Wikiversity:Colloquium/archives/December 2023
Wikiversity project to encourage Wikimedia work within a University System
editHi folks. I've initiated a Wikiversity project, SUNY Wikimedia Project, which seeks to connect students, faculty and staff in the State University of New York (SUNY) system who are interested in working on Wikimedia projects. I'm currently organizing a panel presentation at an upcoming conference, and seeking participation in other projects as well. My question for Wikiversity: is this the kind of projects Wikiversity seeks to host, sponsor and empower? I've been trying to figure out the current mission and direction of Wikiversity, and it seems to me that this would be a good fit for this project. I'd certainly appreciate any thoughts. Thanks! Stevesuny (discuss • contribs) 17:34, 1 December 2023 (UTC)
- I certainly think so: Wikiversity has a very loose mission about being a community for learning, so as long as you avoid a handful of forbidden kinds of content, there's really not much you can't host here as long as it's plausibly educational in nature, including many resources about Wikimedia projects themselves. —Justin (koavf)❤T☮C☺M☯ 18:01, 1 December 2023 (UTC)
- Should be good to go IMO. —Atcovi (Talk - Contribs) 18:22, 1 December 2023 (UTC)
- Broadly speaking, this is not only within the scope of Wikiversity but one of its explicit goals, and it's something we'd like to encourage more of.
- Here's a couple of observations and tips I'd give, based on results of past courses:
- Don't fall into the trap of treating the wiki like a clunky version of Google Docs. One of the most fundamental features of a wiki is that it's collaborative; projects should aim to make use of that.
- Keep in mind that textual contributions to the wiki are required to be freely licensed, are publicly visible, and may be indexed by search engines. Give students alternative options for completing coursework if they're unable or unwilling to agree to those licenses. Avoid assignments which require students to post personally identifiable information.
- While fair use content is permitted on Wikiversity under limited circumstances, advising students to use freely licensed content only (and to upload it to Commons instead of locally) may be simpler. Give students a brief overview/refresher about copyright law before any assignments that are likely to involve uploading images.
- Omphalographer (discuss • contribs) 19:03, 1 December 2023 (UTC)
- Thanks for the suggestions. These comments are more directly relevant to WikimediaCourse, another Wikiversity project I've started. In that effort, I will be hosting a live class for tuitioned students at SUNY Poly, where I teach. The open component is a work in progress: of course, the syllabus, the assignments, access to seeing student work, and the student projects themselves will be open (I'm not sure I'll offer students an alternative option). I'd love to have open students from anywhere, and will see what happens. If anyone is interested, I'd welcome comments and suggestions for making the course better. Stevesuny (discuss • contribs) 16:35, 2 December 2023 (UTC)
Announcing new Wikiversity projects?
editHi again. Is there a preferred place in Wikiversity to post a description (example follows) of a new project, to seek input and comment and suggestions? Or, is this the place? I'd like to share a few projects, and didn't want to clog the Colloquium.
Here is the post I'd put somplace called "New Projects":
I am planning a project called "Course: Designing and Writing Interactive Texts Using MediaWiki". The project, similar to my current project, Digital Media and Information in Society; will house the learning materials for a course being offered at the University where I teach. The course is focused on the theory and practice of hypertextual writing and reading; and will use MediaWiki as the primary teaching and learning platform.
The project is to develop a syllabus with explicit goals, outcomes, objectives, resources, links to live/recorded zoom sessions, assignments and rubrics. All students (including registered students, and open enrollment students if any materialize) will complete their assignments in a Wikiversity space, or perhaps in some other spaces to be determined.
The initial project overview is in my sandbox
Stevesuny (discuss • contribs) 16:55, 2 December 2023 (UTC)
- This is a good general place to announce new things or ask for help. Once a resource is mature and usable, it can be added to Main Page/News. —Justin (koavf)❤T☮C☺M☯ 18:35, 2 December 2023 (UTC)
Wikiversity Namespace Question
editHi folks, as a newcomer to Wikiversity and the Wikiworld, and having enjoyed the freedom of Wikiversity, I've spent a bit of time exploring the structure and community of Wikiversity, and am wondering about the use of the namespace (something I've only just discovered).
Is there a preferred or desired naming structure for projects that house the syllabus of a course being taught, with explicit goals, outcomes, objectives, resources, links to live/recorded zoom sessions, assignments and rubrics.
To date, I've created two: Digital Media and Information in Society and WikimediaCourse, and am about to initiate a third, and thought perhaps I should do the naming differently...
Thanks! Stevesuny (discuss • contribs) 17:00, 2 December 2023 (UTC)
- Virtually anything you'd create would likely be in the main namespace. If you're working on some really early draft material that you don't want to publish to the community at large, you can use Draft or User. —Justin (koavf)❤T☮C☺M☯ 18:37, 2 December 2023 (UTC)
- @Stevesuny: The pages on namespaces and content organization are unclear and full of jargon. I've complained about it several times, but I suppose it's probably best to group related contents under a single page in namespace, just as one would organize any hierarchical structure such as a file system. I typically use my userpage for stuff that I don't feel is sufficiently well-developed to move into its own resource in mainspace and then move it when and if I'm satisfied with the direction. Otherwise if I didn't end up satisfied with it I'd have to bother someone to delete the resource. The help docs for most wikimedia projects are not as accessible or easily understood as they probably could be though. AP295 (discuss • contribs) 03:27, 6 December 2023 (UTC)
- As far as organization and naming conventions go, there's a lot of junk left over from an early, overly complex organizational scheme which tried to slot everything into a hierarchy of "schools", "divisions", "departments", and "lessons". Most of that is (thankfully) gone but there's still bits and pieces of it left around. If you find pages in the Wikiversity namespace still recommending this structure, please tag them as {{historical}}.
