[jameslk]

It was just launched? I remember when Wikipedia was pretty useless early on. The concept of using an LLM to take a ton of information and distill it down into encyclopedia form seems promising with iteration and refinement. If they add in an editor step to clean things up, that would likely help a lot (not sure if maybe they already do this)

[9dev]

Nothing about that seems promising! The one single thing you want from an Encyclopedia is compressing factual information into high-density overviews. You need to be able to trust the article to be faithful to its sources. Wikipedia mods are super anal about that, and for good reason! 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. On Wikipedia, at least there’s lots of people checking on each other. There are no such guardrails for an LLM. 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.

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.

[jameslk]

> 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

[LexiMax]

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

If I want an AI summary of a Wikipedia article, I can just ask an AI and cut out the middle-man.

Not only that, once I've asked the AI to do so, I can do things like ask follow-up questions or ask it to expand on a particular detail. That's something you can't do with the copy-pasted output of an AI.

[jameslk]

The good news is that you don’t have to use it. I see ways this idea can be improved, some of which I already mentioned in this thread. It just launched recently so judging solely by what it is today is missing the point

[madeofpalk]

It’s just a different class of problem.

Human editors making mistakes is more tractable than an LLM making a literally random guess (what’s the temperature for these articles?) at what to include?

[jameslk]

I recall a similar argument made about why encyclopedias written by paid academics and experts were better than some randos editing Wikipedia. They’re probably still right about that but Wikipedia won for reasons beyond purely being another encyclopedia. And it didn’t turn out too bad as an encyclopedia either

[xg15]

Yeah, but that act of "winning" was only possible because Wikipedia raised its own standard by a lot and reined in the randos - by insisting on citing reliable sources, no original research, setting up a whole system of moderators and governance to determine what even counts as a "reliable source" etc.

[rsynnott]

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

I would say more that it’s one of the biggest illusory values they think they are getting. An incorrect summary is worse than useless, and LLMs are very bad at ‘summarising’.

[mixedump]

> 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.

[f33d5173]

It really isn't a promising idea at all. Llms arem't "there" yet with respect to this sort of thing. Having an editor is totally infeasible, at that point you might as well have the humans write the articles.

[jameslk]

> 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?

[jerf]

For the same reason you don't modify autogenerated files in your source code base. It's easy to get an LLM to just regen the page but once someone tries to edit it you're even farther down the road of what an LLM can't do right now. I wouldn't even trust it to follow one edit instruction, at scale, at that size of document, and if we're going to have humans trying to make multiple edits while the LLM is folding in its own improvements... yeah, the LLMs aren't even remotely ready for that at this point.

[jameslk]

That’s a good point. I think it’s a similar problem of why you wouldn’t let a model go wild in your codebase though. If good solutions to how we handle AI models making code changes are found, it seems reasonable to expect they also may be applicable here

[drysart]

There's a significant difference between a site being useless because it just doesn't have the breadth yet to cover the topic you're looking for (as in early Wikipedia); versus a site being useless by not actually having facts about the topic you're looking for, yet spouting out authoritative-looking nonsense anyway.

[jameslk]

> 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

[LexiMax]

Wikipedia early on wasn't competing against Wikipedia, it was competing against hardcover encyclopedias. There was clear value-add from being able to draw from a wider range of human expertise and update on a quicker cadence.

In a world where Wikipedia already exists, there's no similar value-add to Grokipedia. Not only is it useless today, there is nothing about the fundamental design of the site that would lead me to believe that it has any path to being more authoritative or accurate than Wikipedia in the future - ever.