the_af

Excuse my ignorance, but how does context memory provide an insight into a model's inner workings? I mean this:

> It is given sets of paramaters/data and infers a response based on it's acceptable confidence level. The bottleneck for this is compute power and model/prompt/rails/integrations.

or whatever the LLM claims its confidence is on any given output?

Spoiler: turns out the commenter I was asking about this had an issue with Wikipedia re: marxism and the Middle East / Israel.

The mention of marxism is puzzling, but re: the Middle East / Israel, I think it's impossible for any kind of written publication to report anything relatively recent on this topic that won't make anyone enraged about bias. So it seems a bit unfair to claim this is how Wikipedia is "being lost to ideology", because this is a property of the various Middle East conflicts, not of Wikipedia. Let me emphasize: nothing that could be written anywhere about the Middle East or Israel, today or in recent years, can possibly avoid enraging a large number of people all over the world.

Thanks for the honest reply.

Based on what you're saying, I think your real issue is that your own ideology and biases clash with whatever you perceive about Wikipedia. If Wikipedia had a less "marxist perspective of history" (!) or relied more on whatever you think is more reliable about the Middle East and Israel, then we'd have another commenter claiming Wikipedia is hopelessly biased in the opposite direction.

Unfortunately this boils down to "I don't like it", which is very subjective and very hard to evaluate in a neutral way that everyone agrees must mean "Wikipedia is being lost to ideology", as you claimed.

I think we all agree political and recent-ish history are mired in controversy and these divisive topics are problematic in Wikipedia, but ultimately hard to report on neutrally anyway. I guess they could settle for a insincere "neutral" tone that left everyone unsatisfied; Wikipedia supposedly tries to do that.

> The middle east articles involving Isreal use a lot of sources known to be unreliable like middleeasteye and Al-jazeera.

Good luck finding any written publication of any kind that everyone will agree is unbiased and reliable on this topic, and that won't upset a significant number of readers.

Re: specifically Middle East Eye, Wikipedia's own article on it is critical and highlights controversies. As for using it on articles about the Middle East, it would only be a problem if it was the only reference, but this doesn't seem to be the case in a cursory search I did. So aren't you effectively asking they ban some sources but left others, which ironically would introduce bias? Same with Al-jazeera or any other news outlets you dislike.

I find "sealioning" a deeply unsatisfying excuse for not providing a single example.

> I feel like that is the kind of response that is an example that could be used for "sealioning" on Wikipedia.

Not at all. Mainly because this is asking for the first line of examples, not repeatedly bogging down the conversation with constant requests for additional evidence.

Also it matters because in this kind of vague remarks that Wikipedia is unreliable, or that articles are wrong, almost always the people claiming this fail to provide specific examples, instead relying on some general feeling of frustration and shared belief that it must be true. But if they provided a specific example of a factually incorrect article, maybe other people would disagree. Or if they provided an example of ideological bias, maybe it would turn out the complainer is in the minority in believing that was biased. And if it turned out that this were controversial topics that most encyclopedic articles struggle with in some form, maybe it isn't a problem with Wikipedia. So examples matter!

What's the ideological bias you identify in Wikipedia? If you name it, we can discuss whether we agree with you or not.

Edit: wow, downvoted for asking this. Really, it's very simple: one should be able to make it explicit what bias they mean. If they don't, maybe they fear to show their own biases and be called out? Downvoting is the coward's way out.

Edit 2: I'm rate-limited, but I find this answer very puzzling:

> Rather, it's a recognition that it's not worth it to engage when the odds of the question being honest are slim.

That goes directly against HN guidelines. Assume the best version of a comment, not the worst. My question should be fairly straightforward to answer.

Are other encyclopedias or sources of information less or more likely to fall prey to this kind of bias?

Maybe in Wikipedia it's just more open, so you notice it more?

I've also noticed bias, especially in anything political (or that admits different political or ideological interpretations, such as history). However, I think you have one thing backwards: the more obscure and niche a topic is, the more likely you'll find it biased, since it has fewer editors looking at it.

---

Edit: reply to breezybottom, as I'm rate-limited:

> Encyclopedias in general are not primary sources, and should never be cited.

Note this is not what I'm arguing. I mentioned Wikipedia as a source of information, which it is, not as a primary source to be cited.