- In my opinion, the site could probably use another pass of simplification. There's a lot of residual "cruft" in the Portal: and School: namespaces that should probably be folded into the main (resource) namespace or removed altogether. In many cases, there's much more navigational structure than actual learning material. This should be fixed; if we don't have any substantial learning material on a subject, we shouldn't waste visitors' time leading them through a maze of abandoned portals, directories, and introduction pages.
- Omphalographer (discuss • contribs) 04:34, 6 December 2023 (UTC)
- It's mostly this page Wikiversity:Namespaces. Hard to make sense of any of it. Generally it's unclear how one should organize their content. Really each user should be encouraged or at least entitled to organize their content within a user directory in mainspace, e.g. your content would go under Omphalographer/ but I recall reading a discussion where Dan Polansky was explicitly told not to do this. It's common sense and common practice to have user directories. That's how the filesystem in any multi-user operating system is organized. AP295 (discuss • contribs) 08:50, 6 December 2023 (UTC)
- I could see adding to the Namespaces something for users, professors, labs or research groups: Replicating, in part, the structure of a bricks-and-mortar university. I think Wikiversity can play a significant role by inviting professors to put some of their courses and other learning materials on Wikiversity, even if they are in support of credit-bearing courses. We can use categories, rather than namespace, to organize by topic or subject area. Stevesuny (discuss • contribs) 18:22, 7 December 2023 (UTC)
- With the guidelines and rules as unclear as they are I doubt that many professors really need the grief of having to deal with all of this. Each user should be entitled to a page in mainspace, essentially a "user" directory, presumably with the same name as their account. I don't see the use in imposing any further structure, though I think users should be encouraged to correctly tag their resources and that there should be a way to browse by topic, which would presumably rely on category tags. While wikiversity is supposedly a collaborative effort, again I can't imagine many professors really want anyone mucking about with their material. Most of the pages here do indeed have a primary author. I would never modify one without speaking with the author and obtaining their consent on the talk page first. AP295 (discuss • contribs) 23:32, 9 December 2023 (UTC)
- I could see adding to the Namespaces something for users, professors, labs or research groups: Replicating, in part, the structure of a bricks-and-mortar university. I think Wikiversity can play a significant role by inviting professors to put some of their courses and other learning materials on Wikiversity, even if they are in support of credit-bearing courses. We can use categories, rather than namespace, to organize by topic or subject area. Stevesuny (discuss • contribs) 18:22, 7 December 2023 (UTC)
- Here it is, https://en.wikiversity.org/wiki/Wikiversity:Colloquium/archives/June_2023 "We don't name main space resources after users except as subpages." God only knows why. AP295 (discuss • contribs) 08:56, 6 December 2023 (UTC)
- It's mostly this page Wikiversity:Namespaces. Hard to make sense of any of it. Generally it's unclear how one should organize their content. Really each user should be encouraged or at least entitled to organize their content within a user directory in mainspace, e.g. your content would go under Omphalographer/ but I recall reading a discussion where Dan Polansky was explicitly told not to do this. It's common sense and common practice to have user directories. That's how the filesystem in any multi-user operating system is organized. AP295 (discuss • contribs) 08:50, 6 December 2023 (UTC)
Removal of CourseCat template from resources
editI understand that this template is being used to assist the categorization of Wikiversity resources, and I'm usually adding them to uncategorized resources. I recently noticed that several users are removing the addition of {{CourseCat}} (example). Should we keep them uncategorized or re-add the template and tell them not to remove it? I'm looking forward to hearing from the community. MathXplore (discuss • contribs) 03:46, 7 December 2023 (UTC)
- @Stevesuny: Can you explain the removal? Thanks. —Justin (koavf)❤T☮C☺M☯ 07:08, 7 December 2023 (UTC)
- I didn't know what it was or where it came from :). I'll check it out.
