[HN Gopher] Ibis: Federated Wikipedia alternative
       ___________________________________________________________________
        
       Ibis: Federated Wikipedia alternative
        
       Author : mostcallmeyt
       Score  : 124 points
       Date   : 2024-10-27 17:39 UTC (5 hours ago)
        
 (HTM) web link (ibis.wiki)
 (TXT) w3m dump (ibis.wiki)
        
       | schoen wrote:
       | It's super-cool to see an alternative approach to dealing with
       | Wikipedia content disputes.
       | 
       | In the past, when there have been Wikipedia forks, they haven't
       | generally tried to stay in sync with Wikipedia, at least not in
       | both directions. Do we have an example of long-term forks of
       | collaborative software or text editing projects that did manage
       | to keep sharing productively in multiple directions? Maybe the
       | BSDs to some extent?
       | 
       | I wonder how much work people are willing to do to keep actively
       | collaborating with people whom they have big ongoing
       | disagreements with (at least in areas where those disagreements
       | don't have an impact). Or can such collaboration be made
       | relatively seamless with appropriate tooling?
        
       | woodruffw wrote:
       | I don't have any particular trust (or distrust) in Wikipedia as
       | an institution, but this seems to be putting the cart before the
       | horse: is there a reason to believe that a federated wiki would
       | be _more_ accountable, rather than less?
       | 
       | (This isn't meant to be a jab; I like federated stuff. But I'm
       | also not sure the order of operations is right here.)
        
         | whatshisface wrote:
         | The answer to your question is basically the same as for
         | federation in real life: I don't care if they mayor of a city
         | 300 miles away is accountable to me. I only care if my mayor is
         | accountable to me. Making mayors into presidential appointees
         | would represent a slight increase in my influence in the city
         | code of places I don't know the name of, but that's not a great
         | trade for my influence in the city I live in.
        
           | SiempreViernes wrote:
           | Yeah, but the question here is how _your mayor_ being
           | accountable to you helps with the problem that _everyone 's
           | president_ is not being held accountable?
           | 
           | Exactly how does it help that there's a bunch of small
           | Wikipedia mirrors with like four different pages each that
           | differ from the main one.
        
             | whatshisface wrote:
             | Imagine if there was only one soft drink brand, or a single
             | giant book.
        
           | pessimizer wrote:
           | When you join a federated instance, the owner of it still
           | isn't accountable to you, and is likely to be more eccentric
           | than a centralized instance that is at least run by a larger
           | number of people.
           | 
           | Speaking of federation in real life reminds me of people
           | yelling about state's rights. I'm pro-distribution, but none
           | of these e-institutions are democratic, they're all little
           | fiefdoms. Governance needs to be distributed, not just
           | bandwidth.
        
         | __MatrixMan__ wrote:
         | I don't know about accountability on the part of the
         | participants, but in theory it ought to be harder to attack
         | because success means coordinating a consistent (maliciously-
         | altered) experience across infra maintained by a wider variety
         | of people, some of whom might be harder to coerce than others.
         | 
         | In practice, I don't know how much that would matter. If I were
         | the kind of powerful actor that federation is supposed to guard
         | against, I'd use DNS poisoning and crooked CA's/ISP's to work
         | at the network level rather than attempting to corrupt each
         | server admin separately.
         | 
         | So I see it as a good start, but really only meaningful if we
         | de-root-of-trust those things also.
        
           | woodruffw wrote:
           | Serious question: why are we talking about Wikipedia like it
           | has the same shape of trust problem as the Web PKI? With
           | Wikipedia, you or I or anybody else can go and edit out the
           | obvious misinformation. That leaves subtle misinformation,
           | but I don't think there's a technical solution to that.
           | 
           | It's already "hard" to attack Wikipedia, in the sense that
           | there's an army of pedantic dorks ready to argue about
           | _anything_ already on it. Which is a different kind of hard
           | than attacking a PKI.
        
