Wikipedia:Wikipedia Signpost/2024-09-04/News and notes
WikiCup enters final round, MCDC wraps up activities, 17-year-old hoax article unmasked
The WikiCup gears up for its final round
The 2024 WikiCup, hosted by users Cwmhiraeth, Epicgenius and Frostly, is entering its final phase, after Round 4 ended on 29 August. A total number of 135 users, including the late Vami IV, joined the contest at the start of this year; however, just eight of them have made it to the ultimate showdown. Here are the finalists, ranked from first to last as per their scores in the latest round:
- Generalissima (1150 points);
- Arconning (791 points);
- AirshipJungleman29 (718 points);
- BennyOnTheLoose (714 points);
- Hey man im josh (601 points);
- BeanieFan11 (579 points);
- AryKun (499 points);
- Sammi Brie (472 points).
Since its creation back in 2007, the WikiCup has strived to "encourage content creation and improvement and make editing on Wikipedia more fun", and this year's edition is no exception: according to the official data, competitors have so far contributed to 44 featured articles, 72 featured lists, 385 good articles, 94 In the News credits, and over 300 Did You Know credits; thanks to their efforts, 38 articles were also added to featured topics and good topics.
On behalf of The Signpost, we would like to thank the judges and every participant in the 2024 WikiCup, and wish good luck to the eight finalists.
– O
Journals cited by Wikipedia compilation now tracks free DOIs
As of 18 August, the Journals cited by Wikipedia (JCW) compilation (see previous Signpost coverage) now tracks the number of distinct DOIs present on Wikipedia, and how many are flagged with |doi-access=free
. Several of these are automatically tracked and tagged as free to read by templates and bots (see previous Signpost coverage). As of the 1 August dump, the compilation kept track of 3.70M citations, of which 2.41M had DOIs. Of the citations that had DOIs, 661,103 were identified as free to read, or about 27.44%.
The 17–18 August 2024 update of the CS1/CS2 modules further identified the Leibniz International Proceedings in Informatics (doi prefix 10.4230
) and the Living Reviews journal series (doi prefix 10.12942
) as free-to-read registrants, as well as 11 individual journals that can be identified by the starting pattern of DOIs (like 10.1046/j.1365-8711...
, 10.1093/mnras..
, and 10.1111/j.1365-2966...
for the Monthly Notices of the Royal Astronomical Society). Citation bot will automatically flag those with |doi-access=free
when it runs on the article (see our guide on how to use Citation bot yourself).
If you notice a DOI link that takes you to a free-to-read article that wasn't flagged by the bot, you can flag the citation manually with |doi-access=free
. You can also try to use WP:OABOT (see our guide on how to use OAbot yourself). If you are aware of fully free-to-read journals/publishers that aren't already kept track of by the CS1/CS2 templates (see CS1/2 FAQ), leave a note at Help talk:CS1 and User talk:Citation bot.
Following the 20 August dump, the compilation kept track of 3.72M citations, of which 2.42M had DOIs. Of the citations that had DOIs, 663,976 were identified as free to read, or about 27.46% (up from 27.44%). It took a few days for the server cache to clear and tracking categories to be populated. I estimate that the 'true' count should have been about 666K, mostly due to MNRAS and MNRAS Letters being identified as free to read, but we'll see in the 1 September dump what the exact effect of the update was.[a]
Related to the JCW update, all CS1/2 templates (like {{cite journal}} and {{citation}}), and the standalone templates {{doi}} and {{doi-inline}}, now support the flagging of free-to-read DOIs with |doi-access=free
. The standalone versions, however, are not currently supported by any bot, nor do they have tracking categories.
Thanks to Trappist the monk for their efforts on templates and the identification of free-to-read publishers/journals (I was also involved), as well as the maintainers of Citation bot, JL-Bot, and OAbot (particularly AManWithNoPlan, JLaTondre and Nemo bis) for facilitating the mass-tagging of free-to-read articles.
- ^ Note: I will update this article once the 1 September numbers become available.
– H
AI policy positions of the Wikimedia Foundation
In a blog post, the Wikimedia Foundation provides an overview of several statements it has submitted since last year in response to
[...] governments and international organizations [...] seeking stakeholder feedback about how [AI] policies should be formulated in order to best serve the public interest. [...] The Foundation’s comments have fallen into two categories. Some are directly relevant to the work being done by volunteer Wikipedia editors around the world, such as on copyright and openness of foundational AI models. Others applied our values and the valuable lessons we have learned from our AI/ML work to benefit public interest projects focused on free knowledge and the online information ecosystem—i.e., decentralized community-led decision-making, privacy, stakeholder inclusion, and internet commons.
— "AI for the people: How machines can help humans improve Wikipedia" (Wikimedia Foundation)
For example, in a response to the US Copyright Office's Request for Comments on AI and Copyright, the Foundation states that it "generally supports uses of Wikipedia content for purposes including AI model development", but (as summarized in the blog post) argues that
At a minimum, AI developers who include Wikipedia in the training data used to create large language models (LLMs) should publicly acknowledge that use and give credit to Wikipedia and the volunteer editors who made this rich source of raw materials for LLMs.
