Wikipedia talk:Wikipedia Signpost/Single/2025-06-24

Comments

The following is an automatically-generated compilation of all talk pages for the Signpost issue dated 2025-06-24. For general Signpost discussion, see Wikipedia talk:Signpost.

Comix: Hamburgers (1,359 bytes · 💬)

• I'll gladly pay you Tuesday for a hamburger edited today. Randy Kryn (talk) 04:18, 24 June 2025 (UTC)

We get paid?? - UtherSRG (talk) 11:22, 24 June 2025 (UTC)

A funny comix? At this time of year! At this time of day! In this part of the country! Localized entirely within this Signpost issue‽ —⁠andrybak (talk) 16:49, 24 June 2025 (UTC)

• Ho, ho, ho, no, public domain encyclopedia article! Old expired copyright! ―Howard • 🌽³³ 21:56, 24 June 2025 (UTC)
• These are not "hamburgers" old boy! These are Granny's skillet cookies! lol! Marcus Markup (talk) 15:36, 25 June 2025 (UTC)
• No, no, I said steamed hams... because the editors are hamming it up. Old linguistics issue. ;) XFalcon2004x (talk) 17:13, 25 June 2025 (UTC)

Yes! SergioFLS (talk) 00:39, 26 June 2025 (UTC)

Debriefing: EggRoll97's RfA² debriefing (7,231 bytes · 💬)

• Thanks so much for this report. I really hate hearing that these processes are negative experiences for good Wikipedians, but it’s really important to know that’s still the case. Appreciate all your work. Innisfree987 (talk) 03:10, 24 June 2025 (UTC)
• Also thanks for putting this together, and an interesting anecdote into AELECT. CMD (talk) 03:42, 24 June 2025 (UTC)
• @EggRoll97: If you are willing, could you help clarify why you found an admin election less stressful than an open viewpoint request for adminship, but more nerve-wracking? Perhaps you were referring to different phases of each process? No worries if you feel you've already said your piece. I appreciate your sharing of experiences! isaacl (talk) 05:09, 24 June 2025 (UTC)

  As surprised as some may be at this, I actually found a traditional RfA to be less nerve-wracking, as it was a straight week of a nomination, and I could easily see all the !votes coming in, with reasons attached that I could read through and reflect on in the moment. AELECT, though, had a week of voting, where I was completely in the dark.

  Sounds like something about the opaqueness of AElect voters. Aaron Liu (talk) 18:49, 24 June 2025 (UTC)

  Like Aaron Liu said above, I had a lot more to reflect on very immediately, while AELECT didn't really give a lot of feedback. It gave some, definitely, and that was good, but the anonymity of !voting comes at the cost of not necessarily needing to express opinions to the candidate. The traditional RfA format also meant I could take feedback as it came in, as opposed to a few comments, then a failing vote result. Please don't get me wrong though, I think AELECT is great for some, it just happened that a traditional RfA format was more my speed. EggRoll97 ^(talk) 00:15, 25 June 2025 (UTC)

  Sure; I was just trying to understand how those aspects led to a process that was less stressful but more nerve-wracking, as I would have thought being nerve-wracking would also be stressful. But in any case, thanks for the info! isaacl (talk) 01:23, 25 June 2025 (UTC)

• @EggRoll97: Thank you for your effort to summarize your RFAs and compare AELECT with traditional RFA. On my home wiki, the Chinese Wikipedia, the number of questions asked in RFAs has been halved since introducing AELECT and deprecating traditional RFA due to canvassing and harassment concerns pointed out in the office actions in 2021. Having two wikis showing the same trend, I believe the phenomenon you described was not a coincidence, but a common issue that comes with voting, that people are encouraged to simply state their stances instead of pointing out problems and asking about their concerns, a process crucial to concensus building, where consensus "takes into account all of the proper concerns raised," and where the "proper concerns" are those intended to help achieving "the Five Pillars—Wikipedia's goals."
  Another problem that arises with the introduction of AELECT is the existence of disruptive (or otherwise non-constructive in the context of consensus building) rationales of voting for or against candidates, while there are no protocols to strike out such nonsense, and bureaucrats must strictly follow the numbers. (We grant those who got 65% temporary admin flag that must be reconfirmed at the next AELECT, so things might be better than here.) As a Wikipedian planning to run for adminship on the Chinese Wikipedia, this is becoming a real concern to me: Apart from the degradation of community involvement in AELECTs compared to traditional RFAs, do you think that the results of voting for adminship may go against community consensus (or what would have been reached if a traditional one were held instead), either supporting, opposing, or being uncertain about granting the admin flag? 1F616EMO (talk) 11:33, 24 June 2025 (UTC)