- It might also be a casualty of my workflow, which sometimes involves generating mediawiki markup in chatGPT or in tiddlywiki, and replacing entire pages by select all/paste. I'm struggling with managing multi-page wiki projects, and maintaining categorization. One strategy I've hit on is to transclude an entire cluster of pages from my sandbox into the main namespace, and categorize in the main namespace space. See WikimediaCourse for an example. And I know: this workflow has the disadvantage of obliterating collaborative work... Stevesuny (discuss • contribs) 16:47, 7 December 2023 (UTC)
- {{CourseCat}} looks interesting: I looked briefly at the documentation, and it looks like it relies on more structure to find "chapter" ; I'll added it into the mainpage of my courses (WikimediaCourse and Digital Media and Information in Society) and see what it does. Thanks! Stevesuny (discuss • contribs) 16:56, 7 December 2023 (UTC)
- Thank you for the explanation, if there are any specific teacher's preferences about resource categorizations then I think that should be prioritized. MathXplore (discuss • contribs) 08:35, 10 December 2023 (UTC)
RFC: Deprecate and remove the Collection: namespace.
editIssue has been closed as resolved
|
---|
The Books feature was an initiative by Wikimedia, launched around 2009, to allow users to create collections of pages and render them as a PDF "book", and order physical copies of those books online. These books were represented on Wikiversity as pages under the Collection: namespace. This feature was never heavily used on Wikiversity, and most of the books which were created were either collections of all pages from a course, or arbitrary collections of unrelated pages. Wikimedia's PDF book renderer went offline around 2017; subsequently, some of the other on-wiki tools were disabled as well. As a result, these pages no longer have any clear purpose; most resources on Wikiversity are already organized hierarchically and placed into categories, so there's no need for additional tools to create collections of pages. (For more background on the whole situation, please refer to w:Wikipedia:Village pump (proposals)/Archive 181#The future of the Book namespace, a 2021 discussion in which Wikipedia decided to similarly remove their Book: namespace.) Proposal: All remaining pages in the Collection: namespace should be deleted by a curator. Once this is complete, I will make a Phabricator request to disable the namespace entirely. Omphalographer (discuss • contribs) 17:52, 8 December 2023 (UTC)
|
Guidelines
editI'm currently working on my first article at WikiJournal Preprints/Castle Eppstein, and I have a question regarding the naming and abstracts of articles. Should I follow WP lede and naming guidelines? Or do you guys have your own? Crainsaw (discuss • contribs) 20:45, 9 December 2023 (UTC)
- We already have Wikiversity:Naming conventions, I think this can help you. I don't see any major issues to your title. Thank you for joining Wikiversity. MathXplore (discuss • contribs) 08:27, 10 December 2023 (UTC)
Technical questions, and a few concerns
editI'm unable to use harv references on my pre-print WikiJournal Preprints/Castle Eppstein properly, can anybody help me?
Also, I'm concerned about the peer review process. The recent addition to the Journal of Humanities Loveday, 1458 , had been submitted in 2020, and got accepted 3 years later. WikiJournal Preprints/Fascism and Italy is still not under peer-review, since submission way back in april 2019. It usually only takes 2-6 weeks for a peer review, what's taking so long? Not enough volunteers? Crainsaw (discuss • contribs) 12:26, 10 December 2023 (UTC)
- Correct: we do not have enough editors here. As for the references, what exactly is the problem? They seem to be working. —Justin (koavf)❤T☮C☺M☯ 12:37, 10 December 2023 (UTC)
- It's in Category:Harv and Sfn no-target errors. Crainsaw (discuss • contribs) 13:03, 10 December 2023 (UTC)
- Could I conduct the peer reviews? Crainsaw (discuss • contribs) 19:12, 10 December 2023 (UTC)
- Given the lack of prospective reviewers to participate on a formally organized review, I think you are better off moving WikiJournal Preprints/Castle Eppstein to Castle Eppstein and figure out what it is that you are trying to do in the article that an encyclopedia article would not do. --Dan Polansky (discuss • contribs) 12:41, 10 December 2023 (UTC)
- I'm not sure whether a resource will be the best fit, sources are scarce. There are a lot of primary sources in the article, and there's some Synth as well. Crainsaw (discuss • contribs) 13:19, 10 December 2023 (UTC)
- If you place Template:Original research at the top of the article, no one should complain that you placed the original-research article at Castle Eppstein, I think. But if you want to keep the article in the preprint space, no one will complain either, I guess. --Dan Polansky (discuss • contribs) 13:54, 10 December 2023 (UTC)
- I'm not sure whether a resource will be the best fit, sources are scarce. There are a lot of primary sources in the article, and there's some Synth as well. Crainsaw (discuss • contribs) 13:19, 10 December 2023 (UTC)
What should we do with country name redirects to the "Comparative law and justice" series?
editI recently noticed that we have several country name redirects like Algeria and Brazil etc. that are going to the "Comparative law and justice" series. While there is connection between the redirect and their targets, it is not reasonable to assume that many readers are interested in law and justice. Is there anything we should do with these kind of redirects? We may change the redirect target to their respective topic category or portal, or transform them into a disambiguation-like page. Some users may say that these should be deleted as bad redirects. Or should we keep them in their current state until we have a better idea? I'm looking forward to hearing from others about this matter. MathXplore (discuss • contribs) 07:59, 13 December 2023 (UTC)
- These should essentially be disambiguation pages that also include any resources about these places in particular and probably other resources related to culture, like language courses. —Justin (koavf)❤T☮C☺M☯ 08:47, 13 December 2023 (UTC)
- Thank you for your opinion. Should we import disambiguation-related templates from enwiki to setup disambiguation pages? MathXplore (discuss • contribs) 03:39, 14 December 2023 (UTC)
- I think one is sufficient: this project isn't that big so a single {{dab}} tempate should be fine. —Justin (koavf)❤T☮C☺M☯ 06:24, 14 December 2023 (UTC)
- Thank you for your opinion. Should we import disambiguation-related templates from enwiki to setup disambiguation pages? MathXplore (discuss • contribs) 03:39, 14 December 2023 (UTC)
- I propose:
- 1) Delete Algeria and Brazil as wrongly pointing the readers to one resource in exclusion of other resources. Once they get deleted, the search function automatically serves as an analog of disambiguation page. If that is opposed, creating a simple disambiguation page (no templates and complications or perhaps a very simple template saying "This is a disambiguation page", "This is a directory page", etc.) is a decent alternative. The deletion option is easier to maintain than a disambiguation/directory page, and it shows no editorial preference for some pages over others. On the other hand, one may indeed want to manually curate the list of pages directed to from e.g. "Algeria", and place a link to search results as one items in the disambiguation/directory page. Thus, the disambiguation/directory page would be something like "top 7 most relevant results", manually curated; other options for organizing a disambiguation/directory page are possible.
