Wikipedia talk:Wikipedia Signpost/Single/2023-10-03

Comments

The following is an automatically-generated compilation of all talk pages for the Signpost issue dated 2023-10-03. For general Signpost discussion, see Wikipedia talk:Signpost.

i would be lying if i didn't say this is really silly, but i would also be lying if i didn't say that i am very interested. ltbdl (talk) 01:17, 4 October 2023 (UTC)

The turbo-encabulator is a technological wonder, but when will Wikipedia dare to expose the Alemeda-Weehawken Burrito Tunnel? Jim.henderson (talk) 00:44, 6 October 2023 (UTC)
Eagerly antler-cipating the next instalment. ~~ AirshipJungleman29 (talk) 11:40, 6 October 2023 (UTC)
Lol. Presumably Octopodan Earth has an analogous policy on tentacles. Thanks for another fun Concept article. Quercus solaris (talk) 00:06, 7 October 2023 (UTC)

...should I be disappointed that WP:NORACK is not among the shortcuts listed Rotideypoc41352 (talk · contribs) 20:36, 7 October 2023 (UTC)

Kudos to the signpost and the wielders of this new device for such pointed research. If possible could we use this method to find out how communities in Cthulhu led parallel worlds run their policies on editing whilst sane? ϢereSpielChequers 08:58, 12 October 2023 (UTC)
Nice. You have Bazingerzones here... whereas Wikivoyage uses QANZATS (wikivoyage:Wikivoyage:Joke_articles/Interdimensional_travel#Destination_classification), Maybe Wikidata could eventually store the translation between the two systems? ShakespeareFan00 (talk) 17:42, 15 October 2023 (UTC)

Featured content: By your logic, (0 bytes · 💬)

Wikipedia talk:Wikipedia Signpost/2023-10-03/Featured content

Netanyahu and Musk

The Atlantic had a really interesting article on this meeting the other day, focusing on a different part of the proceedings: the part where Netanyahu and Musk discussed the societal impact of AI.

The article is called The Unlikely World Leader Who Just Dispelled Musk’s Utopian AI Dreams (archive). Recommended. Andreas JN466 22:27, 3 October 2023 (UTC)