       | mzajc wrote:
       | This points out several valid issues with Wikipedia, but does not
       | explain how the replacement addresses them.
       | 
       | Centralized moderation, for instance, is replaced by moderating
       | every instance separately. But doesn't that simply shift the
       | problem? The largest instance(s) can still moderate maliciously,
       | while the rest are insignificant.
       | 
       | Also, are there any plans to import existing articles from
       | Wikipedia? I find it hard to imagine an alternative gaining
       | traction by disregarding decades of edits on Wikipedia itself.
        
         | whatshisface wrote:
         | > _But doesn 't that simply shift the problem? The largest
         | instance(s) can still moderate maliciously, while the rest are
         | insignificant._
         | 
         | The effect of this on Mastodon was to establish walls between
         | the big servers over moderation disputes, controlling their
         | growth.
        
           | mzajc wrote:
           | I think this works much better for social media than it would
           | for an encyclopedia - while a Mastodon user I'm interested in
           | is likely on a single instance, a topic an encyclopedia would
           | cover might be present on many instances, each hosting a
           | different version. I don't think it's feasible to expect
           | users to read every single one.
        
             | whatshisface wrote:
             | That would only be likely to happen when communities
             | identify by their bias (see: "conservapedia.") Existing
             | wiki communities tend to divide based on topic, without a
             | lot of overlap. (See: fandom). A federated alternative to
             | fandom.com is what I would imagine when someone says
             | "federated wiki."
        
           | wkat4242 wrote:
           | The problem there was also that it attracted some pretty
           | seriously incompatible ideologies. Like neonazis on gab.com
           | and tankies on lemmy.ml for example. That was never going to
           | end well. How are they supposed to come to an agreement on
           | moderation :)
           | 
           | And then in the middle of that you have the German instances
           | that demand everyone else pledges to follow their peculiar
           | local laws precisely to the latter or get banned.
           | 
           | Fwiw I'm with none of the above groups, I'm pro
           | LGBT/progressive for which there's also a lot of instances
           | but I'm just sad that the fediverse didn't come out.
           | 
           | Having said that, some instances have really made a name for
           | themselves like beehive <3
        
             | whatshisface wrote:
             | They're not supposed to come to an agreement on moderation.
             | The whole idea behind federation is that a company is not
             | trying to navigate a middle ground between Nazis and
             | chinese communists! :-)
        
               | wkat4242 wrote:
               | Hmm yeah but the idea was not to duplicate moderation a
               | thousand times. And then where is your community?
               | 
               | My problem is that the fediverse is tearing itself apart
               | on tiny differences. Lots of instances have banned others
               | from the same ideology because one little time someone
               | said something one of the founders didn't agree with.
               | 
               | It's just not working like this. I think ideological
               | choices should be made by the end user.
               | 
               | I think the instance should be agnostic like an internet
               | browser. And moderation done by the groups the user
               | subscribes to. That's the model that works. We had
               | exactly this in Usenet which is just not really used
               | anymore due to complex technical implementation.
        
           | blackeyeblitzar wrote:
           | That's what I don't get about mastodon. Big servers use
           | commonly shared block lists for others that don't abide by
           | their censorship rules. Now we're back to something similar
           | to what existed anyways - big powerful centralized social
           | media separated from smaller ones that don't benefit from
           | networks.
        
         | dietr1ch wrote:
         | > Centralized moderation, for instance, is replaced by
         | moderating every instance separately. But doesn't that simply
         | shift the problem? The largest instance(s) can still moderate
         | maliciously, while the rest are insignificant.
         | 
         | I've always thought about modeling this as vote based system
         | votes support claims.
         | 
         | Like, "Israel's IDF is committing genocide in Palestine" would
         | be maybe supported by Wikipedia, but would have opposing votes
         | by the US and Israel, and it's up to each client to decide who
         | to trust. Non controversial claims would be generally
         | supported, and on controversial ones you could have them scored
         | given your prior trust in voters and have the UI mention it's
         | controversial and show the underlying votes if you want to dig
         | into that.
        
