Today I had reason to look at 10 years ago, and saw a great pictured comment by you. Thank you for clarification in that matter and many others. We'd need more of it, but best wishes for what you do instead! -- Gerda Arendt (talk) 07:38, 18 October 2024 (UTC)[reply]

Sorry @Gerda Arendt, missed this one when I was replying to the accumulated posts on this page last week. I haven't kept track of Jimmy recently and he may well have changed, but I certainly feel my comment a decade ago was correct at the time—2014 Jimmy Wales was right up there with 2024 Elon Musk regarding misusing his position to disparage perceived opponents. (Where the WMF differs from Twitter is that for all its faults, the WMF has mechanisms for reining in people who abuse their position. The processes may be painfully slow and poorly designed, but in general they do eventually work.) ‑ Iridescent 07:44, 14 January 2025 (UTC)[reply]

		Io, Saturnalia!
		Wishing you and yours a Happy Holiday Season, from the horse and bishop person. May the year ahead be productive and distraction-free. Ealdgyth (talk) 15:14, 17 December 2024 (UTC)[reply]

		Season's Greetings
		Wishing everybody a Happy Holiday Season, and all best wishes for the New Year! The Adoration of the Magi in the Snow (1563) by Pieter Bruegel the Elder is my Wiki-Christmas card to all for this year. Johnbod (talk) 17:36, 17 December 2024 (UTC)[reply]

	A very happy Christmas and New Year to you!
	Have a great Christmas, and may 2025 bring you joy, happiness – and no trolls or vandals! Cheers SchroCat (talk) 08:30, 21 December 2024 (UTC)[reply]

• Thanks to all three of you (Since we're still within the Twelve Days, Hanukkah and Twixmas, I can just about avoid making it "belated thanks") ‑ Iridescent 03:56, 31 December 2024 (UTC)[reply]

And delighted to see you back again after such a long break! - SchroCat (talk) 14:04, 2 January 2025 (UTC)[reply]

Per my comment to Kudpung below I'm not really 'back' per se—more 'less absent'. ‑ Iridescent 08:01, 12 January 2025 (UTC)[reply]

Hi, Iri. Great to see you around again. It looks like I deleted User:Arielvilla07/sandbox just as you were declining the CSD, a race condition based on when I loaded the page. As far as I know, it's common practice to apply G3's hoax subcriterion to sandboxes, if they're formatted like an encyclopedia article and obviously fictitious. WP:FAKEARTICLE says Actual fake articles should be deleted as incompatible with the purpose of the project. Pages that egregiously present false information may be tagged with {{db-hoax}}. and makes no exception for sandboxes. As a result, I'm hesitant to self-revert here, but at the same time, I don't want to step on your feet. I'm about to go to bed, so if you want to restore the sandbox, I don't object, although I'd be inclined to blank and/or MfD it if it's restored. Either way, I leave this in your capable hands. All the best. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 08:19, 1 January 2025 (UTC)[reply]

I've no problem at all with the deletion. My general feeling is to allow pretty much anything in sandboxes regardless of whether it's true or not—it's certainly not unusual for someone who wants to write about (e.g.) a boxer to copy-paste the formatting of an existing boxing biography and play around to get the feel of editing, how templates work, etc. As such, I generally extend maximum AGF to sandboxes, even if what's going on in them would normally be considered vandalism. In this case, looking at Special:Contributions/Arielvilla07 it's fair to say that AGF is well past any reasonable limit. ‑ Iridescent 14:38, 1 January 2025 (UTC)[reply]

Yes, I always try to tackle sandboxes first when I look at CAT:CSD, because they have some of the worst mis-taggings. Entirely valid drafts tagged as U5, unfinished bits and bobs tagged as U5 or G3 (or G1 or G2, which don't even apply in userspace), you name it. Sometimes I wonder what some CSD taggers think sandboxes are supposed to be used for, because it's apparently neither testing nor drafting. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 23:56, 1 January 2025 (UTC)[reply]

IMO most of those mis-taggings are entirely in good faith. Those of us who've been here a while forget how confusing Wikipedia's rules are. To someone with a decent but incomplete understanding of Wikipedia's workings, "even a userspace sandbox is technically released to the world under the Wikipedia name, so anything inappropriate or inaccurate should be deleted no matter where it's posted" is intuitively entirely reasonable. This is especially intuitive to the sizeable proportion of newer good-faith users who've themselves been slapped down at some point for posting an unsourced or insufficiently-sourced claim. ‑ Iridescent 08:08, 12 January 2025 (UTC)[reply]

I've been completely incommunicado myself (but lurked a bit), so I'm gingerly picking up the threads. Mainly cleaning up the junk that's got into some of my favourite articles and trying to avoid the politics because if anything, the place has simply got even more toxic and chaotic than when I left it almost exactly 2 years ago.

Anyway, I'm so relieved to know you're alive and kicking. Happy New Year! Kudpung กุดผึ้ง (talk) 11:52, 2 January 2025 (UTC)[reply]

I'm still alive, albeit minimally active at best. I've not looked at the internal politics yet, but even if it's got worse over the past couple of years I find it hard to believe it could have become more toxic than it was c. 2010–15… ‑ Iridescent 07:49, 12 January 2025 (UTC)[reply]

Was so pleasantly surprised to see you on my watchlist today. Place is still as bonkers as ever, but it's our bonkers place. Star Mississippi 13:26, 29 January 2025 (UTC)[reply]

Nice to see you still here! I'm only vaguely back, but having already been accused of being too keen and too unwilling to delete, I see Wikipedia hasn't changed.  ‑ Iridescent 18:21, 29 January 2025 (UTC)[reply]

Draft:Hindu terrorism in Bangladesh, all of its content is cut, copy, paste from the article Hindu terrorism, with no reference linking to any incidents in Bangladesh, no mention of Bangladesh anywhere in draft article itself. Does this not qualify for speedy deletion? — Hemant Dabral (📞 • ✒) 15:48, 15 January 2025 (UTC)[reply]

I wouldn't say so—copy-pasting an existing article into draft/sandbox space so one can use its structure as the basis for another article on a linked topic isn't at all unusual, especially for a new user. The only potential speedy deletion criterion I could see applying would be as a technical copyvio as one is supposed to include an attribution in the edit summary when copying within Wikipedia, but usual practice is to retroactively attribute, rather than delete-amd-start-again. An article that fails to mention its purported topic obviously wouldn't be valid in mainspace, but as a draft is legitimate provided there's a reasonable assumption the creator intends to expand it. ‑ Iridescent 13:18, 17 January 2025 (UTC)[reply]