Why unlikely? Netanyahu has undergrad and graduate degrees from MIT and is certainly qualified to speak about societal impacts of technology. ☆ Bri (talk) 22:48, 3 October 2023 (UTC)
It has nothing to do with his qualifications but with the fact that Netanyahu took an almost socialist view in the discussion. Here is a relevant passage: The Israeli leader interrogated OpenAI’s Brockman about the impact of his company’s creations on the job market. By replacing more and more workers, Netanyahu argued, AI threatens to “cannibalize a lot more jobs than you create,” leaving many people adrift and unable to contribute to the economy. When Brockman suggested that AI could usher in a world where people would not have to work, Netanyahu countered that the benefits of the technology were unlikely to accrue to most people, because the data, computational power, and engineering talent required for AI are concentrated in a few countries. “You have these trillion-dollar [AI] companies that are produced overnight, and they concentrate enormous wealth and power with a smaller and smaller number of people,” the Israeli leader said, noting that even a free-market evangelist like himself was unsettled by such monopolization. “That will create a bigger and bigger distance between the haves and the have-nots, and that’s another thing that causes tremendous instability in our world. And I don’t know if you have an idea of how you overcome that. There is more in that vein in the discussion. (See also this issue's Recent Research on ChatGPT as a Wikipedia competitor potentially cannibalizing our unpaid jobs.) Andreas JN466 23:19, 3 October 2023 (UTC)
  • Elon Musk may or may not have an interest in assisting Wikipedia with a donation, but may find it within his dreams for humanity to further build Commons and related projects. Being creative with requests and meetings, Musk and WMF might just find much to collaborate on and plan. Randy Kryn (talk) 00:19, 4 October 2023 (UTC)
    And Elon Musk, in case someone points out these comments to you, please realize that you can edit Wikipedia (if you haven't already) to experience the process and improve the encyclopedia. Maybe only make one or three short edits, on points or topics that you care about and believe either need correction, improvement, or readability. Then, if you do enough edits and are inevitably reverted, assume good faith and discuss your position on the article's talk page. That should make that particular talk page sightseeable and educate some of the public about the interesting back-room discussions on Wikipedia. If you make just one edit you will be both a Wikipedian as well as an articulate critic of Wikipedia. And improve the encyclopedia while you're at it. Randy Kryn (talk) 02:42, 4 October 2023 (UTC)
Thanks, good information. Is the two million annual or a one-time donation? What I'm talking of are the kind of donations that are world-changing. Two million to the endowment or to the WMF operating budget is a fine and appreciated start. The importance of Wikipedia to humanity, combined with Musk's sincere intent to improve societal advancement in many fields, hopefully will someday lead to the kind of partnering that 23rd century historians will take note of when writing the definitive history of the project. Randy Kryn (talk) 11:07, 4 October 2023 (UTC)
@Randy Kryn: The way I read the page, the Musk Foundation has given the Endowment grants totalling between 2 and 5 million dollars. (There is a reference to two 1-million dollar gifts from the Musk Foundation in m:Fundraising/2019-20_Report.) Andreas JN466 06:55, 6 October 2023 (UTC)
Respectable, appreciated, helpful, but not spectacular. I want to see Elon Musk WMFoundation spectacular. Randy Kryn (talk) 11:04, 7 October 2023 (UTC)
Donations of tens of millions to the WMF don't seem to be enough to fix critical infrastructure. A donation of 1,000 hours of skilled volunteer labour is world-changing, at least when a few thousand of us do it. — Bilorv (talk) 20:29, 8 October 2023 (UTC)
Each does what they can. I would think WMF letting Wikipedians and other project members decide where 1/3rd of those tens of millions goes yearly would allow some interesting projects to play out. Short descriptor: dance with who brung ya. Randy Kryn (talk) 22:29, 8 October 2023 (UTC)
  • Musk and Netanyahu, like all right-wing authoritarians, distrust and fear a decentralized, democratic digital encyclopedia because it might educate the populace, and undermine their power base. Such individuals maintain power only through a strict control of information. Andre🚐 03:34, 4 October 2023 (UTC)
    like all right-wing authoritarians Are you suggesting an edit to Elon Musk's biography? CurryCity (talk) 08:16, 5 October 2023 (UTC)
    Eh, Musks politics are so mercurial you can say just about anything and find a source from some era which backs it up. For what its worth I think Musk is very fond of wikipedia, but not as fond of wikipedia as he is of winning (which he is by all accounts incredibly driven by, one of the most driven people on the planet). All in all Musks biggest gripe with us appears to be over the whole Tesla "co-founder" thing which while I understand why it would be massively significant to him its not exactly an indictment of wikipedia as a platform. Horse Eye's Back (talk) 19:51, 5 October 2023 (UTC)

vile edits

I agree that the vandalism could better be described as "sophomoric," but it's not appropriate for a journalist to interject that opinion. Better practice would be to find an editor who feels that way, and then quote that source. I'm sure it wouldn't have been difficult, and it would have reflected journalistic standards.~TPW 13:31, 4 October 2023 (UTC)