           | EGreg wrote:
           | I have a different approach -- rather than keeping people
           | apart in echo chambers, I force them all together in a
           | centralized site where they have to see the best arguments
           | for and against each claim
           | 
           | Building this to clean up online discourse:
           | https://rational.app
           | 
           | Anyone want to join me? Reply or find my email in my profile
        
           | mzajc wrote:
           | Wikipedia generally avoids resolving disputes democratically,
           | instead opting to establish consensus based on sources where
           | possible.
           | 
           | With voting systems you end up walking the tightrope between
           | making your system susceptible to Sybil attacks and
           | sockpuppetry on one end, and giving well-established users
           | undue voting rights on another.
        
             | iterance wrote:
             | Obligatory - Wikipedia is not a democracy - https://en.m.wi
             | kipedia.org/wiki/Wikipedia:What_Wikipedia_is_...
             | 
             | (Or the related / more in depth "Polling is not a
             | substitute for discussion" - https://en.m.wikipedia.org/wik
             | i/Wikipedia:Polling_is_not_a_s...)
        
             | mirekrusin wrote:
             | Sybil and sockpuppetry is solvable by "following" (possibly
             | truth-weighted, can nicely express negative weights). The
             | downside is that there is no global truth view, only
             | individual ones based on who you grant trustworthiness and
             | how it recursively expands - but arguably this reflects
             | reality better.
        
               | yorwba wrote:
               | "No global truth view" also makes discoverability
               | difficult. How do you know whom to follow? Just people
               | you already know from elsewhere?
               | 
               | Then if you don't know any subject-matter experts, will
               | your individual view of the wiki be devoid of subject-
               | matter expertise?
               | 
               | That arguably reflects reality better, but also isn't
               | terribly useful.
               | 
               | Edit: I guess the situation of an expert correcting an
               | article written by laypeople is similar to an expert
               | replying to a layperson's toot or whatever. So maybe you
               | could solve this by letting people see new edits made by
               | other people on top of their own edits, and giving them
               | the option to block or accept the change, or even follow
               | the user in question.
        
               | mirekrusin wrote:
               | It would be neutral which is current status quo, not that
               | bad?
               | 
               | The nice part is that it would work the other way around
               | as well - from article/subject back to people who endorse
               | or object it - in case you want to follow (attract/repel
               | their pov).
               | 
               | In reality who to follow should be easy as it would be
               | recursive - if you follow MIT and MIT follows ie. Noam
               | Chomsky - you wouldn't have to directly follow Noam as
               | your trust would already exist as transitive one from MIT
               | etc.
        
               | yorwba wrote:
               | The status quo is not neutral. It is extremely biased,
               | particularly against spam, vandalism and low-quality
               | content. Its defense against sybil attacks and
               | sockpuppetry is a single global truth, collectively
               | patrolled by all users.
               | 
               | If you tried to do a neutral wiki anyone can edit, it
               | would be full of ads and porn.
        
               | mirekrusin wrote:
               | It's neutral in the sense that it's not trust-weight-
               | filtered by observer.
               | 
               | It can be created yes, just like any web page can be
               | created, but won't surface for anybody because nobody
               | will endorse it.
               | 
               | If it's endorsed by sybil accounts - it'll be visible to
               | sybils and their followers - ie. nobody real.
               | 
               | To rephrase it - following somebody means it alters your
               | view and nothing else - it's uni-directional (but
               | recursive).
        
             | blackeyeblitzar wrote:
             | Yea but the list of sources accepted is itself
             | editorialization. And furthermore the articles are
             | unbalanced and incomplete on anything controversial,
             | choosing to quote from sources supporting the view espoused
             | and ignoring others.
        
           | mirekrusin wrote:
           | Exactly, votes + who you trust-follow (ie. "MIT" with high
           | degree, "Flat Earth Society" with high negative degree) -
           | recursively; creates your own landscape of truth-weighted
           | information.
        
           | drdaeman wrote:
           | This requires web-of-trust/friend-to-friend or similar
           | system, and despite numerous attempts none of such things had
           | ever succeeded so far.
        
           | reshlo wrote:
           | That model doesn't work because there are plenty of issues on
           | which a vote by the US would increase my confidence in the
           | truthfulness of a claim, but others where it would not make a
           | difference, or even decrease my confidence.
        