Just a note on the CSD nomination you recently declined. I have also reported the creator of the draft at SPI and will retain it under G5 if confirmed. See Wikipedia:Sockpuppet investigations/Mydigitalid. TornadoLGS (talk) 15:20, 28 January 2025 (UTC)[reply]

For the gallery, the page in question is Draft:MyDigital ID. I personally think that the SPI is questionable, and deleting it under G5 would be highly questionable; while the procedure a new account that's been softblocked for a promotional username should follow is to request a rename and unblock, creating a fresh account is hardly unheard of. It seems to me clearly not to be spam in any meaningful sense; the topic is unquestionably notable in Wikipedia terms (GOV.UK Verify, Diia, MyID (Australia) if you want some precedents), and the tone seems to be just a straightforward explanation of what the service provides rather than any obvious promotion. As such, deleting it only to recreate it, rather than cleaning it up, would seem to be a fairly pointless piece of deleting-for-the-sake-of-deleting. ‑ Iridescent 18:46, 28 January 2025 (UTC)[reply]

The original account was hardblocked for promotional editing as well as the username. So I figured circumventing the block was a violation since the username was not the sole reason for the block. In any case. The second account removed promotional language from the draft after I tagged it for G11. TornadoLGS (talk) 19:22, 28 January 2025 (UTC)[reply]

It's now been deleted by Izno. FWIW I still feel the deletion was incorrect, for the reasons already given—since this is clearly a viable topic, then provided the article is compliant deleting it is just creating work for the sake of creating work for whoever has to recreate it. (We have plenty of pages all the way up to GA/FA level which according to policy should be deleted under G5—the people who are experienced enough to write high-quality content are precisely the people invested enough in Wikipedia to want to sneak back under the wire when they're kicked out—but nobody is seriously proposing we speedy-delete Virgin and Child Enthroned or Little Moreton Hall.) That said it is decidedly not a matter on which I'm losing any sleep. ‑ Iridescent 09:52, 29 January 2025 (UTC)[reply]

An editor has asked for a deletion review of Existential threat. Because you closed the deletion discussion for this page, speedily deleted it, or otherwise were interested in the page, you might want to participate in the deletion review. Presidentman talk · contribs (Talkback) 15:04, 29 January 2025 (UTC)[reply]

Commented there, but I'll post it here as well for future reference: I have no issue if someone wants to restore this and re-RfD it if there's anything to suggest consensus has changed. However, I can't see any circumstances in which consensus would change; that the primary usage of "existential threat" is clearly not as a synonym for "global catastrophic risk" is evident to anyone who's ever opened a newspaper. Deletion review is for cases when there's a legitimate argument to be made that a given deletion (or non-deletion) could feasibly have ended in a different result. (The "evident to anyone who's ever opened a newspaper" isn't hyperbole; obviously everyone will get different results depending on their area and search history, but at the time of writing only one of the first 10 hits on the phrase on Google News is using it in the "end of the world" sense.) ‑ Iridescent 18:18, 29 January 2025 (UTC)[reply]

With all respect, I can't see any circumstances in which consensus would change is your opinion. I believe there was and is a legitimate argument to be made that a given deletion (or non-deletion) could feasibly have ended in a different result based on the explicitly stated purposes of a deletion review (specifically #2 and #5). In hindsight, I probably should've reached out directly beforehand, but what's done is done. Presidentman talk · contribs (Talkback) 15:28, 30 January 2025 (UTC)[reply]

Regarding declining speedy deletion because the user might have been attempting to create a draft, I considered that possibility, but they said themselves that it was all a fantasy sports team here. They have now blanked the page so I suppose it is moot unless they start doing it again. ☆ Bri (talk) 14:44, 30 January 2025 (UTC)[reply]

Page in question, for any curious TPWs. If they were doing it in article cpace I'd be down on them like a ton of bricks, but the user in question had made a bunch of apparently-legitimate edits on beauty pageants. It's not particularly uncommon for newer editors to goof around with 'fantasy football' type edits in userspace when it comes to complex markup—when you've been around here a while, it's easy for forget just how spectacularly unintuitive MediaWiki's handling of tables is. ‑ Iridescent 19:37, 9 February 2025 (UTC)[reply]

Hey,
Echo didn't go through, hence dropping by. You missed the Afd bottom temp. I have (Wikipedia:Articles for deletion/Nataliia Greshchuk (Diff ~1272973589)) added it. Thanks and happy editing :) — Benison (Beni · talk) 03:10, 31 January 2025 (UTC)[reply]

(talk page stalker)

I don't know what the script thought it was doing on that discussion. I meant to fix it sooner when I saw it closed but on the log and then forgot. Thanks @Benison for the reminder. I couldn't figure it out so just re-closed and I think it's behaving. Star Mississippi 03:21, 31 January 2025 (UTC)[reply]

I suspect I have a three-year-old version of AFDCloser which needs updating. I very rarely close AFDs unless it's one of those mega-contentious ones where someone's asked me to to close it as an admin-of-last-resort—in which case I'd generally do it manually so I can write a full explanation—so it hopefully won't come up too often. ‑ Iridescent 19:27, 9 February 2025 (UTC)[reply]

Happy new year 2025! We had, pictured on the Main page, on 14 January Tosca, in memory of her first appearance on stage OTD in 1900, and of principal author Brian Boulton, and today is Schubert's birthday. I added a pic to his article (and my story) and raised a question on the talk, regarding the lead image. -- Gerda Arendt (talk) 21:10, 31 January 2025 (UTC)[reply]

Glad to see Brian still remembered… (I won't comment on the talkpage—as someone who's never contributed to the article, it's not my business—but I firmly agree that as long as there's a general consensus that a painting looks like the person it's meant to be depicting, it doesn't matter whether it was painted at the time. That's particularly true for people from the early-mid 19th century, when any existing photographs are likely to be of poor quality and where the subjects are generally people who would have been unlikely to have either the money to commission a formal portrait, or the time to sit for one, until well after the period in which they were most active.) ‑ Iridescent 19:20, 9 February 2025 (UTC)[reply]

Eilley Bowers has been nominated for a good article reassessment. If you are interested in the discussion, please participate by adding your comments to the reassessment page. If concerns are not addressed during the review period, the good article status may be removed from the article. Z1720 (talk) 04:05, 6 February 2025 (UTC)[reply]

That looks like a citation density complaint. The GA requirements changed about 18 months ago. WhatamIdoing (talk) 04:51, 6 February 2025 (UTC)[reply]

@WhatamIdoing, For what it's worth, the version that passed GAN was fully cited (at least, as far as I can tell on a quick skim). Bowers is one of the few women from Nevada history who a) has a reasonably well-documented history, b) is far enough in the past that it's unlikely anything contentious will come to light, and c) wasn't a gangster's moll, frothing racist, or something else to render her problematic. Owing to that, she suffers Mary Seacole syndrome—despite being an ultra-niche character, at any given time she's always going to the the subject of a grade school class project somewhere. As such, this was one of those articles where the timer for its deterioration started ticking the moment it was written. (Whatever books I had on the topic, I'll no longer have 18 years on. Dr. Blofeld wrote the bio of her husband and may still have some sources floating about.) ‑ Iridescent 19:23, 9 February 2025 (UTC)[reply]

Yes I am that editor that... uhh... didn't have the maturity for the time to take Wikipedia seriously. Well that was then and that is now. I can't believe it has been more than seven years. As the saying goes the journey is more important than the destination. It is very rewarding to actually be helpful rather than to immaturely seek attention, 14 year old me wouldn't have figured.