I did get one editor opinion in there - that "I'm baffled as to how one IP editor's childish vandalism could be considered remotely newsworthy" from WindTempos. Beyond that I've come to view this column as more of a journalism review than a straight recitation of the facts, which would sound something like a plot summary. Pretty boring IMHO with this subject matter, and looking back over how it was done in the past almost all the authors have done some interpretation of what's going on with the stories we present here. If I were to write a section here about my story on the Express (me writing a story about my own story), I might write that the reporter (me) might have just as well left out the story - as the saying goes "There's no use getting into a pissing contest with a pisser." But some things really suggested to me that the Express reporter must have known that his story was unfair. For example: What was he doing minutes after the actor's death looking at a Wikipedia article (for material on the actor's death?). And how did he just happen to arrive during the right 1.5 minutes when he could have seen it? Or did he go trawling through the edit history for say 20 minutes until he found something like what he was looking for? But why would they want to do that? Well, I've seen some cases (not to say that it was in this case) where journalist (or involved parties) actually make (or encourage others to make) the edits themselves. There was a story like that a few years ago where the Detroit Tigers put up a suggestion on the big scoreboard during a game to vandalize Wikipedia. That type of stuff just strikes a nerve with me. So it's better just to be straightforward about it and say what I think. Sorry for the rant! Smallbones(smalltalk) 16:16, 4 October 2023 (UTC)
@True Pagan Warrior: See "Gobbler of the month" at Wikipedia:Wikipedia Signpost/2019-05-31/In the media. Smallbones(smalltalk) 16:26, 4 October 2023 (UTC)
The rant helps me understand your process, and I certainly support the goal. As a practicing journalist, I'll have to disagree with you about the tactics, but I suppose it's really a question of whether one should consider the Signpost to be journalism or not. If it's just a newsletter, then you're 100% in the right. If not, I'll give you 85%. ~TPW 16:33, 4 October 2023 (UTC)
For what it's worth, I find the the "gobbler of the month" raises no similar questions from me, because there's no sense that opinion is injected by the reporter. ~TPW 16:35, 4 October 2023 (UTC)
I scan the news for interesting Wiki-relevant things for pretty much every Signpost issue. There's usually a smattering of "so-and-so's article got vandalized, isn't that fun". We don't usually run any of these. There's a pretty high bar in the for why it would interest the community, and sometimes they get discussed beforehand in the SP newsroom like this. ☆ Bri (talk) 18:52, 4 October 2023 (UTC)
Thanks @Bri and True Pagan Warrior: Bri certainly does add a lot to this column, which includes warnings to me on a fairly regular basis not to get sucked into these types of articles. Part of our objective here is to let our readers know what the outside world is saying about us. Part of that is some very ignorant blather such as the Express story above. So I'll get caught on this type of thing a bit too often, but it does serve our goal. Our readers should not be surprised when they see that type of story in the press. But I do try to limit the coverage of these nonsense stories.
TPW: I'm glad you brought up journalistic standards. We don't dicuss this enough. So I'll include a link here to the SPJ Code of Ethics and note that they are written more like enWiki's Guidelines than our Policies. Common sense needs to be applied. Feel free to comment here or via email anytime on The Signpost's standards or any ethic lapses. Everybody needs a reminder every now and then. I have thought a long time about adding a notice here about being a journalism review rather than straight news. Dear editor-in-chief @JPxG: what do you think? Smallbones(smalltalk) 22:09, 4 October 2023 (UTC)
If this were marked as a review, then yes I think that would do it, but why not just find a source and leave the writer's opinion out of everything that isn't specifically an editorial? I don't think it's inappropriate to cover this issue, because it surely is of interest to the community. I'm just not clear why it's difficult to cover it in a journalistic fashion, particularly by sourcing every opinion to someone other than the writer of the article. ~TPW 13:19, 5 October 2023 (UTC)

On Nellie Biles and checking Wikipedia

A while ago my mother couldn't remember what year she and my late father got married. I couldn't remember either, so I checked my father's article, which told us it's 1991. She said that sounds right.

It did make me wonder how many people there are who do that, and whether there are any cases of citogenesis into someone's own narrative of their life. -- Tamzin[cetacean needed] (she|they|xe) 06:35, 6 October 2023 (UTC)

Silly games

I submit that any effort to determine who can go the fastest from "Jimmy Wales" to "Stroopwafel" must certainly involve Drmies, a well-known stroopwafel expert. Cullen328 (talk) 07:09, 7 October 2023 (UTC)

