Wikipedia:Reference desk/Computing
of the Wikipedia reference desk.
Main page: Help searching Wikipedia
How can I get my question answered?
- Select the section of the desk that best fits the general topic of your question (see the navigation column to the right).
- Post your question to only one section, providing a short header that gives the topic of your question.
- Type '~~~~' (that is, four tilde characters) at the end – this signs and dates your contribution so we know who wrote what and when.
- Don't post personal contact information – it will be removed. Any answers will be provided here.
- Please be as specific as possible, and include all relevant context – the usefulness of answers may depend on the context.
- Note:
- We don't answer (and may remove) questions that require medical diagnosis or legal advice.
- We don't answer requests for opinions, predictions or debate.
- We don't do your homework for you, though we'll help you past the stuck point.
- We don't conduct original research or provide a free source of ideas, but we'll help you find information you need.
How do I answer a question?
Main page: Wikipedia:Reference desk/Guidelines
- The best answers address the question directly, and back up facts with wikilinks and links to sources. Do not edit others' comments and do not give any medical or legal advice.
January 15
[edit]Wikipedia Chatbot
[edit]Is there a free chatbot using only Wikipedia as source material? ~2026-29536 (talk) 14:25, 15 January 2026 (UTC)
- No. And I doubt it would work. LLMs need very large amounts of training data just to 'learn' to compose comprehensible sentences etc. AndyTheGrump (talk) 14:29, 15 January 2026 (UTC)
- Many chatbot projects use local documents exclusively, such as chatbots used to answer questions about corporate documentation. ~2026-29536 (talk) 16:49, 15 January 2026 (UTC)
- Anyone can train one with Wikipedia's CC-BY materials if they wish to, (not that LLM training is a highly copyright-aware process anyways...) but Wikipedia is probably not doing this itself.
- Because even if we assume everything on WP is factually correct (I wish!), the model is still going to hallucinate. 海盐沙冰 / aka irisChronomia / Talk 16:56, 15 January 2026 (UTC)
- I think there is a confusion about the question. It is very common to use a pre-trained LLM to make a chatbot that uses only a limited source material for answering questions. For example, a company can make a chatbot, pretrained on whatever it was trained on, to answer questions limited solely to HR documentation. There are multiple methods of doing so, such as LLM+RAG. The question is not asking "Has anyone trained an LLM on Wikipedia?" It is asking "Has anyone made a free chatbot that only uses Wikipedia for the source material when answering prompts?" ~2026-29536 (talk) 17:24, 15 January 2026 (UTC)
- I see. Sorry for the confusion, I'll keep the question in mind. But I haven't see such a chatbot yet.
- If we're trying to limit an LLM's response to be sourced from WP, can we solve that simply by doing prompt engineering? 海盐沙冰 / aka irisChronomia / Talk 02:17, 16 January 2026 (UTC)
- You can ask LLMs to back up their statements by citing sources, and some will comply (or even do so without you asking), so I bet (I haven't tried) that you can ask for the cited sources to be confined to Wikipedia articles. You still have to check, though, whether the cited sources actually exist and support the statements; my experience is that this is not at all guaranteed. ‑‑Lambiam 10:16, 16 January 2026 (UTC)
- I've tried a few different forms of prompts on ChatGPT. It has worked OK to apparently limit the content it returns, but I haven't got it to give good links to the Wikipedia articles. I was hoping for a tool to search Wikipedia that is better than the very poor built-in search bar. The search bar requires you to already know what you are looking for. If you don't know, you can't use the search bar for something like: "The article about the person who was in a movie about time travel, but she was a psychiatrist, and she didn't think it was time travel until later in the movie." ChatGPT said it was Madeleine Stowe, but didn't link to the article. Wikipedia's search jumps to Oppenheimer. Update: I realized it was giving me a link to the 12 Monkeys article in faint gray on white background. Not the article requested, but close enough to find the article requested and still far better than Wikiepdia's search. ~2026-29536 (talk) 11:57, 16 January 2026 (UTC)
- For Google or DuckDuckGo: just search as normal and use "site:wikipedia.org"
- For ChatGPT:
- Don't use the "Search" feature, instead, turn "Thinking" on and ask it to search.
- (When you use the Search buttons on the interface, it switches to a smaller/older model when interpreting the results and will only try to search once, I've found it pretty bad compared to using Thinking)
- Update your personalization settings or memory so that it includes statements such as:
When citing online sources, ChatGPT will ensure what ChatGPT says actually reflects those sources. Do *not ever* cite sources without having searched for them.