At the same time, the Foundation's statement indicates that this attribution might not always be legally required, depending on whether courts decide that the unauthorized use of copyrighted content in training of such AI models is covered by fair use (in which case the attribution requirements of Wikipedia's CC BY-SA 4.0 license would be moot). The Foundation refrains from taking a categorical position on this legal question: "Based on our analysis, we do not believe that training AI models should either be categorically fair use or categorically not fair use. Rather, the particulars of the training process and the way courts view the purposes of a use should inform whether a particular training process is fair or not." The analysis does however offer some detailed if speculative observations on how courts might evaluate the four fair use factors in this context. For example, it is argued that because "the vastness of the datasets used in training mean that any single copy [of a copyrighted work] is barely a drop in the ocean of the whole", judges may want to focus on "the extent to which a work is weighted in the development of a model": "Hypothetically, if a copyright protected work was manually weighted to have an outsized impact in model development, then one could argue that although the uses of other full works may be fair, the amplification of one particular work in the training set is not." (Various LLMs are known to have weighted Wikipedia more highly than other parts of their training dataset, for example GPT-3.)
On the other hand, the Wikimedia Foundation's statement also urged the Copyright Office to take not only the perspective of copyright owners into account, but also that of the users of copyrighted works and of AI-based tools – noting that "The Foundation is somewhat uniquely positioned as both the host of a primary source of training material for generative AI and a user of many AI and ML tools that aid human editors with the creation of free knowledge." In particular, it cautions to keep public interest in mind in possible future changes to copyright laws and AI regulations, e.g.
On the use of data, specifically, we encourage regulators and legislators to align their approaches with existing models, such as the European Union’s inclusion of an exemption for text and data mining in the Directive on Copyright in the Digital Single Market, that enable public interest research and other beneficial uses of protected works.
[...] we encourage the Office to consider the potential impacts that changes to copyright law could have on competition among AI developers. If copyright law changes are enacted such that the acquisition and use of training materials becomes more expensive or difficult, there is a risk that dominant firms with greater resources will become further entrenched while smaller companies, including nonprofit organizations, struggle to keep up with mounting development costs.
– H
The Farewell of the MCDC
Chosen by communities, selected by affiliates, and appointed by the WMF, the Movement Charter Drafting Committee (MCDC), a committee of 15 Wikimedians, first took on the job of drafting a Charter for the Wikimedia movement in November 2021.
There were multiple feedback rounds, a lot of conversations, more discussions and a final ratification vote where the community and affiliate support was overwhelming (albeit with a low turnout in both cases due to the voter eligibility criteria), but the WMF's Board of Trustees decided the draft was not good enough (not safe to try). As reported in the previous issue of The Signpost, the Foundation published three pilot projects to take the work forward.
In August 2024, the committee (which still included 11 people), shared their process and ratification reflections pre-Wikimania. Before dissolving on 30 August, they also published their recommendations for next steps, including a response to the three pilots proposed by the WMF.
– Ciell, former MCDC member
Brief notes
- China blocks Wikimedia Foundation's WIPO accreditation for the fourth time: In what has become an annual tradition, an application by the Foundation to be accredited as a permanent observer at the World Intellectual Property Organization (WIPO) failed because of opposition by the Chinese government, despite support from several other countries. As explained by the Foundation's Global Advocacy team, the WIPO is "the specialized United Nations (UN) agency that determines global policies on copyright, patents, and trademarks. Observer status would enable the Foundation to participate in and contribute to WIPO committees where intellectual property norms are set." The Foundation first applied in 2020, when Beijing's delegate objected to "a large amount of content and disinformation in violation of [the] ‘One China’ principle" (see Signpost coverage: "Beijing blocks WMF from World Intellectual Property Organization, citing Wikimedia Taiwan"). China has since also blocked several Wikimedia chapters from gaining permanent observer status.
- Wikimedia ESEAP Conference: Wikimedia Community User Group Malaysia has uploaded a video with highlights of the May 2024 Wikimedia ESEAP Conference to YouTube.
- Articles for Improvement: This week's Article for Improvement is Cancel culture. Please be bold in helping improve this article! Next week's Article for Improvement (beginning 9 September 2024) is Caroline Islands.
- New record low count for active administrators: A new low point of 427 active administrators was reached on 26 August, marking a new decline after the slight increase in July. Meanwhile, the first RfA since June was initiated just a few days before we went to press with this issue.
- The curtain rises on a new edition of WLM: The fifteenth edition of Wiki Loves Monuments kicked off on 1 September. For this year's photo contest, which will focus on built heritage, 53 countries have enrolled. The respective national campaigns will be hosted until 31 October, and then each nation will send their Top 10 to the international jury for the Grand Finale in December.
- Long-lived hoax article removed: "Pratylenchus dulscus," a ten word article without references, was one of the longest-lived hoaxes on Wikipedia. It was deleted on 22 August, after existing in the encyclopedia for over 17 years.
- The Campaigns teams would like to learn about how you collaborate online: Editors can take this Google Form survey or share examples of successful collaborations on Meta Wiki.
Discuss this story