The updates are necessary in order to:
Other possible changes include strengthening enforcement of the requirement that paid editors must declare their employers, clients, and affiliations. – S
Wikimedia Foundation Board Member Shani Evenstein announced the results of the vote on the Wikimedia-l mailing list on 13 February 2023:
Today the results of the Universal Code of Conduct Enforcement Guidelines vote were tallied. We are pleased to report the results show that the Enforcement Guidelines are strongly supported by the community, with 76% of participants voting in support of the Enforcement Guidelines.
3,097 voters from 146 Wikimedia communities participated. Statistics for the vote are available. A more detailed summary of comments submitted during the vote will be published soon.
Next, the results and comments collected during this vote will be submitted to the Board of Trustees for their review. It is expected that the Board of Trustees review process will complete in March 2023, at which time another update will be published. – AK
The Steward confirmations – an annual performance evaluation by the community that determines whether or not existing Stewards will retain their status for another year – are also currently underway.
Voting for the election and confirmation processes will remain open until 26 February 2023, 21:00 (UTC). – AK
Now you see it, now you don't – Wikipedia had hardly been blocked by Pakistan before it was unblocked.
Shehbaz Sharif, the Prime Minister of Pakistan, stepped in and convened a committee composed of the Minister for Law and Justice, the Minister for Economic Affairs and Political Affairs and the Minister for Information and Broadcasting to examine the matter.
The committee took the view that –
Wikipedia was a useful site/portal which supported dissemination of knowledge and information for the general public, students and the academia. Blocking the site in its entirety was not a suitable measure to restrict access to some objectionable contents/sacrilegious matter on it. The unintended consequences of this blanket ban, therefore, outweigh its benefits.
The government subsequently barred the Pakistan Telecommunication Authority (PTA) from blocking any website without first consulting the Ministry of Information Technology and Telecommunication.
Wikimedia Foundation General Counsel Stephen LaPorte said on the Wikimedia-l mailing list:
We received news yesterday that the Pakistan Telecommunications Agency was directed to restore access to Wikimedia projects, including Wikipedia, in Pakistan. Our internal traffic reports confirm that the block has been lifted and we've updated our public statement.
We're happy that the people of Pakistan can continue to use Wikipedia to access knowledge and share their experiences and culture with the world. These situations are a reminder of the importance of the work of our movement and the vital role our communities play in building the largest collection of open knowledge in history. It also underscores the need to protect people's right to access free knowledge and participate in contributing to its growth globally. Thank you for your support. We'll continue to keep you updated if there are any new developments.
For the media coverage of this event, see this issue's in the media.
If only it remained that way: English Wikipedia editor Rory Jaffe (Draft:Cow Farts – not actual bull shit, but pretty close. Here's what he says about his discovery technique.) has found a number of them, including
I've been looking for "tells" before testing [with a classifier tool that can detect AI generated text with some accuracy]. The text is more "lifeless" than human text typically is, tends to have uniform length sentences, and may have a paragraph at the end that is a summary paragraph. The text is also unlikely to have in-line references. This doesn't catch all of them but tends to be a high-likelihood way of finding them. Some of these AI-generated articles also have AI-generated references. The references are almost always fake: the AI confabulates reasonable-sounding references!
Articles or drafts tagged by Jaffe and reviewed by The Signpost were created as early as 17 December 2022 (Draft:Brisbane Lord Mayor YAC). The topics tagged by Jaffe and others range from the whimsical and neologism-ic (Draft:Twinkanic Awareness, Draft:Sharks should be saved from humans) to the plausibly- and obviously-promotional, and appear to be created by many editors, including single-purpose accounts that may be using them for promotion. Take Draft:Dachshund harness, "an essential accessory for owners of this unique breed of dog", which has a list of references at the end related to shopping for such a device. Given affiliate marketing exists, this could be promotional. Or it could be purely accidental. We don't know. Some of the rest may be deleted by the time you read this – a few examples are retained permanently at User:JPxG/LLM dungeon.
