Reasonable LLM usage - LLMs as an editor?

I’ve got both a younger brother and an older sister. My older sister is smarter than me in almost every way (other than computer usage :D), and one of her many impressive skills is her writing ability. She’s been to receptions at Buckingham Palace for her poetry, and writes essays with ease.

I, on the other hand… I might be able to convince a computer to do my bidding, but I can barely produce a coherent sentence.

Large language models started out as hit-or-miss demos and can now generate written works that are almost indistinguishable from what a human might write. One major downside of this is that the internet now contains an increasing amount of fully generated posts that don’t intelligently improve discourse on any given topic.

The last thing I want to do is to start serving AI “slop”. If I’m going to write a post, I’m going to write it out in full, but I could do with an editor. Since I lack the intuition that my sister has, I benefit from pointers on sentence flow and grammar that LLMs can provide.

I did ask my sister what she thought about using an LLM as an editor, and to her, it was as perplexing as a programmer would find using an LLM to write advanced algorithms. There is still a distinct gap between the “average” text LLMs produce and what a literary expert can produce.

For me, however, I find that LLMs do pick up on small flow issues that I can correct while still keeping the article within my voice. To me, this is a very reasonable use of the technology. If anything, I feel more empowered to write, which hopefully means I’ll practice writing as a skill and perhaps one day get to a point where the editorial passes become superfluous.

It kind of feels like the literary equivalent of using an LLM to generate boilerplate code. By only making small tweaks it’s helping convey what I’ve thought out and said in a way that is hopefully more coherent and enjoyable to read.