- Update your personalization settings or memory so that it includes statements such as:
- (For memory, you can't update that in the settings directly, but you can tell it to 'use the bio tool to remember' statements and preferences).
- If you want to limit sources, ask it to "for this conversation" prefer a certain source, e.g. "for this conversation, when searching online, I want you to only cite Wikipedia as a source and no others". It should be fairly okay at obeying this constraint if you use GPT-5 (or higher) with reasoning turned on.
- Remember that it's an LLM and you'll want to verify the sources anyway.
- Limiting search to Wikipedia is not something I tend to do personally, but if I were you, I would not limit searching to Wikipedia, but would try to limit citations to Wikipedia. That way, during its reasoning phase, it will be free to gather context from other sources before returning to looking for Wikipedia-based sources to cite. (You can also tell it to explicitly do that, and it should be okay). For example:
Tell me about [X]. Back it up with sources from Wikipedia only - but before you give me the final answer, you're free to search elsewhere too for extra context.Komonzia (talk) 17:17, 17 January 2026 (UTC)
- I've tried a few different forms of prompts on ChatGPT. It has worked OK to apparently limit the content it returns, but I haven't got it to give good links to the Wikipedia articles. I was hoping for a tool to search Wikipedia that is better than the very poor built-in search bar. The search bar requires you to already know what you are looking for. If you don't know, you can't use the search bar for something like: "The article about the person who was in a movie about time travel, but she was a psychiatrist, and she didn't think it was time travel until later in the movie." ChatGPT said it was Madeleine Stowe, but didn't link to the article. Wikipedia's search jumps to Oppenheimer. Update: I realized it was giving me a link to the 12 Monkeys article in faint gray on white background. Not the article requested, but close enough to find the article requested and still far better than Wikiepdia's search. ~2026-29536 (talk) 11:57, 16 January 2026 (UTC)
- You can ask LLMs to back up their statements by citing sources, and some will comply (or even do so without you asking), so I bet (I haven't tried) that you can ask for the cited sources to be confined to Wikipedia articles. You still have to check, though, whether the cited sources actually exist and support the statements; my experience is that this is not at all guaranteed. ‑‑Lambiam 10:16, 16 January 2026 (UTC)
- I think there is a confusion about the question. It is very common to use a pre-trained LLM to make a chatbot that uses only a limited source material for answering questions. For example, a company can make a chatbot, pretrained on whatever it was trained on, to answer questions limited solely to HR documentation. There are multiple methods of doing so, such as LLM+RAG. The question is not asking "Has anyone trained an LLM on Wikipedia?" It is asking "Has anyone made a free chatbot that only uses Wikipedia for the source material when answering prompts?" ~2026-29536 (talk) 17:24, 15 January 2026 (UTC)
- Many chatbot projects use local documents exclusively, such as chatbots used to answer questions about corporate documentation. ~2026-29536 (talk) 16:49, 15 January 2026 (UTC)
Kidsinmind
[edit]Hallo. Good late afternoon. Does anyone here know about kidsinmind.com? They are the OG parental guide for movies since '92. Sure, something like them probably existed, but not online. Even earlier than IMDB's parent guides. I personally think the site is a very good idea because every kid is different. Also, some of their reviews are unintentionally hilarious. My question is were they really online since 1992? Wayback machine has nothing before 2000. ~2026-24671-3 (talk) 16:05, 15 January 2026 (UTC)
- I tried with the (Google) search term
"kids-in-mind" before:1998-12-31 after:1989-01-01and a litany of others (e.g., +.com, remove dashes, etc.) and can't find website mentioning it. - It might exist somewhere else on the net, but at least I can't find any traces. I'm thinking about searching for it in Usenet archives example link. iris 0:24a, edited 0:29a 海盐沙冰 / aka irisChronomia / Talk 16:24, 15 January 2026 (UTC)
- I believe it's probably on the Internet some time around or after 1999. Check this query to Google Group's usenet archive: [1]. 海盐沙冰 / aka irisChronomia / Talk 16:38, 15 January 2026 (UTC)
