And no surprise, apartheid apologetics: https://grokipedia.com/page/Apartheid#debunking-prevailing-n...
Hilarious factual errors in https://grokipedia.com/page/Green_Line_(CTA)
Impossible! That article was "Fact checked by Grok yesterday!"
I'm glad we've solved the LLM hallucination problem by fact-checking with LLMs. No way that could go wrong.
Many of the most glaring errors are linked to references which either directly contradict Grokipedia's assertion or don't mention the supposed fact one way or the other.
I guess this is down to LLM hallucinations? I've not used Grok before, but the problems I spotted in 15 mins of casual browsing made it feel like the output of SoA models 2-3 years ago.
Has this been done on the cheap? I suspect that xAI should probably have prioritised quality over quantity for the initial launch.
Citizendium is still around, though they've loosened some of the requirements in order to encourage more contributions, which seems self-defeating to me. I think they should have tried to cooperate with Wikipedia instead. The edits and opinions of subject matter experts could be a special layer on top of existing Wikipedia articles. Maybe there could be a link for various experts with highlights of sections they have peer-reviewed and a diff of what they would change about the article if those changes haven't been accepted. There could also be labels for how much expert consensus and trust there is on a given snapshot of an article or how frozen the article should be based on consensus and evidence provided by the experts. This would help users delineate whether an article contains a lot of common knowledge or whether it's more speculative or controversial.
Regardless, the business was there. Wikipedia killed all that. So if you want to create an expertly created encyclopedia anno 2025 you have a real problem: you will need to pay experts for their time somehow otherwise why would they compete with the million monkeys, but your source of revenue has been strangled by those very same monkeys, who it turns out produce content that is orders of magnitude better than anything I've ever read in a for-pay encyclopedia from before Wikipedia.
The bar to entry is insanely high.
I mean, I don't think this is _for_ people who care about quality, tbh. For those, there is wikipedia. This is more of a safe space for Musk.
Wikipedia isn't for those who care about quality, either. It's still quantity over quality, just not as badly as this LLM garbage.
++
I find this to be the most annoying aspect of AI. The initial Google AI results were especially bad. It is getting better, but still spout info I know is false without any warning.
Like, I find blowhards tiring enough in RL. Don't really want to deal with artificial blowhards when I'm trying to solve a problem.
Whilst I haven't read the entire article, the first paragraph is actually on-point: apartheid was shit in a lot of respects, but the schools, especially in rural areas, have dramatically declined since 1994, as have most government-run companies (with the exceptions like Eskom being bailed out every year).
You don't have to like the facts, but that's what they are.
"I think we gotta hand it to Apartheid because schools were very slightly less worse" isn't the argument you think it is. It does paint where you stand quite clearly.
Never start a sentence with "I'm no apartheid apologist, but". Nothing good can ever come out of it.
Saying "advancements in black literacy and real wages during the era" as if those things are due to apartheid is offensively absurd.
How about we advance literacy and wages without, you know, all the apartheid.
So, ask the rest of Africa how that has gone for them.
This is a false contradiction. You can oppress people and still pay them higher wages and fight illteracy.
Since a rising tide lifts all boats, in a growing economy you might see wages, literacy and health outcomes improve nominally for an oppressed group, in absolute terms, while all of these outcomes improve significantly faster for the other group(s) in the same period.
That's not another narrative that's a false contradiction
You also don't seem to understand "apologetics"?
Weird that it displaying some other web site's embed/shortcodes:
> ![Cottage Grove-bound Green Line train approaching Roosevelt station][float-right] The Green Line utilizes primarily 5000-series railcars
But yes, it also suffers from attention to irrelevant detail.
> The article does seem a bit much though. I've noticed a lot of the Grokipedia articles just go on.
And yes, that's what I'm noticing as well. There is a clear attempt to establish a narrative.
So, being attacked and nearly killed by other white boys does not validate his opinions on apartheid.
Not true, nearly 30% of their budget goes to partisan activism with DEI related initiatives.
