Tim Bray on Grokipedia

> Why on earth would we want a technology that’s as good at summarisation as it is at hallucinations to write encyclopaedia entries?? You can never trust it to be faithful with the sources.

Isn’t summarization precisely one of the biggest values people are getting from AI models?

What prevents one from mitigating hallucination problems with editors as I mentioned? Are there not other ways you can think of this might be mitigated?

> You would need to trust a single publisher with a technology that’s allowing them to crank out millions of entries and updates permanently, so fast that you could never detect subtle changes or errors or biases targeted in a specific way—and that doesn’t even account for most people, who never even bother to question an article, let alone check the sources.

How is this different from Wikipedia already? It seems that if the frequency of additions/changes is really a problem, you can slow this down. Wikipedia doesn’t just automatically let every edit take place without bots and humans reviewing changes

> If there ever was a tool suited just perfectly for mass manipulation, it’s an LLM-written collection of all human knowledge, controlled by a clever, cynical, and misanthropic asshole with a god complex.

It’s painful to watch how many people (a critical mass) don’t understand this — and how dangerous it is. When you combine that potential, if not likely, outcome with the fact that people are trained or manipulated into an “us vs. them” way of thinking, any sensible discussion point that lies somewhere in between, or any perspective that isn’t “I’m cheering for my own team no matter what,” gets absorbed into that same destructive thought process and style of discourse.

In the end, this leads nowhere — which is extremely dangerous. It creates nothing but “useful idiot”–style implicit compliance, hidden behind a self-perceived sense of “deep thinking” or “seeing the truth that the idiots on the other side just don’t get.” That mindset is the perfect mechanism — one that feeds the perfect enemy: the human ego — to make followers obey and keep following “leaders” who are merely pushing their own interests and agendas, even as people inflict damage on themselves.

This dynamic ties into other psychological mechanisms beyond the ego trap (e.g., the sunk cost fallacy), easily keeping people stuck indefinitely on the same self-destructive path — endangering societies and the future itself.

Maybe, eventually, humanity will figure out how to deal with this — with the overwhelming information overload, the rise of efficient bots, and other powerful, scalable manipulation tools now available to both good and bad actors across governments and the private sector. We are built for survival — but that doesn’t make the situation any less concerning.

> Llms arem't "there" yet with respect to this sort of thing

Yes, nothing about this is “there yet” which was my point

> Having an editor is totally infeasible, at that point you might as well have the humans write the articles.

Why?

> versus a site being useless by not actually having facts about the topic you're looking for, yet spouting out authoritative-looking nonsense anyway.

You just described Wikipedia early on before it had much content, rules around weasel words, original research, etc

Any condition which causes the individual to self-sterilize or not have progeny is maladaptive from an evolutionary perspective. Some traits like sickle cell adaptive against malaria, but those who are homozygous suffer from the disease of sickle cell anemia. I struggle to imagine how trans is adaptive in any way, seems to only cause problems. The leftist narrative is that such individuals must engage in costly medical procedures to avoid committing suicide, so by their own framing they basically consider it a mental disease.

> "[...] conservative and anti-woke content can reach wide audiences, particularly when Elon puts his thumb on the scale."

No shit; it's always been that way since mass media became a thing. Besides, there is no such thing as quality conservative and/or "anti-woke" media. The very concept represents a contradictio in adiecto. And Elon's just the modern version of an industrialist of yesteryear. Back in the day they owned the mass media of their time: radio and television. Today its "AI"-enshittified parasocial media and ideally the infrastructure that runs those dumps.

> "Don't forget that public opinion and the media landscape are quite different in 2025 from what they were in the 2010s when most prior studies on WP bias have been written."

Bias studies have been written since Wikipedia became a staple in hoi polloi's info diet. And there's always been a whole cottage industry of pathological and practised liars (e. g. the Heritage Foundation, amongst others) catering to right-wing grievance issues. The marked difference is that the right's attacks against Wikipedia as an institution are more aggressive since Trump... completely in line with the more aggressive attacks on human rights, reason, science, and democratic institutions on part of conservatives world wide.

