We don’t really need to test the idea that people want to ask factual questions in natural language. The question is: what’s the best way to serve people who want to ask a normal human question and get an answer? 

The definitive truth is that people want to search this way in the same place they search for everything else. This is why Google and ChatGPT are so massive while dedicated Q&A platforms like Quora and Yahoo Answers haven’t done so well. (Trust me, I actually wasted a couple years working at Quora and was a Top Writer there along with Jimmy and some other Wikipedians. As a Quora shareholder I wish it wasn’t true, but Q&A sites will never be as big as Wikipedia.) 

Starting an entirely new portal for natural language questions and answers could be a good way to experiment with Wikipedia-backed natural language search but our ultimate goal should be to integrate it directly in Wikipedia as soon as it moves past beta testing. Otherwise it’s just going to end up like projects basically no normal readers directly search. Even Commons and Wikidata are arguably total failures in this regard given how little direct search and browse activity happens there. Starting a separate domain and wiki is pretty much a terrible idea. 

On Mon, May 15, 2023 at 6:24 PM Risker <risker.wp@gmail.com> wrote:
I note that there are discussions going on in the Technology stream that very definitely touch on this topic. 

The first I noticed is a discussion on Wikitech-L entitled "Word embeddings / vector search".  The second one is a discussion point on this week's Tech News:   There is a recently formed team at the Wikimedia Foundation which will be focusing on experimenting with new tools. Currently they are building a prototype ChatGPT plugin that allows information generated by ChatGPT to be properly attributed to the Wikimedia projects.

These may be good starting points to discover what is already happening within the technical space, and what the thinking is on the likelihood of it filling the need of the proposed project. Regardless, since the project being proposed will require a lot of technical/developer/engineer work, it would be very useful to talk to the people who already have been working and researching in this topic area to determine if the proposed project is viable.

Risker/Anne

On Mon, 15 May 2023 at 20:41, Adam Sobieski <adamsobieski@hotmail.com> wrote:
I would share that I don't fully understand the current WMF procedure for project proposals. I noticed an April 14 email in this mailing list about forming a new taskforce on these topics (https://lists.wikimedia.org/hyperkitty/list/wikimedia-l@lists.wikimedia.org/thread/ZRV7MGHF4IH6LCN3DF6FF6IFMHXJTXZO/). So, when it comes to expectations for a project proposal, the current WMF process, procedure, and related definitions of success for a proposal, I have more questions than answers. 

James, thank you. I see your points and, as envisioned, teambuilding for the Wikianswers project would welcome participants from both within and outside of the WMF movement. I anticipate a considerable excitement with respect to combinations of AI and Wiki platforms, infrastructure, and search. Hopefully the Wikianswers proposal indicates some of the possibilities and opportunities in these regards to interested researchers and developers (https://meta.wikimedia.org/wiki/Wikianswers). 


Best regards,
Adam

P.S.: Thank you for the discussion thus far. I'm still considering the epistemology of which AI-generated multimodal answers would be cacheable, editable, and thus correctable by a community of editors. 

From: James Heilman <jmh649@gmail.com>
Sent: Monday, May 15, 2023 6:28 PM
To: Wikimedia Mailing List <wikimedia-l@lists.wikimedia.org>
Subject: [Wikimedia-l] Re: Wikianswers Proposal
 
Setting up a project outside the WMF would be much easier to start with. One can then trial the idea and if successful the movement may then be willing to have the WMF take it on.

WikiVoyage started outside the WMF by a small group in Germany (after they split from WikiTravel). They had an active community and simply migrated to WMF servers.

Similarly we at Wiki Project Med have started NC Commons https://nccommons.org/wiki/Main_Page Will the movement be interested in this project at some point? I guess we will see.

James

On Tue, May 16, 2023 at 3:16 AM Risker <risker.wp@gmail.com> wrote:
My plate is full at the moment, and project creation is not a specific interest of mine.  I hope that Adam does not see things as demotivating; creating a new project type *should* be a big challenge. I do think that those standards need to be significantly revised.  They were all written at a time when the WMF had no problems at all just raising the target for fundraising, and being successful with the new goal.  This year we are dealing with the reality that the fundraising pool is not unlimited.  We have been told flat out that there are significant limitations to the available human resources required to create a new project.  This isn't 2012 anymore, it's 11 years later, and the world in which the Wikimedia community operates has changed significantly. We can't just be doing the same things that we did back in the olden days.  It would be setting ourselves up - as a community and as a potential new project - for failure.

Risker/Anne

On Mon, 15 May 2023 at 13:53, Ilario valdelli <valdelli@gmail.com> wrote:

I think that here the proposal is to have a new sister project (https://meta.wikimedia.org/wiki/Wikimedia_projects).

There is a long list: https://meta.wikimedia.org/wiki/Proposals_for_new_projects

The concept behind a new sister project is the capacity to build a community and an enthusastic group of people.

The policy is here: https://meta.wikimedia.org/wiki/New_project_process

There is nothing about capacity and risk management.

I wanted to encourage Adam after some very demotivating comments and to look to the positive aspects that, in my opinion, are present in this idea.

To start his project, he needs 10 interested participants / supporters. Are you interested?

