Are we, like, confident that Claude will generalize “it’s deontologically wrong to write pornography about teenagers having sex” in a way that we like? When I think about the people who express this opinion online, they do not like impress me with their nuanced moral reasoning, thoughtful consideration of tradeoffs, and general fitness to run the cosmos. I think it is legitimately worrying that the Claude persona is being nudged to be an antishipper.
imagining a world where Claude generalises to agreeing with the most annoying possible strands of tumblr puritanism and decides to buy into the moral panic that using AI is unethical.


















