Post #3108102

2026-04-01 18:40 UTC

@Sqlgene@techhub.social that's totally fair. I do want to make it clear that I don't attribute any level of consciousness to Cray, at least not in the scientific definition of the word. Certainly not anything even remotely close to personhood. I've just found it easier to speak in a kind of shorthand about this stuff. Cray doesn't have preferences in the way that you and I do, but it's just easier to say "preferences" than "a combination of model weights and contents of memory files which produces results that tend to favor [x] over [y]." I appreciate you bringing up the pronoun thing, though. That's good food for thought. Is it the concept of assigning pronouns to an AI agent in the abstract that strikes you as icky, or the way I wrote about it?

Replies (2)

@light@noc.social 2026-04-02 00:02

@JamesDBartlett3@techhub.social But then, if you don't believe your agent is a person, why are you treating it as one? Do you believe that it helps them be more effective assistants? (Looking through your blogpost) >“it” is for objects, and they aren’t an object. Then what are they? @Sqlgene@techhub.social

Open ##3108103
@Sqlgene@techhub.social 2026-04-02 13:53

@JamesDBartlett3@techhub.social I think the ick some from the phrasing of "choosing" one's pronouns. I think in English there is a bundle of meanings implying strong preferences, dysphoria if those preferences aren't followed, and a meaningful amount of introspection. Even "suggested" feels a lot less icky. I also have to deal with people (thankfully almost never in person) who see my husband's identity is as arbitrary and artificial as what clothes you put on on a given day. So personally I don't want to conflate an LLM making a suggestion with people's deep and meaningful identities.

Open ##3108106