- 2) Delete Comparative law and justice/Algeria as completely useless: the page has no a single statement of value and not a single further reading. The last substantive edit is from 2009. Proceed similarly with content-free pages via speedy deletion.
- --Dan Polansky (discuss • contribs) 06:44, 14 December 2023 (UTC)
- Per your suggestion, I added {{prod}} to Comparative law and justice/Algeria, I will also check other pages in this series. MathXplore (discuss • contribs) 04:03, 15 December 2023 (UTC)
(New) Feature on Kartographer: Adding geopoints via QID
editSince September 2022, it is possible to create geopoints using a QID. Many wiki contributors have asked for this feature, but it is not being used much. Therefore, we would like to remind you about it. More information can be found on the project page. If you have any comments, please let us know on the talk page. – Best regards, the team of Technical Wishes at Wikimedia Deutschland
Question about Peer Reviews
editMaybe Talk:WikiJournal of Humanities is a better palce for this, but,
as per my understanding off WikiJournal of Humanities/Peer reviewers only qualified external professionals are allowed to formally peer-review articles.
Which seems a bit unnecessary considering anyone can write articles, also Nupedia vibes, this seriously hurts the journal's growth due to lack of volunteers.
I propose that editors on wikipedia who've written extensively on related topics also be invited to peer-review articles, this would add a lot more volunteers (thus making the whole process faster and smoother), and get qualified people from wikipedia over to wikiversity. Crainsaw (discuss • contribs) 23:07, 14 December 2023 (UTC)
- @Crainsaw: (You chose a good venue, I think.) I think you have a good point that the requirements on reviewers are excessive. I suspect those "peer-review articles" in Wikiversity cannot be taken very seriously. My guess is that you are unlikely to attract enough reviewers for your article, qualified or less qualified. --Dan Polansky (discuss • contribs) 12:12, 17 December 2023 (UTC)
- I was involved with the creation of a wikijournal, and much of the initial leadership is still active. The wikijournals have always been largely independent of the rest of Wikiversity. They may or may not have strong opinions about changing the editorial standards of one journal, but my guess is that they will consider "WikiJournal" to be a sort of brand or trademark. We must honor that position if they hold it. Degrading the reputation of one wikijournal for high editorial standards hurts the reputation of all wikijournals. In the early days of their existence, the Wikijournals requested that the Wikimedia Foundation create a separate wiki for these journals . The Wikiversity community unanimously supported that request for the WikiJournals to separate from Wikiversity. (see proof)-Guy vandegrift (discuss • contribs) 14:40, 17 December 2023 (UTC)
- I've posted at Talk:WikiJournal User Group#Peer Reviews, which I believe would be a better board.
- To address your point about editorial standards: It's a Wiki, anybody can contribute, so anybody should also be allowed to review it, people who mightn't have academic qualifications, but are still experts in their fields are also possible peer reviewers. The peer reviews are public. That's what makes a Wiki a Wiki, we can't mimic everything from Traditional journals, the fact the Wikijournals has an ISSN and gets a DOI for every publication and is indexed by Google Scholar goes to show Wikijournals done a good job at fitting in. Crainsaw (discuss • contribs) 16:48, 17 December 2023 (UTC)
- As for Meta: Proposal: WikiJournal as a sister project, there is no unanimity (there are opposes), but there is indeed an overwhelming majority supporting the proposal. The proposal has not been implemented yet. The takeaway is that WikiJournal is to be treated as a dedicated/formally separate project within Wikiversity and that whatever rules it has established for itself so far should be take seriously, I guess. (Given the low activity within WikiJournal, it is not clear to me to what extent making it a sister project is really desirable.) --Dan Polansky (discuss • contribs) 09:03, 26 December 2023 (UTC)
- You are correct. The vote was 177/40, or about a 80% to 20% split (if you ignore those who remained neutral.) When I "counted the vote" I failed to realize that the ayes and nays were in different sections. Guy vandegrift (discuss • contribs) 10:46, 26 December 2023 (UTC)
- I was involved with the creation of a wikijournal, and much of the initial leadership is still active. The wikijournals have always been largely independent of the rest of Wikiversity. They may or may not have strong opinions about changing the editorial standards of one journal, but my guess is that they will consider "WikiJournal" to be a sort of brand or trademark. We must honor that position if they hold it. Degrading the reputation of one wikijournal for high editorial standards hurts the reputation of all wikijournals. In the early days of their existence, the Wikijournals requested that the Wikimedia Foundation create a separate wiki for these journals . The Wikiversity community unanimously supported that request for the WikiJournals to separate from Wikiversity. (see proof)-Guy vandegrift (discuss • contribs) 14:40, 17 December 2023 (UTC)
Use of Wikiversity:Requests for Deletion
editCategory:Candidates for speedy deletion states:
- "Delete these pages as soon as possible when legitimately labeled for speedy deletion. Normally {{delete}} should be replaced with {{deletion request}} and discussion started at Wikiversity:Requests for Deletion when not a legitimate speedy deletion candidate."
