|This article is of interest to the following WikiProjects:|
Just an idle thought, but if there is any way Cyc could understand Wikipedia content, this would solve much of the database creating content. Wikipedia's goal is, after all, to contain all of the world's knowledge. --Piotr Konieczny aka Prokonsul Piotrus Talk 20:22, 24 Apr 2005 (UTC)
- Yeah, that was my fantasy too, that Cyc would one day attain basic understanding and the next day devour Wikipedia and understand everything! Oh well. There is a Simple English Wikipedia whose articles are supposed to be easier to understand, it would be great if Machine translation and Cyc could in fact parse these. The meta:Semantic MediaWiki project allows you to express relations between articles and give attribute values to articles, which you can export as RDF and then computers can infer things from them. This is really cool, but it's a long way from a computer having any sort of common-sense knowledge. -- Skierpage 08:50, 11 June 2006 (UTC)
Here is a story about computers learning from wikipedia. I am not sure if it applies to cyc. http://www.physorg.com/news87276588.html
What is referencing all those references? It just seems like there is a lot of them. Cmouse 02:42, 25 July 2005 (UTC)
It appears to be a copy-and-paste from Cycorp's publications page, which I believe is a fairly complete list of everything the company has published. Given the automatic nature of the list's selection, and the fact that the content is purely bibliographic entries for publications, I do not think there is a copyright issue. On the one hand, this looks like a good set of further reading about the Cyc project. On the other hand, the information could be used almost as easily simply by linking to Cycorp's publications page. Disclaimer: I appear as a co-author in this list. Bovlb 05:40:13, 2005-07-25 (UTC)
- I think the quantity of references is overkill. They should be weeded down to a handful of the most pertinent/general/accessible ones, with a note to the effect that many more are available at the Cycorp website. (I'll probably do this myself if no one protests.) Dbtfz (talk - contribs) 07:14, 10 February 2006 (UTC)
The criticisms section
The criticisms section reads "Scalability problems from widespread reification, especially as constants". I'm not a native English speaker, but somehow this doesn't make a sentence. And even if it does I find it hard to understand. Probably it could be rephrased?
- I am no expert on this subject, but it appears to mean that it can be difficult to add new "facts" which do not contradict or introduce logic errors in the existing content, without rescanning the whole database to look for contradictions anytime anything new is attempted to be added to the database.
- Based on the reification article, a possible solution is that nothing in the knowledge base should be considered a fact, and everything is an opinion. All data entries in the knowledgebase need a tag from who entered the data as well as a reliability rating for the factualness of the opinions entered by this person, as well as perhaps a chain of backup ratings of this person from others who have their own reliability ratings. This would allow it to juggle multiple and potentially contradicting opinions about the same data sets entered by different people, and to perhaps return a number of results with different factualness/reliability levels with some cutoff point for the speculative results of low reliability.
- This would allow entry of data from young children for example, but as a group the reliability of the children might be demoted by a parent of high reliability, and so cyc wouldn't put as much weight to their entries.
- DMahalko 05:28, 28 October 2007 (UTC)
--Ivant 15:36, 26 June 2006 (UTC)
I think the criticisms section in general is a bit weak. Cyc is indeed a controversial project as noted, but the reasons then listed aren't really why. Sure, its documentation is kind of weak, but that's not why it's controversial among AI researchers---that's an implementation/usability issue, not a fundamental issue. The fundamental controversies are over whether compiling a large ontology represented in formal logic is the right direction to be going in in the first place. --Delirium 14:49, 28 June 2006 (UTC)
The same criticism has been aimed at the semantic web, that it is doomed to uselessness because its way of understanding the world is tragically limited. As a society we've learned the limitations of many technologies and learned to work within those limits to do interesting and useful stuff. I think we'll do that with both the semantic web and Cyc. Neither will become the AI that steals your job, both will become useful. -- WillWare 01:02, 1 September 2006 (UTC)
- Found a small problem as well:
The lack of any meaningful benchmark or comparison for the efficiency of Cyc's inference engine
- ... so that means, they object to it because they have nothing to compare it to? Somehow, this seems rather illogical. --Melissia 17:08, 2 October 2006 (UTC)
The biggest criticism is that there are exactly zero systems that have used Cyc for anything remotely useful.
Are you sure? Data mining is big these days. Perhaps Cyc is being used to put people on the terrorist watch list as we speak. Remember to buy pork with your credit card and order it with your airline meal. 18.104.22.168 07:27, 5 January 2007 (UTC)John Creighton
The criticism section has only one reference. Point to references of discussions of these criticisms, or remove them from the section. 22.214.171.124 03:03, 1 November 2007 (UTC)
What Cyc is up to now
Here is a talk that Lenat gave at Google.
These days, Cyc's killer app seems to be homeland security. Perhaps that's worthy of mention in the article, since it is likely to strongly influence the direction of Cyc's future development, and possibly the question of whether it will become more substantially open-sourced than it currently is. -- WillWare 01:02, 1 September 2006 (UTC)
(#$isa #$BillClinton #$UnitedStatesPresident)
- Does it really matter what language you choose to communicate information to the AI system. If the system can parse Lojban then great. I think they built a natural language interface for Cyc but if you enter the information into the system via CycL, then the system should be more likely to understand what you mean.
- The system doesn’t actually need any language interface. If the authors wanted they could have created forms for the user to fill out like people normally do in data entry. The advantage of a lisp like language is it is easy to parse. With systems devoted to translation, it should be easy to communicate between a wide variety of systems regardless of the language interface each system uses..
- ~John Creighton
I noticed that there is no mention of the FACTory "game" (there is a link on the Cycorp homepage, I believe). This serves as a GUI between it and internet users, and aims to make adding to CYC fun. This is rather significant, so I am surprised that it is not even mentioned. Unfortunately, I don't have time to add anything now. Aero-Plex (talk) 13:12, 7 October 2011 (UTC)
Stewart Russell is listed as a notable employee but the link associated with the name is to a wiki page for someone that appears to have no computer science training. I suspect the person who included Stewart Russell in the list must have meant the Stewart Russell who is a famous AI researcher. https://en.wikipedia.org/wiki/Stuart_J._Russell However, I could not find any mention of Cyc in his bio on that page. So, I suspect the link for Stewart Russell should either be changed to Stuart_J._Russell, or his name should not be included at all.
Jeffteeters (talk) 17:18, 9 January 2015 (UTC)