The topmost bullet in subsection "The future that needs to be prepared today" correctly describes the limit on Wikipedia size. Simple English Wikipedia, though far from the biggest WP, illustrates this, as obvious vandalism often goes uncorrected for days or months because we don't have enough article watchers. If we can't keep up with quality problems, growth becomes unprofitable as it makes the encyclopedia more misinformative and untrustworthy. As for a catalog of a billion stars, or of everyone who ever lived, or every road or building everywhere, they should not even get a Wikidata item each, since specialized Web sites can take care of them. Wikipedia need not say all that is known about everything. Being an overall rough map and epitome of the universe of knowledge is a goal we can approach; we should not try to be the universe. Jim.henderson (talk) 01:39, 2 May 2025 (UTC)[reply]
I found the comment above by Jim.henderson meaningful and useful, went on to share its meaning with Wikimedia Language Diversity Community Russian. The last sentence Being an overall rough map and epitome of the universe of knowledge is a goal we can approach; we should not try to be the universe. seems a vote for Federation of Wikimedia Wikis, all describing different universes, as all language and cultural communities understand boundaries of valuable knowledge differently, whilst Wikidata with its elements is what's actually keeping us together. Frhdkazan (talk) 07:35, 3 May 2025 (UTC)[reply]
One issue that has never been given sufficient thought is what should the purpose of this encyclopedia be? By this, I am defining two possible purposes. The first is that an encyclopedia should contain all information, thus making all other reference works redundant. (This is a goal many encyclopedias have striven for, as documented by Richard Yeo in Encyclopaedia Visions:Scientific Dictionaries and Enlightenment Culture (Cambridge: University Press, 2001) ISBN978-0-521-15292-1.) This goal is obviously unobtainable: it's a well-known fact that knowledge expands faster than it can be recorded; no citation is needed for this assertion, although I can provide one if demanded. The other purpose is to be a springboard for further research, a purpose that is achievable, yet is often baffled by Wikipedia policies. (For example, policy is to rely on secondary sources, not primary, which means the reader or user is not alerted to all of the sources.) If Wikipedia should be a place to begin researching or learning on a subject, then policy needs to be revised to accommodate that purpose. -- llywrch (talk) 06:38, 11 May 2025 (UTC)[reply]
In the "how big can Wikipedia be?" section, though this content amount is intriguing, I fail to see how AI could help with this. As previously mentioned in Signpost articles, AI draws most of its knowledge from online sources such as Wikipedia itself, and with such a limitation to online sources and many sources being offline (especially ones regarding events in big Indian cities between 1975 and 2025, due to India being a country with a high poverty rate and the fact that the Internet did not exist in such a form in '75 and afterwards). Humans could better write these articles since they have access to these sources. Not to mention the many so-called "hallucinations" of artificial intelligence. ISometimesEatBananas (talk) 00:25, 7 May 2025 (UTC)[reply]
+1 to the recognition that Wikipedia is actually the largest source of training data for mainstream LLMs (e.g. GPT), the second largest are newspapers by the way. This only works because each Wikipedia article is hand-checked by authors. If Wikipedia articles and news sources were mass-written by LLMs, the entire AI ecosystem would implode. So this is a terrible idea.--Nordostsüdwest (talk) 20:25, 7 May 2025 (UTC)[reply]
Rather than something like a ban, we could try an extension of our old bot license system. Designate responsible operators of generative AI and mark their work, with either the simple B in the watchlist or with its own maker. Set AI to work searching for edits by unlicensed generatives, and template them for closer inspection. Jim.henderson (talk) 15:28, 10 May 2025 (UTC)[reply]
Actually, there are kinds of bot work that I would like to see done, though I don't know whether "generative" would describe them. I work the geocoordinates of historical sites and buildings, and wish Wikidata items, with coords where available, were made for all the items in list articles like National Register of Historic Places in Essex County, New Jersey. I would also like WD items for every marker in the Historical Marker Database even though almost none of them ought ever be the subject of an article. But, that's all Wikidata, where errors and duplications already abound and clumsy AI can make it more comprehensive while probably not making it much dirtier. WP has to be much cleaner. Jim.henderson (talk) 15:26, 10 May 2025 (UTC)[reply]
Wikipedia talk:Wikipedia Signpost/2025-05-01/In focus
← Back to In focus