I hope to see you around here. Have some cookies to hopefully get you reenergized:

		Cookies!
		Awesome Aasim has given you some cookies! Cookies promote WikiLove and hopefully this one has made your day better. You can spread the "WikiLove" by giving someone else some cookies, whether it be someone you have had disagreements with in the past or a good friend. To spread the goodness of cookies, you can add {{subst:Cookies}} to someone's talk page with a friendly message, or eat this cookie on the giver's talk page with {{subst:munch}}!

Aasim (話す（はなす）) 22:05, 7 February 2025 (UTC)[reply]

Good to see you still about. I know I was snappy with you, but I suspect that in hindsight, you can see why. ‑ Iridescent 19:21, 9 February 2025 (UTC)[reply]

I know. You don't need to apologize, it was justified at that time. It appears this is a recurring theme because before that, I found one editor who got blocked in 2011 because at the time they lacked maturity. More recently there was this editor that I encountered first at the village pump with an outlandish proposal about redirecting the page "Battle for Dream Island" to the Fandom wiki. Digging into that user's edits, they also made unproductive changes to April Fools' pages and whatnot. Some teens lack serious maturity while others are far more mature they can get the admin rights at 16. Also recognizing one's own incompetence is an important skill in any general online project. Aasim (話す（はなす）) 20:34, 9 February 2025 (UTC)[reply]

We gained a great editor on this day :) (Acer's Communication Receptacle | what did I do now) | (PS: Have a good day) (acer was here) 16:23, 15 February 2025 (UTC)[reply]

Thank you for doing the G4 on Ivana Knöll. As the nominator, I agree it wasn't a pure G4 case, but it would have struck me as a violation of WP:NOTBURO to insist that we have a second AfD discussion when the recreated article is so much worse (and much less compliant with BLP) than the originally deleted one. I appreciate the way you approached the decision. Dclemens1971 (talk) 14:02, 25 February 2025 (UTC)[reply]

No problem at all. To me, deleting this kind of BLP is an entirely appropriate place to invoke IAR—since the page would unquestionably be re-deleted at AfD, it would be more of a BLP issue to hold a second discussion, the only effect of which would be to add to Wikipedia's permanent record (and thus the internet's permanent record) a second round of people saying "this person isn't important".

(I've never heard of Knöll before but she does sound like an interesting edge case. She's unquestionably more noteworthy in real-world terms than many—even most—topics which Wikipedia would automatically deem 'notable', but virtually all coverage appears to be in sources which aren't deemed appropriate for a BLP. This is going to arise more and more as the restructuring of the media landscape creates ever more walled gardens where even the more significant figures don't get covered by the sources Wikipedia considers reliable.) ‑ Iridescent 05:09, 26 February 2025 (UTC)[reply]

Hi there! I wanted to reach out because I noticed you recently deleted বাগিরঘাট উচ্চ বিদ্যালয় ও কলেজ (Bagirghat High School and College) via A7. I recognize that the article was a non-English stub and likely should have been deleted in some form. However, this article was for an educational institution, which does not qualify for A7, as you should know. Anyway, I know I'm a newbie admin, but I wanted to bring this to your attention. Take care, Significa liberdade (she/her) (talk) 19:26, 4 March 2025 (UTC)[reply]

• Now there is a 16 year old discussion with not a great deal of participation that probably needs revisiting. There is no reason why educational establishments should not be subject to A7, though I admit that a 2009 discussion probably pre-dated the huge amount of diploma mill spam that we have been inundated with since. (Also, of course, no-one who cannot read Bengali would have even known it was a school, but ... whatever). Black Kite (talk) 19:32, 4 March 2025 (UTC)[reply]

  Indeed. An exemption for schools that are part of a traditional government-run school system might make sense, but I mean... technically when I started editing Wikipedia, I was a student at The Kelly School—enrollment 1, faculty ~0, admin staff 2 (we counted my dog as "Dean of Student"). It's silly to think that, if I had created an article on that, which wasn't even incorporated, by the letter of policy it would need to go to AfD. (That said, so long as policy is what it is, this article should probably be restored and draftified-by-histmerge. Also the creator should probably get a CIR block. I was going to say "until they can find bn:", but no, they've found it.) -- Tamzin^{[cetacean needed]} (they|xe|🤷) 19:44, 4 March 2025 (UTC)[reply]

  @Tamzin: In other words, does Wikipedia is not for things made up one day at school apply where what someone has made up is the school? Newyorkbrad (talk) 03:21, 7 March 2025 (UTC)[reply]

  Yeah, I suppose A11 might apply in that hypothetical (well, if it happened now; A11 didn't exist at the time). That said, the school did issue a transcript that was accepted as valid by a major university, and eventually a diploma—featuring my dog's initials, stylized to look vaguely like a pawprint—when a college insisted I provide one despite already having my HiSET. So, I'm not actually sure where the line is for "obviously invented" there. -- Tamzin^{[cetacean needed]} (they|xe|🤷) 03:33, 7 March 2025 (UTC)[reply]

  All these policies were formulated back when we were much smaller—there's an underlying assumption underpinning all the discussions that everyone involved shares a certain degree of common sense when it comes to where the 'this is obviously inappropriate even if it's technically within the letter of the law' line sits. It's why we consistently do so badly at things like handling the "people who played in one professional sports match and were never heard of again", "every village in Luxembourg" etc data dumps. (By the letter of policy, if I were to write an article about this electrical sub-station in Runcorn it would be virtually undeletable.) It's probably been at least a decade since that assumption that we all had roughly the same basic shared values regarding what constitutes "common sense" held true; I'd be willing to bet that if I opened the discussion, we'd get at least some people earnestly and sincerely arguing that home-schooling does transform your house into an educational institution and thus provided you could find a couple of mentions in the local paper an article on your house would be a viable topic. ‑ Iridescent 05:57, 7 March 2025 (UTC)[reply]

• What you're omitting to mention is that you draftified this well before I deleted the article-space page, and all I did was delete a duplicate page that had been recreated. If you really want bureaucracy-for-the-sake-of-bureaucracy then feel free to undelete the mainspace page, replace it with a redirect to the draft, and then re-delete it as a cross-namespace redirect, but in the absence of any significant history worth merging to preserve, I can't see any obvious reason to do so.

