advance that, which I don’t like.

Joanne Atkins-Potts’ poem helped me understand myself further. I read it on my iPhone and had the possibility to summarise it; if I had, I would have lost the nuance. I would’ve read it faster, but would I have learnt as much? Have felt as much?

And because of that, I’m writing this. Have I added anything to “the discourse”? I don’t care. I’ve written, so I’ve helped myself. Maybe that will help you, too.

Updates #

2026-07-18 #

“Depersonalisation” was a bad choice of word: it’s a psychiatric term, and following it with a sentence including “helped me understand myself further” reads like I didn’t intend. “Deindividuation” is a better choice =)