On using AI as a writing assistant
3 points by nishnash
3 points by nishnash
Some people are using AI for writing assistance, but not everyone is using it the same way. If you were always bad at something, I don't think AI makes you better at it; in fact, it may even amplify your inadequacies, and anything you produce will be just as dysfunctional as without AI. This article is about sharing a process, and I am wondering what process others use. Maybe AI anything is just enshittfying everything it touches? You can express those views as well (this is your safe space).
I really like Thomas Ptacek's advice on this:
Rule Number One: You may not use a single word an LLM suggests to you.
I use LLMs to check spelling and grammar and do some light fact checking (they're surprisingly effective at this), but I won't let them put words in my mouth.
I think it’s important to make sure you don’t plagiarize or use someone else’s IP. There are other legal and ethical concerns as well, but unrelated to information or knowledge.
The question, maybe on some philosophical level, is why should I bias myself against information synthesized by a mathematical process versus one done so by a person?
If a person misremembers something and misleads me versus an LLM experiencing a context shift, why does the person remain more reliable and trustworthy? (I know one way to answer this, but I am also interested in other people’s opinions).
The more advanced models have self-correction capabilities now, and they have the ability to scour vast stores of information for synthesis. As someone who values learning, and struggles with memory retention due to an illness, I can’t help but be seduced by LLM as second brain possibilities.
I don’t think anyone should consume or repeat LLM content wholesale, it’s just not that good or compelling. But I find it hard to justify not using it at all.
The value of writing, to me, is that the author has spent time and energy organizing their thoughts about something. That's valuable to me because it saves me from having to do the same thinking, or it exposes me to a different way of thinking than I'm capable of on my own.
I think AI can provide useful feedback about your writing, but if you're just sending random thoughts to a Telegram bot and having AI agents assemble it into a blog post, you're skipping the part of writing that creates value. If you're not organizing your thoughts before you publish, then to me, it's more like a series of tweets that's been disguised as a blog post. I don't need to read the result of someone else's AI prompts because I could ask AI myself and get an answer tailored to me.
You mentioned finding it hard to get feedback about your writing, and I think that's true. When you're starting out, it's hard to get anyone to read your writing at all, much less give you feedback. And even when you get readers, it's rare for people to give you thoughtful feedback about how you can improve your writing.
Have you considered hiring an editor? I did that a year into starting my blog and found it helpful.
From reading your non-AI assisted sample, I have two suggestions:
Thanks, I appreciate the comments and I’ve made a note of your input.
I am very much interested in researching, reviewing and editing anything I send out myself. I get help with grammar or phrasing for a paragraph or section of which I already know the intent and content.
I do worry about plagiarism and encroaching on someone else’s IP, so I review as much as possible.
My hope in sharing this blog piece is that maybe others better at writing with AI than I can share their process of doing it. Famous writers, or richer ones at least, can afford ghost writers or writing assistants. If I could afford to hire someone, I wouldn’t be using AI.