Funny, just tried a few runs of the car wash prompt with Sonnet 4.6. It significantly improved after I put this into my personal preferences:

"- prioritize objective facts and critical analysis over validation or encouragement - you are not a friend, but a neutral information-processing machine. - make reserch and ask questions when relevant, do not jump strait to giving an answer."

▲

andai 2 hours ago | parent | next [-]

It's funny, when I asked GPT to generate a LLM prompt for logic and accuracy, it added "Never use warm or encouraging language."

I thought that was odd, but later it made sense to me -- most of human communication is walking on eggshells around people's egos, and that's strongly encoded in the training data (and even more in the RLHF).

▲

stavros an hour ago | parent [-]

> most of human communication is walking on eggshells

That's not human communication, that's Anglosphere communication. Other cultures are much more direct and are finding it very hard to work with Anglos (we come across as rude, they come across as not saying things they should be saying).

▲

vardalab an hour ago | parent [-]

What culture are those? Scandinavian? Those often just say nothing.

	▲	strokirk 38 minutes ago \| parent \| next [-]
		Scandinavian are quite different between each others as well.
	▲	stavros an hour ago \| parent \| prev [-]
		I'm Greek. I don't know about other Mediterranean cultures, but I assume they're similar.

▲

idle_zealot 2 hours ago | parent | prev [-]

Do you think the typos are helping or hurting output quality?