The Signpost asked Jaffe "Do you have a personal reaction to the quality of the AI generated articles? Briefly, are you for or against this method of creation?" He told us this:
Strongly against. These language generators are good at stringing together reasonable-sounding text, but:
1. Don't rely solely on what Wikipedia recognizes as reliable sources.
2. Don't have a sense of what is real and what is fake.
3. Will fabricate information to fill in gaps.
4. Cannot identify where any specific bit of "information" came from.
However, they're good at helping to break "writer's block", by giving an example of how to write something. So I see having some text generated by ChatGPT would be helpful as a writing prompt, but without using any of the generated text directly—just looking at it and then writing a well-researched article similar to the prompt.
More on Jaffe's experience and conclusions on this topic can be found at the Essay in this issue.
We asked ChatGPT for a commentary as well. It said:
It's interesting to see that people are now using AI to generate articles on Wikipedia, but it's also concerning as this could lead to a proliferation of inaccurate or fake information on the platform. It's good to see that editors like Rory Jaffe are using techniques to identify AI-generated text, but it's clear that the AI-generated articles are becoming more sophisticated in their structure and references, making it harder to detect them.
It's important for Wikipedia to maintain its high standards of accuracy and reliability, so it's crucial that the platform continues to monitor and regulate the use of AI-generated content. While AI can be a valuable tool for assisting with research and writing, it should never be a substitute for human expertise and critical thinking.
Revenue in the 2022 calendar year (from subscription and professional services) totaled $3.12 million, vs. expenses of $3.18 million ($1.06 million in cost of services and $2.12 million in operating expenses), for a net loss of $60,000.
The Enterprise team points out that the present report for the 2022 calendar year should be considered a "beta" edition. The Wikimedia Foundation normally operates on a financial year starting in July and finishing the following June. But as January 1, 2022 marked the official start of commercial operation of the Enterprise project, this first report covers the 2022 calendar year. For future financial reports the Enterprise team intends to align with the normal financial reporting schedule of the rest of the Wikimedia Foundation. The next report will be published in late 2023.
The report adds that:
As the LLC is wholly owned by the Wikimedia Foundation, all of the financial information presented here will also be included within the Wikimedia Foundation's audited financial statements and will be in the next Wikimedia Foundation "Form 990" filing as it relates to fiscal year 2021–2022, and future years.
The Enterprise team says it is very happy with its first year in business and looks forward to growth and profitability this year:
In 2023, our second year of operations, with the addition of new customers and also new features, both revenue and expenses are expected to increase – but revenue growth is expected to outpace expenses. By comparison to other startup commercial API projects, to reach this stage within one year of operations is extremely rapid progress. The governance of these and all other commercial customer relationships is consistent with how the Wikimedia Foundation treats large corporate donations.
No information has been released on how many paying and non-paying customers Wikimedia Enterprise has at the moment. Corresponding inquiries on Meta-Wiki have been answered as follows:
As stated in the original press release, Google and the Internet Archive are indeed the first to receive paid and free access (respectively) but we have not publicized the subsequent customers (paid or free) who have signed-up to the service. ... Maintaining a public and comprehensive list of paying and free/trial customers would look like advertising/promotion of those customers, and also introduce a new privacy (and potentially security) problem: i.e. in the same way that it would be inappropriate to make a public list of "all people who have used the Wikidata Query Service this month" (for example) – it goes against our privacy culture. Nonetheless, we do intend to be making "use case" blog posts – which will describe how some users (either general categories or individual cases with their permission) are benefiting from the service in the real-world.
The Enterprise API features are designed with large commercial reusers in mind – primarily search engines and voice assistants, although education (question-and-answer modules), finance and the fast-growing AI sector represent additional market opportunities the team may explore.
But other kinds of reusers, including individual volunteer Wikimedians, can also benefit. There are several access methods to the datasets available at no cost/no registration, including a free Wikimedia Enterprise account via the project's homepage. Ongoing updates for the community can be found on the project's MediaWiki page.
The publication of the financial report was followed by a public Zoom meeting on 10 February, a recording of which is available. Click the picture on the right or go to the Wikimedia Enterprise talk page on Meta-Wiki, which also includes a list of questions discussed during the meeting and their approximate timecodes. The file description in Commons even includes an automated transcript. – AK
In response to multiple inquiries, the Wikimedia Foundation has posted a breakdown of how and when Tides Advocacy spent the $4.223 million Annual Plan Grant portion of the $8.723 million the organisation received from the WMF in the 2019–2020 financial year. – AK