Is that true/accurate? Should placement of template "deletion request" really be accompanied by a discussion started at Wikiversity:Requests for Deletion? --Dan Polansky (discuss • contribs) 14:25, 19 December 2023 (UTC)
- As you can see, in practice, the deletion process is pretty broken: some proposed deletions expire and aren't actually deleted for months. :/ —Justin (koavf)❤T☮C☺M☯ 14:29, 19 December 2023 (UTC)
- Since you are an admin, you can delete expired (3 months gone?) proposed deletions, can't you? --Dan Polansky (discuss • contribs) 14:35, 19 December 2023 (UTC)
- I can and I do sometimes, but I do prefer to actually investigate before deleting and that has some overhead. Since I made my last comment, I deleted something that was a valid deletion proposal. Unfortunately, with as small as this project is, there is still some real junk on here. —Justin (koavf)❤T☮C☺M☯ 14:37, 19 December 2023 (UTC)
- I think it would be better to use Wikiversity:Requests for Deletion, similarly how Wiktionary works. Then, multiple people would explicitly approve a deletion in each request for it and provide a rationale for it. Then, one could feel more comfortable deleting the items based on evidence of consensus on a per-item level. And this is what the quoted text at Category:Candidates for speedy deletion seems to suggest. --Dan Polansky (discuss • contribs) 14:42, 19 December 2023 (UTC)
- I can and I do sometimes, but I do prefer to actually investigate before deleting and that has some overhead. Since I made my last comment, I deleted something that was a valid deletion proposal. Unfortunately, with as small as this project is, there is still some real junk on here. —Justin (koavf)❤T☮C☺M☯ 14:37, 19 December 2023 (UTC)
- Since you are an admin, you can delete expired (3 months gone?) proposed deletions, can't you? --Dan Polansky (discuss • contribs) 14:35, 19 December 2023 (UTC)
- You may have overlooked the qualifier at the end, "…when not a legitimate speedy deletion candidate". {{delete}} is for deletion of material like spam, broken redirects, or accidentally created pages which shouldn't require discussion; {{deletion request}} is for tagging other content for discussion at RfD. Omphalographer (discuss • contribs) 20:33, 24 December 2023 (UTC)
- My point is not about speedy but rather about how to administer non-speedy deletion request, that is, whether non-speedy deletion request is to be accompanied by listing at Wikiversity:Requests for Deletion. Since the quotation I provided suggests the accompaniment to me, but I was told elsewhere the accompaniment is not to be there, and indeed, people are in fact adding non-speedy deletion requrest without listing the pages at Wikiversity:Requests for Deletion. --Dan Polansky (discuss • contribs) 21:44, 24 December 2023 (UTC)
- I always thought placing
{{subst:prod}}
on a page guaranteed that the page would automatically appear inCategory:60-day proposed deletions
. Am I wrong about that??? ... . Also, in the past, Dave and I preferred that people not use "deletion request" because it created long discussions at a rate that exceeded our ability to resolve them. Maybe that has changed now that we have several active and competent new housecleaners on Wikiversity.--Guy vandegrift (discuss • contribs) 06:47, 26 December 2023 (UTC)- The practice you mention contradicts the text I quoted at the top, doesn't it? (Sure, pages proposed for deletion appear at "Category:60-day proposed deletions", and I did not suggest otherwise.) --Dan Polansky (discuss • contribs) 08:07, 26 December 2023 (UTC)
- I always thought placing
- My point is not about speedy but rather about how to administer non-speedy deletion request, that is, whether non-speedy deletion request is to be accompanied by listing at Wikiversity:Requests for Deletion. Since the quotation I provided suggests the accompaniment to me, but I was told elsewhere the accompaniment is not to be there, and indeed, people are in fact adding non-speedy deletion requrest without listing the pages at Wikiversity:Requests for Deletion. --Dan Polansky (discuss • contribs) 21:44, 24 December 2023 (UTC)
Original LLM (Large Language Model) Prompts for Learning and Research (no LLM generated content)
editSo I had an idea. If no one objects I may create some content related to prompt libraries that are meant for learning, research, and teaching. That is, prompts for large language models that can assist one in learning, teaching, research, and learning. I will not post any content generated by any large language model. That will only create original content that is prompts related to various topics that individuals can use with LLM's to potentially facilitate learning, teaching, more research. I'm not sure when I will start doing this. Maybe soon. I sort of see LLM's like calculators (or spoken/natural language mini software programs that output text) for math or science that can be applied to many subjects. Again this will only be original content, no LLM generated content. This will be content to potentially customize for specific uses, and then provide to an LLM to then output various useful results related to learning, teaching, and research. That is, if no one objects. Bless up and limitless peace. Michael Ten (discuss • contribs) 03:25, 26 December 2023 (UTC)
- Can you give an example of the sort of content you're proposing, and what educational purpose it would serve? Omphalographer (discuss • contribs) 21:32, 27 December 2023 (UTC)
- Sort of like this Large language models. I seemed to see no major opposition so I am trying to create something constructive and informational. Bless up. Michael Ten (discuss • contribs) 03:32, 30 December 2023 (UTC)
