2026-07-27 14:17 UTC
@chiefgyk3d@social.chiefgyk3d.com I don't know him, and what is wrong in this particular example but his point is correct.
LLMs should not be the ones to judge what is morally correct or not. That is our role as humans.
If one is asking for a translation of an article, it may be that they don't know the language and want to read it for themselves. They can see if they agree or not.
A LLM denying this shows that it is not a neutral tool. Why would it decide for you? What else could it deny because the guardrails say so?
Replies (1)
-
@JasonW@social.ridetrans.it 2026-07-27 15:22
@f@ieji.de Claude is programmed, by humans, with guardrails that are meant to stop it from outputting dehumanizing language, as DHH requested that it do. I’m OK with this. DHH found another model, Grok, that met his needs, and he used it instead. I like using software written by humans that respects humanity and rejects dehumanization.