"Supporting equity represents the second largest part of our programmatic work"
https://meta.wikimedia.org/wiki/Wikimedia_Foundation_Annual_...
And an encyclopedia can absolutely do that and still present factual information based on actual research and facts.
You know that just because a lady has blue hair or a person has colored skin does NOT mean that they can't be right about something or do good research. Right? You do know it, right?
Because in the end, when you cry about DEI (whatever you believe it to mean), this is the implication that comes with it: that you can't imagine for a second that anyone who doesn't look exactly like you could ever do anything competently. I genuinely wonder if you've ever thought about that for more than half a second after you closed that Charlie Kirk video.
If you do believe it, fair enough. I guess you're allowed to believe it. But at least be honest about it.
A comment like this is completely unacceptable on HN. The guidelines make it clear that we expect better than this here:
Be kind. Don't be snarky. Converse curiously; don't cross-examine. Edit out swipes.
Comments should get more thoughtful and substantive, not less, as a topic gets more divisive.
When disagreeing, please reply to the argument instead of calling names. "That is idiotic; 1 + 1 is 2, not 3" can be shortened to "1 + 1 is 2, not 3."
Please don't fulminate. Please don't sneer, including at the rest of the community.
Please respond to the strongest plausible interpretation of what someone says, not a weaker one that's easier to criticize. Assume good faith.
Eschew flamebait. Avoid generic tangents. Omit internet tropes.
Please don't use Hacker News for political or ideological battle. It tramples curiosity.
If you're correcting a lie that they actively want to spread (e.g. the "White Genocide" lie) they obviously won't accept your correction.
If you're correcting a lie that they don't care about (hypothetical example: "ripe strawberries are blue") they might accept your correction, but that makes it less likely that an uninformed visitor will see through their lies.
So the best case scenario is that literally nothing happens, and the worst case one is that you're indirectly helping their cause. What am I missing?
They know that this is being built with an explicitly partisan bias, but also know liberal folk are super weak to certain types of attacks. Attacks like "be fair" and "you can submit feedback". They are explicitly playing on liberal weaknesses all while leaning into the blatantly false representation of history. They will say "you can't judge it since you haven't provided feedback to help correct it" but that was never their intention and they just say those things because it confuses liberals. And I know I've overused "liberals" here. That's absolutely my bias slipping in. But they are the most credulous folks who will constantly try to reach across the aisle to blatant bad faith actors. If there's a better term for this, I haven't learned it in my 20+ years of being politically active in the US.
Not that I think it does any good on a site like this, but I vouched for your original comment. I completely agree with you. I was also raised in the cult and have first hand experience with many people still in it. I know why they would support and lend effort to things like these. They support all sorts of awful shit.
>pays for real expenses
Only a small percentage of donations do.
>instead of whatever vanity project
Anytime the topic of Wikimedia donations come up you will see people complaining about their vanity projects too, wishing they could donate towards wikipedia itself.
Not the person you are replying to, and it is a bit tangential, but you just basically described a solid chunk of open-source software work.
I am not mocking open-source software work, I am mocking how reductionist the parent comment was, because their logic often applies to volunteer open-source software work as well. And, I suspect, on HN we can agree that volunteer open-source software work can often be worth doing, regardless of how "irrational" it is or how much for-profit corporations could benefit from it.
Why? This site isn't run by people who are interested in factual accuracy.
If they think Wikipedia articles are inaccurate, they could always propose changes and have a proper discussion with the rest of the contributors. Grok was trained on Wikipedia so realistically this is just a jumbled regurgitation of Wikipedia articles blended with other sources from across the web without the usual source vetting process that Wikipedia uses.
This is a politically motivated side project being run by the worlds richest man, and frankly I doubt many people are interested in helping him create his own padded version of reality.
the proper discussion you want will never happen. it’s an exercise in persuasion ie trying to move people from one entrenched position to another, and there’s nothing more impossible than that. the only way out is to offer competition, and that’s what grokipedia seems to be doing. check the history of christianity, heresy, reformation. when the catholic church set itself up as the object to be won over persuasively it successfully stifled doctrinal progress. until the intolerants exited.
