Thoughts on writing with generative AI
I’ve been attempting to use generative AI to help me write documentation because there’s so much pressure to be able to say you’re using it right now, and I honestly think the outputs are unusable without essentially rewriting the whole thing.
I find the outputs are:
- Excessively verbose (please get to the point)
- Subtly incorrect, and I’d rather just research and write correct information than spend time fact checking AI
- Not quite what I want, because the machine can’t read my mind, and I feel like prompting it correctly and then fact checking and editing the output would be more time consuming than just writing the damn thing
- Just like, tonally weird
I don’t know, it’s a weird experience. I will give it this:
- Having text to look at (and then mostly throw out) is much easier to start from than a blank page
- It’s good for extremely basic instructions (“go to X menu and click on the Y button type instructions) and those can essentially be used verbatim
- It does actually do a much better job at writing coherently than some of the people I’ve worked with (the bar for writing skills in the workplace is uh, not very high in some cases, clearly)
Overall, my take on “writing” with generative AI is that if you’re an incompetent writer, it can get you about 70% of the way to sounding somewhat competent with not much effort. If you actually know what you’re doing, you can maybe borrow basic ideas or boilerplate text from it, but actually attempting to use it might drive you insane.