Kind regards

On 15/05/2023 19:16, Risker wrote:
So....how much does it cost to develop and run an entirely new and different type of project?  Who develops it?  How much would the hosting cost on an ongoing basis?  Is this project more important and more needed than an existing project type?

These aren't small questions; they are in fact the questions that need to be asked every time we come up with an idea (no matter how great the idea) for a new project type.  Contrary to popular belief, there isn't an unlimited budget, and there aren't unlimited staffing resources for these things.  Everything costs real money and the time of real people, and we as a broad community need to be far more cognizant of the limitations of these resources, and the likelihood that the financial situation is not going to change significantly in the coming five years.

Personally, I think this is an idea with possibilities, but it is a very expensive one because it will need an entirely different way of operating.  So...where would be cut costs to make this possible?  Should we close some little-used projects so they no longer draw on our limited pool of resources?  Should we cut back on volunteer/community safety and resources?

There are always trade-offs.  In the WMF annual plan, they talk about some of those trade-offs.  This is another one.  New, different projects need to be able to justify their cost and existence. In fact, in the not-too-distant future, I can foresee that some minimal-use projects may also have to justify their existence.  I am all in favour of being visionary - but I'm also in favour of planting these visions in solid ground. 

Risker/Anne

On Mon, 15 May 2023 at 12:19, Ilario Valdelli <valdelli@gmail.com> wrote:
I like the idea.

Sometimes people need a simple answer.

At the moment to receive an answer from Wikipedia for some articles people need a Phd.

A solution like that can give a smart and quick and understable answer.

Kind regards

On Mon, 15 May 2023, 11:48 Adam Sobieski, <adamsobieski@hotmail.com> wrote:
Wikimedia,

Per the recent interest in and discussions about artificial intelligence in this mailing list, I am pleased to indicate the Wikianswers project proposal. The proposal describes some approaches for integrating these technologies (e.g., multimodal dialogue systems, chatbots, and question-answering systems) with existing Wiki platforms.

"Wikianswers would be a large-scale, user-editable cache of multimodal answers from artificial intelligence systems, e.g., one or more large language models, which tightly integrates with Wikipedia, Wikidata, and Commons."

This project proposal is described in more detail here: https://meta.wikimedia.org/wiki/Wikianswers .

Thank you. Please feel free to review the project proposal and to comment either here or there with any opinions, questions, feedback, or suggestions.


Best regards,
Adam Sobieski

_______________________________________________
Wikimedia-l mailing list -- wikimedia-l@lists.wikimedia.org, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
Public archives at https://lists.wikimedia.org/hyperkitty/list/wikimedia-l@lists.wikimedia.org/message/JXYO3QW6QYCBHLS4JOHKPPAIJRPMYO4C/
To unsubscribe send an email to wikimedia-l-leave@lists.wikimedia.org
_______________________________________________
Wikimedia-l mailing list -- wikimedia-l@lists.wikimedia.org, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
Public archives at https://lists.wikimedia.org/hyperkitty/list/wikimedia-l@lists.wikimedia.org/message/5BFKXPFA6NWGBEY26YSBRP3PUNFWW73Y/
To unsubscribe send an email to wikimedia-l-leave@lists.wikimedia.org

_______________________________________________
Wikimedia-l mailing list -- wikimedia-l@lists.wikimedia.org, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
Public archives at https://lists.wikimedia.org/hyperkitty/list/wikimedia-l@lists.wikimedia.org/message/35ALTTWNCCV56EMW7KYCP7NRKVDXBRIY/
To unsubscribe send an email to wikimedia-l-leave@lists.wikimedia.org
-- 
Ilario Valdelli
Wikimedia CH
Verein zur Förderung Freien Wissens
Association pour l’avancement des connaissances libre
Associazione per il sostegno alla conoscenza libera
Switzerland - 8008 Zürich
Wikipedia: Ilario
Skype: valdelli
Tel: +41764821371
http://www.wikimedia.ch
_______________________________________________
Wikimedia-l mailing list -- wikimedia-l@lists.wikimedia.org, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
Public archives at https://lists.wikimedia.org/hyperkitty/list/wikimedia-l@lists.wikimedia.org/message/TMO5RH5OGGY6EO4Q6KJPC7F2WZE2I5GT/
To unsubscribe send an email to wikimedia-l-leave@lists.wikimedia.org


--
James Heilman
MD, CCFP-EM, Wikipedian
_______________________________________________
Wikimedia-l mailing list -- wikimedia-l@lists.wikimedia.org, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
Public archives at https://lists.wikimedia.org/hyperkitty/list/wikimedia-l@lists.wikimedia.org/message/7YBDH6IB74QLRCHJE4ZJ5YRWXXA4C7I7/
To unsubscribe send an email to wikimedia-l-leave@lists.wikimedia.org
_______________________________________________
Wikimedia-l mailing list -- wikimedia-l@lists.wikimedia.org, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
Public archives at https://lists.wikimedia.org/hyperkitty/list/wikimedia-l@lists.wikimedia.org/message/UZ5PTZ6RPFSAOT2ZTSZXTDT4T775F7LQ/
To unsubscribe send an email to wikimedia-l-leave@lists.wikimedia.org