> that is like skim reading or basic introductions

No? How do you learn stuff you don't know? Are you really telling me you enroll in a graduate course or buy a textbook for everyone one?

Like, can you give an example of a "deep dive" research project of yours that does not begin with an encyclopedia-style treatment? And then, maybe, check the Wikipedia page to see if it's actually worse than whatever you picked?

Again, true domain experts are going to read domain journals and consult their peers in the domain for access to deep information.[1] But until you get there, you need somewhere you can go that you know is a good starting point. And arguments that that place is somehow not https://wikipedia.org/ seem... well, strained beyond credibility.

[1] Though even then domains are really broad these days and people tend to use Wikipedia even for their day jobs. Lord knows I do.

I don't know if this is why, but: he's in a unique position of having an article on himself on Grokipedia, and thus being able and willing to compare it with the reality as he remembers it.

That's in contrast to other topics, the nuances of which even seasoned experts could disagree about. Any discussion on that could devolve into the nuances of the topic rather than Grokipedia itself. But it's fair to assume the topmost expert on Tim Bray is Tim Bray, so we should be getting a pretty unbiased review.

As such it could be a useful insight into how Grok and Grokipedia and its owners operate.

I doubt a post saying it was so boring he was unable to finish reading the page about himself is going to bring in many readers.

That's kind of been my impression too. Not that it's terribly biased or anything but just rather boring to read.

What percent of edits on Wikipedia do you think are done by LLMs presently? It looks like there is a guide for detecting them https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing . The way Wikipedia functions, LLMs can make edits. They can be detected, but unless you are saying they are useless I don't know what point you are making about an LLM contribution versus a human. That LLMs aren't good enough to make meaningful contributions yet?? That Grok is specifically the problem?

I fully agree. Even assuming no forced ideological bias from Elon, I doubt it would be nearly as good. I still thought it could be an interesting concept, even if I had very low hopes from the start.

"Warts and all" says it all really. What are those warts? Who's responsibility are they?

Wikipedia is really not ideal for the LLM age where multiple perspectives can be rapidly generated. There are many topics where clusters of justified true beliefs and reasonable arguments may ALL be valid surrounding a certain topic. And no I am not talking about "flat earth" pages or other similar nonsense.

And the idea of it being controlled by any one entity makes it less interesting and less "good" when compared to Wikipedia

My expectations were extremely low, as were, and are, my expectations of Grok in general. Was just making an actual devil's advocate case.

I am not using hyperbole or speculating. I absolutely mean it.

"Biggest" is tough to quantify, but "most significant" and "most effective" is what I meant. I use Twitter way too many hours a day basically every day and have a morbid fixation on diving deep into right and far-right rabbit holes there. (Like, on thousands of occasions.)

Grok is without a doubt the single most important contributor to convincing believers of right-wing conspiracy theories that maybe the theories aren't as sound as they thought. I have seen this play out hundreds of times. Grok often serves as a kind of referee or tiebreaker in threads between right-wing conspiracy theorists and debunkers, and it typically sides overwhelmingly with the debunkers. (Or at least used to.) And it does it in a way that validates the conspiracy theorist's feelings, so it's less likely to trigger a psychological immune system response.

https://www.reddit.com/r/GROKvsMAGA/ contains some examples. These may seem cherry-picked, but they generally aren't. (Might need to look at some older posts now that Elon has put increasingly pressure on the Grok and Grokipedia developers to keep it """anti-woke""".)

When a right-wing conspiracy theorist sees some liberal or leftist call them out for their falsehoods, they respond with insults or otherwise dismiss or ignore it. When daddy Elon's Grok tells them - politely - that what they believe is complete horseshit, they react differently. They often respond to it 3 - 20 times, poking and prodding. Of course, most still come away from it convinced Grok is just compromised by the wokes/Jews/whatever. But some seem to actually eventually accept that, at the least, maybe they got some details wrong. It's a very fascinating sight. I almost never see that reaction when they argue with human interlocutors.