If I understand correctly, the claim there is that during the investment time period, the S&P 500 went up 13%, while the Wikimedia Endowment through Tides investment went down by 8%. Is that the claim? If that is the claim, does anyone have enough financial expertise to know what is normal / safe / expected for a nonprofit investment strategy? Bluerasberry (talk) 23:02, 3 October 2023 (UTC)
No, here are the salient parts:
Fiscal year ending June 2017 2018 2019 2020 2021 2022 2023 Mean return
WMF Endowment ESTIMATED* return on investment 7.1% -8.0% 18.7% 11.6% 14.7% -17.9% 9.4% 5.1%
Median Nat'l Assoc. of College and University Business Officers endowment returns 12.5% 8.0% 5.1% 1.8% 30.1% -8.7% TBD 8.1%
S&P 500 net total return 17.1% 13.7% 9.8% 6.9% 40.1% -11.0% 19.0% 13.7%
 *assuming half of all contributions were realized on the first day of the fiscal year, and the other half on the last day.
I don't think it's outside typical values, but I'm no institutional investment expert. Sandizer (talk) 00:39, 4 October 2023 (UTC)
  • 22 million lost in 2022? Why not just put the endowment in the bank and stop gambling on the markets? As I've said elsewhere, am totally in favor of WMF having as much money on hand and "banked" as possible, and hopefully quite a few billionaires will donate 25 to 100 million each. Hopefully they will also stipulate, as the WMF ought to be doing on its own, that projects put forward by the Wikipedia community, Commons, and the rest of the volunteer core that the public actually believes that they are funding, will be given, and decide among themselves, how to spend 1/3rd of all donations. Randy Kryn (talk) 23:23, 3 October 2023 (UTC)
    They did not actually lose 22 million. Its misleading to think of an unrealized loss in money that is invested in the long term as being the same as actually losing money. Bawolff (talk) 15:40, 4 October 2023 (UTC)
    In the short term, "gambling on the markets" is a bad idea. For long term investments such as this, it's the prudent choice; financial markets have a long history of rising value, even if occasionally impacted by recessions and crises, whereas cash in the bank will lose value over time due to inflation. — The Anome (talk) 08:04, 5 October 2023 (UTC)
    Generally, things like index funds have had a very good and consistent rate of return, because they represent the whole stock market, which in turn represents the whole economy (or at least the whole economy as can be measured and quantified by finance). Of course, there are incidents (1929, 2008, etc) in which every single stock eats it at the same time, but in these cases, the economy as a whole also eats it; in 1929, for example, I don't think there's any place you could have put money where it wouldn't have been substantially affected by the Depression. I believe, although I'm pulling this out of my hat, that the historic rate of return on index fund investment has been somewhere around five to seven percent. Meanwhile, average year-over-year inflation in the US has been somewhere around three to four percent -- if you just put a bunch of cash into a savings account that pays 1% interest you're actually losing money in the long run because the value gets eaten by inflation. The only real way to offset this is by investing it in something with low variange and average RoI, hence throw it in index funds and forget about it. This part, at least, makes sense to me. I don't know why the Endowment didn't do this, but then again, I'm not a finance guy, so I won't presume to know how this works or what a good idea for investing this amount of money looks like. jp×g 18:59, 5 October 2023 (UTC)
    Jimbo has provided some further information on his talk page, based on his conversation with the Endowment President: [1] Andreas JN466 19:25, 5 October 2023 (UTC)

Poetry: "The Sight" (1,654 bytes · 💬)

I think there is something worthwhile being expressed here in Tamzin's poem about content work vs. drama boards, and human nature seen through the prism of Wikipedia collaboration.

It's worth noting that Wikipedia can easily lead to an unduly jaundiced view of human nature. Sometimes, it magnifies flaws and hides redeeming features that would be readily apparent if one sat around a table.

Long story cut short: I enjoyed the poem and hope I have understood at least part of what it was about.   --Andreas JN466 19:20, 5 October 2023 (UTC)

Funny, @Andreas, I almost started this with an epigraph about jaundice, the same line I quote at User:Tamzin/Disclosures and commitments § Blocks, and what we can learn from them. Instead I decided I'd let the poem speak for itself. I'm glad you enjoyed it. :) -- Tamzin[cetacean needed] (she|they|xe) 19:32, 5 October 2023 (UTC)

What a feat, adding spots to the liars without dripping ink on the lies! :) -- SashiRolls 🌿 · 🍥 14:52, 7 October 2023 (UTC)

FlaggedRevs

There seem to be some errors in The Risks, Benefits, and Consequences of Prepublication Moderation: Evidence from 17 Wikipedia Language Editions (https://arxiv.org/abs/2202.05548) papers assumptions on how FlaggedRevs works. For example:

  1. ... FlaggedRevs is a prepublication content moderation system in that it will display the most recent “flagged” revision of any page for which FlaggedRevs is enabled instead of the most recent revision in general... - This is depending on wiki where user is. For example German Wikipedia uses pre-moderation, and as different setup example is Finnish Wikipedia which uses post-moderation.
  2. ... Although editors without accounts remain untrusted indefinitely, editors with accounts are automatically promoted to trusted status when they clear certain thresholds determined by each language community.... - This also depends on wiki. For example dewiki promotes automatically, but huwiki has rather small number manually selected trusted reviewers. Fiwiki has large number manually selected trusted reviewers.