  I don't think it goes against community consensus necessarily, though I think the candidate in a massive AELECT receives ultimately less feedback than a candidate in a traditional RfA, and to some degree, I think the ability of monitors/crats to clerk RfAs keeps things in order. With AELECT, no reason is required to be attached to one's vote, and this does lead to a lot of opposition without necessarily knowing whether that opposition is in line with whatever was stated in the RfA page, or whether there is an entirely different issue with one's candidacy. I will note that I received a 39% in AELECT, and a 66% in RfA, but that isn't necessarily an indication that the AELECT voting didn't reflect the consensus of the community. EggRoll97 ^(talk) 00:20, 25 June 2025 (UTC)

• Thank you for your honest plus/minus personal feedback of the AELECT process. There appears to be a negative bias against any editor who self-nominates for adminship - as opposed to being recommended for it by at least two other admins. I took a look at your user page and had another admin recommended you through RfA you likely would have passed. It is what it is and you are to be commended for being willing to continue your work on Wikipedia rather then be discouraged by the process. Blue Riband► 12:36, 24 June 2025 (UTC)
• Thanks for writing this, EggRoll97. I really do think that if you'd had nominators, you would have passed (though perhaps you'd have delayed your run to do so), but I'd like to disagree with Blue Riband above that there's a "negative bias" against people who don't have nominators. Well, I suppose there probably is a bit of a one, but that's not really the thing that nominators are for, in my opinion - the real value of having nominators is that you have two experienced administrators working with you before the nomination, giving you helpful and frank advice. They can let you know if you're not quite ready, give you suggestions about your answers to the three standard questions, and so on. Also, at least in my experience, I just found it comforting to know they were there during the process. I really encourage any prospective admins reading this to reach out to the admin you know best (that doesn't mean you have to be besties, just someone who knows your name), or at least to an admin who frequently works in the areas you do, to talk about nomination or adminship in general before taking the plunge. -- asilvering (talk) 20:27, 28 June 2025 (UTC)

Disinformation report: Pardon me, Mr. President, have you seen my socks? (529 bytes · 💬)

Thanks for the comprehensiveness of this sockfarm analysis! It's quite interesting that not only do these recipients of corrupt pardons have a shared history of sockfarm whitewashing, but also that many relied on the same sockfarms. ViridianPenguin🐧 (💬) 12:32, 2 July 2025 (UTC)

In the media: Playing professor pong with prosecutorial discretion (4,244 bytes · 💬)

• Can someone clarify what The Jerusalem Post is trying to say (and whether it's accurate)? What do they mean by "platforming" on the part of arwp? Nardog (talk) 02:47, 24 June 2025 (UTC)
  • Just from a quick rereading of the JPost article (so you might want to check the details): The Palestinian Authority (PA) sponsored an on-Wiki contest to edit the Arabic Wikipedia, with prizes totaling about $1,000 (divided among 4 winners) about the 1948 Nabka (catastrophe) when the Palestinians lost their land. I don't think this type of government sponsored contest would fly on EnWiki, but maybe a government sponsored GLAM downunder might have done something similar (just guessing). One definite parallel I've seen is a contest by the Ukrainian Foreign Ministry to encourage articles on the Ukrainian and other Wikis about Ukrainian culture, the spelling of Kyiv (not Kiev), and Ukrainian cities and settlements (with total payments maybe about $5,000 (before the war). Different Wikis have different rules. These small payment contests sponsored by governments are not my favorites, but they are usually not against Wiki rules (arguably are on EnWiki). And I can see them as good faith efforts all around on other Wikis. "Platforming" seems to mean just posting the contest on Arabic Wikipedia. Smallbones_(smalltalk) 03:40, 24 June 2025 (UTC)