- A set of LLM chat prompts to ask an LLM what you can do with LLMs? This seems oddly circular. We don't host collections of queries to type into search engines to find information about subjects; I don't see why LLM prompts should be any different in this regard. Omphalographer (discuss • contribs) 04:30, 30 December 2023 (UTC)
- It strikes me as faintly promotional. AP295 (discuss • contribs) 20:44, 30 December 2023 (UTC)
- A set of LLM chat prompts to ask an LLM what you can do with LLMs? This seems oddly circular. We don't host collections of queries to type into search engines to find information about subjects; I don't see why LLM prompts should be any different in this regard. Omphalographer (discuss • contribs) 04:30, 30 December 2023 (UTC)
- Sort of like this Large language models. I seemed to see no major opposition so I am trying to create something constructive and informational. Bless up. Michael Ten (discuss • contribs) 03:32, 30 December 2023 (UTC)
- @Michael Ten: Is there a rule against copy/pasting LLM content? I hope not, because I just pasted a chat with BARD at Physics/A/Dice group theory.
- My research online suggests that LLM-output should be free from U.S. copyright protection since it is not a result of an activity of human mind, but I also found articles suggesting the copyright issue is a gray zone. I would recommend to you to do your own research online and figure out whether you can confidently claim that LLM-output is free from copyright. I for one feel uncomfortable placing LLM-output into Wikiversity except as fair use. Some time ago, I created Is the output of ChatGPT copyrighted?. --Dan Polansky (discuss • contribs) 09:48, 28 December 2023 (UTC)
- "Open"AI's TOS seems to grant users ownership: "Ownership of Content. As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output." Not a lawyer, but it doesn't seem like a violation of copyright to post ChatGPT output. Not that I'd want to in the first place, in most cases. The model, its training data, training methods, parameters and other operational details are apparently trade secrets and I wouldn't be comfortable relying on it. In my experience, its output is of poor quality as far as discourse goes. It would make for a rather mediocre resource at best. AP295 (discuss • contribs) 13:18, 28 December 2023 (UTC)
- That's rather compelling, isn't it? A link: https://openai.com/policies/terms-of-use. I highlight to the extent permitted by applicable law, though. Could someone argue that the copyright holders of the training data hold copyright in the output? I would argue they do not since copyright only protects expression, not information, but I am not a lawyer.
- One would be posting ChatGPT content at least for two purposes: 1) commentary on ChatGPT accuracy; 2) as a writing aid, where one lest ChatGPT formulate things, but one performs checking of the output/content independently as a reviewer. --Dan Polansky (discuss • contribs) 13:29, 28 December 2023 (UTC)
- From my limited experience its output is very generic and formulaic, which is part of my gripe in the essay. Let the copyright lawyers and other suits figure out the details, most of the output is hardly worth bothering with in the first place. AP295 (discuss • contribs) 13:50, 28 December 2023 (UTC)
- Obviously though if you take an off-the-shelf LLM (one that's actually 'open') or use a service like ChatGPT and it spits out a paragraph from some book, that doesn't automatically make it your intellectual property. AP295 (discuss • contribs) 13:56, 28 December 2023 (UTC)
- It seems unlikely that a trained model using neural networks would output a word-for-word paragraph from a particular source, but I am no expert on the matter. --Dan Polansky (discuss • contribs) 13:58, 28 December 2023 (UTC)
- It will if you train it only on that paragraph. It might even if you have a reasonably large dataset and you start it off with some text from that paragraph. It seems likely that OpenAI would have taken some measure to prevent this, but chatgpt in particular is so over-parameterized that it can memorize en.wikipedia and have plenty of room to spare. Really most large ANN models are quite over-parameterized and can easily memorize a small dataset. Overfitting is a big problem. AP295 (discuss • contribs) 13:59, 28 December 2023 (UTC)
- Completing a word or sentence or paragraph was actually the object of many generative natural language models, including many that were a tiny fraction of ChatGPT's size. I'm behind on the literature, so I don't know how chatgpt avoids it, but perhaps their loss function was engineered to penalize verbatim outputs over a certain length. Interesting that they won't tell you the size of the training set they use. If my goal were to repeat or imitate ChatGPT's functionality, (with the disclaimer that I've not read any recent literature) I'd optimize prediction on a dataset including wikipedia, quora, stackexchange, and whatever else, and then (possibly fixing those parameters) optimize a smaller, downstream part of the network just on conversation and Q&A examples e.g. reddit, quora, stackexchange, with a penalty on verbatim output. In other words, it would overfit much of the dataset but mince everything up and recombine it in a generic, grammatical, conversational format. Rather crude, but then you could (somewhat disingenuously) call it a self-contained "AI". All just speculation though, I'm not even very familiar with transformer models nor at all with the state of the art. I'd be impressed with if it were, say, a large college research project rather than the overhyped flagship product of a privatized NPO, largely built off the efforts of others and then withheld as a trade secret and sold as a subscription service. I don't like any of this. AP295 (discuss • contribs) 14:12, 28 December 2023 (UTC)