--Zache (talk) 09:11, 4 October 2023 (UTC)

In Russian Wikipedia, as well as in Russian Wikinews, FlaggedRevs is a disaster. You say Germans are guilty in that? --ssr (talk) 06:09, 27 October 2023 (UTC)

Wikidata

Wikidata as Semantic Infrastructure: Knowledge Representation, Data Labor, and Truth in a More-Than-Technical Project.
  The article has a lot of critiques of Wikidata, some interesting, some less relevant in my opinion. Some of them were a big part of the discussions around adopting Wikidata on frwiki, interestingly, so might be a little bit out of date. The copyright issues, for example, was and is still I guess a show stopper for some. Interestingly I think this point is significantly weakened with the emergence of the LLMs like ChatGPT as a major products of the Gafams : they don’t really care about the copyright issues or the attribution of datas they learn from, and they really can’t track back the generated factoids origin or reliably source anything … Without any restraint to date from copyright laws. Copyright does not seem to slow automated fact extraction from any public source in any way … (for the rest of the arguments covered for or against Wikidata on wikipedias, this has already been extensively discussed and argued on the projects, I won’t return on this because I don’t really feel like it)
I don’t quite understand the upper ontology part, as I think Wikidata as is is difficult to really use as an upper ontology because of the lack of formalisation. There is no reasoning rules in Wikidata to capture some of the meaning, and this is a crucial part of how an upper ontology is supposed to be used. The upper tree is indeed, or as long been as there is ongoing effort to clear this up, globally a hard to manage mess. But I doubt anyone actually used in a meaningful way the classes they show up like bad, or actually used Wikidata as an upper ontology. I’d like this claim of the article to be given more substance or proof.
It’s a critic I would actually make for the whole article, this is a discussion of the critiques but there is not really a lot of datas to give substance, this is mostly qualitative, and as such arguable. TomT0m (talk) 11:38, 4 October 2023 (UTC)

ChatGPT v. Wikipedia

The study authors comment on prose quality. I happened to ask ChatGPT yesterday to explain what government shutdowns in the U.S. are and what effects they have. I got the following answer:

Government shutdowns in the United States occur when Congress fails to pass a budget or a continuing resolution to fund the government. This can happen due to disagreements between political parties or issues related to specific policies. The shutdown results in the temporary suspension of non-essential government services and the furlough of federal employees.
During a government shutdown, essential services such as national security, law enforcement, and emergency activities continue to operate. However, many government functions, agencies, and national parks may be affected. Federal employees who are deemed non-essential may be furloughed, meaning they are placed on temporary unpaid leave.
The economic impact can be significant, affecting various sectors, including tourism, government contractors, and the overall economy. Additionally, government shutdowns can lead to a loss of public confidence in the government's ability to function effectively.
It's worth noting that my information is up to September 2021, and I recommend checking the latest sources for updates on this topic.

I then compared that to the lead of Government shutdowns in the United States:

In the United States, government shutdowns occur when funding legislation required to finance the federal government is not enacted before the next fiscal year begins. In a shutdown, the federal government curtails agency activities and services, ceases non-essential operations, furloughs non-essential workers, and retains only essential employees in departments that protect human life or property. Shutdowns can also disrupt state, territorial, and local levels of government.
Funding gaps began to lead to shutdowns in 1980, when Attorney General Benjamin Civiletti issued a legal opinion requiring it. This opinion was not consistently adhered to through the 1980s, but since 1990 all funding gaps lasting longer than a few hours have led to a shutdown. As of September 2023, ten funding shutdowns have led to federal employees being furloughed.
The most significant shutdowns have included the 21-day shutdown of 1995–1996, during the Bill Clinton administration, over opposition to major spending cuts; the 16-day shutdown in 2013, during the Barack Obama administration, caused by a dispute over implementation of the Affordable Care Act (ACA); and the longest, the 35-day shutdown of 2018–2019, during the Donald Trump administration, caused by a dispute over funding an expansion of barriers on the U.S.–Mexico border.
Shutdowns disrupt government services and programs; they close national parks and institutions. They reduce government revenue because fees are lost while at least some furloughed employees receive back pay. They reduce economic growth. During the 2013 shutdown, Standard & Poor's, the financial ratings agency, said on October 16 that the shutdown had "to date taken $24 billion out of the economy", and "shaved at least 0.6 percent off annualized fourth-quarter 2013 GDP growth".