> The situation were formed by pretty predictable and easy to model factors

If what you claim were true, "pretty predictable and easy to model factors", it would have been modeled easily in a way that worked accurately to predict the future. Just the fact that this didn't happen should tell you it wasn't all that easy.

> if the science have failed even with that, how can we right now trust it with the things that are obviously way harder to predict?

What alternative to science and the scientific method do you propose?

It is sad because this is how science is meant to work, but propagandists will use it as evidence to dismiss it as some sort of liberal conspiracy. This is what the commenter you're replying to means by "doubt".

Science uses models to make falsifiable explanations and predictions. That these predictions are falsifiable and can be rejected, superseded or require fixing is how it's supposed to work, not an indictment against them.

> Isn't all of this true for a human-used internet as well?

Yes, but if you read this "semantic" info directly from a site, you can judge the site and its potential biases, and decide for yourself "this sells Louis Vuitton, of course they will praise their handbags". Or to pick a more extreme example that nobody would ignore, if you somehow land on an extremist website spewing racist crap, your alarms will surely trigger by anything they say.

But through an LLM this gets whitewashed and presented as authoritative and "neutral", when it isn't. You might know not to trust the LLM either (if you're paying attention; we all have our unconscious biases) but I bet you most people will be caught off guard.

There's nothing left for your imagination to fill in if there's official content (or expanded universe stuff) making it explicit for you.

As an example, nothing George Lucas did compares to what I wondered about the mysterious clone wars back when I watched (and rewatched) Star Wars. Lucas effectively "filled in the gaps", leaving nothing to my imagination. That what he did envision turned out to be pedestrian only added insult to injury, but now I cannot unwatch his prequels.

There's a similar phenomenon to when you read a book and imagine the characters and even situations mentioned off hand in the text, and then someone makes a movie out of it and makes it explicit and locks in specific actors and CGI-filled situations for whatever you imagined in your mind. You may or may not like it, but your own imagination gets stifled.

For things like "The Hunt of Gollum" it becomes downright bizarre. Whatever became of Gollum as he crawled towards Mordor, tracked by Aragorn and later mentioned by Gandalf, is such a minor footnote; definitely something for your imagination. Tolkien definitely didn't think it warranted even a short story. But no, they must make a movie out of it, because why not; the fans will lap it up even if it garners bad reviews.

As I mentioned in another comment, and there's no polite way of saying this, the nerdy need to know "what really happened" in excruciating detail (not to speculate, but to know) is a sign of immaturity. Art, even pop art, thrives as much in the known as in the unknown. To want to cast light into every single detail and "make it official" is a mistake. Though of course publishers and studios salivate at the chance of milking the proverbial cash cow.

Good point on Tolkien, though I would argue it's different because the Silmarillion was written for his own satisfaction, not his fans, and actually it was The Lord of the Rings which was initially written as "another Hobbit story" (though the story later "grew in the telling").

I think a more suitable example of what I find annoying is the extra material for the small and big screens, such as The Hobbit movies, suffering from movie elephantiasis; and The Rings of Power, which (over)explain and show the past of characters such as Galadriel, Elrond, and Sauron's rise to power, which in my opinion are best left as the ancient past of LotR (and maybe belong in the Silmarillion, but do note the latter is not as expository as the TV show). And does anybody really think The Hunt for Gollum was necessary?

Yet, if they make them, it's because they are sure fans will lap it up. To be honest I consider this need for endless "filling in the gaps" a sign of immaturity.

> We have to be careful about letting fiction inform our worldview too much.

Agreed. But would our AI entrepreneurs, so busy building the Torment Nexus, pay heed?

Agreed, and also: what's with the obsession with filling every possible gap in stories, leaving nothing to the imagination?

I remember when the Butlerian Jihad and the Clone Wars from Star Wars were complete mysteries. We readers and watchers were left mostly with our imaginations. My own feelings about the Clone Wars, mentioned off hand in the original Star Wars, turned out to be way better than the pedestrian stuff that George Lucas later concocted. And the less said about Brian Herbert's books, the better.

I get why the industry must rehash and fill every possible gap: it makes business sense, and they are risk averse for blockbusters.

But the fans? Why do nerds want everything spelled out for them? It boggles the mind.