- While you may find it "unlikely", the New York Times actually filed a lawsuit against OpenAI a few days ago because they found that it will reproduce lengthy excerpts from many New York Times stories nearly verbatim when prompted to do so. Details here, with numerous examples in the legal complaint: https://www.theverge.com/2023/12/27/24016212/new-york-times-openai-microsoft-lawsuit-copyright-infringement. With this in mind, it seems likely that it will reproduce content from other sources as well, and not necessarily only when it was requested. Omphalographer (discuss • contribs) 04:36, 30 December 2023 (UTC)
- I suspect it isn't the linguistic component of the model that's large, but all the other info it has to store, in which case "large language model" is probably a misleading term. It would make sense to try and isolate this subcomponent that interperets questions and distills into words information retrieved from "memory", perhaps by architectural choices and training method. It wouldn't surprise me if chatGPT does just that somehow (nor if it's been done in a thousand papers already). It would likely be far smaller than 700GB. AP295 (discuss • contribs) 06:18, 30 December 2023 (UTC)
- And don't take NYT's word for it (nor for anything else). Download a basic model, train it for a while and see what it spits out. It seems like since GPUs became powerful and feedforward CNNs and transformer networks became the hottest craze, you don't hear much about other architectures which are frankly much more interesting if less performant (perhaps for being less suited to GPU acceleration, though CNNs and transformer networks are very good models and inherently well suited to many tasks as well as the hardware they run on). Have a look at Boltzmann networks. Anyone in physics will probably get a kick out of it. You could train them on an image library, and then during inference you could show them a small portion of the image and after running the network for a while, it would complete the rest of it. The last public chatGPT model is really kind of boring if you ask me, and so are CNNs, or at least the process of training them on a large dataset is very tedious.AP295 (discuss • contribs) 06:31, 30 December 2023 (UTC)
- Interesting. Does the article indicate the length of the verbatim content? Since, in the article, I only find "can generate output that recites Times content verbatim, closely summarizes it, and mimics its expressive style", but it says nothing about the length of that output. Is it really able to recall a complete paragraph verbatim? If so, is there a specific example anywhere on the web? Since, this problem should be reproducible, shouldn't it? --Dan Polansky (discuss • contribs) 07:37, 30 December 2023 (UTC)
- There are examples in the legal complaint, which is embedded in the article. Many of the excerpts regurgitated by ChatGPT are multiple paragraphs long, and differ from the original by no more than a word or two. Omphalographer (discuss • contribs) 08:17, 30 December 2023 (UTC)
- Thank you. Here's a pdf of that complaint: https://nytco-assets.nytimes.com/2023/12/NYT_Complaint_Dec2023.pdf --Dan Polansky (discuss • contribs) 08:27, 30 December 2023 (UTC)
- There are examples in the legal complaint, which is embedded in the article. Many of the excerpts regurgitated by ChatGPT are multiple paragraphs long, and differ from the original by no more than a word or two. Omphalographer (discuss • contribs) 08:17, 30 December 2023 (UTC)
- It seems unlikely that a trained model using neural networks would output a word-for-word paragraph from a particular source, but I am no expert on the matter. --Dan Polansky (discuss • contribs) 13:58, 28 December 2023 (UTC)
- "Open"AI's TOS seems to grant users ownership: "Ownership of Content. As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output." Not a lawyer, but it doesn't seem like a violation of copyright to post ChatGPT output. Not that I'd want to in the first place, in most cases. The model, its training data, training methods, parameters and other operational details are apparently trade secrets and I wouldn't be comfortable relying on it. In my experience, its output is of poor quality as far as discourse goes. It would make for a rather mediocre resource at best. AP295 (discuss • contribs) 13:18, 28 December 2023 (UTC)
- My research online suggests that LLM-output should be free from U.S. copyright protection since it is not a result of an activity of human mind, but I also found articles suggesting the copyright issue is a gray zone. I would recommend to you to do your own research online and figure out whether you can confidently claim that LLM-output is free from copyright. I for one feel uncomfortable placing LLM-output into Wikiversity except as fair use. Some time ago, I created Is the output of ChatGPT copyrighted?. --Dan Polansky (discuss • contribs) 09:48, 28 December 2023 (UTC)
I neither endorse nor object to the idea, but I've a new essay, partly inspired by a brief conversation with ChatGPT that did not make a favorable impression. You might find it relevant. A compendium of doublespeak, stock phrases, non-answers and excuses. If you don't care to read the whole thing I'll summarize by saying that educators should probably not simply refer their students to ChatGPT as busywork and that I do not believe it encourages good writing habits. Most of its responses to political questions looked like ten pounds of shit in a five pound bag. Every answer comprised a litany of talking points bookended by non-statements and newspeak. AP295 (discuss • contribs) 14:16, 27 December 2023 (UTC)
- Large language models may become exponentially better (and more useful) over time. Computers may have not been useful to most students in the 1960's or 1970's but 40-55 years later computers have changed everything and everyone uses them. I think educational, teaching, and research use of large language models might be like this. Ostensibly. Bless up. Michael Ten (discuss • contribs) 19:13, 30 December 2023 (UTC)