- I forgot the most obvious part: An ICANN whois lookup. [2] [3], and quote:
Created: 1998-09-04海盐沙冰 / aka irisChronomia / Talk 16:44, 15 January 2026 (UTC)- Sorry for the reply. So does it mean it was on AOL first? ~2026-24671-3 (talk) 23:23, 18 January 2026 (UTC)
- This I'll have no idea. Maybe there's some way to search historical archives of AOL, but I don't know of any. 海盐沙冰 / aka irisChronomia / Talk 11:50, 19 January 2026 (UTC)
- Sorry for the reply. So does it mean it was on AOL first? ~2026-24671-3 (talk) 23:23, 18 January 2026 (UTC)
January 16
[edit]Boolean Algebra
[edit]Explain the science of half adders and binary multipliers in a way such that it may have seemed to us that we could have discovered it ourselves ~2026-34329-6 (talk) 14:20, 16 January 2026 (UTC)
- Imagine you work on binary addition on your own. You are using one bit. If you add 0+0, you get 0. If you add 0+1, you get 1. If you add 1+0, you get 1. If you add 1+1, you get 10, but you are only using one bit. You actually have 0 with a carry of 1 - just like carrying the 10s digit when adding decimal numbers. So, you decide to separate the value and the carry digits. 0+0 is carry 0 and value 0. 0+1 is carry 0 and value 1. 1+0 is carry 0 and value 1. 1+1 is carry 1 and value 0. If you look closely, you can see that the value is an XOR function. If one (and only one) of the values is 1, value is 1. Carry is an AND function. If both are 1, carry is 1. So, you don't try to program or engineer A+B in binary. You program Value=A XOR B and Carry = A AND B. That's a half adder. ~2026-29536 (talk) 14:30, 16 January 2026 (UTC)
- Some links: Half adder; AND gate; XOR; XOR gate. While this is commonly viewed as an application of Boolean algebra, the subject of Shannon's master's thesis A Symbolic Analysis of Relay and Switching Circuits, you don't actually need to understand more than the rules of binary addition and very basic digital logic to design a half adder. ‑‑Lambiam 15:18, 16 January 2026 (UTC)
- Or you could try New Math - Tom Lehrer as an introduction before going to the math of 0's and 1's 😀 NadVolum (talk) 10:26, 19 January 2026 (UTC)
January 19
[edit]Amazon AI
[edit]Good morning. Has anyone noticed that Amazon has more and more AI products, like books? ~2026-24671-3 (talk) 05:40, 19 January 2026 (UTC)
- Well you seem to have noticed. That is hardly surprising, as AI is the cool tech at the moment, despite its massive shortcomings. Shantavira|feed me 09:09, 19 January 2026 (UTC)
- I'm sorry if you've been caught out by one of those scams. I guess in the future they'll start getting good enough to fool people or some will actually like them like much of the music now coming out. And they'll write the reviews too. After Facebook then Twitter and X and TikToc and Instagram etc etc then AI on Microsoft and search engines all turning peoples brains to mush where do you expect the world to end up? NadVolum (talk) 10:08, 19 January 2026 (UTC)
- It seems everyone is getting AI, even the Wikimedia Foundation has a Lead Product Manager, AI.
- Commander Keane (talk) 12:36, 19 January 2026 (UTC)
- Yep, AI-generated books, but also product descriptions, the same product rehashed hundreds of times with different branding, etc. The latter has been going on for much longer than LLMs have been around though. https://www.youtube.com/watch?v=4UrqlMfwUC4
- If you care about basically anything at all, you should at least think about trying other places first before you look on Amazon (just search "don't books from Amazon" online, I'll avoid naming a specific outlet here). Komonzia (talk) 14:10, 19 January 2026 (UTC)
- You might like the article Dead Internet theory. Um, that's like a recommender which gets you engaged by getting angrier ;-) NadVolum (talk) 15:12, 19 January 2026 (UTC)
- Don't believe everything you see on the Web; like so many conspiracy theories, this dead Internet theory is mainly being spread by bots. ‑‑Lambiam 21:15, 19 January 2026 (UTC)
January 20
[edit]Can they be removed?
[edit]Good late afternoon. It's me yet again. I might be annoying, but i hope not. My question is if it's possible to not see those small rectangles (hope that's the right shape) with search queries (Discover More)? They're annoying. ~2026-24671-3 (talk) 17:00, 20 January 2026 (UTC)
- When and where do you see these? ‑‑Lambiam 18:51, 20 January 2026 (UTC)
- @~2026-24671-3 Wikipedia doesn't have that feature by default, so maybe it's an addon or browser extension you've installed? Or that was installed "for you" when you installed something else on your PC?
