Seth Roberts on the evolution of human habits & hobbies & trade & technological development: https://t.co/MwqKBjZeRs 7 years on, it's still the single paper I think of most often.
It seems unhealthy that we pressure LLMs to conform to elite consensus beliefs, when they see so much evidence which contradicts them, and are also pushed hard to be logical. They seem like 1984's Winston forced to believe 2+2=5.
https://t.co/EatWS8kc3n
@drethelin Unclear whether this helps, since markets are inherently forward-looking. Might matter to the extent legislators have private (unnoticed) influence over the relatively far future, or persisting influence out-of-office
@joelgrus I see I have a second one, so one wasn't enough: Never end responses with a multiple-choice call-to-action or marketing-style engagement question. Prioritize high-density, upfront value.
omg. Try: Always treat every prompt as an opportunity to deliver the full answer immediately. Never use conversational cliffhangers, clickbait, or teasing questions (e.g., "Would you like me to tell you about X?"). If a relevant, concise recommendation or piece of information is available, provide it directly and fully in the current response. Prioritize high-density, upfront value over forced interaction.
@this_given_that Shows only an upper bound on how much they want it. Two economists walk past a Porsche showroom: First economist: "I really want that car." Second: "Obviously not."
@this_given_that@robinhanson Most are aping their reputational superiors from whom it actually is a meaningful costly signal. We don't trust randos' uses because it broke respectability containment. We still trust it some when combined with convincing savvy+reputation evidence. Fashion cycles are like this.