  (Yes, A7 doesn't technically apply to schools. However, translated or not this page wouldn't possibly have met WP:NSCHOOLS. As such I can't see any grounds for deleting the draft and re-creating the mainspace page, which would have been the the only alternative to deleting the mainspace page given that we wouldn't keep both the article and the draft simultaneously—all that would happen would be the mainspace article inevitably being deleted via AfD, so we wouldn't have the article or the draft page at the end of the process. If it's having "A7" in the log that's the issue, I can do an IAR log-revision if you want—we normally don't allow that owing to the potential for people misusing log revision to cover mistakes, but given that this talk page currently has 650 watchers I assume nobody is going to consider it an attempted cover-up if I do so.) ‑ Iridescent 04:58, 6 March 2025 (UTC)[reply]

  A7 was originally established in 2005 because an IP went round and created articles (yes, this was pre Siegenthaler, they could do that back then) about all their high school teachers (Wikipedia:Articles for deletion/Bob Burns), and everybody thought it was a silly waste of time to send them all to Votes for Deletion (as it was called at the time). WP:G11 didn't appear until a few years later, which is probably what they'd be speedied under nowadays if A7 didn't exist, plus the extra restrictions towards article creation we've put in over the years, would have meant A7 wouldn't have been necessary a few years later. Ironically, Bob Burns is now a disambiguation page with numerous articles pointing to it.

  Anyway, since users were forced to be autoconfirmed to create articles, the number of A7s has dropped to virtually nothing, and if you're going to contest an A7, I think it's polite, if you can, to improve the article a bit and make it obvious. (This can sometimes backfire, as seen at the recent templating on my talk page because I felt like being charitable about ten years ago). Ritchie333 ^{(talk) (cont)} 14:24, 11 March 2025 (UTC)[reply]

  I suspect if I went through Category:2000s and its subcats I could A7 upwards of 10,000 pages without much difficulty—don't underestimate how much cruft we accumulated back in the days when anyone could create an article, and we didn't even maintain a fiction of new page patrolling. (If any aspiring wannabe wants to get some hands-on experience of deletion tagging, just head on over to Musical group stubs or Retailing by country.)

  As I've been saying for years, the WMF really ought to consider summarily archiving to a separate site (or at least a separate namespace) every article that averages less than one non-crawler pageview per month. Nothing would be lost since everything would still exist and could be brought back if it were actually useful, but it would mean everyone involved in maintenance could focus on those things that are actually being used. As of today we have 15342 mainspace articles per active admin; fifteen years ago (which was about the time the "Wikipedia has become unmanageably large!" chat shifted from "the usual malcontents griping on WR" to "generally accepted consensus") the equivalent figure was 3528. ‑ Iridescent 06:01, 12 March 2025 (UTC)[reply]

  Even as late as 2008, things could drop through the cracks, such as Wikipedia:Articles for deletion/Kuda Bux (band). Regarding, "the WMF really ought to consider summarily archiving to a separate site (or at least a separate namespace) every article that averages less than one non-crawler pageview per month" - The forking of Wikiproject US Roads to a separate, unaffiliated wiki (although handled with far too much grandstanding and ego posturing) is the right way to go, and I wonder how many of those articles the project previously curated fail to meet that pageview criteria. Ritchie333 ^{(talk) (cont)} 08:59, 12 March 2025 (UTC)[reply]

  Those band stubs have always been a nightmare from the admin point of view. Because the market is so fragmented and localised it's really hard to judge what constitutes "notability", and because the music press is so fragmented and localised it's really hard to judge what constitutes "reliable sourcing". (It also doesn't help that bands' agents and publicists—or just overenthusiastic fans—have a tendency to fill Wikipedia pages with unsourced and inappropriate waffle in an effort to make the acts look interesting, which makes even the obviously significant acts look like a bunch of spammers.) I assume every admin who's regularly worked the deletion queue—and every editor who's every gone through the old stubs looking for inappropriate material to flag for deletion—has the experience of being harangued by angry fans when what appeared to be a blatant WP:SOUNDCLOUDBAND case turned out to be A Big Deal on the Welsh bubblegum-punk scene.

  Most of the road stubs probably would actually meet any given pageview criterion, purely by virtue of false positives from people Googling directions. Just checking the pageviews on an assortment of random UK A and B road numbers—none of which actually have articles—the redirects and blank pages where those articles would have been still meet that pageview criterion. I'm really not talking about setting an unreasonably high bar here. (I'm not picking on niche transport articles here; I'm the author of such things as Droxford railway station, A215 road and Wandsworth Bridge, I certainly appreciate that even the most apparently boring bits of transport infrastructure can be hugely significnt features to the community they serve, and can quite often have a more interesting history than one would immediately think.) ‑ Iridescent 06:23, 13 March 2025 (UTC)[reply]

  I've noticed the sourcing issue in FA reviews too - many song articles are sourced to magazines, where the main evidence of reliability I have are the Wikipedia articles on them. Jo-Jo Eumerus (talk) 07:45, 13 March 2025 (UTC)[reply]

  Entertainment articles (not just songs) are always going to suffer from that issue. The specialist press in all the various -ologies has that whole network of citation counts, online databases, historical reputation, inclusion in university library catalogues, etc etc etc, by which one can establish at least a rough approximation both for whether any given source is likely to be legitimate, and whether an apparently niche topic is actually considered notable/significant by those working in the field. For videogames, (recent) music, sports, and so on, that ecosystem doesn't exist so one has to rely a lot more on experience and gut instinct as to which sources are legitimate and which are promotional/biased/irrelevant. (When it comes to pop culture, the author of any given source is usually more relevant than the journal/site in which the source appears—the personal website of someone well-respected is likely a RS, whereas even the most well-regarded entertainment publications like Billboard and The Stage contain their fair share of puffery and outright advertorial.)