         | jonstaab wrote:
         | For a more relativistic approach, built on nostr, see
         | wikifreedia.xyz, which allows everyone to maintain their own
         | version of any article, fork, vote, and merge as needed.
        
           | labster wrote:
           | So move the moderation decisions onto the end user? Somehow
           | this doesn't feel like progress.
        
           | FireInsight wrote:
           | Seems a bit bloated. I opened the front page and could only
           | see a list of recently edited pages, with the count
           | constantly ticking up. Trying to open any of them by clicking
           | them didn't do anything, seems like the page was ground to a
           | halt. Tried to search 'wikifreedia' to see if there's a wiki
           | page explaining what the project is about, but could not get
           | the search to go through.
        
         | Levitating wrote:
         | I am also confused how this project improves upon Wikipedia's
         | moderation issues. Especially calling Ibis "federated", is it
         | more "federated" than Wikipedia is?
         | 
         | It's also not like Wikipedia is unusable because of these
         | issues. As with any source of information it is necessary to be
         | skeptical with wikipedia articles. Wikipedia has the major
         | advantage over any other source that the edit history is
         | public.
         | 
         | And even though there are many incidents with Wikipedia's
         | moderation, the website has also existed for a long time now
         | and contains millions of articles. With the scope and size of
         | Wikipedia some incidents are expected.
         | 
         | Wikipedia also compiles their own backup collections.[1] If at
         | any point Wikipedia becomes actually unusable anyone could use
         | the existing article to produce a new alternative database.
         | 
         | [1]: https://en.wikipedia.org/wiki/Wikipedia:Database_download
        
       | iterance wrote:
       | Federation of a wiki is... well, it's a bit strange, isn't it?
       | Imagine if only one library at a time held a copy of a given book
       | you need, and the only way to access that book was via an inter-
       | library loan via library partnerships, or by visiting the library
       | that has the book. This is assuredly the situation many academics
       | are in for key reference texts, but it is not what I'd call
       | ideal. It is in fact very fragile at times. There is a book I
       | need right now, for example, but if I want to read it I will have
       | to drive to see it... The library that has it won't send it for
       | ILL. In a physical sense they "declined to federate."
       | 
       | The difference between a wiki and a social media network is that
       | anyone can spin up a template social media site; the fundamental
       | _user-side_ barrier to entry is pretty small. The same is not
       | true of wikis - at least not high quality ones. Documentary
       | standards, tone, quality, reviewership, consistency, policy,
       | moderation, accountability, leadership, thoroughness, these are
       | all qualities that take time and commitment to develop. They are
       | hallmarks of centralization for a reason: arguably the innovation
       | of human governance is centered around qualities like these. They
       | take a long time to develop.
       | 
       | As a counterpart to Wikipedia, well... fragmentation is often a
       | death knell for efficient knowledge transfer. We are already
       | losing massive swathes of our early Internet history due to
       | fragmentation, attrition, and destruction. The thought that any
       | piece of knowledge stored in a safehouse could go offline at
       | once, without replication or warning, it scares me a bit. The
       | thought that we don't really know who we're trusting as stewards
       | of human knowledge in a federated model disturbs me too. You can
       | have your issues with Wikipedia but at least you know who they
       | are. You know their biases.
       | 
       | That's not to say there aren't use cases for this... but man,
       | this seems like an easy way to lose or destroy important parts of
       | our shared history on accident.
        
       | exitb wrote:
       | I get the concerns towards Wikipedia governance, but this feels
       | almost like slacktivism. It's been done numerous times, to slap
       | ActivityPub onto a communication mode and call it a ,,federated
       | foobar alternative". It never works, it always fails to gain
       | meaningful adoption. The federated hosts end up in disputes
       | between themselves and their users. It's mostly a miserable
       | experience.
       | 
       | There needs to be another way to increase accountability for
       | critical online services.
        
         | bbor wrote:
         | Hmm when has it happened in the past? This is literally the
         | Lemmy creator so it seems somewhat unique! But I could
         | definitely just be behind the times.
         | 
         | Are there other federation projects other than mastodon and
         | Lemmy?
        