Personally I found ChatGPT's output a lot more readable than the Wikipedia lead – it is just better written. The English Wikipedia text often required me to go back and read the sentence again.

Take the first sentence: In the United States, government shutdowns occur when funding legislation ... At first I parsed "when funding legislation" as an indication of when shutdowns occur (i.e. "when you are funding legislation"). I needed to read on to realise that this wasn't where the sentence was going.

Next, Wikipedia uses the rather technical expression "when funding legislation ... is not enacted" (which is also passive voice) where ChatGPT uses the much easier-to-understand "when Congress fails to pass a budget" (active voice).

Where ChatGPT speaks of a "temporary suspension of non-essential government services", Wikipedia says the federal government "curtails agency activities and services, ceases non-essential operations", etc. I find the ChatGPT phrase easier to understand and faster to read while providing much the same information as the quoted Wikipedia passage (a point the study authors commented on specifically).

The Wikipedia sentence Funding gaps began to lead to shutdowns in 1980, when Attorney General Benjamin Civiletti issued a legal opinion requiring it. leaves me wondering even now what the word "it" at the end of the sentence is meant to refer to.

I suspect our sentence construction and word use are not helping us win friends. It's one thing when we are the only service available; it's another when there is a new kid on the block. Andreas JN466 13:56, 4 October 2023 (UTC)