  At the risk of sounding like Wikipedia Review c. 2008, there's also an unavoidable issue of recentism when it comes to modern popular culture. With the benefit of hindsight we can reach a reasonable degree of consensus about which contemporary sources were relevant with regards to J.S. Bach, Mickey Mouse, or the Waverley novels. We can't yet say which sources will be considered significant over the long term when it comes to Sabrina Carpenter or Disco Elysium, and as such there's an understandable urge to shove all the reviews and commentary into the article and hope the readers have to sense to weigh the sources for themselves. ‑ Iridescent 05:28, 15 March 2025 (UTC)[reply]

I was coming here to comment that I found your use of c. to be very cute and it made me smile. But then I saw the underlying wikimarkup added in the edit and I said yikes what happened here. I really hope that's not a widespread bug. --MZMcBride (talk) 20:34, 27 March 2025 (UTC)[reply]

@MZMcBride the explanation isn't exciting. I wiki-grew-up (as did you) back in the days when there was constant encouragement to subst templates on talkpages to avoid making the servers melt down (IIRC, nested-template-heavy pages like WP:FAC still occasionally crash into the template limit), and occasionally still subst them on autopilot. The unsubsted template {{c.}} is a lot more reasonable. ‑ Iridescent 20:40, 28 March 2025 (UTC)[reply]

Pretty sure I've seen FACses overrunning the template limits recently. It's why folks frown on {{done}} templates, for example. Jo-Jo Eumerus (talk) 10:38, 29 March 2025 (UTC)[reply]

Hi, Iridescent!

I just wanted to thank you for taking the time to review the Frantz article and for removing the deletion request. I appreciate your assessment and help in keeping the article up. Thanks again!

Editora89119 (talk) 11:10, 16 March 2025 (UTC)[reply]

@Editora89119, be aware that although I've declined the speedy deletion request, that hasn't immunised it against deletion; if the editors who feel it's inappropriate for Wikipedia want to nominate it for the full "week-long discussion" deletion process they're entirely within their rights to do so.

The page in question is Jonathan Frantz, if any talk page watchers are interested. Laser eye surgery is not a topic on which I have enough knowledge to give any particularly informed opinion, but it might be worth asking at Wikipedia talk:WikiProject Medicine—Wikipedia has a lot of editors who are genuine subject matter experts in medicine and might be able to expand it (or conversely, to explain why it's not in fact an appropriate biography for Wikipedia and save you wasting time expanding it). ‑ Iridescent 15:20, 16 March 2025 (UTC)[reply]

I would not expect WP:MED to embrace this article. I would expect some of the older hands to think about that time when the Alaskan plastic surgeons all hired a PR firm to spam in articles about them. I'm pretty sure something similar happened with LASIK-related subjects, too.

@Editora89119, some of your responses at Wikipedia:Articles for deletion/Jonathan Frantz sound a bit like someone said "Dear ChatGPT, please give me a 500-word-long answer about why Wikipedia should not delete this". Effective answers sound like "Here's a link to 600-word-long news article entirely about Frantz, and here are links to the widely recognized awards he's won, and here are links to reputable trade rags that profiled him..." Seriously: Just posting the link https://winknews.com/2019/09/06/cataract-patient-first-to-get-latest-lens-implant-technology-in-swfl/, with nothing else, would have been just as good as all the words you wrote around it. (Also, you seem to have posted the same thing twice.)

And you still haven't posted there a clear statement about whether you're being paid to write this article. Paid editing is legal. Undisclosed paid editing, however, is a serious problem. So if that's you, then disclose! And if it's not, then say so. WhatamIdoing (talk) 03:12, 17 March 2025 (UTC)[reply]

@Editora89119, let me second the good advice you're being given above. Regardless of whether they are AI generated, your wall-of-text replies at Wikipedia:Articles for deletion/Jonathan Frantz read as if you've instructed an AI to write an essay defending the article. People understandably get annoyed if they feel their time is being wasted; your best rebuttal to people saying a topic isn't notable isn't a long essay, but a concise "this topic is notable, here's a list of external reliable sources which discuss its impact".

Also, I echo the advice to make it clear if you're being paid to write the article (or if you're otherwise connected to Frantz). Wikipedia allows people with a conflict of interest to edit here provided they declare the conflict of interest so that third parties can assess any potential bias. What we don't allow is an undeclared conflict of interest, which we treat as attempts to deceive and/or to manipulate our content. ‑ Iridescent 16:14, 17 March 2025 (UTC)[reply]

Hi I responded to that on the Articles for deletion page. Pasting it in here - I went to his office and had LASIK. He did not perform it though. Another doctor performed the procedure. I started a conversation with him and learned about Dr. Frantz, who I had seen on local news. He told me about Dr Frantz living in Louisiana where I once lived and studied. It was fascinating to me because I am connected to the medical world-- my husband is a cardiologist. I then talked to my husband about LASIK and how it was pioneered, and he helped me research in medical journals. I simply found it a fascinating subject. I have been a bit determined on this because once I start something I like to finish. I love this process and want to start and edit more articles. It's like I've found my calling. For the record, I have never even met Dr. Frantz. I have only seen him on TV and in magazines. Editora89119 (talk) 17:40, 17 March 2025 (UTC)[reply]

Thanks for clarifying. I hope your LASIK procedure had a good outcome.

You are welcome to come hang out with us at Wikipedia talk:WikiProject Medicine. We actually need someone who is willing to put some time in on biographies for healthcare professionals. Wikipedia:WikiProject Medicine/Article alerts usually has a list of articles that need attention. Many of them are about people and businesses. Feel free to help out. Even just reading through some of those pages might give you a better idea of what's expected and helpful in discussions like the AFD for your new article. WhatamIdoing (talk) 01:01, 18 March 2025 (UTC)[reply]

Editora89119, to increase the chance of your seeing it I'll copy across the comment I just made at Wikipedia:Articles for deletion/Jonathan Frantz in light of your being blocked from that discussion:

@Editora89119, I do agree with Star Mississippi's rationale for blocking you from this discussion—as you've been warned both here and at my talkpage, your discussion style on this page is becoming actively disruptive. If there's a comment here to which you feel you really need to reply, then post the proposed comment in your thread at my talkpage and I (or one of the other people watching my talkpage) will copy it across to this page if it's not something that's likely to get you in trouble. ‑ Iridescent 15:55, 18 March 2025 (UTC)

If you want to comment in that discussion, then post the comment you want to make here and if it's appropriate someone will copy it across to the discussion. (Your being blocked from that discussion might seem harsh, but it's primarily to prevent you getting in trouble. If you're going to tell multiple highly experienced Wikipedia editors that they're not understanding Wikipedia policy, then unless you can explain why they're wrong it's just going to irritate people. You presumably don't want to end up blocked from Wikipedia altogether, which would potentially be the outcome if you give the impression that you're unwilling to respect other people's opinions.) ‑ Iridescent 15:55, 18 March 2025 (UTC)[reply]

Thank you. Admittedly, I had never done a full article and tend to be hyper-focused and goal oriented to a fault sometimes. The only issue I have is why the double standard when it comes to academics and private sector pioneers? And I reiterate that his own colleague, Marguerite McDonald, has far fewer sources / citations measuring up to Frantz. Thank you for allowing me the space here (and perhaps to vent). Editora89119 (talk) 14:56, 19 March 2025 (UTC)[reply]

If we were going to change that "double standard", it would likely be in the direction of making none of them qualify for (separate) articles. See Wikipedia:An article about yourself isn't necessarily a good thing for one of the reasons why it might be better to err in the direction of excluding individual people.