           | Vivtek wrote:
           | There are one or two.
           | 
           | https://github.com/shleeable/Big-List-of-ActivityPub
        
             | rapnie wrote:
             | Also https://delightful.club/delightful-fediverse-apps/
        
         | input_sh wrote:
         | If by "never works" you mean "someone's side project whose
         | monetization strategy is to put up a donation link never takes
         | down an established business worth hundreds of millions in a
         | fraction of the time it took for such business to establish
         | itself as some sort of a monopoly", then yeah, no shit.
        
       | foxhop wrote:
       | If the maintainers are reading this is cool. I would start with
       | an exporter/importer for pages from Wikipedia format into
       | whatever format you use, it should also deal with the media. This
       | is no small task but what you need.
        
         | kemayo wrote:
         | It's legitimately a pain to do, because most Wikipedia articles
         | use a bunch of templates. So for anything outside the most-
         | simple of articles you suddenly have to implement a lot of the
         | more painful bits of wikitext. Then if you want to import them
         | _back_ to Wikipedia, you 've got to match those things back up
         | or you'll get reverted for messing up the page...
         | 
         | Plus complying with CC BY-SA requirements for the content, of
         | course.
        
       | jll29 wrote:
       | It's legitimate to create alternatives to anything, and to re-
       | think what has been done already.
       | 
       | One thing that would be on my personal wish list for any
       | Wikipedia alternative is ease of machine processing: the
       | MediaWiki format/mark-up and the templates are horribly
       | inconsistent and a nightmare to parse. This should be done better
       | by any serious successor. Wikipedia has got the excuse
       | "historically grown", any successor doesn't.
        
         | richardw wrote:
         | That seems to be the same goal for
         | https://www.wikidata.org/wiki/Wikidata:Main_Page
         | 
         | Do you mean something different, e.g. not just structured data?
        
           | kemayo wrote:
           | For the parser itself, there is work being done on something
           | that's more structured. The new parser lets you round-trip
           | from the wikitext to the generated HTML and back, which means
           | that it's entirely possible to work with wiki content by just
           | manipulating a DOM rather than having to work with the
           | wikitext yourself.
           | 
           | https://www.mediawiki.org/wiki/Parsoid/Parser_Unification
        
         | kemayo wrote:
         | The problem for any "Wikipedia alternative" is that there's one
         | obvious way to bootstrap themselves: importing Wikipedia's
         | content. But that leaves them either needing to work out how to
         | impose structure on it themselves (difficult), or just
         | importing it as-is and leaving it messy.
        
       | richardw wrote:
       | If the top "scandals" in the article are considered reasons to
       | change, just wait until the alt-crew get a hold of wikipedia.
       | Medical, political, science, education, religion discussions
       | would all be even more crazy when people can point to some wiki
       | gospel as proof. We'd find the list of scandals grows a
       | thousandfold.
       | 
       | Wikipedia is amazing, if imperfect. Make it better, don't try to
       | break it by splintering it into some weird "whatever you think is
       | right" solution. That's already what we have in social media.
       | Truth is already federated enough.
       | 
       | Edit: what would convince me more is some demonstration
       | article/cluster of articles and maybe some math proof that shows
       | how this is better, through faster convergence on truth, defence
       | against rando's etc. I think even getting a definition of
       | "better" would be a bit of a battle!
        
       | highwayman47 wrote:
       | What does this term "federated" being thrown around in tech
       | lately mean?
        
         | rrgok wrote:
         | Good question. I too don't understand it.
        
         | fragmede wrote:
         | Meaning it's like email. I can have my own domain name -
         | @example.com , not just Gmail, and they talk to each other so
         | people on Gmail can send email outside of Gmail, and people can
         | send emails into Gmail.
        
         | poszlem wrote:
         | It means that each federated technology creates a federation of
         | "self-governing" entities, for example Mastodon servers. The
         | meaning is very similar to the political meaning of the word.
        
       | somat wrote:
       | I don't intend to be mean but isn't the federated wikipedia, the
       | web.
       | 
       | Wikipedias whole value function, don't get me wrong it's a great
       | value function, is that it is a curated centralized web.
        
