Thank you. In this comment I posted my raw imperfect writing, but otherwise I often procede just like you did. However the questions discussed in the FAQ arise often concerning more developed writings. It is tempting to let the LLM do more work, and it would be stupid to reject assistance just because it comes from a machine. It would be like a lumberjack refusing a chainsaw. But I think that JusticeMills is utterly right. We must be very careful and accept only a very limited amount of assistance.
broken english, sloppy grammar, but clear outline and readability (using headers well, not writing in a single paragraph (and avoiding unnecessarily deep nesting (both of which I’m terrible at and don’t want to improve on for casual commenting (though in this comment I’m exaggerating it for funsies)))) in otherwise highly intellectually competent writing which makes clear and well-aimed points, has become, to my eye, an unambiguous shining green flag. I can’t speak for anyone else.
For what it’s worth, I think that Justis hits the nail on the head with “I think probably under current conditions, broken English is less of a red flag for people than LLM-ese.” In such a global language as English, people naturally give slack. (Also, non-native speakers are kind of in an adversarial situation with LLM-ese, since it’s harder to detect when you aren’t as immersed in standard American/British English.)
Concrete example: my parents, whose English is fairly weak, always say that one of the nice things about America is that people are linguistically generous. They illustrate it like this: “In our country, if people can’t understand you, they think it’s your fault. In America, they think it’s theirs.” I think the same is true of the internet, especially somewhere like LessWrong.
On a practical note, I think spellcheckers like those in Docs and Word are sufficient for these contexts. In academic writing or whatever, when standard English serves more of a signaling function, it’s trickier.
Thank you. In this comment I posted my raw imperfect writing, but otherwise I often procede just like you did. However the questions discussed in the FAQ arise often concerning more developed writings. It is tempting to let the LLM do more work, and it would be stupid to reject assistance just because it comes from a machine. It would be like a lumberjack refusing a chainsaw. But I think that JusticeMills is utterly right. We must be very careful and accept only a very limited amount of assistance.
broken english, sloppy grammar, but clear outline and readability (using headers well, not writing in a single paragraph (and avoiding unnecessarily deep nesting (both of which I’m terrible at and don’t want to improve on for casual commenting (though in this comment I’m exaggerating it for funsies)))) in otherwise highly intellectually competent writing which makes clear and well-aimed points, has become, to my eye, an unambiguous shining green flag. I can’t speak for anyone else.
For what it’s worth, I think that Justis hits the nail on the head with “I think probably under current conditions, broken English is less of a red flag for people than LLM-ese.” In such a global language as English, people naturally give slack. (Also, non-native speakers are kind of in an adversarial situation with LLM-ese, since it’s harder to detect when you aren’t as immersed in standard American/British English.)
Concrete example: my parents, whose English is fairly weak, always say that one of the nice things about America is that people are linguistically generous. They illustrate it like this: “In our country, if people can’t understand you, they think it’s your fault. In America, they think it’s theirs.” I think the same is true of the internet, especially somewhere like LessWrong.
On a practical note, I think spellcheckers like those in Docs and Word are sufficient for these contexts. In academic writing or whatever, when standard English serves more of a signaling function, it’s trickier.