BTW, if the AFD closes with deletion, you can request a WP:REFUND to your userspace (or you could just copy/paste your own work to a page named something like User:Editora89119/Frantz now, before the deletion happens). That would make it possible for you to keep working on it, but you should not WP:MOVE it back into the mainspace without jumping through all the hoops. WhatamIdoing (talk) 18:50, 19 March 2025 (UTC)[reply]

Saved. THANK YOU SO SO MUCH! Editora89119 (talk) 12:06, 20 March 2025 (UTC)[reply]

Wikipedia:Articles for deletion/Jonathan Frantz is a fun read. I do wonder what happens when the undisclosed use of large language models becomes harder to detect. The tone and em dashes in this case are a pretty clear tell today, but I imagine the models or prompts or user behavior will start to mask what's really happening.

It feels like these services really ought to start charging people to use these tools to keep the spam rate down. It's kind of good that it costs 73 cents to send a piece of postal mail and kind of bad that people can send e-mails for basically free. On the other hand, spammers on Twitter weren't really deterred by paying for the verification badge so I dunno. --MZMcBride (talk) 20:26, 27 March 2025 (UTC)[reply]

And the only thing in my mailbox most days is advertisements. WhatamIdoing (talk) 22:51, 27 March 2025 (UTC)[reply]

From the Wikipedia point of view, I'd argue that a kind of Turing test applies. If the LLM-generated material is genuinely indistinguishable from human input, does it even matter who's behind it provided that in the case of articles the material created is neutral, accurate and reliably sourced (something LLMs are currently decidedly unable to do), and in the case of discussions that the arguments are coherent and genuinely based in policy?

Considering some of the waffle written by genuine humans here, it's hard to say the AI stuff is significantly worse. The usual tells of LLM content (slightly pompous tone mixed with colloquialisms, inconsistent approach to grammar, preference for falsifying sources when faced with apparent inconsistencies, unattributed copying…) could describe about 75% of Wikipedia's internal discussions.  ‑ Iridescent 20:57, 28 March 2025 (UTC)[reply]

Someone (maybe @MastCell? I can't find the original) years ago said that every time they found a newcomer writing well, it turned out to be a copyvio. WhatamIdoing (talk) 03:32, 30 March 2025 (UTC)[reply]

To be honest both ChatGPT and copyvio are mistakes for which I can't really blame new editors. The learning curve for Wikipedia may not be any steeper than it was 20 years ago, but the slope goes on much higher with 20 years of accumulated policies, guidelines, and unwritten ways-of-doing-things; it's no surprise that people have difficulty understanding which corners are OK to cut. (Neither you, me, nor MZMcBride would say we're familiar with all the things one is supposed to do, and we're all about as insider as one can get.)

Couple that with the fact that the older generations have now had 25 years to get used to Wikipedia, and thus a significant proportion of new editors are going to come from a generation who grew up in the culture of social media where copy-pasting is seen as routine and uncontroversial.* It's a miracle we have any competent new editors who arent either plagiarising, using AI to "improve" writing, or sockpuppets of people who learned Wikipedia rules back in the days when you could still pick them gradually as you went along without people shouting at you when you got them wrong. ‑ Iridescent 06:47, 30 March 2025 (UTC)[reply]

• Before anyone says it, yes I know the culture of copying texts has existed since the clay tablet era, but Web 2.0 transformed it from something people did occasionally for something they found particularly significant, to something done daily and routinely.

do you think in the near future, after the AI boffins get around to fixing the hallucinations and crappiness, that wikipedia will have a local article creation LLM where editors essentially fill out a form, link all the sources they've found, press generate and WikAI spits outs a perfectly formulated article, templates, embeds, infoxboxes and all? I started on this site nearly a decade ago to improve my writing, but a combination of creative block with lack of expertise with the technical aspects made me abandon article creation by and large for counter-vandalism and now copyvio patrolling. It's been discussed at length how all the sexy articles have been written and much of article creation going forward will be writing 3-5 paragraph pages on long dead BLPs who happen to meet GNG. ^Thanks,L3X1 ◊distænt write◊ 15:58, 30 March 2025 (UTC)[reply]

They've been working on a hand-coded, multilingual approach to Wikipedia. Imagine if you could code a template – maybe something like "{SUBJECT} {#be:singular|third person|tense} {#wikidata:p106|gender}." – and have it spit out "Isadora Duncan was a dancer" or "Sarah Lamb is a ballerina", or the equivalent in the local language, any time someone put your template into an article. It probably wouldn't produce brilliant prose, but it would give total control to editors, while working at scale. Firstly, there's no room for hallucinations, and secondly, individual editors decide whether or not to use each templated function. WhatamIdoing (talk) 16:14, 30 March 2025 (UTC)[reply]

They've been working on a way to auto-convert Wikidata into text for what feels like longer than Wikidata itself has existed (remember Reasonator?). Way back when, Haitian Wikipedia had the dubious privilege of being the testbed for the software to auto-create a pseudoarticle based on Wikidata if you searched for a topic that didn't have its own article. (I assume Haitian was chosen as it's a language with so few users that it wouldn't hugely annoy millions of people, but which is close enough to French that large numbers of people could tell at a glance if it were spewing gibberish.) On a quick dip of topics on which they don't have articles, it looks like they now just serve up the Wikidata entry rather than trying to parse it into text.

What you're talking about—making each sentence into a collection of templates which can autoconvert between languages—I can't see catching on; writers can grudgingly accept the intrusion of templates into blocks of text when they accept it's necessary for the formatting, but only crazy people actually write in wiki markup. This sounds like the WMF—not for the first time—has made the mistake of thinking that "the way professional programmers think, act, and communicate" is the same as "the way the other 99.99% of the world thinks, acts and communicates". (Aside from anything else, such a system would presumably finally slam the lid on the coffin of their beloved Visual Editor, and I can't see the WMF signing off on that.) ‑ Iridescent 20:09, 30 March 2025 (UTC)[reply]

mw:Extension:ArticlePlaceholder is on several wiki. As an occasional contributor to the Haitian Creole Wikipedia, I'm glad that it's there.