       | sureglymop wrote:
       | ActivityPub federation, while maybe a little complicated, seems
       | to be well thought through.
       | 
       | I think many platforms would benefit from implementing that.
       | Especially git forges, though I think _someone_ is already
       | working on that.
        
       | Levitating wrote:
       | How ironic that the link to the mentioned article by Helen
       | Buyniski "Wikipedia: Rotten to the Core" is the victim of link
       | rot.
       | 
       | You can find the article reposted here:
       | 
       | https://helenofdestroy.substack.com/p/49-wikipedia-rotten-to...
        
         | timmytokyo wrote:
         | Fascinating that the first person cited on this site's
         | announcement page is Helen Buyniski, a "journalist" for Russian
         | state media and a COVID conspiracy theorist. It's almost as if
         | this project has an agenda.
        
       | anthk wrote:
       | The only wiki I like it's the rational wiki.
        
       | pona-a wrote:
       | See previous discussion
       | 
       | https://news.ycombinator.com/item?id=39694045
       | 
       | 163 points | 252 comments
        
       | kemayo wrote:
       | It'd be more convincing if the "problems with Wikipedia" section
       | referenced anything from the last decade. (Also, one of the three
       | things that apparently warranted being called out as an example
       | was pretty petty -- the ArbCom member's on-wiki job didn't have
       | anything to do with how they misrepresented their credentials.)
        
       | nightpool wrote:
       | I have the same problem with this that I do with "federated
       | forums" or "federated forges"--what does it even mean for an
       | encyclopedia to be federated? In fact, they already are--if I
       | want to go look up information about Minecraft, I can use the
       | federated internet to go to minecraft.wiki.
       | 
       | So what are these new "federation" features actually giving us?
       | Is it just OpenID login reinvented?
        
       | hiAndrewQuinn wrote:
       | This alternative doesn't load with NoScript on, so I think I'll
       | pass.
        
       | fny wrote:
       | Social problems demand social solutions.
       | 
       | If you're skeptical about Wikipedia, you can easily create your
       | own fork of Wikipedia: the data and code is open source after
       | all. In the end, it's all about whether you can keep a community
       | alive and kicking. No one gives a damn about whether a wiki is
       | built on ActivityPub.
       | 
       | There are many examples: Larry Sanger, who was mentioned in the
       | article, created Citizendium after breaking from Wikipedia.[0] He
       | was then involved with Everipedia, a for-profit venture built the
       | original code base which later morphed into crypto nonsense.[1]
       | 
       | There are many examples of other wikis too. Some are
       | focused[2][3], some are fun[4][5], some are revisionist[6][7][8],
       | and some meet the requirements of totalitarian regimes[9][10].
       | 
       | If Wikipedia's not your style, grab the code, rally a crowd, and
       | make the encyclopedia you want to see--just know it's the people,
       | not the platform, that make it thrive.
       | 
       | [0]: https://ballotpedia.org/Main_Page
       | 
       | [1]: https://rosettacode.org/
       | 
       | [2]: https://citizendium.org/
       | 
       | [3]: https://iq.wiki/
       | 
       | [4]: https://uncyclopedia.com/wiki/Main_Page
       | 
       | [5]: https://edramatica.com/Main_Page
       | 
       | [6]: https://www.metapedia.org/
       | 
       | [7]: https://www.conservapedia.com/
       | 
       | [8]: https://infogalactic.com/info/Main_Page
       | 
       | [9]: https://www.qiuwenbaike.cn
       | 
       | [10]: https://ruwiki.ru
        
         | f33d5173 wrote:
         | A social problem is by definition one with a social solution.
         | If we can find a technical solution to a problem, it turns into
         | a technical problem. Wikipedia is itself a technology. It's not
         | outrageous to suggest that new technologies can improve on it.
        
       | blackeyeblitzar wrote:
       | How does this overcome power struggles and bias? That's the most
       | important question for me, as bias in Wikipedia is rampant these
       | days, and even acknowledged by its founder.
        
