This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
You are conflating two vastly different things. Perhapse this is an issue of poor translation between Indian and English but what GPT has is better described as a notion of "consensus" or "correlation". A degree to which [token a] is associated with [token b] which is emphatically not a concept of true vs false. To illustrate, if you feed your LLM a bunch of Harry Potter fan fiction as a training set your going to get a lot of Malfoy/Potter gay sex regardless of how Rowling may have written those characters and this is not an aberration, this is the system operating exactly as designed.
Liberals assume minorities can't speak English well, conservatives don't. Maybe you were the blue-tribe progressive all along?
More options
Context Copy link
Even after seeing your explanation below, it's hard to read this comment in good faith. (What language is "Indian", even?) You're either being profoundly ignorant or just antagonistic.
You know what, the more I think about it the more absurd this warning feels.
Are we really going to pretend that acknowledging a potential cultural/linguistic difference that may in fact be relevant to the discussion at hand is somehow more "ignorant", "antagonistic" or *chuckles* "racist" than the typical HBD or Joo post that you guys routinely let slide?
Really?
You're talking to someone who has established, with a long history here, that his English proficiency is native level, even on scientific and medical subjects. I have never seen you pull some kind of "Well, maybe the concept just doesn't translate well from Russian" when arguing with Daseindustries. I'm not buying it.
I'm gonna be blunt: I don't believe you actually believed there was a linguistic misunderstanding. I think you wanted to imply self_made_human is ignorant, and do it with an added dose of condescension. "Translation between Indian and English" indeed - if you know that there are multiple Indian national languages (it's actually more like 20+, not five), then you know perfectly well how stupid and ignorant it is to refer to "Indian" as a language. I'm extending you as much charity as I can to assume you were being ignorant and not intentionally insulting and - yes - racist.
More options
Context Copy link
More options
Context Copy link
I had to restrain myself from reporting this myself, or calling him out (more than I did) for the boorish reply. My (near-angelic) patience has been rewarded.
More options
Context Copy link
Whichever one the dude speaks natively, India has like 5+ national languages and IME confusing Urdu, or Bengali for Hindi is often a good way to start a fight.
More options
Context Copy link
More options
Context Copy link
I don't think the example suffices. What a model is trained on is really important for what it does. An LLM trained more on text describing reality should more frequently make true statements.
Yes, but where do you find that text? The problem is that contra what several other users here keep saying, LLM's are not reasoning engines they are pattern generators. Which in turn brings us back to my post from a month ago.
You can't ask an LLM to restrict itself to only giving "true answers" because LLM's don't actually have a concept of true vs false.
More options
Context Copy link
More options
Context Copy link
actually kind of racist comment
More options
Context Copy link
What the hell is that supposed to mean? I speak English just as well as you do Hlynka, and likely better. You can condescend to someone else.
Humans develop their sense of truth and falsity from comparing/correlating new evidence to previous evidence and their environs, biased by whatever intrinsic priors they were born with. The only difference here is that GPT-4 has no appendages to interact with the world and seek out further evidence, merely what we've trained it with.
I fail to see how this has any bearing on the "truth" of it all. Depending on how big the model is, it can when asked almost certainly tell you that in the primary text Harry wasn't casting spells on Malfoy's wand.
That is so clearly obvious to me I'm not even going to burn the 100 joules or so of energy it would need to spin up a GPT-4 instance to confirm it.
Do you think an AI who hadn't read the original HP and had a knowledge base entirely of yaoi fanfic has any way of knowing better? Or a human for the matter.
Exactly what it says. I know of at least a couple SE Asian languages (EG Tagalog and Malay) where the distinction between Yes/No, True/False, and Agree/Disagree, is significantly less distinct than it is in English and I was wondering if something similar may be going on here, as by my reading your second statement doesn't follow from the first at all, nor from anything in Open AI's White Paper as far as I recall. (Assuming we are both referring to the same paper)
If you set aside for a moment your pre-existing knowledge that Harry Potter and Draco Malfoy are both fictional characters "how this has any bearing on the Truth" ought to be immediately apparent. Stop for a moment and reflect. Ask yourself WHY you believe that GPT's description of Harry and Draco's relationship would in any way resemble that of the "real" people (or in this case that of the characters as originally written).
A statement being "True" is not the same thing as agreeing with a statement, or that statement comporting with the popular consensus and the seeming conflation of these three distinct stances is what initially lead me to suspect some sort of translation issue might be at play. Is it possible that are you are using the word "truth" when you really mean something closer to "popular" or "I agree with"?
With that out of the way, to answer your question, Do I think a human who hadn't read the original HP and had a knowledge base entirely of yaoi fanfic would know any better? Yes I do, because in contrast to GPT I would expect a normal human to display some level of contextual awareness/meta-knowledge IE being aware of yaoi fanfic and it's tropes. Or being able to assign a confidence level to a prediction that was anything other than completely arbitrary.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link