user talk:Jimbo Wales

Hi Jimbo. Over at Meta.Wikimedia, @LWyatt (WMF) and I had an interesting good-faith discussion about Wikimedia Enterprise ([https://meta.wikimedia.org/wiki/Talk:Wikimedia_Enterprise#FY25-26_Annual_Plan_for_Enterprise_&_Tech_Partnerships link]). My opinion: Enterprise, while probably worthwhile, is an area of risk for our core values, and needs to be carefully handled. To summarize, I asked who Enterprise's customers were, and how much they paid. In the course of the conversation, it became clear that the WMF is not currently able to release that information, because large companies usually consider their business contracts proprietary information. The WMF has been able to mention a few Enterprise clients by name in public, but not all, and never how much they paid. We also discussed the distinction between expectations of privacy for charitable donations and privacy in commercial transactions.

My questions for you (and by extension, the Board): do you think that the community has a need to know who Wikimedia Enterprise's customers are? Accordingly, do you think the WMF Board should place a higher priority on making the names of all Enterprise customers freely available, even if that means some customers disengage? —Ganesha811 (talk) 16:24, 10 June 2025 (UTC)

::I think that a balance is needed. I'm curious to know a bit more specifically what kinds of issues you're most concerned about, and what kind of resolution would be satisfactory? For donations, we have a longstanding practice that the WMF board is directly informed of donations above a certain size. It's never raised any problems, but I think it's pretty obviously good practice. I don't know at this moment if there's a specific convention in terms of that kind of policy for Enterprise, but it wouldn't be a problem obviously for the board to know.

::If you have specific concerns, please let me know, and we can think about whether a policy is needed in order to say "we don't want this type of customer" or "we need to have this sort of protection baked into contracts".

::What I don't think makes sense, is the very much not-assume-good-faith concept that it's desirable for the community to have the names of all Enterprise customers. I could be persuaded of course, but I'd be interested to know what the purpose would be.

::We've had the same sort of discussion around other contracts for provision of services. Let me put forward a hypothetical: imagine that a big supplier (hosting service or content distribution network) loves Wikipedia and does a deal with us on extremely favorable terms but would rather not make those terms public as they would cause their more commercial customers to go: hey, I want that big discount too. This would make giving us a sweetheart deal a bad idea for them, even though they'd love to do it.

::The downside of course is that there's no direct way for the community to confirm that we aren't insanely overpaying for services - but that's why we have a majority-community and purely voluntary board of directors and high quality staff who are looking to do things in the best possible way for the projects.Jimbo Wales (talk) 07:05, 11 June 2025 (UTC)

:::My concerns with Enterprise fall into two categories, fundamental and specific. Fundamentally, I am suspicious because it is a for-profit, commercial service dedicated to selling access to Wikipedia data for money ([https://meta.wikimedia.org/wiki/Wikimedia_Enterprise/FAQ as described in the FAQ]). Since every part of that cuts against our basic ethos, I think it needs to be very carefully managed. Specifically, I am concerned that it will do damage to the culture of the WMF, and create a stream of revenue for the organization that is based not on Wikipedia's value to readers, but on Wikipedia's value to major corporations and AI developers. There's now a large and dedicated team of WMF staff who work on meeting the needs of those customers, not on meeting the needs of editors or readers. Their paychecks depend on it. Transparency - sunlight - is the simplest way for everyone, including the WMF, to ensure that Enterprise does not ever lead to issues at the WMF, either for culture or finances. Without the most basic information (who pays for it and how much), how could we really know what impact it is having?

:::I'm not sure why it's not AGF to ask the WMF to be transparent. The WMF is a non-profit organization here to serve the needs of the project and the movement. Transparency has always been central to that work and the WMF is very transparent in most ways, as is Wikipedia. When the WMF spins off a for-profit group to sell a commercial service, the bar should be set higher, not lower. The purpose of transparency is to let third parties (i.e. the community) have full knowledge of the WMF's commercial work, and therefore have a chance of detecting issues as they arise. It's to the WMF's benefit to be transparent - the WMF keeps an eye on the community, and we keep an eye on the WMF - a positive reinforcement loop which depends on openness and trust.

:::As to your hypothetical, I would want the Board to reject such an offer. The WMF should not put itself in a position where part of its financial plan depends on the continued beneficence of a for-profit corporation. What happens if two years from now they decide they hate our article on Zionism, or Kashmir, or Lord of the Rings, and cut us off? Real independence means avoiding these sorts of entanglements and sweetheart deals. (Incidentally, has something like your hypothetical ever occurred?) I don't doubt the good faith of the current WMF staff, and I understand why Enterprise was created; but by its very nature I believe it requires greater transparency than it currently has. —Ganesha811 (talk) 12:23, 11 June 2025 (UTC)

::::And to be explicit about "what kind of resolution would be satisfactory", rather than just implying it; a satisfactory resolution would be the WMF Board adopting a policy that all new or renewed Enterprise contracts must specifically allow the WMF to publish the name of the customer and the dollar amount of the contract, and that the WMF then publishes that information regularly. —Ganesha811 (talk) 12:34, 11 June 2025 (UTC)

Hello Ganesha811, than you for tagging me above. Naturally, I leave it to Jimmy to address your questions to him and the Board of Trustees. Nevertheless - for the benefit of anyone else here who has not read our conversation over on the Wikimedia Enterprise talkpage -  I would like to highlight the responses here to a couple of points from there so all the information is in the one place.

