First, editors can use LLMs to suggest refinements to their own writing, as long as the edits are checked for accuracy.

translation assistance

The former I’m still looking sideways at.

The latter, probably the only truly benevolent use of LLMs.

Eh I think this sounds ok. If you prompt an AI to improve your text, you submit that, and another human reviews that (and maybe asks you to make changes) it should be fine. I can see this giving more people the ability to make edits (e.g. non-native speakers)

The problem is, it doesn’t improve text, it worsens it. And if your grasp of the language isn’t good enough, you can edit a page in your own language, or ask nerds in the discussion section to help you, it will be better written, they will be happy, and you might learn something.
Asking a slop generator to generate some slop about what you wanted to write will make things worse.

I think it’s more nuanced than that. It all depends on what you’re asking it to do (and a bit of luck that it complies as intended). Using a thesaurus can also either improve or worsen a text.

I’m not a native English speaker, but have lived in an English speaking country for many years now. I still make mistakes, but there is no point in me asking for help with English writing as my mistakes are subtle and I don’t realise I made them. Getting an AI to detect clumsy use of English and grammar mistakes has worked quite well for me before publishing reports. While I don’t always use the correct grammar while writing, I’m very capable of judging whether an LLM suggested improvement is actually better.

Of course, letting an LLM rewrite a whole text is much riskier in terms of the original meaning getting lost. But that’s not the only way to use it.