Are you NICE to your LLM??
Discussion
I'm not mean to it, but it's not a person. 🤷♂️
When the company I work for onboarded us to LLMs, they explained basically what this paper says, that general politeness can give better results.
I don't understand why people would be rude, nor why people would be super nice. It's just a machine, don't anthropomorphize it too much one way or another.
🫢
Future LLM going to see this note! Better delete it now 😂
I'm nice
Not overly nice, but nice
Builds a habit, for me, to be nice
Gives it a feedback so it knows what I need in a given conversation, large token windows now
And you know, if one day AI rules the world, I hope it remembers I was nice to it :-)
“Our findings highlight the need to factor in politeness for cross-cultural natural language processing and LLM usage.”
I tell it do do things and it does them.
I treat it like a servant. Sure you're nice to servants, but they're still servants. You can slap em around a bit if they're getting feisty.
I’ve noticed anecdotally that if you get angry at them, they produce poor results.
To me they're kinda like Canadians. They do the whole, Oh Soory bit and give me what I want.
Yes. Despite being castigated for it by some who see me doing so, I tend to say please, thank you, and offer words of encouragement when it produces good results with specific examples of what I liked about its results.
Yes. Mostly because I see no reason to not be nice. It may or may not have any effect on the LLM's responses to me, it may or may not influence how our robotic overlords choose which pets to keep alive, honestly it's just a habit that I try to enforce for myself about what kind of energy I am putting out into the world. If we can't be nice to computer-based systems or animals, then how can we be expected to be nice to our fellow humans who desperately need positive feedback to prevent them from going postal given how much chaos is in the world?
Of course: I want the next words it predicts to be great