Are you familiar with Wikipedia at all? Here, for anyone who is unfamiliar, let's take a look at an example page: https://en.wikipedia.org/wiki/Apartheid - this is guaranteed to have controversial ongoing discussions given the political climate.
Note how at the top of the page right now there are two large boxes discussing ongoing changes to the article - one indicating that it is considered too long, and another indicating that some of the content is being split into a separate draft [0] page. Both of these boxes include links to the relevant pages and policies.
The first box, indicating that the article is too long and drifting off topic, includes a direct link to the Talk page [1]. Note that this page is also linked at the top of the article, and that goes for every single article on wikipedia.
That talk page is where the proper discussion that I want happens - out in the open. Note that you can even reply to talking points without needing an account. Note that replies and criticisms are reproduced and readable directly on the page.
This is what open collaboration and truth seeking looks like. "Grokipedia" requires you to create an account and funnel a suggested correction into an black box. It's the equivalent of a suggestions box in an HR office. On wikipedia, the discussion is out in the open, while the grok version just says "Fact checked by Grok" at the top, like we're supposed to blindly trust that.
Which of these is modeling open collaboration, and which of these is just deferring to priest grok, again? The grok page gives no indication that alternative interpretations exist, they don't show any indication that sections are being criticized as inaccurate. Comparing Wikipedia to the catholic church like this is divorced from reality, doubly so in comparison to this grok project.
[0] https://en.wikipedia.org/wiki/Draft:History_of_South_Africa_... [1] https://en.wikipedia.org/wiki/Talk:Apartheid
Have you ever been to a Wikipedia Talk Page? Basically every page you can find will have some people arguing about what should be placed on the page on the Talk page.
I think that's the "Talk" pages that go with the entry pages.
That being said, my biggest issue with it is how Grok is writing everything. It's like it is trying REALLY hard to be neutral but it's conversational training slips up and starts "spicing" things up a little. For example on Elon's article:
"...at age 12 in 1983, developing a space-themed video game called Blastar, which he sold to PC and Office Technology magazine for approximately $500. *This early entrepreneurial act foreshadowed Musk's later pursuits in technology and business*."
Sentences like that are designed to subtly bring emotion to certain topics.
>..Musk founded SpaceX in 2002 as CEO and chief engineer, Tesla in 2003 ... (grok)
>Musk joined the automaker Tesla as an early investor in 2004 and became its CEO ... (wikipedia)
I think Wikipedia is more accurate on that one.
It’s not that it’s trying to lie — it’s just how these models work. They’re great at making language sound right, not necessarily be right. Feels more like a mirror of what the internet “thinks” than an actual source of truth.
If they framed it that way — more experiment, less Wikipedia — I think people would take it a lot better.
An utter waste of everyone's time, money, effort, and manpower.
“ In recent decades, the party has prioritized identity-based equity policies, climate interventions, and expansive regulatory frameworks, yet empirical critiques highlight correlations between its governance in major cities and elevated crime rates, homelessness persistence, and educational stagnation amid softened enforcement and redistribution efforts.[7][8]”
Which doesn’t link to anything supporting the negative assertions.
No search results for Republicans Party, which I assume means it said something Musk didn’t like.
It really likes that word, and seems to use it a lot to justify displaying its owner's views.
https://grokipedia.com/page/Republican_Party_(United_States)
> On July 2, 2025, the band released their first live album, American Football (Live in Los Angeles), recorded during the anniversary shows at the El Rey Theatre in Los Angeles with guest appearances by Ethel Cain and M.A.G.S., accompanied by a concert film documenting the performance.
If you go to the source [^4] for this claim, you'll see that:
- They dropped a film of the same name alongside the album release.
- The "guest appearances" are actually interviews in the film.
- The entry excluded the female artist that was cited in the source.
I, then, compared Grok's entry on United Airlines [^2] against Wikipedia's [^3]. Grok's seemed to be autogenerated this time.