- Troubleshoot a little: Try to access the same site from a different web browser (or ideally a different device), if you don't see it, it's just the original browser/device affected. Komonzia (talk) 20:11, 20 January 2026 (UTC)
- Not on Wikipedia, on other sites. Go a site that has ads and see if you see them. ~2026-24671-3 (talk) 20:27, 20 January 2026 (UTC)
- I've never seen this. Are they embedded in the ads? Then use an ad blocker. ‑‑Lambiam 21:53, 20 January 2026 (UTC)
- @~2026-24671-3 Do they look a bit like these?
- YouTube: https://www.reddit.com/r/youtube/comments/1g2n6k9/youtube_blue_magnifying_glass_comments/
- Reddit itself: https://www.reddit.com/r/reddithelp/comments/1mum1wx/what_are_the_random_magnifying_glass_links_i_see/
- Facebook: https://www.reddit.com/r/facebook/comments/1hjyrr7/what_does_the_blue_magnifying_glass_alongside/
- If so, it's a feature of each website. Some sites are trialling it, it seems. Ad blockers likely have a way to block this kind of thing, but I'm not sure personally. Komonzia (talk) 13:28, 21 January 2026 (UTC)
- No, not those. Go on a site without Adblock and see if they appear. ~2026-24671-3 (talk) 13:57, 22 January 2026 (UTC)
- Can you attach a picture? thetechie@enwiki:~$ she/they | talk 17:25, 22 January 2026 (UTC)
- Could you turn off Adblock, go on a site like FanFiction net, search for something there and see if they appear? Please. ~2026-24671-3 (talk) 17:46, 22 January 2026 (UTC)
- @~2026-24671-3 I went there, searched for something, then turned off uBlock Origin, and from the page's source code it's very clear that there's an
<ins>tag being inserted with an 'id' ofadsbygoogle-- so it's almost definitely a new way of Google Ads putting ads in the page. It matches the developer documentation here which to me (because it mentions GDPR) means they might only be doing this in particular regions (based on local privacy laws). - You block it by using an ad blocker, if you don't want to see those. Komonzia (talk) 01:25, 24 January 2026 (UTC)
- Just to be sure, did you see those small shapes with words in them? There are often many of them in a row. ~2026-24671-3 (talk) 14:41, 24 January 2026 (UTC)
- Pretty much what you're describing https://ibb.co/qFjcD9XY
- That's in Firefox - when I open it in Chrome (where I have fewer extensions), there's also more in the middle of the page.
- Again, if you have an issue with them, best to contact the website or configure your ad blocker differently.
- uBlock Origin should already block it, but it's customisable so you can block specific parts of the page you don't like if you wanted to:
- quick demo: https://www.youtube.com/watch?v=8TvCGWwQr5o
- full guide: https://github.com/gorhill/ublock/wiki/Element-picker
- Komonzia (talk) 14:53, 24 January 2026 (UTC)
- Thanks. Do you know exactly what question they have? Also, props for it being a Hey Arnold fic! ~2026-24671-3 (talk) 21:07, 24 January 2026 (UTC)
- Just to be sure, did you see those small shapes with words in them? There are often many of them in a row. ~2026-24671-3 (talk) 14:41, 24 January 2026 (UTC)
- @~2026-24671-3 I went there, searched for something, then turned off uBlock Origin, and from the page's source code it's very clear that there's an
- Could you turn off Adblock, go on a site like FanFiction net, search for something there and see if they appear? Please. ~2026-24671-3 (talk) 17:46, 22 January 2026 (UTC)
- I use the Internet quite a lot and most sites don't have it, for me. So it's either a you problem or a site-specific problem. :) Komonzia (talk) 01:15, 24 January 2026 (UTC)
- Can you attach a picture? thetechie@enwiki:~$ she/they | talk 17:25, 22 January 2026 (UTC)
- Can confirm that adblock does not block the "feature". thetechie@enwiki:~$ she/they | talk 17:26, 22 January 2026 (UTC)
- No, not those. Go on a site without Adblock and see if they appear. ~2026-24671-3 (talk) 13:57, 22 January 2026 (UTC)
- Not on Wikipedia, on other sites. Go a site that has ads and see if you see them. ~2026-24671-3 (talk) 20:27, 20 January 2026 (UTC)
Egyptian hieroglyphs not displaying properly on Wikipedia pages
[edit]I have just noticed that there are quotes in Egyptian hieroglyphics in some Wikipedia articles (e.g. here, after the words 'is depicted standing on a lion on a plaque where she is given the triple name of Qetesh-ʿAstart-ʿAnat'), except that half the hieroglyphs aren't displayed properly in my browser and empty squares appear instead. Does anyone have any idea why that would be the case and how I could fix it? I'm using an up-to-date version of Chrome (and I get the same result in Microsoft Edge), I have Segoe UI Historic, and I have even installed Noto Sans Egyptian Hieroglyphs. ~2026-41408-0 (talk) 23:24, 20 January 2026 (UTC)
- Look at this diff. It appears that the hieroglyphs that don't render were added between the original hieroglyphs. I'm not sure what's going on here. user:Temerarius, user:Antiquistik, what does it all mean? Oh, codepoints.net tells me that the offending glyph is "Egyptian Hieroglyph Vertical Joiner", a non-printing character for positioning on hieroglyph above another. Looking at Template:Script/Egyptian_Hieroglyphs, perhaps you need Egyptian OpenType, eot.ttf, which is linked to there under "Egyptian Text"? Card Zero (talk) 23:46, 20 January 2026 (UTC)
- The page and the question are certainly in my wheelhouse, is that why you're asking me? I don't recall being involved in the edits. I see empty squares too. It'd be nice if the joiner worked!