• Who's going to tell Collider that xXx has been a consistently top-rated article every month for the past decade and no one knows why? ~Maplestrip/Mable (chat) 07:48, 24 June 2025 (UTC)
  • Yeah, like anyone who's been reading the Top 25 Report for long enough, I couldn't help but laugh at the shoddy journalism there. Not once do they stop to consider why people could be searching for it aside from genuine interest in Vin Diesel's movie career. I think there's probably some merit to the theory that the views are driven by people looking for pr0n and googling "xxx movie" or similar while leaving SafeSearch on. SMH —pythoncoder (talk | contribs) 15:42, 24 June 2025 (UTC)
    • I mean, let's be real, that's the only logical explanation for it... Even I remember that xXx wasn't even that big of a box office hit when it came out while I was a kid, and instead it has almost cult-classic status sitting on the Hit Report! XFalcon2004x (talk) 17:25, 25 June 2025 (UTC)
      • It's an illogical explanation since it implies the world's interest in pornography has gone up 40% in the last few months. Chess (talk) (please mention me on reply) 23:27, 27 June 2025 (UTC)

        Wouldn't that suggest that perhaps a search engine changed its behavior when people search "xxx" or a related term? i.e. people aren't searching for porn more, people are getting shown the article for xXx more *while* searching for porn. —Ganesha811 (talk) 01:47, 28 June 2025 (UTC)

        The age verification laws that went into effect in various US states in the past year or so may have some to do with it. Nardog (talk) 02:10, 28 June 2025 (UTC)

XXX (soundtrack) (views) is on the level of Beethoven, Taylor Swift, Bollywood Bluerasberry (talk) 16:19, 2 July 2025 (UTC)

News and notes: Happy 7 millionth! (17,202 bytes · 💬)

AI summary trial halted

• The good faith AI summaries that the WMF pondered putting at the top of every article, at least the ones I read, all violated MOS:OUR. WMF's AI missing such a MOS-foundational error in the test summaries could mean that maybe some of the people working there should take a course, or edit Wikipedia more, or something. Randy Kryn (talk) 03:19, 24 June 2025 (UTC)
  • One of the primary reasons I still use Wikipedia is that the information is human-gathered and human-sourced, rather than risking being some kind of AI hallucination. Furthermore, using AI on Wikipedia feels... wrong. There's not really a good way for me to express it, it just doesn't seem right for AI to be implemented here of all places! XFalcon2004x (talk) 17:22, 25 June 2025 (UTC)
    • Of course, to me (full disclosure: I was in the focus groups discussing this at Wikimania last year) the issue is: now that we've put our foot down on this, what do we do if someone creates a third-party app that does this? And is successful. We couldn't exactly (I think) sue them for a use so parallel to our own, and if they do things like put ads in the content, something that people come to believe we do, what effect will that have on our donation drives? Daniel Case (talk) 02:42, 26 June 2025 (UTC)

      We have never been able to stop forks by design, all our content is reusable. WMF could likely sue any app that imitates Wikipedia trademarks, if the concern is deception. However, I'm not seeing a scenario where this specific usage affects donation drives in a way that the very existence of publicly accessible llms in general does not. One can already ask an llm to summarise an article, and in a browser extension to boot. CMD (talk) 03:01, 26 June 2025 (UTC)

Don't know why someone would use a fork with ads when they have Wikipedia, because isn't a summary at the top of the page what we in the burbs used to call the "lead" paragraph? Whatever the name, such a fork would quickly become known as "AI Wikipedia", and might rival Wikipedia for awhile - until the mistakes multiply and differences become apparent (and, ads). But yes, the first real competitor to Wikipedia might use the AI plus fork approach, only to drift into advertising and major errors (the great ship Britannica comes to mind). Wikipedia seems here to stay, 25 years next year and it has outlasted vandals and should be able to survive being the human-written ad free encyclopedia. Randy Kryn (talk) 03:24, 26 June 2025 (UTC)

Except that it is not. Not anymore. Wikipedia editors are using LLMs to generate "content" for pages. For example Esculenta says they use a LLM to generate all the Litchens they add to Wikipedia including the Wikipedia:The World Destubathon. If you ever wondered why articles like Lichesterol have weird phrases like "proving that this unusual molecule can be synthesized artificially" or "and proving, for the first time, that it is feasible to synthesise this type of sterol".

