Ask HN: Can being polite yield better answers from LLMs?

2 points by Marciok ↗ HN
If they were trained using questions from the internet and assuming that well-written inquiries yielded better answers, could this influence the responses of large language models when prompted?

7 comments

[ 4.4 ms ] story [ 22.3 ms ] thread
Can being polite with Sky Net save your life? :)
I gave it a list of John Connors -- and it thought I was joking.
The hypotheses are interesting. “Polite” and “well-written” are different concepts though
Correct, I expressed myself poorly, I'm basically trying to understand if I write something like:

"Please help-me solve the following problem, ..."

VS

"Solve this problem ..."

Would affect in the quality of the answer.

I'd be interested in several types of questions:

- be polite, ask nicely

- ask directly with no politeness or rudeness

- be rude and demanding

- be very disrespectful, call it names

Interesting! I'm considering setting up an experiment.