Wikipedia Virtual Mind

Wikipedia sentences are of 3 kinds:

  • true sentences
  • context dependent sentences
  • some untrue and ambiguous sentences that eventually are going to be removed.
This is a proposal for a new Wikimedia sister project.
Wikipedia Virtual Mind
Status of the proposal
ReasonInactive proposal. --Sannita (talk) 10:43, 19 September 2013 (UTC)[reply]

If true sentences could be extracted from wikipedia, they could be used to create a knowledge database and a virtual mind with the ability to answer to questions.


Sounds groovy. I can't wait. --Larry_Sanger

How do you establish truth?--user:Seb

  • We should establish truth by creating a wiki-Logic-Tree where branches are analyzed based on their validity. If an argument is found to be guilty of a logical Fallacy then it's branch will be labeled as such.

Perhaps you don't have to. You could create an irrational, fallible AI. Surely that would do better on Turing tests?

  • Surely in the process of scanning the data, any inconsistencies would emerge. This could then be used as a source of directions as to places where Wikipedia needs to be checked. Phil 17:12, 6 Apr 2004 (UTC)

Perhaps a feature coud be added to pages presented to browsers that would allow a highlight of a sentence and a vote by non anonymous browsers stating true, false, maybe, a nubmer or weight on scale, etc. The sentence and vote could be forwarded to a virtualmind server for processing or recorded and forwarded in batch or maintained as part of wikipedia. It seems out of focus to me, first we get a successful free encyclopdia.

If this was set up as a separate project we agree to forwad data to then it would have the advantage of easily incorporating data from other sources as well.

Interesting concept but not essential to getting the wikipedia up and running usefully and reliably .... unless we need to interest more highly creatve, capable software people in programming and debugging? user:mirwin

Sounds to me like this proposal is specifically a proposal for an alternativeinterface. The idea is that the people are the minds. But this interface could also use an emergent interconnectivity scheme like uses, or perhaps similiar ideas like that, where the user's activities implicity create pathways. In summary: advanced interface, emergent organization, and advanced search methods. = the virtual mind. user:Kevin baas -2003.03.14

Regarding the 3 kinds of sentences in the original post on this page, I would argue that all sentences, knowledge, facts, words, meanings and truths are contextual. Besides, there is no way for a system without senses to verify truth about the world, which presumably is what is meant by truth here (as opposed to truths only about Internet content). Voting may be a decent heuristic for validation, but I think it would be thoroughly worthwhile to "simply" parse and classify all statements made on a site such as the Wikipedia, and be able to perform queries on it. If contradictory statements are made on the same or similar subjects, perhaps the system can flag it, or the querying party can decide what to believe.

I would be interested to hear if anyone has made any prototypes in this sort of direction.


As we know, There are known knowns. There are things we know we know. We also know There are known unknowns. That is to say We know there are some things We do not know. But there are also unknown unknowns, The ones we don't know We don't know. DonaldRumsfeld has a program that is interesting to talk to. Perhaps we should contact the programmer...

Calmypal 23:34, 21 Jan 2004 (UTC)

(Nota Bene - oliverbot is based upon ALICE and AIML see for more info)

That sounds like something like the Society of Mind-theory? :) // Noone

This is an area I've always been very interested in (I'm studying AI at university, indeed). A couple of quick thoughts that occur are:

  • OpenMind is a collaborative effort similar to MidPxel but somewhat more structured.
  • everything2's "soft links" implement a similar kind of quasi-semantic linkage suggested by Kevin above.
  • This needn't even impact Wikipedia itself - the content is redistributable after all, so the project could simply feed off dumps of the database, and have its own dedicated band of contributors.

Hmm.... - IMSoP 18:53, 27 Feb 2004 (UTC)

From: Jerry P

Why not integrate the cyc knowledge server from, after all, it is open source software. But because it is under the LGPL, I'm not sure if it would be ok to integrate into this project. Cyc is a knowledge base of common sense knowledge and it can generalize from information in its knowledge base. So if you say something it can interpret it as being true or inaccurate. To read more about cyc go to and

Also e-Cyc portal toolkit, but am not not sure if it is open source. e-Cyc was used by lycos and hotbot (in hotbot beta) to help in semantics, that is in deciphering the ambiguity of the meanings of words. An example of e-Cyc in lycos would be, for example, searching for the term vets, it has two meanings, veterans and veterinarians. So, e-Cyc would help by asking you which you meant, and then filter out all the web pages with that meaning.

Why not adapt the googlism mechanism so that it uses the wikipedia information instead of the google information ? 20:20, 19 Apr 2004 (UTC)

This idea seems AI-complete. -- 22:14, 16 Jun 2004 (UTC)

Slotipedia could be usefull for this project.-- 23:39, 26 Jun 2004 (UTC)