3 ms·

> value within their right answers The thing is that LLMs provide plenty of answers where "right" is not a verifiable metric. Even in coding the idea of a "rig

by awongh 1y ago

> value within their right answers The thing is that LLMs provide plenty of answers where "right" is not a verifiable metric. Even in coding the idea of a "right" answer quickly gets fuzzy- should I use CSS grid or flexbox here? should these tables be normalized or not? People simply have an unconscious bias towards the output just like they have an unconscious bias towards the same answer given by two real people they feel differently about- That is, the sort of thing all humans do (even if you swear that in all cases you are 100% impartial and logical). I think the impulse of ascribing intent and meaning to the output is there in almost all questions, it's just a matter of degrees (CSS question vs. meaning of life type question)

mdp2021 1y ago
> LLMs provide plenty of answers where "right" is not a verifiable metric I do not use them for that: I ask them for precise information. Incidentally, that one time in which I had to ask for a clever subtler explanation, it was possible to evaluate the quality of the answer - and I found myself pleasantly surprised (for once). What I said is, some people ask LLMs for information and explanation in absence of better and faster repositories - and it is just rational to do so. Those «answers where "right" is not a verifiable metric» are not relevant in this context. Some people use LLMs as <whatever>: yes, /some/ people. That some other people will ask LLMs fuzzy questions does not imply that they will accept them as oracles. > bias ... all humans do Which should, for the frame and amount in which the idea has some truth, have very little substantial weight and surely does not prove the "worshippers" situation depicted by the OP. You approach experience E in state S(t): that is very far from "wanting to trust" (which is just the twisted personality trait of some). > the impulse of ascribing intent and meaning [...] meaning of life First, of all, no: there seems to be no «intent and meaning» in requests like "what is the main export of Kyrgyzstan", and people who ask an LLM about the meaning of life - as if dealing with an intelligent part instead of a lacking thing - pertain to a specific profile. If you have this addiction to dreaming, you are again requested to wake up. Yes, we know many people who stubbornly live in their own delirious world; they do not present themselves decently and they are quite distinct from people radicated in reality. I am reading that some people as if anthropomorphize LLMs, some daemonize LLMs - some people will even deify them - it's just stochastics. Guess what: some people reify some other people - and believe they are being objective. The full spectrum will be there. Sometimes justified, sometimes not.
mdp2021 1y ago
Addendum, because of real time events: I am reading in a www page (I won't even link it, because of decency): > The authors[, from the psychology department,] also found that ... dialogues with AI-chatbots helped reduce belief in misinformation [...] «This is the first evidence that non-curated conversations with Generative AI can have positive effects on misinformation» Some people have their beliefs, but they can change them after discussing with LLMs (of all the ways). Some people are morons - we already knew that.
- fragmede 1y ago
  We knew that, but that doesn't help it when the moron is the one holding the metaphorical (or even literal) gun.
  - mdp2021 1y ago
    It surely does not help to suggest the lower decile or the median as representative of the whole. "People cannot count past ten - see how difficult it is to visualize eleven". If the mudstuck wants to be «honest» with itself, shall he ask around to some """outliers""" in the right side of the curve and be surprised.

by awongh 1y ago

mdp2021 1y ago

> LLMs provide plenty of answers where "right" is not a verifiable metric I do not use them for that: I ask them for precise information. Incidentally, that one time in which I had to ask for a clever subtler explanation, it was possible to evaluate the quality of the answer - and I found myself pleasantly surprised (for once). What I said is, some people ask LLMs for information and explanation in absence of better and faster repositories - and it is just rational to do so. Those «answers where "right" is not a verifiable metric» are not relevant in this context. Some people use LLMs as <whatever>: yes, /some/ people. That some other people will ask LLMs fuzzy questions does not imply that they will accept them as oracles. > bias ... all humans do Which should, for the frame and amount in which the idea has some truth, have very little substantial weight and surely does not prove the "worshippers" situation depicted by the OP. You approach experience E in state S(t): that is very far from "wanting to trust" (which is just the twisted personality trait of some). > the impulse of ascribing intent and meaning [...] meaning of life First, of all, no: there seems to be no «intent and meaning» in requests like "what is the main export of Kyrgyzstan", and people who ask an LLM about the meaning of life - as if dealing with an intelligent part instead of a lacking thing - pertain to a specific profile. If you have this addiction to dreaming, you are again requested to wake up. Yes, we know many people who stubbornly live in their own delirious world; they do not present themselves decently and they are quite distinct from people radicated in reality. I am reading that some people as if anthropomorphize LLMs, some daemonize LLMs - some people will even deify them - it's just stochastics. Guess what: some people reify some other people - and believe they are being objective. The full spectrum will be there. Sometimes justified, sometimes not.

mdp2021 1y ago

Addendum, because of real time events: I am reading in a www page (I won't even link it, because of decency): > The authors[, from the psychology department,] also found that ... dialogues with AI-chatbots helped reduce belief in misinformation [...] «This is the first evidence that non-curated conversations with Generative AI can have positive effects on misinformation» Some people have their beliefs, but they can change them after discussing with LLMs (of all the ways). Some people are morons - we already knew that.

fragmede 1y ago
We knew that, but that doesn't help it when the moron is the one holding the metaphorical (or even literal) gun.
- mdp2021 1y ago
  It surely does not help to suggest the lower decile or the median as representative of the whole. "People cannot count past ten - see how difficult it is to visualize eleven". If the mudstuck wants to be «honest» with itself, shall he ask around to some """outliers""" in the right side of the curve and be surprised.