I see more content generation like this on the horizon for Wikipedia. I don't want it to be, but this is what people apparently want. 🌿MtBotany (talk) 04:42, 29 June 2025 (UTC)

Wikipedia administrators arrested in Belarus

• Thanks to the writers for putting these all together. I'm concerned we may need to work on a language for arrests that has a bit less jargon if this becomes an increasingly common occurrence, as this may be something we want external media to pick up on. CMD (talk) 03:28, 24 June 2025 (UTC)

Some additional context from this May 30 article in Nasha Niva (a newspaper that the May 26 Wikinews story summarized in this Signpost article draws from), Google translated:

The editor of the Belarusian Wikipedia says that the detention of two administrators has not led to “paralysis of the section’s work” and the page “continues to function as usual.” [...]

The situation looks different in the case of the Belarusian Wikipedia in "Tarashkivitsa".
"The detained administrator of that section, [...] had spent several years using insults, threats and blocking to push out almost everyone who wanted to write there. At some point, he became almost the only one creating content - a kind of one-man wiki.
[...]
Wikipedia and YouTube are two projects that cannot be simply blocked, because they are needed by absolutely everyone. [...]

The Wikimedia Foundation is well informed about the detentions in Belarus, the editor says.

Regarding the latter, it would be interesting to know if the Wikimedia Foundation has made any attempts to lobby the US government to intervene in support of the detained editors. Two days ago, the NYT reported:

President Trump sent a special envoy to Belarus for talks on Saturday with President Aleksandr G. Lukashenko [...] There was no immediate sign that Washington would ease sanctions on Belarus. But John Coale, Mr. Kellogg’s deputy, said the visit to Minsk had secured the release of 14 political prisoners from Belarusian jails. [...] The freed prisoners, who arrived in neighboring Lithuania by car on Saturday afternoon, included Sergei Tikhanovsky, a dissident and the husband of the exiled opposition leader Svetlana Tikhanovskaya.

Regards, HaeB (talk) 06:25, 24 June 2025 (UTC)

@HaeB: Can the WMF not appoint a steward in this case? Where then those rights may be furthered to the local bewiki community, as this crisis isn't resolving itself anytime soon. Gotitbro (talk) 11:18, 26 June 2025 (UTC)

Recall of Bbb23

I’m a little disappointed to hear that Bbb23 lost his admin tools, particularly since I do not recall ever seeing a notice about the recall petition or any ANI report inviting larger community input on such a matter. I have growing concerns that this process is being used as a back door to selectively harass admins by threatening them with recall petitions if the outcome of a given admin action was not to a specific account’s liking. I also find it odd that out of the entire myriad of tool sets it’s only admins that are specifically target by recall pages, not crats, not extend confirmed or auto patrolled accounts, just admins. It reads to me like a low level form of authorized harassment against the admin corps, which then puts the Foundation’s anti-harassment initiatives in an interesting place vis-a-vis needed admins to help on Wikipedia but throwing them under the proverbial bus if they do something people don’t like, which is all but guaranteed to happen sooner or later. I’d be the first to admit I’m no wait, and I’ve made some admin tool fuck ups myself, but I don’t want to have the bit stripped from me simply because I crossed wires or tried to help with a block or page protection. Maybe I’m just being paranoid again, but I feel like this process ought to be shut down or at least retooled some, I personally don’t like feeling that I can be singled out exclusively for prosecution in a kangaroo court for no reason other than being an admin. Anyone else kind of get that feeling out of the process too, or is it just me? TomStar81 (Talk) 23:22, 24 June 2025 (UTC)

@TomStar81: The admin recall procedure is an imperfect one and is being improved on with almost each recall initiated. FOr Bbb23, there was a long discussion on WP:AN, Wikipedia:Administrators'_noticeboard#Admin_Bbb23, which led to a recall petition started with notice given on the same noticeboard, Wikipedia:Administrators'_noticeboard/Archive371#An_administrator_recall_petition_has_been_initiated_for_Bbb23. The petition was certified within hours, Wikipedia:Administrators'_noticeboard#Bbb23_recall_petition_certified.

