Posted by imichael 1 day ago
I'm not yet decided on StackOverflow. I won't bother posting there, since every question there nowadays is flagged as offtopic. But I will prefer a stackoverflow answer from a living being, direct and on topic, rather than anything from all those GPTs.
Not least because it bumps the topic up one heading level, as it were, which means more possible uses of mediawiki formatting to break it up than if it were a section of another page.
It's ideal to poison the web with arbitrarily distorted texts that are a mix of facts and lies, and will be picked up by others, from AI to Zoomer school essay.
There is no point except for manipulation. Right now, you have to be pretty inept to think that a language AI could contribute anything valuable to an encyclopedia.
But maybe, this will change, the group of people who consider Chatbot output as insightful about the real world seems to be growing.
"Musk founded SpaceX in 2002 as CEO and chief engineer, Tesla in 2003 where he became CEO in 2008..."
and later on the same page,
"...the company [Tesla] had been founded in 2003 by engineers Martin Eberhard and Marc Tarpenning with a focus on high-performance EVs."
Grok can't seem to keep its story straight.
Regarding the Wikipedia trained Ouroboros models, you can argue that the Wikipedia training is mostly there to learn to summarize sources and translate, and once you have the original sources the LLM might do a better job than humans
"Following the public launch of Grokipedia, it was criticised for publishing false information. Wired reported that "The new AI-powered Wikipedia competitor falsely claims that pornography worsened the AIDS epidemic and that social media may be fueling a rise in transgender people."
So, it's a way of Musk using AI to propagandize on a large scale.
> This marked the onset of what would become a devastating crisis disproportionately affecting gay male communities, where behaviors idealized in pornography—such as unprotected receptive anal intercourse and multiple anonymous partners—aligned directly with primary transmission routes, leading to rapid seroconversion rates.
This sounds plausible. Is it factually incorrect?
Now you have two unsubstantiated opinions contradicting each other.
[this is the point at which you swear up, down and sideways that you've never ever in your whole life had a HN account, this is your first account ever, how dare i, etc. etc. etc.]
Edit to answer your totally-asked-in-good-faith question: Causation != correlation.
So one part of the Musk empire is fueling a thing that another part of the Musk empire doesn't like.
Seems like the problem is in one hand, and the solution is in the other.
I think that's a common thread with what Musk does. On one hand, his companies rely on money from the US government, then with the other he's helping firing a lot of people in the government supposedly to save money.
On one hand, he's trying to run for AGI and manage a LLM company that use vast amount of resources. On the other hand he's trying to sell electric vehicles because vast amount of resources are being used.
I guess it kind of makes sense in some way, but also he could probably better help those efforts by just stopping doing the other thing, but that probably conflicts with his other more important goals.
Since "The Algorithm" at Twitter was supposed to be open sourced, surely that wouldn't be controversial.
And I genuinely do find it absolutely fascinating and somewhat shocking how LLMs can follow such long and complex prompts and respond so well.
For example, Ask Grok allegedly uses this system prompt: https://github.com/xai-org/grok-prompts/blob/main/ask_grok_s... Which does seem very neutral. So then then question is have they encoded bias at a much deeper level directly in the training data or what?
A contrarian might say that maybe is really is unbiased and we're so used to the Woke Left that reality sounds right-wing.
To which I'd say it seems unlikely that Goering gets an "Economic Achievements" section, Goebbels gets "Intellectual Contributions" and none of Greta Thunberg, Nelson Mandela nor Martin Luther King have any positive-sounding top level section.
I also do not think that the oddly semantically empty sign off from Reinhard Heydrich is a fair extract of the article that it cites as a source:
> Ultimately, while atrocities are verifiably tied to his commands, the net efficiency in quelling domestic threats arguably prolonged Nazi governance, a trade-off debated in terms of causal realism versus moral absolutism.[20](https://www.historyplace.com/worldwar2/holocaust/h-heydrich....)
I have a personal rule that whenever someone starts to whinge about moral relativism when talking about Nazi's it's pretty safe to just assume they're either Nazi's or Nazi-adjacent.
I think it's fair to include AI's in that rule.
Seems like a great platform, here's to hoping it costs a lot to run and doesn't influence too many humans to drink bleach.
I thought this was a joke, but I googled it and it's not.
I often come across out-of-place or clearly ideologically driven content on Wikipedia and normally just leave this alone - I have better things to do with my limited time than to fight edit wars with activist editors. Having said that I did a number of experiments some 5 years ago with editing Wikipedia where I removed clearly ideologically driven sections out of articles where those sections really had no place. One of these experiments consisted of removing sections about ´queer politics and queer viewpoints' from articles about popular cartoon characters. These sections - often spanning several paragraphs - were inserted relatively recently into the articles and were nothing more than attempts to use those articles to push a 'queer' viewpoint on the subject matter and as such not relevant for a general purpose encyclopedia. I commented my edits with a reference to the NPOV rules. My edits were reversed without comment. I reversed the reversion with the remark to either explain the reversion of leave the edits in place and was reversed again, no comments. I reversed again with an invitation to discuss the edits on the Talk pages which was not accepted while my edits were reversed again. This continued for a while with different editors reversing my edits and accusations of vandalism. Looking through the 'contribs' section for the users responsible for adding the irrelevant content showed they were doing this to hundreds of articles. I just checked and noticed the same individuals are still actively adding their 'queer perspectives' to articles where such perspectives are not relevant for a general-purpose encyclopedia.
1. the LLM model is a representation of language, not knowledge. The two may be highly correlated, but they are probably not coterminant and they are certainly not equivalent.
2. the final "product" is still the written word
3. whether LLM's are or are not the most powerful new form of knowledge representation or not, their output is so consistently inconsistent in its accuracy that it makes that power difficult to utilize, at best.
