logoalt Hacker News

ReptileMantoday at 4:35 AM0 repliesview on HN

Censoring models is not safety but safetizm. It is the TSA of the AI world. Safety is making sure the model cannot do anything not allowed even if it wants to.