       | ErrantX wrote:
       | As with all Wikipedia competitors, this misses two key points
       | 
       | Firstly the most critical part of competing with WP is not the
       | technology, it's having the critical mass of people willing to
       | write article. Scandals aside over decades hundreds of thousands
       | of people have built out the content and continue to do so - that
       | is not easy to emulate.
       | 
       | Also for all the scandals, the toughest problems in WP early days
       | were spammers and trolls. Moderation is a niche community problem
       | (which till you have a community is moot). Stopping bot armies
       | and countless trolls is a day 1 issue.
        
         | aljgz wrote:
         | Out of genuine curiosity, what are the scandals?
        
           | GuB-42 wrote:
           | https://en.wikipedia.org/wiki/List_of_Wikipedia_controversie.
           | ..
           | 
           | One good thing about Wikipedia is that it is pretty open with
           | regards to internal affairs. If you want to criticize
           | Wikipedia, there is no better source than Wikipedia itself.
        
           | ErrantX wrote:
           | As always, there is an article about it :)
           | 
           | https://en.m.wikipedia.org/wiki/List_of_Wikipedia_controvers.
           | ..
        
         | ComplexSystems wrote:
         | > Firstly the most critical part of competing with WP is not
         | the technology, it's having the critical mass of people willing
         | to write article.
         | 
         | This is much less of an issue now that there are LLMs, even
         | with the occasional hallucination.
        
           | lionkor wrote:
           | LLMs do not write information dense content
        
           | ErrantX wrote:
           | Actually writing the content is usually the easiest part.
           | 
           | 98% of writing these articles is research. Current LLMs are
           | nowhere near good enough.
        
           | harrisi wrote:
           | One must always assume anything an LLM outputs is false until
           | proven otherwise, not the opposite.
        
             | birriel wrote:
             | Ok, but that criterion is true for human-generated output
             | as well.
        
               | fragmede wrote:
               | False. Please prove otherwise.
        
               | harrisi wrote:
               | If you truly never believe anything anyone has ever told
               | you until you prove it yourself, I'm amazed you've
               | somehow learned to type in English on a website with
               | accounts.
               | 
               | We trust people all the time for many reasons. Authority,
               | past experience, logic (why would they lie about
               | something in context), etc. and we get by alright.
               | Knowing it's another conscious, mortal human existing in
               | society makes it much easier to know when someone is
               | likely being truthful.
               | 
               | Obviously there are all sorts of caveats to that. The
               | main difference is that LLMs don't lie. They don't tell
               | the truth either. They just generate stuff with no
               | meaning. There's no way to ever put any trust in that,
               | the same way I wouldn't trust my ice maker to make sure
               | my dog gets enough water while I'm on vacation.
        
       | openrisk wrote:
       | Some sort of federation has already been discussed in the
       | Wikimedia universe [1]. It aims to operate at a lower level,
       | federating _structured_ data between different specialized
       | wikibases [2]. AFAIK it has not (yet?) seen much development and
       | /or traction.
       | 
       | ActivityPub style federation opens up some interesting technical
       | possibilities for exchanging data but if the main motivation is
       | to "fix" the consensus/moderation processes of Wikipedia one
       | should start with outlining how _that_ would work.
       | 
       | [1] https://www.mediawiki.org/wiki/Wikibase/Federation
       | 
       | [2] https://wikiba.se/
        
       | groceryheist wrote:
       | This is the latest in a long line of attempts to make a non-
       | centralized Wikipedia. A Wikipedia (not me) maintains a record
       | 
       | https://en.wikipedia.org/wiki/User%3AHaeB%2FTimeline_of_dist...
        
       | zozbot234 wrote:
       | > This means the same topic can be treated in completely
       | different ways. For example geology.wiki/article/Mountain may be
       | completely different different from poetry.wiki/article/Mountain.
       | 
       | Doesn't MediaWiki (and other common wikis) support this kind of
       | thing already via the InterWiki-links system? You can set up a
       | MediaWiki install such that [[geology:Mountain]] and
       | [[poetry:Mountain]] go to the appropriate places. What does Ibis
       | add that's not accounted for by this?
        
       ___________________________________________________________________
       (page generated 2024-10-27 23:00 UTC)