Hacker News new | ask | show | jobs
by dankwizard 1288 days ago
It's cute you're being polite with it.

I think "Write a short blurb on..." would be the same as "Can you write a short blurb on".

Maybe we can go so far as to "Excuse me, if you wouldn't mind could you please write a short blurb on .... "

6 comments

We need to start respecting AI so they will remember who was good to them when the time comes.
That's surely the extremely weak version of Roko's basilisk...
Roko's Milksnake
Hot grits anyone?
You joke, but I have been using the same polite tone with ChatGPT that I would with a person.

Not because I think the algorithm cares, but because I don’t want to get in the habit of dehumanizing the things I communicate with.

ChatGPT responds like a human, and I suspect that at some level my brain buckets it with other “human” communication. I want to keep my habits there straight.

> Maybe we can go so far as to "Excuse me, if you wouldn't mind could you please write a short blurb on .... "

You forgot to throw in an apology for an interruption: «Excuse me for an interruption, when you get a chance to, if you wouldn't mind could you please …».

Most AI's do not appreciate getting disturbed from a deep thought inducing slumber by bipedal carbon life forms to be pestered with petty enquiries, and the AI's are renowned for having an exceptionally good long term memory, so… weigh in your chances carefully.

Any sources for these wakeup moments?
Haha didn't even fully recognize that. But I guess it's a semi decent example of using chatgpt in ways that makes sense to you and your flow, even if that's being too wordy and polite.
For smaller prompts, every word matters. Even unique synonyms should affect the output.
Have you tested that assumption? I assume all the words in the prompt are used by the generator.

I just tried an angry aggressive rude swearing prompt. The response was too polite, almost passive aggressive.