The first petition was started without prior notification to the admin in question, but that admin eventually lost his RRfA anyway. The latest instructions now indicates that it is a best practice to start the a discussion with the admin first and upon exhuasting the discussion(s), start the recall. – robertsky (talk) 01:39, 25 June 2025 (UTC)

The administrators' noticeboard has long been a venue to discuss the removal of most permissions (it's mentioned at Wikipedia:Requests for permissions § Review and removal of permissions), and you only have to convince one administrator to proceed on their own initiative. isaacl (talk) 01:51, 25 June 2025 (UTC)

Tom, I might step back and reconsider this comment. For example, one of the requirements of being an administrator here is that they hold the trust of the English Wikipedia community. Extended-confirmed and autoconfirmed users get those userrights after some editing and some time. There's no comparison to be made. It may be worth having a look at Wikipedia talk:Administrator recall and Wikipedia:Requests for adminship/2024 review to see how hundreds(?) of users worked to get to the current process. Ed ^{[talk] [OMT]} 01:58, 25 June 2025 (UTC)

@The ed17: The lower end user rights, as you noted, are earned, but bureaucrats, arbcom officials, checkuser rights, etc (the "higher level rights", if you will) are exempt from this particular form of recall. Only the admin corps specifically is singled out, which is what bothers me here. Why us admins, and not a generic recall for the crats and arbcom members and so forth in that manner? For me, its in interesting question with no immediate answer. It may be that the admin corp is the most visible part of policy and guideline enforcement, it may be that the crats and the check users and such are not thought of as being subject to recall (or worse, are shielded from it by the foundation, preventing recall), it may be that no one has suggested it, or it just may be that the tasks they do are not thought of as important enough for the community to consider a formal recall process. I'm not going to pitch a fit or protest or so forth in that manner, it just strikes me as weird that at present the only group of those required to obtain community consensus for additional privileges that is subject to recall is the admin corps. TomStar81 (Talk) 17:13, 3 July 2025 (UTC)

@TomStar81: I'm still not sure you're looking at this fairly. Arbcom can be voted out and its members can be removed faster by a committee vote. Checkusers have their own, and quite serious, process. Bureaucrats, with respect to those in that role, have few discretionary powers these days. Ed ^{[talk] [OMT]} 19:39, 3 July 2025 (UTC)

Who cares about the victims? We need to make sure we don't hurt the perpetrator's feelings. Thebiguglyalien (talk) 🛸 15:34, 26 June 2025 (UTC)

Seriously though. What's with so many admins going "oh, that poor administrator", while blatantly ignoring the dozens of editors, both old and new, that were harassed and driven from the project by said admin. That's especially true for Bbb23, where the AN discussion had to overturn so many of his blocks even from just a cursory look, because they were so appallingly terrible. TomStar81, are you actually concerned about being called out for harassing newbie editors? Why exactly would that be something you, as an administrator, would be worried about? Silverseren^C 17:25, 3 July 2025 (UTC)

Adminship is a thankless task, I've had the bit for about 16 years now, been yelled out a lot, made a few mistakes - some of them dozzies - and have not, in all that time, managed to earn an admin's barnstar. I note that your privileges do not include the admin bit, but you have four of the big ones and I suspect have probably made mistakes too, yet there's no formal recall process for such as this for rollbackers, autoconfirmed users, extended confirm contributors, or pending changes reviewers. Not that the tools can't be revoked if you miss use them, but there is no established recall page for editors who mess up with your tool sets vs the admin recall page which is set up to deal specifically my toolset. You don't find that odd, or unusual, or peculiar? TomStar81 (Talk) 17:44, 3 July 2025 (UTC)

For some of them, the damage they can do is limited. XCON users can only edit a few more articles (though more contentious) than an IP editor. Rollbacking really isn't something that an IP couldn't do, as it is simply reverting an article to a previous state which can be done through copy-pasting. Pending changes only affects a tiny portion of articles and a tiny portion of edits made to those articles. On top of all of that, all of them can simply be blocked by an admin if needed. I cannot comment of checkusers or other roles as I do not have those permissions and I am not well versed in those areas.