• selling access to Wikipedia data for money''”: as discussed before, anyone may access that API for no cost, and the data within that service is still the same freely-licensed Wikipedia (etc) content. It is just structured in data-formats more applicable for third-party organisations. In accordance with the project’s principles there are no exclusive contracts nor exclusive content. The “for money” part is when the volume/speed of that usage becomes extremely high and/or if the user needs a contractual guarantee of uptime and customer support. The distinction between a contract for access to a service vs. a content license, is indeed quite significant – legally (both in terms of contract and copyright law) and culturally terms of the way that the whole Wikimedia ecosystem operates. You can read [https://enterprise.wikimedia.com/pricing/ the specific pricing structure here] and [https://enterprise.wikimedia.com/docs/ all the technical documentation here].
• “There's now a large and dedicated team of WMF staff who work on meeting the needs of those customers, not on meeting the needs of editors or readers”: This implies there is a zero-sum approach, where a fixed/finite number of staff and budget exists - which is not the case. No engineer has been taken away from editor projects to do this, and ALL the revenue raised by it goes into the general WMF budget to be allocated in accordance with the annual plan. By becoming profitable and by trying to get the very high volume commercial organisations use this service ([https://diff.wikimedia.org/2025/04/01/how-crawlers-impact-the-operations-of-the-wikimedia-projects/ rather than by intensively scraping/crawling which is very expensive to WMF infrastructure]) this raises more revenue to be used for the WMF Annual Plan and also lowers expenses for the existing infrastructure.
• “When the WMF spins off a for-profit group to sell a commercial service, the [financial reporting transparency] bar should be set higher, not lower.”: As we previously discussed, and for precisely the reason you describe, the project’s principles include "The publication of overall revenue and expenses, differentiated from those of the Wikimedia Foundation in general, at least annually". Obviously the Enterprise revenue/expenses are also included in the overall WMF audit report etc too, but this extra detail is is specifically to allow everyone to see what is happening about this specific part of WMF finances differentiated from the rest. So, yes, to your point – this IS more transparent than other parts of the WMF, not less. [https://foundation.wikimedia.org/wiki/Wikimedia_Enterprise All those annual financial reports are here]. Relatedly, the next technical/strategic objectives for the team for the forthcoming year are on Meta here.
• “The WMF should not put itself in a position where part of its financial plan depends on the continued beneficence of a for-profit corporation.”: Absolutely agreed. This is precisely why the financial oversight rules of the WMF Board of Trustees for commercial contracts match exactly those for philanthropic gifts, or any other kind of money. ALL large gifts or contracts or grants must be reviewed by the Board. It does not matter how money is received - the Board has the same responsibility to monitor it. Furthermore, as stated both in that Board document and in the project principles - commercial revenue to the WMF from all sources combined (which includes both this but also things like selling merchandise) shall not exceed 30% of the total. This means that the WMF can never become too dependent upon any individual [person or organisation] financially.
• And finally – we have always, and will continue to, state our preference for public acknowledgement of any customer contract of the Wikimedia Enterprise API service - [https://enterprise.wikimedia.com/blog/tag/partner/ Here are some example case studies that have been published] for example. This strong preference is consistent with how major donors to the WMF [some are individuals, some are philanthopic organisations/foundations, some are commercial companies] are also encouraged to publicly show their support. But some of them chose not to, and that is their right.

LWyatt (WMF) (talk) 11:32, 12 June 2025 (UTC)

:Thanks for adding that context, Liam. I sense your distaste for the phrasing that the WMF is "selling access to Wikipedia data for money" - and that's natural! But that is precisely what you go on to describe. Of course that data is also still available (in a different format/technical structure) to others for free - I don't think I've ever implied otherwise. But that's what the WMF is doing - selling a specific type of access to a subset of organizations who use Wikipedia data for profit. And, as you say, providing them with dedicated technical support. I don't believe that WMF staffing is a zero-sum game, but it is undoubtedly reflective of an organization's priorities. An issue I've seen develop at other nonprofits is the expansion of staff dedicated to fundraising, with the same justification that staffing is not a zero-sum game. As long as any individual fundraiser with a salary of $X brings in $(X+1) each year, why not keep hiring them? Eventually, though, the organization distorts itself away from its original mission. To that point, a 30% ceiling for commercial revenue is far too high - the WMF budget is substantial and if commercial revenue ever reached that share and then unexpectedly dropped, it would have dramatic impacts on WMF budgeting. Finally, I will say that the level of transparency (as well as the revenue ceiling) are, practically speaking, set by the Board, which could choose to require customers to waive their "right" to commercial privacy, but has not done so. Hence my question for Jimmy. —Ganesha811 (talk) 12:39, 12 June 2025 (UTC)