I agree that ChatGPT produces better prose than most humans do (myself included), and as far as reliability is concerned, I'd honestly trust it to give me a general overview on any non-controversial topic, no matter how obscure (I tested it earlier on "the Black sermonic tradition of whooping" and it passed with flying colours). I think where it fails and starts to hallucinate is when you ask it to drill down into the detail. This is Wikipedia's strength – we may not be good at summarizing complex topics, but we're very good at detail.
So I don't think it has to be "machines vs. Wikipedians". I used ChatGPT this morning to write a lead for the article Khlysts. As leads go, it's maybe not perfect, but it's decent. But when I ask the bot for more in-depth information, like "Who founded the Khlysts?", it starts to get vague, and if I press it for specific answers, that's when it starts lying rather than admit it doesn't know. Chatbots have got a long way to go before they can write entire articles (at least without a lot of coaching and cross-checking, at which point you're not really saving any time), and I personally don't believe that AI will ever reach that level. But I can easily foresee a not-too-distant-future in which Wikipedia has an integrated LLM interface that allows editors to auto-generate summary-style content, or perhaps provides it to readers on the fly via a "Summarize this article/section" button.
In my view, the alarm of the study authors is misplaced. They seem to agree that the AI-generated content they used in the study is, in fact, trustworthy and competent, so there's no reason to be "unsettled" that people rate it as such. I suspect the participants would give a different rating to the vague, generic answers that ChatGPT gives you when you try to go beyond a summary. I don't think there's any competition here – when people want pedantic, granular detail, they'll always know where to find it. Sojourner in the earth (talk) 21:52, 4 October 2023 (UTC)
The fact that ChatGPT doesn't know when it doesn't know something means it's "intelligent" only in a restricted sense. To go on and on about a topic, all the while presenting itself as an "intelligence", also does not seem ethical. CurryCity (talk) 08:29, 5 October 2023 (UTC)
Quoted from above, "I'd honestly trust it to give me a general overview on any non-controversial topic, no matter how obscure" — that's extremely problematic. Recently I read about someone who is a critic with a dim view of a certain thing, a fact that is well known in his field (albeit not by the general public), that when he asked ChatGPT who he is, it smoothly and confidently claimed that he is a proponent of it. The devil is in details of the black box of "where it fails and starts to hallucinate is when you ask it to drill down into the detail" — but clearly the upshot is that you absolutely cannot trust GPT not to mix any bits of plausible-sounding-but-definitely-incorrect hallucination into any kind of "explanation" that it gives you about what something or someone is or isn't, when that something or someone is any more obscure than extremely widely known and understood. When I read in this Signpost article the sentence, " […] even if the participants know that the texts are from ChatGPT, they consider them to be as credible as human-generated and curated texts [from Wikipedia]," the first thought that popped into my head is, "Well, that's it, we're boned." Duck and cover and kiss our present standard of living goodbye, because some bad shit will be happening if that shit continues very long in its current configuration. You know how McNamara said that "the indefinite combination of human fallibility and nuclear weapons will destroy nations"? Like it doesn't even matter if the annualized risk is somewhat low, if you run the runtime for enough years? That's most likely what's going to happen with the combination of human [epistemologic] fallibility [incompetence] plus LLMs, unless the "indefinite" part can be massively shortened in duration by chaining today's type of LLMs in series with some kind of downstream rapid-acting bullshit-detector software that can act like a filter. If that shit doesn't get invented within the next N years, where time N is pretty fucking short, then there is going to come a time in the future where some Homer Simpson who works at a nuclear plant is going to ask an LLM how to run the reactor properly and the LLM is going to reply with some super-smoothly-worded and speciously/superficially-plausible-but-yet-idiotically-wrong answer, and there's going to be another Chernobyl-type steam explosion of a reactor, because even though the plant has a strict rule against asking any LLMs for any cheatsheet help, Homer is going to carry his smartphone into the bathroom and ask it while inside the toilet stall because he thinks he's getting away with a clever clandestine lifehack and sticking it to The Man. Or some librarian is going to ask an LLM which library books to ban for age-inappropriate content and it's totally going to confabulate a bunch of imaginary but plausible-sounding horseshit and get books banned for containing things that they don't even contain at all (that one already happened, by the way). Or a lawyer is going to ask an LLM to write a legal argument and cite a bunch of 'definitely real and please not fake' references that support it, and he's going to get fired because those smoothly cited plausible-sounding references are all made-up imaginary horseshit anyway (that one already happened, by the way). "Oh, no, don't be silly, Quercus, that reactor-explosion thing's not gonna happen." Yeah. We all sure better just hope not, right? Running on hope and faith but no guardrails at all, what could go wrong, right? This is why the likes of Sam Altman are begging governments to regulate their industry (albeit begging with crocodile tears in their eyes while they shovel wads of money into their bank accounts). Also, just to clear up a misconception implied on this page, among some of these other comments — regarding the notion that, "Well, how can you blame people, because a supposed-to-be-nonfiction answer that's factually wrong but super-smoothly worded and easy to read is preferable to one that's factually correct but somewhat clunkily worded." Jesus, Lord help us with the stupidity. If you have to ask what's wrong with that notion, then goto Homer's steam explosion above. Quercus solaris (talk) 23:37, 6 October 2023 (UTC)
Thing is, there are plenty of Wikipedia articles that also contain "plausible-sounding-but-definitely-incorrect" information, so for LLMs and Wikipedia to be rated as equally credible sounds about right. Of course I wouldn't use either as a source if I were writing my own article, but for satisfaction of personal curiosity, sure. As for the rest of your comment, I'm not convinced that "human fallibility plus LLMs" is a bigger danger to society than, say, human fallibility plus the invention of the printing press, or human fallibility plus firelighters. Pretty much anything that benefits humanity can be misused, and every Next Big Thing gets people prophesying the imminent collapse of civilization, but somehow we keep trucking along. Sojourner in the earth (talk) 05:40, 7 October 2023 (UTC)
In this case success and failure are asymmetric. It takes only one bad enough catastrophe once to wipe out civilisation, if not humanity. Even ones that are "too small" to destroy the rest of the biosphere or planet with us. CurryCity (talk) 06:33, 10 October 2023 (UTC)

Even if ChatGPT or its successor becomes the predominant internet search tool, that doesn't mean Wikipedia will be obsolete. It likely means that Wikipedia will go back to its theoretical origin as a reference work rather than the internet search tool many readers use it as. Thebiguglyalien (talk) 16:11, 4 October 2023 (UTC)

Ah, the rise of AI. I've used it to get ideas for small projects in the past, but people prefer LLMs over Wikipedia? That's, just... sad. The Master of Hedgehogs is back again! 22:09, 4 October 2023 (UTC)

  • It's quite understandable because articles often have a ponderous, pedantic style. As a fresh example, consider the current FA, ministerial by-election. Its first sentence is

    From 1708 to 1926, members of parliament (MPs) of the House of Commons of Great Britain (and later the United Kingdom) automatically vacated their seats when made ministers in government and had to successfully contest a by-election in order to rejoin the House; such ministerial by-elections were imported into the constitutions of several colonies of the British Empire, where they were likewise all abolished by the mid-20th century.

