Hacker News Reader

Comment by cm2012 | Hacker News Reader

cm2012

This is cute, but kind of an example of Wikipedia's off-mission bloat. It irks me that they constantly fundraise when most of it is not needed for Wikipedia proper, but rather used for initiatives people know less about and may not fund if they knew.

View on HN

I don't begrudge them the odd party, anniversary, meetup.

And some of their subprojects are a great idea and could go much further -- it'd be fantastic to have a Wikipedia atlas, for example. The WikiMiniAtlas on geolocated articles is nice but it could be so much better.

But as per https://en.wikipedia.org/wiki/WP:CANCER it's a huge concern that they're blowing money pretty much at the rate they get it, when they should be saving it for the future, and be pickier and choosier about what they're funding at any given time.

I made “Wikidata Atlas” several weeks ago. [1] [2]

[1] : https://wd-nearbyitem.toolforge.org/

[2] : https://rtnf.substack.com/p/wd-nearbyitem

That is a nice start, a rendering of GIS wikidata. Perhaps ask Wikimedia Foundation for funding :)

What I'd like to see is a more intimate marrying of OSM data and Wikipedia data. For example, if I go to zoom level 12 centred on London, UK on your page, there are about 80 text labels on the OSM layer itself. At minimum this is going to need OSM vector tiles. I'd expect to be able to click any of the OSM labels for the corresponding Wikipedia article, as well as adding in POIs for articles that don't have corresponding OSM links. And then you need OSM rendering style rules about which POIs you show at each zoom level, based on whether labels will run into each other or not.

The problem right now is that the WikiMiniAtlas treats all things, whether large areas or individual POIs, as POIs.

Thank you for this link, I was looking for something with that data in a clean format for some time!

i feel like that's a bit silly, the other projects are listed on the donation page (https://donate.wikimedia.org/wiki/FAQ) and tbh you are unlikely to be donating to the wikimedia foundation without being aware of (at least some of?) the rest

I promise you that 99% of normal people have no idea what the Wikimedia foundation is and think that they're just donating to "fund Wikipedia".

are normal people donating to wikipedia tho

Yes, the ads are essentially a guilt tax on normies who remember Wikipedia helped them in high school

Honestly, most of the other projects get almost none of the resources (other then maybe wikidata and commons, but both of those are directly used by wikipedia)

I wonder whether the emergence of a single, true Wikipedia competitor would actually put an end to this never-ending fundraising criticism (since people could simply donate to the competitor as a form of protest)

Projects like Wikipedia never have meaningful competition, because the social dynamics invariably converge to a single platform eating everything else.

Wikipedia is already dead, they just don't know it yet. They'll get Stackoverflowed.

The LLMs have already guaranteed their zombie end. The HN crowd will be comically delusional about it right up to the point where Wikimedia struggles to keep the lights on and has to fire 90% of its staff. There is no scenario where that outcome is avoided (some prominent billionaire will step in with a check as they get really desperate, but it won't change anything fundamental, likely a Sergey Brin type figure).

The LLMs will do to Wikipedia, what Wikipedia & Co. did to the physical encyclopedia business.

You don't have to entirely wipe out Wikipedia's traffic base to collapse Wikimedia. They have no financial strength what-so-ever, they burn everything they intake. Their de facto collapse will be extremely rapid and is coming soon. Watch for the rumbles in 2026-2027.

Wikipedia is not even in the game you are describing here. Wikipedia does not need billions of users clicking on ads to convince investors in yet another seed. They are an encyclopedia, and if fewer people will visit, they will still be an encyclopedia. Their costs are probably very strongly correlated with their number of visitors.

SO was supposed to be much the same, though. I guess you really do have to get directly funded by users for the model to work.

If we kill all the platforms where content for training LLMs comes from, what do LLMs train on?

This. I'm really bothered by the almost cruel glee with which a lot of people respond to SO's downfall. Yeah, the moderation was needlessly aggressive. But it was successful at creating a huge repository of text-based knowledge which benefited LLMs greatly. If SO is gone, where will this come from for future programming languages, libraries, and tools?

Newspapers, scientific papers and soon, real-world interactions.

News is the main feed of new data and that can be an infinite incremental source of new information

You talk about news here like it's some irrefutable ether LLMs can tap into. Also I'd think newspapers and scientific papers cover extremely little of what the average person uses an LLM to search for.

This always feels to me like, an elephant in the room.

I’d love to read a knowledgeable roundup of current thought on this. I have a hard time understanding how, with the web becoming a morass of SEO and AI slop - with really no effort being put into to keeping it accurate - we’ll be able to train LLMs to the level we do today in the future.

Most people went to SO because they had to for their job. Most people go to Wikipedia because they want to, for curiosity and learning.

LLMs will use Wikipedia the same way humans use it

And they have a huge endowment fund now too that more than covers the cost of Wikipedia...

They, just like some newspapers, try to present themselves as neutral, not tied to any interest.

Like Hackernews, supposedly neutral.

The main issue with neutral people is that we do not know in which camp they are.

> The main issue with neutral people is that we do not know in which camp they are.

And that's a good thing, 'cause it means they're living to their standards.

Reply on news.ycombinator.com