::Apologies that I've been slow answering your response to my questions! I'll hopefully get to it tomorrow or Friday - don't be shy about chasing me, I don't want to ignore you! Jimbo Wales (talk) 15:28, 18 June 2025 (UTC)

:::That's ok, this isn't an urgent issue. I'll ping you next week if I haven't heard back then. Thanks for considering it thoughtfully! —Ganesha811 (talk) 15:33, 18 June 2025 (UTC)

File:AI proofreader2.png

File:AI Source Verification.png

Hi Jimbo. Can you get a WMF JavaScript coder to contact me?

• User:Polygnotus/Scripts/AI Proofreader: This AI proofreader suggests improvements to Wikipedia articles. Gemini (free), Claude and ChatGPT are supported.
• User:Polygnotus/Scripts/AI Source Verification: Check if a claim is supported by a source. Uses Gemini/Claude/ChatGPT to extract the relevant text from the source and compare it to the claim in the article. It also shows the user the relevant text so they can confirm that it supports the claim.

This is what we want. AI that supports editors. The human makes the decisions, the AI can only propose improvements but is incapable of doing anything. Wikipedians know how to deal with unreliable sources (like AI).

The idea is to use these scripts to prove to the Wikipedia community that not all AI is bad.

I am not a JavaScript coder. These scripts should be improved and deployed. It would be nice if the WMF could play a role in that. Thank you, Polygnotus (talk) 12:23, 12 June 2025 (UTC)

:For what it's worth, I believe verifying sources with AI has been attempted before, and it was found that even when directly given the linked page, LLMs still hallucinated enough to make it not worth pursuing. Things might have changed, of course, but worth keeping in mind. — Qwerfjkltalk 17:53, 12 June 2025 (UTC)

::The point is not "verifying sources with AI", the point is having an AI try to find the text that supports the claim in the article, and show that to the user who can then determine if the source supports the claim. The script also tries to make a determination, but it is the user who makes the decision.

::And the thing you are referring to is from 2023, back when AI was a completely different beast. Polygnotus (talk) 18:00, 12 June 2025 (UTC)

:::AIs are the ultimate unreliable source, known to make shit up (euphemized as "hallucinate") and based on a randomized set of texts of unknown origins and veracity, which they have been known to misattribute. I know a number of writers far more notable than myself whom AI results assert that they wrote things they didn't write, or didn't write things they did write; and that garbage has gone into articles banged together by lazy "writers." (Jeeze, we thought it was bad when "journalists" didn't follow our sourcing and just cut-and-pasted our articles!) --Orange Mike | Talk 18:08, 12 June 2025 (UTC)

::::@Orangemike True, but random AI bashing about hallucinations makes little sense when the scripts are unable to add text to Wikipedia articles, and you haven't bothered to give 'em a try. Polygnotus (talk) 18:11, 12 June 2025 (UTC)

:::Polygnotus, it's true the user makes a decision, but the script very much uses AI to verify sources, as you said yourself. I suppose the problem is whether the user will do due diligence checking the AI (see for example the problems with BrandonXLF's ReferenceExpander script). Re {{Tq|1=the thing you are referring to is from 2023,}} plus ça change... — Qwerfjkltalk 19:42, 12 June 2025 (UTC)

::::@Qwerfjkl But the difference is that ReferenceExpander actually added stuff to Wikipedia articles (even worse, plausible looking but bad stuff). But these scripts add nothing to Wikipedia articles. And they cannot be used to confirm any random claim; only claims that already are in a Wikipedia article (added by a human, not a script/tool/AI). And I assure you that in AI-land things have changed since 2023. Polygnotus (talk) 20:08, 12 June 2025 (UTC)

:::::I just wanted to say that I support your direction here. AI directly adding things to Wikipedia, even with a human looking over the shoulder, seems too risky to me. But an AI making suggestions is not risky and really the question of whether it's good enough to be useful, or a waste of time, is something that we can gain experience with over time. My guess is that we'll find that this is improving over time, and that even so, some areas it will be better than others. Jimbo Wales (talk) 21:21, 12 June 2025 (UTC)