To be clear, it was never perfect. For example, if you word things in just the right way and ask leading questions, then like with any LLM (especially one that needs to respond in under 280 characters) you can often eventually coax it into saying something close to what you want. I have just seen many instances where it cuts through bullshit in a way that a leftist arguing with a Nazi can't really do.

I think there is a problem sometimes that "debunkers" are often more interested in scoring points with secondary audiences (i.e. people who already agree with them) than actually convincing the people who believe the misinformation.

Most people who believe bullshit were convinced by something. It might not have been fully rational but there is usually a kernel of something there that triggered that belief. They also probably have heard at least the surface level version of the oppising argument at some point before. Too many debunkers just reiterate the surface argument without engaging with whatever is convincing their opponent. Then when it doesn't land they complain their opponent is brainwashed. Which sometimes might even be true, but sometimes their argument just misses the point of why their opponent believes what they do.

One of the rallying cries of the right is "facts don't care about your feelings", but it's interesting how the facts either get distorted or ignored.

On one hand, yes, you're completely right.* On the other hand, there is an obligation for something or someone to do the job of pointing out the info is wrong, and how and why. Even if it makes most of them believe it even more strongly afterwards, it's still worse for it to go constantly unchallenged and for believers to never even come across the opposition.

*(The same is true of left-wing conspiracy theories. It's silly to pretend that right-wing conspiracy theorists aren't far more common and don't believe in, on average, far more delusional and obviously false conspiracy theories than left-wingers do, but it's important not to forget they exist. I have dealt with some. They're arguably worse in some ways since they tend to be more intelligent, and so are more able to come up with more plausible rationalizations to contort their minds into pretzels.)

The problem of debunking left-wing misinformation is that it doesn't seem to matter. The consumers of that misinformation want it and those of us who think it's bad for society already know that its garbage.

[flagged]

So by your logic, there's only room for one gatekept source of general purpose information on the web? No one besides entrenched progressive interests can do anything to Wikipedia. It's like reddit. We're tired of your guys' spaces. So sorry if this offends.

The ADL was caught in a campaign making edits. I remember more details in the past but I simply can't find them now with any search engine.

https://forward.com/news/467423/adl-may-have-violated-wikipe...

But also the ADL is accusing others of covert campaigns: https://wassermanschultz.house.gov/news/documentsingle.aspx?...

So I am sure this is a thing among corporations/NGOs. Note that I picked the ADL because I happened to know this and not because I am trying to make a point about the ADL's purpose. Also I am not really answering the part about progressives although the ADL is arguably a progressive NGO. I think there are astroturfing campaigns on Wikipedia whether progressive or not.

How many more than a couple do you need? 20 anecdotes? 40 anecdotes? 100? How much bias is okay for you and the world?

Although I'm sure it's been a blast, we don't need to play by your rules any more.

Crucial to distinguish between knowledge, fact, claim and allegation. Compare:

https://en.wikipedia.org/wiki/Charlie_Kirk#Assassination

https://grokipedia.com/page/Charlie_Kirk : Assassination Details and Investigation

This is an active case that has not gone to trial, and the alleged text messages and Discords have not had their forensics cross-examined. Yet Grokipedia is already citing them as fact, not allegation. (What is considered the correct neutral way to report on alleged facts in active cases?)

> The Wikipedia process is created to manage bias. LLMs are created to repeat the input data, and will therefore be quite biased towards the training data.

The perception of bias in Wikipedia remains, and if LLMs can detect and correct for bias, then Grokipedia seems at least a theoretical win.

I'm happy with at least a set of links for further research on a topic of interest.

That's effectively impossible to prove, especially if you don't believe statements made by the only organization that has access to the underlying evidence.

I actually think that it's funnier if it was an emergent behavior as opposed to a deliberate decision. And it fits my mental model of how weird LLMs are, so I think unintentional really is the more likely explanation.

It’s amazing the see the credulity of Elon stans. It’s the exact same reason grifting is so profitable among the right wing. It literally doesn’t matter how much evidence there is. If dear leader gives an excuse, they all believe and repeat the excuses. They are conditioned to it at this point. Any sources that refute their position is just leftist bias. This world fucking sucks.