In light of the recent Crowdstrike crash revealing how weak points in IT infrastructure can have wide ranging effects, I figured this might be an interesting one.
The entirety of wikipedia is periodically uploaded here, along with many other useful wikis and How To websites (ex. iFixit tutorials and WikiHow): https://download.kiwix.org/zim
You select the archive you want, then the language and archive version (for example, you can get an archive with no pictures, to save on space). For the totality of the english wikipedia you’d select the “wikipedia_en_all_maxi_2024-01.zim”
The archives are packed as .zim files, which can be read with the Kiwix app completely offline.
I have several USBs I keep that have some of these archives along with the app installer. In the event of some major catastrophe I’d at least be able to access some potentially useful information. I have no stake in Kiwix, and don’t know if there are other alternative apps and schemes, just thought it was neat.
The text version of Wikipedia*
The images and other media are a hell of a lot more.
I presume this is images directly hosted on English Wikipedia and not the entirety of Commons where the vast majority of images are kept, right?
Wikimedia Commons is 373TB images. https://commons.m.wikimedia.org/wiki/Special:MediaStatistics
The 100Gb version mentioned above does only have thumbnails/lowres pictures, yeah. Better than nothing for some types of articles, but not everything. The true text-only version is actually only ~53Gb though.
Some of the high res photos are ridiculous.
Like a 8000x9000 uncompressed image of someone’s hand and weighs about 22mb.
I know that because I use a lot of royalty free images.
Without images Wikipedia is a “mere” 22.14gb.
So something akin to this joke image I saw the other day is actually feasible for Wikipedia?
Probably a lot less, keep in mind that whenever it answers a question the whole model is traversed multiple times, going through multiple GBs is not possible in the matter of seconds the model answers.
I’d be surprised if it was significantly less. A comparable 70 billion parameter model from llama requires about 120GB to store. Supposedly the largest current chatgpt goes up to 170 billion parameters, which would take a couple hundred GB to store. There are ways to tradeoff some accuracy in order to save a bunch of space, but you’re not going to get it under tens of GB.
These models really are going through that many Gb of parameters once for every word in the output. GPUs and tensor processors are crazy fast. For comparison, think about how much data a GPU generates for 4k60 video display. Its like 1GB per second. And the recommended memory speed required to generate that image is like 400GB per second. Crazy fast.
I mean, you can self-host your own local LLMs using something like Ollama. The performance will be bound by the disk space you have (the complexity of the model you’re able to store), and the performance of the CPU or GPU you are using to run it, but it does work just fine. Probably as good results as ChatGPT for most use cases.
We do this at work (lots of sensitive data that we don’t want Openai to capitalize on) and it works pretty well. Hosted locally, setup by a data security and privacy sensitive admin, who specifically runs the settings to not save any queries even on the server. Bit slower than chatgpt but not by much
Aside from the text clarification, this is also only the US version of Wikipedia.
What worries me though is that most videos linked on Wikipedia are hosted on YouTube. That’s a pretty dangerous choke point.
Ten year old me would beg to differ.
Videos turned Encarta 95 from being an encyclopedia to the encyclopedia!
I jest - a multimedia experience helps but I agree that the text knowledge is the big draw.
My brain immediately thought archive.org but after the last incident, I kinda feel like archive org is going to get lawsuited into oblivion
This saved my ass at my engineering chemistry exam (still a requirement, even for software engineers) where only offline tools were allowed. Love Kiwix!
DYK that Kiwix was actually created by Wikipedia? Back in the late 2000s there was this gigantic effort to select and improve a ton of articles to make an offline “Wikipedia 1.0” release. The only remains of that effort are Kiwix, periodic backups, and an incredibly useful article-rating system.
- There is a set of criteria to rate an article B, C, Start or Stub. These are called classes. Similarly, articles can be rated to be of 1 of 4 importance values to a particular WikiProject.
- There’s a banner on every article’s talk page. Any editor can change an article’s rating between one of the above classes boldly; if a revert happens, they discuss it according to the criteria.
- Some WikiProjects have their own criteria for rating articles. Some of them even have process to make an article A-class.
- Before this system, Wikipedia already had processes to make an article a Good Article or Featured article.
- With GAs, a nominator should put a candidate onto backlog. Later, a reviewer will scrutinize the article according to criteria. Often, the reviewer asks the nominator to fix quite a bit of issues. If these issues are fixed promptly, or the reviewer thinks that there are only nitpicks, the article passes. If they aren’t fixed in a week or the reviewer thinks that there are major problems, the article fails.
-
- As with other processes, the nominator and reviewer can be anyone, though reviewers are usually experienced.
- With FAs, a nominator brings the candidate to a noticeaboard. Editors there then come to a consensus about whether the article should pass.
- Both processes display a badge directly on passed articles.
- Both processes have an associated re-review process where editors come to a consensus whether the article should fail if it were nominated today
- There’s also an informal process called “peer review”, where someone just puts an article at a noticeable and anyone can comment about its quality.
- Articles are automatically sorted into categories by their rating and importance. Editors usually look at these to decide which articles to focus on nowadays.