::::::@Jimbo Wales Excellent, but can I use that support to get a WMF JavaScript coder for a couple of hours? I'm more of an ideas person and good at getting things to the 80% mark. But I am really gonna need someone who actually knows JavaScript to turn this into a proper gadget anyone can use (accessibility, following the [https://design.wikimedia.org/style-guide/ style guide], testing with various themes, i18n, et cetera). I know some other computer languages but my JavaScript is very rusty. The WMF needs a quick win, especially on the AI front, we all need to convince the sceptics that AI isn't evil, and I need someone who is used to JavaScript. If you don't know which specific person then you'll know who to ask. When [https://en.wikipedia.org/wiki/User_talk:Jimbo_Wales#c-Orangemike-20250612180800-Polygnotus-20250612180000 Orangemike] returns to his computer he'll find I improved some of the articles he wrote using AI. Polygnotus (talk) 22:32, 12 June 2025 (UTC)

:::::::Well, I don't actually work at the WMF so I can't just assign someone to you of course. I'd stick my hand up to volunteer myself, as I can code in Javascript (though not to a professional level, although mucking around with AI and online courses as a hobby means I'm kind of ok) - but there's zero chance I wouldn't disappoint you with my lack of time! Jimbo Wales (talk) 11:42, 18 June 2025 (UTC)

::::::::True, but if you ask someone they'll find someone. Polygnotus (talk) 19:46, 18 June 2025 (UTC)

Hello, Jimmy! I love editing Wikipedia. By the way, why did Larry leave it's association to Wikipedia?Rafaelthegreat (talk) 17:42, 14 June 2025 (UTC)

:@Rafaelthegreat, while it's not really a thing that gets talked about often—Larry Sanger has had a sort of ideological opposition to Wikipedia's coverage of political subject matter, with Wikipedia reflecting a more Liberal tilt in its articles that cover political subjects, which is reflective of a liberal-majority media; and Larry believing that balanced coverage of such subjects can only be achieved by giving equal weight to each of the relevant viewpoints involved in those said subjects.

:That's the big, glaring issue in why Sanger doesn't really agree with Wikipedia—how to balance coverage—but I'm sure there are other reasons that Jimbo probably doesn't fancy discussing. And we don't generally prod him about it because he is a respected person around here and we're all pretty lucky to be able to chat with him at random. BarntToust 21:39, 14 June 2025 (UTC)

:::Just FYI as a point of history, Larry didn't leave because of any disagreement about NPOV/bias. That's a concern he raised many many years later. The way you present it - I believe very very strongly in the view that you attribute to Larry: Wikipedia should NOT reflect a more liberal tilt in articles that cover political subjects and should take great care with sources that are biased to avoid it. Balanced coverage does demand strong intellectual commitment to the hard work of giving equal weight to relevant viewpoints, as opposed to accepting anything like counting up number of sources, etc.--Jimbo Wales (talk) 11:40, 18 June 2025 (UTC)

::::Oh, thanks for clearing up my goofy recency bias, Jimbo. I only looked at it from there. Frankly, I never thought either of you would say anything specific much to why Larry left one random day in the early 2000's, so I offered the obvious conflict—which wasn't right-on. I also appreciate that you've took a stand here to side with Mr Sanger's pretty objective reading of neutrality. Hope you are well! BarntToust 15:19, 18 June 2025 (UTC)

::As Larry Sanger undoubtedly could explain his viewpoints much better than a random could, I'm linking [https://larrysanger.org/2021/06/wikipedia-is-more-one-sided-than-ever/ this article] he wrote back in 2021 that offers his perspective on encyclopedic neutrality. He gives a key example about Biden scandals in explaining his viewpoint. BarntToust 21:46, 14 June 2025 (UTC)

:::thanks Rafaelthegreat (talk) 12:58, 15 June 2025 (UTC)

Some Wikipedia articles about very popular subjects get big. I figured Jimbo's talk page was as good a place as any to put a thought out there: if a Wikipedia article gets too possibly large, and there is simply too much content that ought to be covered about the subject, would it be an acceptable idea to split the singular subject's article into a Part 1 and Part 2? Anyhow, I'm speaking hypothetically and in disregard to the idea of sub-articles, but what if there was just too much important stuff to cover—even presented in cut-down summary style—for one possible Wikipedia page? IDK. Just a shower thought I came up with. Hope Jimbo is doing well considering all the bumps in the road that the figurative "Wikimedia car" is running over of late. BarntToust 21:32, 14 June 2025 (UTC)

:@BarntToust The procedure used is described over at WP:SPLITTING. People either use a split on size or on content. This system works well so I don't think we really need a part1, part2 alternative. Polygnotus (talk) 10:47, 15 June 2025 (UTC)

::yeah, yeah. I kinda said my idea here was ignoring the idea of sub-articles and whatnot. I was just looking for some discussion of maybe the ups and downs up that other approach, for craps and giggles. BarntToust 14:56, 15 June 2025 (UTC)