    When this was a candidate, a reviewer observed that

    ... you have a tendency to write very long sentences, exemplified by the lead sentence, which is 67 words long. A quick google search tells me 25 words is the optimal sentence length (and 30 the maximum), but here you can routinely see double that.

    By my count, this sentence now has 68 words so it's only getting longer!
Andrew🐉(talk) 11:47, 5 October 2023 (UTC)
I really hope we can get a guideline on readability. My young sister-in-law complains that we adults "Explain things like Wikipedia", rather than in a way that a teenager can understand. I started to draft readability guideline once, but never found time to get it past the first couple of sentences. —Femke 🐦 (talk) 19:59, 6 October 2023 (UTC)
There are plenty of guidelines such as WP:NOTJARGON, WP:NOTJOURNAL and MOS:LEAD. The challenge is getting them followed. There are also plenty of metrics for this such as the automated readability index. We just need a bot or tool to assess all our articles and so highlight cases in need of attention. Andrew🐉(talk) 22:46, 6 October 2023 (UTC)
What is missing from these guidelines is a) the need to make non-technical text easy to understand b) more practical advice or what sentence and paragraph lengths make text difficult. I've been playing with the idea of creating a contest, similar to WP:the core contest, to tackle lead difficulty. Not sure yet what the best set-up is, as misguided attempts at simplifying can be quite bad. —Femke 🐦 (talk) 06:50, 7 October 2023 (UTC)
You have your work cut out for you as TFA is now featuring another good example: Affine symmetric group. ChatGPT struggles with the topic too. I reckon that it's possible to do better but it would be a thankless task currently. Andrew🐉(talk) 08:55, 8 October 2023 (UTC)
@Femke and Andrew Davidson: For what it's worth, you can ask ChatGPT or DeepL (which has a beta feature for this) to rewrite texts to make them easier to understand. Fed the Wikipedia lead section of Government shutdowns in the United States mentioned above and asked to improve its readability, ChatGPT produced the following:
In the U.S., government shutdowns happen when the funding needed to run the federal government isn't approved before the new fiscal year begins. During a shutdown, the government cuts back on agency activities and services, stops non-essential operations, puts non-essential workers on leave, and keeps only essential employees in departments that protect people or property. Shutdowns can also cause disruptions at the state, territorial, and local government levels.
Funding gaps started causing shutdowns in 1980, following a legal opinion from Attorney General Benjamin Civiletti. While this opinion wasn't consistently followed in the 1980s, since 1990, any funding gap lasting more than a few hours has led to a shutdown. As of September 2023, ten funding shutdowns have resulted in federal employees being placed on leave.
Major shutdowns include the 21-day shutdown in 1995–1996, during the Bill Clinton administration, due to disagreements over significant spending cuts; the 16-day shutdown in 2013, during the Barack Obama administration, caused by a dispute over implementing the Affordable Care Act (ACA); and the longest one, the 35-day shutdown in 2018–2019, during the Donald Trump administration, triggered by a disagreement over funding for expanding barriers on the U.S.–Mexico border.
Shutdowns disrupt government services and programs, leading to the closure of national parks and institutions. They also result in reduced government revenue as fees are lost, even though some furloughed employees eventually receive back pay. Furthermore, they negatively impact economic growth. For instance, during the 2013 shutdown, financial ratings agency Standard & Poor's reported on October 16 that the shutdown had already taken "$24 billion out of the economy" and "reduced annualized fourth-quarter 2013 GDP growth by at least 0.6 percent."
I wouldn't endorse every change, but would argue that the overall result is indeed improved readability. Andreas JN466 11:51, 8 October 2023 (UTC)

Interesting data! The Master of Hedgehogs is back again! 22:13, 4 October 2023 (UTC)