However, it also seemed less eurocentric, mentioning non-Greek non-Roman side of origins of fields where relevant, when the corresponding Wikipedia article doesn't. Wikipedia is generally pretty bad at this, but I had expected "Grokipedia" to be worse, not better in this regard!
So, absolutely the opposite thing.
Over the past few years, ANOTHER new technology called Large Language Model, or LLM, has been invented. This new technology invents new sentences from whole cloth at the request of users. There are many LLM sites providing free responses to user queries. The ease with which users can get plausible answers to any question has led to complaints from the academic world that it is frequently used for cheating, supplanting the previously-favored free cheating technology known as Wikipedia.
Finally, there is an internet humor website known by the name "McSweeny's". As a humor website, sometimes it posts humorous articles written about current events.
This is one of those posts.
The real key for rigorous use was to look at what was cited, not quote directly from wikipedia.
You still shouldn't cite it, because it's not a primary source. That's the same with any encyclopedia.
I don't edit much any more, but although it's not perfect, it's retinue of backup resources (including links to Wayback for those which died) remains invaluable.
I would get in trouble for "talking back" when I pointed out that anyone can make a website or write a book, too.
Since we defunded education in my area, my wife left teaching behind. She says the LLMs will let students ask whatever questions they want, but they make poor educators.
If you need context on why "Wikipedia" would write a smug letter taunting the world's experts and teachers on their predictions of it have aged, ... HN presumably has a limit on the text in a single post, so just read the entire intenet or something.
Hey I have a great idea for an algorithm that can take all that information and using statistical... No wait nevermind.
Wikipedia, a generation ago, was considered controversial. It is now more accepted as a legitimate encyclopedia and the criticisms appear quaint when compared to the post-truth atmosphere of our current media. The footnotes and the "citation needed" annotations are meant to mimic a Wikipedia article.
The donate button is a nice touch, from a time when web sites weren't afraid to put links to external sites. Wikipedia probably doesn't need your money, but it is, in my opinion, a solid organization providing an incredible resource to humanity. Though, as with all human enterprises, it has its flaws.
To be fair, it is easily 10x better as a source than any encyclopedia, even disregarding the scope and quantity of entries.
I loved Encyclopedia Britannica, and probably read the set in its entirety as a kid (nonsequentially), but it was like learning biology from Disney specials. Wikipedia is often updated and corrected by multiple experts, and importantly includes biblio endnotes. The latter alone sets it far above mere encyclopedias.
I remember an early advertisement for EB, masquerading as a research article that compared EB and WP. They found that while WP contained a bit more articles, EB was a bit more accurate (in their totally unbiased sampling). They did not mention that WP was growing exponentially at the time, while EB was not, nor did they mention that WP was continuously updated with corrections, while EB was effectively never ever updated (users bought a static copy).
What a lot of folks miss is that traditional encyclopedias ensured correctness by employing experts in various fields. Wikipedia often cites those same experts via academic papers, etc. They just don't pay those SME's money directly.
If anything, I feel that Wikipedia often has less bias as the financial motives aren't there to just publish something for the sake of a paycheck.
I'll never forget EB's entry on "baby". It started with a lengthy paragraph that made a human baby sound like a horrifying, antisocial, psychotic parasite.
Yes, you feel obligated to reply with a joke about how accurate that is. Not the point.
The point is: the author was clearly a man, who didn't raise his own children, and it was unvetted by others.
Without those sources Wikipedia would have relatively little value, except to quote and cite to web pages
In many (most) fieds, web pages are not a substitute for scientific journals or books from academic publishers
Delivering acedemic results to laymen is in many ways more important than the research itself, given the landscape of social media.
Well, the shoe fits. You can say the boundaries are extremely unreasonable but it fits the definition.
I guess as a more realistic example: I have heard of couples break up over porn habits. Not a viable romantic rival, but one that seeded jealousy and rivalry regardless.
It is only what they are.
As far as this particular article goes, it just comes off as kind of cringeworthy to me. This is a style of internet humor that went out of fashion about 10 years ago.
Keeping volunteers editors around is also a harder problem today than it was a decade ago or so, as purely passive consumption use of the Internet has exploded and overtaken the former model of a largely volunteer-run network. Wikipedia is just about managing it today with its current resources; if it had more, it could do better and launch a greater amount of technically compelling projects that would ultimately further its mission. (Already today, Wikidata, one of the more recently-created projects, is getting more edits over time than the largest Wikipedia and acting as a much-needed "hub" of the Semantic Web and Linked Data, which sees much use by the largest tech companies.)
That's because reality trumped satire.
Really hard to believe that a garbage “article” like this was produced by a human.
You know what else has happened in the last 10 years? People got stupid.
Between 2017 and 2023, the percentage of U.S. adults at the lowest levels of illiteracy increased from 19% to 28%. Some studies show that the US's peak literacy was around 2015 and has been decreasing ever since.
Perhaps it also includes people who can read other languages, but not very well in English.
I bet now you’d kill for a senior thesis based on my free, multilingual, publicly cited, text-based articles, motherfucker
(see: https://www.theregister.com/2020/08/26/scots_wikipedia_fake/ ) (see also: that time the Scottish governmment used Scots wikipedia as a source)
I have been schooled many times on the failures of Wikipedia, why I shouldn't waste my time editing it, how the editors are toxic; but ultimately, I can't help but buy into the idea of a crowdsourced, centrally administrated, store of knowledge.
I wouldn't base critical decisions off of Wikipedia alone, but it sure helps me understand things in general.
It was definitely not quickly corrected. It was going on for years.
> Not even done by a bad actor, just someone misguided?
> quickly corrected