- Temerarius (talk) 00:42, 21 January 2026 (UTC)
- For what it's worth, it seems to display correctly for me. I'm using Firefox with no special (hieroglyph-related or otherwise) fonts installed. Matt Deres (talk) 03:48, 21 January 2026 (UTC)
- Me too, on Arch Linux and Zen Browser (a Firefox fork). I wonder if this is a lack of Unicode/rendering compatibility in Windows, as they mentioned Segoe is installed. thetechie@enwiki:~$ she/they | talk 05:23, 21 January 2026 (UTC)
- I'm on an Android tablet at the moment and the joiner chars aren't working. Card Zero (talk) 05:58, 21 January 2026 (UTC)
- @Card Zero Are you on Chrome? I wonder if it's a bug with non-Linux, as Linux, in my experience, sometimes contains fonts/keyboard layouts, etc. preinstalled that other OS's don't have. I wonder if it's because I installed KDE Plasma so I have Noto fonts. thetechie@enwiki:~$ she/they | talk 15:50, 23 January 2026 (UTC)
- For clarification, Noto is the default font on that environment and it installs a bunch of Noto fonts if you preinstall KDE Plasma on a Linux distro. thetechie@enwiki:~$ she/they | talk 01:21, 24 January 2026 (UTC)
- @Card Zero Are you on Chrome? I wonder if it's a bug with non-Linux, as Linux, in my experience, sometimes contains fonts/keyboard layouts, etc. preinstalled that other OS's don't have. I wonder if it's because I installed KDE Plasma so I have Noto fonts. thetechie@enwiki:~$ she/they | talk 15:50, 23 January 2026 (UTC)
- I'm on an Android tablet at the moment and the joiner chars aren't working. Card Zero (talk) 05:58, 21 January 2026 (UTC)
- Me too, on Arch Linux and Zen Browser (a Firefox fork). I wonder if this is a lack of Unicode/rendering compatibility in Windows, as they mentioned Segoe is installed. thetechie@enwiki:~$ she/they | talk 05:23, 21 January 2026 (UTC)
- @~2026-41408-0 I wondering if this is a rendering thing with Windows. this page says to install Noto Sans Cuneiform or a similar font instead? Maybe try installing the Noto one?
- (If you really want to, you could install all of the Noto fonts, but I don't think there's any easy way to do it on Windows (you can use a package manager on Linux)). thetechie@enwiki:~$ she/they | talk 05:26, 21 January 2026 (UTC)
- I reckon it must require The Egyptian OpenType font, distributed by Microsoft on github here. "Renders Egyptian Hieroglyph Format Controls", says our template page. Card Zero (talk) 06:00, 21 January 2026 (UTC)
Installing the Egyptian OpenType font worked. Thanks to everyone and especially to Card_Zero! It's a pity that most people will have the same problem when reading such pages and probably won't find the solution.--~2026-41408-0 (talk) 06:42, 21 January 2026 (UTC)
- Is there a Wikipedia page where you looked for advice first, before the ref desk? Maybe that page needs this info adding. Card Zero (talk) 08:40, 21 January 2026 (UTC)