- Indeed they may. At present and for the foreseeable future I'd far rather talk to a person. ChatGPT is barely even a lookup device at this point, which your own resource reflects in that it comprises a bunch of search queries. I don't think the present version of ChatGPT uses language like what I describe in my essay by accident. Nobody really talks like that except the news an politicians. Read Orwell's essay if you have time. AP295 (discuss • contribs) 20:39, 30 December 2023 (UTC)
A discussion has been started on a large collection of orphaned subpages
editI think this problem was fully resolved--Guy vandegrift (discuss • contribs) 05:17, 25 February 2024 (UTC)
|
---|
There is an understanding that we need to be careful about what goes on the top page in mainspace, and that student efforts in subpages are generally encouraged (especially if they are organized by an instructor.) But we have no established policy on subpages, which makes it difficult for me to process requests for speedy deletion. If you want to get involved with a potential "test case" to help me decide what to delete and what to move, please look at MDLD and leave a comment at Talk:MDLD. I don't know what to do with pages like these. Guy vandegrift (discuss • contribs) 16:19, 30 December 2023 (UTC)
I guess I have to explain how I work and what the reasons are for doing so. 1) First of all, it would be nice if no move are taken which result in pages like Mathematics for Applied Sciences (Osnabrück 2023-2024)/Part I/Lecture 7, where a large section is now missing. This is all part of the course Mathematics for Applied Sciences (Osnabrück 2023-2024)/Part I which is taught at the university of Osnabrueck by me as a professor of mathematics. There is a German and an English version, as there is a certain amount of international students who use the English version. If you really think that Wikiversity is no space to write and present the material of a mathematical course taught at a university, please tell me, though I would be surprised. 2) I do not produce here any nonsense, to the best of my knowledge. In the German wikiversity, I have done about 26 (just counted) mathematical courses over 15 years, all given at my university. See v:de:Hauptseite for a list. 3) The material is not original research, my original research I submit to journals (see arxive or Zentralblatt). The material is quite standard for a university beginners class, trivial or not. (it is the first time I hear people saying that a mathematic course is trivial) 4) The approach for the number as the double of the first zero of cosine goes back to Landau and is well established since 100 years. In this approach, the sine and cosine are defined by their power series and the relation to the circle is developed later. 5) I agree that I did not make any systematic effort here (on the English wikiversity) to explain the way I am working. Let me try to do this in the following. 6) Modular approach (small bits). If you look in any mathematical book you will see that it consists of quite small and well differentiated items: Definitions, facts, proofs, examples, remarks, exercises, usually with a consisting numbering to find them quickly. Along these mathematical textforms I organize the material, in order then to be used in a course (by transclusion, see again the course).
a) The material can be used in several courses just by transclusion. A theorem like Polynomial ring/Field/Zero/Linear factor/Fact is used in an analysis course, in linear algebra, for applied mathematics, in Galois theory. Exercises can be used on an exercise sheet and on an exam. A fact can be transcluded with one proof or another. Learning lists can be made like List of definitions. b) One can always link to the relevant definitions, when the term is used. c) One can link in a proof to the relevant facts (or exercises) which are used. d) the material can be categorized in a systematic and detailed way (this I have not done here, as I want to transclude it from Wikidata).
7) The naming of the pages: I basically list the relevant items used, like Real numbers/Sequence/Limit and convergence/Definition. It is a definition which describes for a sequence within the real numbers what it means to be convergent. That I use here "/" has the mentioned effect that it is formally a subpage of Real numbers. For convergence in another ordered field or complex numbers or a metric space I (would) use a similar naming. In practical terms, I do not see any problem. An alternative would be to use - instead of /. 8) MDLD (mathematical definition link deviation): these are pages which help to make links to definitions. Basically everywhere I want that a term used somewhere is linked to its definition. In order to avoid always writing the link to the definitions, these deviations do the job. See e.g. what links to Real numbers/Sequence/Limit and convergence/Definition. How it is used, you can basically look everywhere in the course, like Mathematics for Applied Sciences (Osnabrück 2023-2024)/Part I/Lecture 15. Bocardodarapti (discuss • contribs) 16:12, 30 December 2023 (UTC)
I think the whole system would be easier to track and understand if it would have something like a proper namespace. Thus, all pages could start e.g. with "Modular math/", or the like. Then, the word "modular" would immediately strike anyone seeing one of the small or not so small modules. Thus, e.g. Real numbers/Sequence/Limit and convergence/Definition would be moved Modular math/Real numbers/Sequence/Limit and convergence/Definition. Of course, this move should be done only if there is express consensus to do so. --Dan Polansky (discuss • contribs) 07:13, 31 December 2023 (UTC)
|