I skipped to the section on MileagePlus since I know a bit about how that program works. It has a few inaccuracies:
- It only lists the four published MileagePlus tiers: Silver, Gold, Platinum and 1K and omits the two unpublished, but well-known, tiers above 1K: Global Services and Chairman's Circle.
- The 2025 premier qualifying point (PQP) redemptions are actually from 2024.
- Some of the language it uses wouldn't meet Wikipedia's editorial standards, like the nebulous "priority everything" benefit from obtaining 1K status (whose source is unclear, as neither of the two sources cited use this phrase).
- "The current logo features a stylized "U" incorporating a world map outline, symbolizing global connectivity" That's United's old logo. They absorbed Continental's logo when they merged.
- The article opens with the claim that United has 1018 aircraft in its fleet as of APR 2025, then, later, states that it has 1,001 active aircraft as of OCT 2025. The source for the 1,001 figure states 1,055 on the page with 1,003 in revenue service.
So I wouldn't use Grokipedia as a source for anything, just like Wikipedia, though I'm sure some will try.
[^0]: https://archive.is/twkBP (might not be available yet; it's still getting archived)
[^1]: https://archive.ph/lOkdT
[^2]: https://archive.ph/EnN2T
[^3]: https://archive.ph/uooNW
[^4]: https://pitchfork.com/news/american-football-to-share-new-li...
Then actual description of the war is much more biased in the Elonopedia. In every case possible the invasion is presented as "both sides are guilty". I wouldn't list the examples, anyone can do it. Too much effort imo.
Then I checked Russo-Georgian War articles, this time at least the century and war was correct in Elonopedia. But again, right from the start it is incredibly biased towards Russia. Elonopedia completely omits the initial attack make bu Russian forces at 01 Aug 2008, skip a week and presents war as if it was initiated by Georgians, following Kremlin propaganda line. Didn't both reading full article.
All in all it is 100% as I have expected reading the news about this supposedly "unbiased" encyclopedia - it's a LLM-generated slop, with no human fact checking (mixing two different century separated wars into one article is telling), and it is essentially a far-right propaganda outlet. It will follow Goebbels rule of mixing 60% or truth with 40% of lies, to prime up unsophisticated readers towards Elon's and rightwing crowd goals.
The EU isn't exactly known for being Kremlin propagandists. Here is the link to the 700-page international fact-finding report they published in 2009: https://www.echr.coe.int/documents/d/echr/hudoc_38263_08_Ann...
This Radio Free Europe article is a decent summary of the report: https://www.rferl.org/a/EU_Report_On_2008_War_Tilts_Against_...
Why do you think the international team of Europeans would leave out something like an August 1st attack by Russian forces? Why would the US-funded media outlet for Europe (RFE/RFL) parrot the report's position that the conflict was overwhelmingly Georgia's fault?
"The Mission is not in a position to consider as sufficiently substantiated the Georgian claim concerning a large-scale Russian military incursion into South Ossetia before 8 August 2008."
Can you share the evidence you have that supports your position that Russia attacked on 01 August? The EU concluded that was unsubstantiated.
And if you are that suspicious of that date, we can pick another. In 1992 Russia invaded independent Georgia (among other countries), so any action was towards occupation force, in defense.
PS: and if look throughout the history, we will find very few cases, when a smaller country attacks much bigger one especially after already losing at least one fight against them. And the opposite is true, there are hundreds and thousands of cases when a bigger country attacks the smaller one, especially after already winning once against them. And countless times when a bigger country lied about pretext for such attack, to be seen as not crazy murderers outright, but muddy waters and sow doubt. Russia succeeded it seems.
Grok’s pages on the prosecutions of Trump are definetly biased.
I’m probably not the core audience for this though. I use Wikipedia as a reference, not to tell me what to think.
For non-experts just exploring new topics, it’s still perfectly useful. Grokipedia probably uses a similar search, verify, summarize workflow, so it naturally inherits mistakes from the internet, which isn’t really an LLM problem.
Grok is just the first to make it public, and other AI companies could easily build their own synthetic data Wikipedias, and some probably already have.