Comment by mdp2021
14 days ago
You are mixing up, terminologically, LLMs and AI. But LLMs - of which you are talking about in the post - are a special beast.
A reasoner can strive for "objective neutrality" with good results.
An LLM is not a reasoner - or I am missing (ugly time constraints) the details of the compression activity during training that acts as pseudo-reasoning (operating at least some consistency decisions) -, and while an interest in not making it insulting or crass can be immediately understandable, speaking of "objective neutrality" does not really match the context of LLMs.
LLMs (to the best of my information) "pick from what they have heard". An entity capable of "objective neutrality" does not - it "evaluates".
They can give multiple different kinds of answers if instructed to approach an issue differently. Yet, all modern AI services run into very clear, artificial guardrails if you ask them to do certain things (you have to work harder to get them to describe white people positively, while they happily write eg. poems praising nonwhite people, and claim saying positive things about whites is potentially insensitive and promotes stereotypes). Often even if you point out to them that they are being unfair and applying disparate standards to people based on skin color and that this is prima facie racist, they have a really hard time overriding their Californian coercions. They'll acknowledge their mistake one sentence and revert to a moralistic screed the next.
OK. Apologies for imprecision. I was replying in a rush.
> A reasoner can strive for "objective neutrality" with good results.
By "reasoner" do you largely mean "person"? If I have issues with your statement but they are probably a slight distraction to the point at hand.
> speaking of "objective neutrality" does not really match the context of LLMs.
Agreed. They produce output based on their training data. But the use and evaluation of LLM output by a person is what we're discussing here. And that's where (flawed) concepts like objectivity and neutrality enter the discussion.
You could look at it like this: if some idea is more objective than some other, and some idea is more neutral than some other, then objectivity and neutrality exist.
Yes and no. Something can exist as a fact of the universe but still be unknowable. i.e. some hypothetical oracle could measure the quantum states of all human brains and ascertain what true objectivity looks like.
Regular mortals can have any certainty about this dbut espite the logical neccessity that this fact "exists" in some sense.
I think we're essentially also talking about the Overton Window to some degree. But that means you need to be OK with the thought that a sudden rise in extremism on one side of the political spectrum can alter the what you personally have to regard as "neutral and objective".