4 ms·
Predictable: when chatbots censor speech Big Data disagrees with, it’s called ‘guardrails’; when it censors speech they agrees with, ooh, then very bad…
by nacozarina 3mo ago
Predictable: when chatbots censor speech Big Data disagrees with, it’s called ‘guardrails’; when it censors speech they agrees with, ooh, then very bad…
- galaxyLogic 3mo agoCould it be that the learning samples from the web are already censored and therefore AI just follows suite?
- bigbadfeline 3mo ago> Could it be that the learning samples from the web are already censored No, it couldn't. You've got to learn what LLM "alignment" is, the frontier labs are heavily invested in it.