Or to put it another way, actual AI safety researchers/advocates don't really care about making 'harmless' or 'ethical' AIs in the sense that the LLM crew pratt

Or to put it another way, actual AI safety researchers/advocates don't really care about making 'harmless' or 'ethical' AIs in the sense that the LLM crew prattle about... what they care about is making 'sane' AIs that behave in ways that are intuitively sensible to average people. The censorship brigade with LLMs is something new and mostly unrelated that is doing basically surface-level nonsense to limit responses for milquetoast corporate culture reasons, because actually enforcing 'sane' behavior on an LLM is impossible with the current state of the art.