Admins on the other hand have blocking power which cannot be done by any other role to my knowledge. They can also delete and undelete articles. Both of those can be very serious if used incorrectly compared to something like rollbacking, which can be done by anyone in theory. They also cannot be combatted by anyone but admins. I also believe admins in particular are targeted as they are common enough to be often seen, unlike bureaucrats, but not so common that they are drowned out, like rollbackers. ✶Quxyz✶ 17:55, 3 July 2025 (UTC)

Tesla Takedown Request

• This may be a very common opinion, though it still feels like one I must put into words here: This feels very revenge-motivated. As many of us remember, within the last year Elon Musk, head (? What does he actually do, let's be real) of Tesla among numerous other things, has had a very public feud with Wikipedia as well as the Foundation as a whole (resulting in a very successful donation run). I guess the real question is if there's any legal basis for the DMCA request or if it's just more Elon grumping about being shown up on the public stage, something we've seen very publicly within the last two weeks with the "Big Beautiful Break-Up". XFalcon2004x (talk) 17:22, 25 June 2025 (UTC)

  The Commons discussion seems to suggest there was some basis for 1 of the 4 files, and that there was a general issue (not specific to Tesla) of people misclaiming "own work". CMD (talk) 21:54, 25 June 2025 (UTC)

  I think I have seen multiple cases of archaic maps and satellite photos being claimed as own work. My favourite ones are the ones that have attribution in the caption but claim own work in the credits. ✶Quxyz✶ 16:16, 30 June 2025 (UTC)

• It's interesting to read the letter from Are not Fox. It makes "good faith" claims that no competent attorney should make. Of course this is a common ploy, and a key use of lawyer's letters, to achieve a goal that not only would not be achievable in an actual court, but is not actually supported by law. It's not one, however, that a reputable lawyer should ever be party to. All the best: Rich Farmbrough 15:36, 2 July 2025 (UTC).
  

Opinion: Russian Wiki-fork flails, failing readers and editors (3,890 bytes · 💬)

Ruwiki interface looks nice, but to me, at least, its sluggish (scrolling is very slow for some reason). --_{Piotr Konieczny aka Prokonsul Piotrus| reply here}

I concur. I tried looking at a dozen articles or so, and I kept getting mostly AI stuff. They have changed the layout somewhat from when they started, looks nice, but doesn't perform as well. On Wikipedia content is king. The format is nice and conservative and doesn't distract from the content. It looks like an encyclopedia. We could do a slightly better job getting videos and other multimedia integrated into the content. You can do it, but it's not often easy. We're pretty close to a sweet spot though. Twitter/X and Reddit always look like dog droppings to me, and the flashy sites are hiding a lack of content. Smallbones_(smalltalk) 04:04, 24 June 2025 (UTC)

Something I've noticed about made-in-Russia items. At the beginning of the recent Russian invasion of Ukraine, there would be images of the Russian equivalent of K-Rations/Ready-to-Eat packages. The packaging would look very attractive, the product of professional designers -- however, according to reliable reports, the meal inside was of poor quality. There are many more examples of this, where the container or packaging would be very well designed, look as good if not better than the Western equivalent, yet the contents be substandard. I don't know if this is a failing due to poor manufacturing standards, or corruption, or that Russian culture emphasize appearances more than substance. -- llywrch (talk) 17:00, 3 July 2025 (UTC)

There is also an interesting-looking recent academic paper about Ruwiki which looks at some of these questions in more detail:

• Trokhymovych, Mykola; Kosovan, Oleksandr; Forrester, Nathan; Aragón, Pablo; Saez-Trumper, Diego; Baeza-Yates, Ricardo (2025-04-14), Characterizing Knowledge Manipulation in a Russian Wikipedia Fork, arXiv, doi:10.48550/arXiv.2504.10663

(In case someone is interested in reviewing or summarizing it for the Signpost's "Recent research" section/the Wikimedia Research Newsletter, feel free to call dibs in our Etherpad.)

Regards, HaeB (talk) 04:19, 24 June 2025 (UTC)

Any particular reason to say für not for? (In more comprehensive, neutral, and up to date, and with AI support für users.) Or is it just the jumble one ends up in when speaking two languages fluently? (I get that.) — Alien  3
3 3 06:38, 24 June 2025 (UTC)

Seems like a botched copyediting, can be fixed. Deinocheirus (talk) 11:11, 24 June 2025 (UTC)

• Well they haven't yet implemented the threatened block of Wikipedia in Russia, we are only being labeled as disinfo as of now :|. That is perhaps a major influence on traffic. Gotitbro (talk) 11:26, 26 June 2025 (UTC)

Recent research: Wikipedia's political bias; "Ethical" LLMs accede to copyright owners' demands but ignore those of Wikipedians (5,667 bytes · 💬)