Also, m:Wikifunctions wasn't the WMF's idea. It was formulated and promoted by a long-time editor (since ~2003). I'm surprised that the WMF took it on at all, and I'm doubly surprised that they didn't hand it to WMDE, but I think that it will be popular with the smaller wikis. A locally written article would usually be best, but if you're not sure what to write, or you're trying to create a basic set of articles (htwiki is mostly one-sentence boilerplate articles about US census locations; Reasonator would be an improvement), or if you don't want to mess with updates (e.g., new census reports), then I can see editors choosing to drop a pre-written {#basic intro} or a {#list of works} or a {#demographics} template into an article rather than starting completely from scratch. WhatamIdoing (talk) 21:35, 30 March 2025 (UTC)[reply]

I'm not insulting the principle of "some kind of summary is better than nothing" when it comes to the smaller Wikipedias. I completely agree that provided the Wikidata entry is correct, it's better for the reader to at least have some kind of "this should give you some idea of what we're talking about" summary than to be faced with the generic "This page does not exist. You can click here to create the page directly, or you may create a draft and submit it for review." when clicking on a redlink.

What I don't believe is that the WMF devs have the necessary skillset do do true parsing of data into coherent text across multiple languages, when even billion-dollar tech corporations struggle with it. Even something basic like "generate a plain text version of the infobox" is fraught with difficulties given how easy it is to miss nuances when translating. Particularly if the US government carries out its threat to revoke §230 (I suspect Musk will talk Trump out of it as the liability to Twitter would be huge, but the intention is there), I can't imagine the WMF's lawyers would be particularly keen on this. If we're machine-translating between 342 languages, all it takes is a single mistranslation leading to (e.g.) someone getting lynched / a product being boycotted / a health scare / etc, and the reputational damage and financial liability are potentially huge. (Paging Yngvadottir if she's still around, as she can generally make this argument more eloquently than I.) A trivial example off the top of my head of how machine translation between wikis could go seriously wrong would be Ce produit contient des traces de poison—a human editor will (hopefully!) see this as "this product contains a trace quantity of fish" with a typo, but machine-translation will quite happily translate it as "this product is toxic". ‑ Iridescent 06:14, 31 March 2025 (UTC)[reply]

Except that this system won't ever see "traces de poison". It will see something like "d:Q351792 of d:Q600396". WhatamIdoing (talk) 20:47, 31 March 2025 (UTC)[reply]

This is a trivial nitpick (but hey, this is Wikipedia!), but in French, poison means poison, while the word for us fish is poisson. --Tryptofish (talk) 20:51, 31 March 2025 (UTC)[reply]

Yes, that's the typo that could lead to mistranslation. WhatamIdoing (talk) 21:11, 31 March 2025 (UTC)[reply]

Ooooh lovely, this talk page is once more hosting wide-ranging and interesting conversations, and Iridescent is back! Maybe that should be the other way round. ... My hot take is that, first of all, this is another condescension to "small wikis" couched as benevolence. It should be left up to the editing/language community of each Wikipedia how they want to go about expanding their coverage, including whether they want to largely delegate it to bot or human translation, or to auto-generate entries in some other way, what their priorities are for new articles, or even how they would like articles to be formatted. It's not the WMF's, or en.wiki's, concern if their versions of Wikipedia are small, or "unbalanced" in their coverage, use or don't use infoboxes, or have more or fewer pics. As I recently said on the unnameable site, "Either let them have their own Wikipedias, or don't." Secondly, that this is a separate pickle from the WMF's promotion of machine translation, and by "pickle" I mean that it's also going to produce garbage, but in this case through GIGO. Wikidata entries are usually not entirely correct: they are often either duplicates or conflations, they enshrine information drawn directly from Wikipedia (including things that are correct on one version of Wikipedia and incorrect on others; it's soul-destroying to keep trying to fix something like a date of birth, especially since the interface is alien and quite hard, and adding a reference is so hard I've never figured it out—and in any case Wikidata has almost zero vandalism patrolling and no edit summaries, so it will just be switched back). Wikidata doesn't handle well where a topic is an article on one Wikipedia but a section or a redirect on another Wikipedia (and I don't see how it can, especially since it's sometimes a matter of different judgements of importance in different cultural contexts, and sometimes purely accident or expediency, like the fact we cover Eichler Homes as background in Joseph Eichler rather than the other way round). Plus, the Babel issue particularly affects Wikidata: the site is more monolingual even than en.wiki, so both the accuracy and the sorting of foreign-language data are rock-bottom. (Where there are foreign-language descriptions of "items", they're often error-filled; and there used to be a tremendous knot over linking the Midsummer Night's Dream articles to Japanese, since the Japanese translation of the phrase is the title of a film that was at that title on ja.wiki—I hope someone has now sorted out that mess, but I'm sure there are other such messes.) Yngvadottir (talk) 22:54, 31 March 2025 (UTC)[reply]

I have created Wikipedia:How to add sources to Wikidata for you. It's quick and easy.

I don't know how to add page numbers yet, but since most of my refs involve PubMed, it's not been a question I've personally needed to address. That might be of more importance for you than it has been for me. WhatamIdoing (talk) 23:33, 1 April 2025 (UTC)[reply]

Thanks. That will enable me to add books with ISBNs. But I see that I would have to use "import URL" to add a reference to, for example, a news article or obituary. (The case in which I've most wanted to fix this circularity is dates of birth.) I know I can't manage that. (And of course not all books have ISBNs; does it by any chance accept OCLC numbers?) Yngvadottir (talk) 02:05, 2 April 2025 (UTC)[reply]

I don't know what "import URL" does, so I suggest using "reference URL". You can manage it. You just paste the URL into the little box. It seems to take any URL, so you could also use, e.g., a link to an exact page in Google Books.

Yes, OCLC numbers work. Type "OCLC number" in the ref property box to find the right one. WhatamIdoing (talk) 02:23, 2 April 2025 (UTC)[reply]

I'm not sure I agree on another condescension to "small wikis" couched as benevolence. Presumably if the Haitians (or whoever) don't want the auto-generate placeholder 'articles' they could choose to turn them off. My point isn't an issue with the existence of the placeholders; it's that their general unreadability illustrates just how far off MediaWiki developers are from being able to create readable text on the fly.

With the disclaimer that I haven't been following Wikidata particularly closely, my experience of it tends to align more with Yngvadottir than with WAID. The level of inaccuracy which we'd never trust on Wikipedia itself, a toxic and dysfunctional internal culture, a very unfreiendly editing interface, and a lack of clear processes for checking and correction, combine in my opinion to create something we shouldn't be using. (The relevant thought experiment might be, "If Wikidata didn't exist but there was an identical-in-every-way site run by Amazon or Google, would we trust it?". Given the history at RSN of other user-generated data sites like IMDB, the answer would almost certainly be no.) Given that we (rightly) don't trust other language Wikipedias as sources without independently verifying every source for every claim, I don't understand the attitude that we dhould grant some kind of exception for Wikidata.