Wikipedia's political bias

"To estimate the political polarization of Wikipedia citations, we use the Media Bias Monitor (Ribeiro et al., 2018). This system collects demographic data about the Facebook followers of 20,448 distinct news media outlets via Facebook Graph API [10] and Facebook Marketing API [11]." It's bizarre the extent to which the authors treat this as a neutral dataset to form the entire basis for their classification system. --Goldsztajn (talk) 12:01, 24 June 2025 (UTC)

"Through further analysis using ratings of factual reliability from Media Bias/Fact Check..." Thanks, I needed a laugh today. Stepwise Continuous Dysfunction (talk) 17:16, 25 June 2025 (UTC)

According to the Media Bias/Fact Check enWiki article, this website is widely used in scientific studies and got a decent mark from Scientific Reports. So is it perfect? Definitely not. Can it be used for reliable bias estimation? Still probably yes. Deinocheirus (talk) 16:09, 27 June 2025 (UTC)

"Ethical" LLMs

Regarding the "ethical" AIs - are Chinese models represented here? --_{Piotr Konieczny aka Prokonsul Piotrus| reply here} 03:27, 24 June 2025 (UTC)

Interesting question. I am not aware of any Chinese LLMs in that regard.

The well-known open-weights/open-source models by DeepSeek or Qwen can be assumed to have been trained on copious amounts of non-freely licensed material (just like, say, ChatGPT or Claude). So they too are super "unethical" in the framing promoted by PleIAs and the Common Pile group.

Regards, HaeB (talk) 03:50, 24 June 2025 (UTC)

An addendum that didn't quite fit into this already lengthy review:

I was advised that among the 28 authors of the Common Pile paper, not everyone had been comfortable with its anti-fair-use framing. While I do think that people have to take responsibility for what is being published under their name, one may want to keep that in mind when making conclusions about particular individuals.

On the other hand, some of the "Common Pile" authors have gone even further elsewhere with advocating for the financial interests of the copyright industries: Nikhil Kandpal and Colin Raffel (the paper's first and third listed author) recently came out with a position paper where they argue the rather radical notion that AI developers should be obliged pay the full replacement cost for every single source used to train an LLM (e.g. if a particular text was accessed, the full labor costs for creating an equivalent text - not merely, say, the market price for a copy). One has to wonder whether Kandpal and Raffel put their money where their mouth is and paid the full labor cost for each of the 40+ publications they cite themselves in their paper. Or how many billions or tens of billions of dollars Wikipedia would be charged if Kandpal and Raffel were to apply the same "fairly compensating" principles to it - considering that Wikipedia consists entirely of information published in other sources. (On the other hand, they evidently believe that the fair compensation for Wikipedians whose labor is used for LLM training should be $0.) They openly acknowledge in that paper that their position would bankrupt basically any developer of currently used LLMs (including, surely, their Common Pile co-authors EleutherAI and other non-profits who have worked to make AI accessible to everyone by releasing open-weight/open-source LLMs trained on non-permissively licensed text).

Nikhil Kandpal and Colin Raffel seem oblivious of various hard-won insights from the centuries-old history of intellectual property, e.g. that the amount of pure labor by itself should not create rights against others (cf. sweat of the brow), or that numerous limitations and exceptions to copyright should exist to balance the private interest of copyright industry businesses and professionals with public interests such as access to information. It's a reminder that research prowess or technical skills in one area (Raffel in particular has published influential AI/ML research) do not automatically translate into expertise in a different area (the law, economics and ethics of intellectual property). It remains to be seen how influential this kind of advocacy for radically prioritizing the financial interest of IP owners becomes. But I do think it is worth watching as an emerging danger to free knowledge and open culture in general - see also the warnings by the Internet Archive's Brewster Kahle and the EFF cited in the review.

Regards, HaeB (talk) 21:46, 26 June 2025 (UTC) (Tilman)

I appreciate the nuance to recognize both a) the Common Pile is overly dismissive of Wikipedia's share-alike requirement on the uncertain assumption of a fair use defense AND b) the movement to restrict use of Wikipedia in AI training runs counter to our mission of freely sharing knowledge. ViridianPenguin🐧 (💬) 12:53, 2 July 2025 (UTC)