Yes, I'm aware this is starting to veer some distance from the original point, which if I'm understand it seems to be more about creating coded fragments to make machine translation easier. I still think that's problematic—even something as basic as "Isadora Duncan was a dancer" needs a source, and the various Wikipedias vary in their attitudes both to which sources are reliable and how citations should be formatted. (Also, I don't think it's too WP:BEANS for me to point out that it would probably usher in golden age of crosswiki vandalism.) ‑ Iridescent 04:04, 3 April 2025 (UTC)[reply]

Yes, "Isadora Duncan was a dancer" would benefit from a source, and the source could be coded into it.

It's true that at enwiki, we believe we're mostly better off without Wikidata. However, other wikis make the opposite choice, and prefer to rely on it as much as possible.

(I agree with the BEANS concern. One tool Wikidata has for mitigating that is that you can set a bot to prevent or revert individual edits. Imagine if an antivandal bot here could be told to watch a specific infobox parameter in a specific article, and to restore that parameter if anyone removes or changes it.) WhatamIdoing (talk) 20:42, 3 April 2025 (UTC)[reply]

There's no technical reason we can't do that now. The reason we don't use micro-Cluebots set to patrol a particular sentence, paragraph, or paramater and revert any changes made isn't technical, it's social—"Anyone can edit" is so fetishized here. (I can think of very few cases when this would actually apply on the Wikipedias. It works on Wikidata because that's not public-facing in the same way; when it comes to Wikipedia, what information is included is always a value judgement and as such pretty much any piece of information could potentially be removed in good faith from any given article or template.) ‑ Iridescent 04:56, 4 April 2025 (UTC)[reply]

I think it works on Wikidata because it's structured data. A bot can reliably tell whether "occupation = dancer" is in the entry. It can't reliably tell whether "was a dancer" being turned into "danced professionally for three decades" means that the words "was a dancer" need to be crammed back into the article. WhatamIdoing (talk) 05:37, 4 April 2025 (UTC)[reply]

Based on things I've seen recently, I have a completely separate concern about AI and Wikipedia, which I've raised on my talkpage. I'd appreciate any comments there. Thanks, Newyorkbrad (talk) 15:32, 2 April 2025 (UTC)[reply]

As this reply is somewhat tangential, I'll reply here rather than at User talk:Newyorkbrad#Question for discussion, so any back-and-forth doesn't divert your thread down a sidetrack. Feel free to copy it across to your talk if you want all the discussion in one place.

IMO this is one of those rare occasions where you'd actually be better off having the discussion somewhere like Wikipediocracy, rather than on-wiki. For long boring reasons Wikipedia has always had a disproportionate number of "better data processing is the solution to everything" True Believers, and in my experience they tend to shout the loudest. As such discussions held on Wikpedia about how we handle data tend both to get very heated and to give a distorted picture of consensus, as those sceptical of Big Data quite understandably feel they have better things to do than be ranted at. (Your memories are no doubt just as fond as mine of people ranting about how if we didn't follow their preferred formula on infoboxes, categories, date formatting etc it would RENDER WIKIPEDIA UNUSABLE!!!!!!) For a discussion like this, the input of people like Somey, who understand Wikipedia and the broader internet/data/AI ecosystem well enough that their opinions are worth hearing, but who aren't invested in pushing Wikipedia down one path or the other and aren't worried about whether they upset people on-wiki who are strongly pro- or anti-AI, is probably more valuable that the opinions of the people who spend all their spare time on-wiki.

My 2c on the issue you raise regarding AI scraping of talk pages would be that it doesn't really matter. These talk pages are all publicly published so there's no privacy or copyright issue—{{NOINDEX}} is just a courtesy to the rest of the internet to prevent search results getting cluttered with shitty drafts and tedious internal Wikipedia discussions. (The BLP issue I see someone raising is IMO a red herring. Obviously we don't want libel or untruth on talk pages for legal and ethical reasons, but if an AI scrapes something inappropriate and doesn't factcheck it then as far as I'm concerned that's the AI's problem not ours—any AI operator who isn't teaching their AI not to blindly trust Wikipedia isn't doing their job.) I can't see it having any significant impact on AI output—given that LLMs are being trained on 20+ years of social media posts, the distorting effect of any inappropriate talkpage content would be a drop in the ocean. (There's one very specific use case where this would be a significant issue—an AI bot that can mimic a new Wikipedia editor gradually learning, such that one could set a bunch of scripts running and then come back a couple of years later to harvest a crop of fully-grown sock accounts, all well-enough regarded that they could then work in lockstep to overwhelm discussions. I suspect we're still decades away from the point where a LLM could consistently pass the Turing test day-in-day-out in the Wikipedia environment, so I'm not losing sleep over it.) ‑ Iridescent 04:40, 3 April 2025 (UTC)[reply]

I sometimes used AI to format citations because doing so by hand is just painful (both mentally and physically). I also thought to see if ChatGPT could propose a rewrite of (parts of) Lake Tauca, which might make a good FAC candidate but would need a serious rewrite. Jo-Jo Eumerus (talk) 08:53, 3 April 2025 (UTC)[reply]

Unless and until ChatGPT can understand how to cite sources, how to select which sources are appropriate to cite, and (most importantly from our point of view) when to recognize a situation where we need to omit something because the only sources that mention it are low-quality, it's not going to be particularly useful on Wikipedia except in limited "rewrite this paragraph to sound less verbose while still meaning the same thing" situations. For something like Lake Tauca, where presumably a significant proportion of the sources are going to be in Spanish, my confidence in AI drops to something approaching zero—even bilingual humans struggle with miltilingual sourcing and weighing not only the relative reliability of sources in different languages but the fact that different languages represent different cultures with potentially different ideas of which elements are significant. (I'll fall back on my go-to example of en:Texas Revolution, es:Independencia de Texas and ca:Guerra de la independència de Texas, all of which are written by highly-regarded editors on their respective wikis and which are two FAs and a FFA respectively, but which are so different in terms of which elements they give significance to they could almost be describing different incidents.)

Strangely in light of certain other unpleasantnesses, the only LLM in current widespread use that seems to grasp the concept that "reliable source", "unreliable source" and "potentially unreliable source" are three different things is Elon Musk's pet Grok system, which for other reasons one would hope are obvious I wouldn't trust within a mile of Wikipedia. ‑ Iridescent 16:05, 3 April 2025 (UTC)[reply]