The Los Angeles Semantic Web Meetup Message Board › link to article about semantic web vs. A.I.

link to article about semantic web vs. A.I.

A former member
Post #: 2
Per Alex's request at our meet up group on 7-14-09, here's the Clay Shirky article I thought was great. It shows great insight, especially how the semantic web is basically merely an A.I. application. My conclusion is that since no breakthrough in A.I. has been made in 50 years, the semantic web cannot become the reality that most people envision (computer understanding of text) until a breakthrough is made in A.I., which might take another 50 years. The article's insights into the worthlessness of syllogisms and overly-optimistic hiding of the hard parts of problems are variations of the same complaints I've had for years.

http://www.shirky.com...­
Hung T.
temojin
Santa Monica, CA
Post #: 4
Thanks for posting - interesting read. Worthy of a reread in depth. At a glance this document might be taking an extreme reductionist view of the semantic web is simply meta-data used to solve syllogisms. Not being an expert in this field I've always thought if it as complementary to the field of AI. Would love to hear more thoughts on this article and the topic in general.

In the meantime hurray for Semtantic Web and hurray for AI!
A former member
Post #: 3
I agree, Hung, that the author seemed to assume from the start that "The Semantic Web is a machine for creating syllogisms", and has no other purpose. What I would like to find, and one of the reasons I joined this group, is a good summary of how people envision the semantic web. Some people think the browser should bear the burden of making sense of our text documents, such as via Latent Semantic Analysis, others think the document writers should bear the burden of intelligence by making documents in RDF. Within this last category some people think only official experts should be allowed to write ontologies, others think that anyone should be able to contribe an ontology via an unofficial "folksonomy." Some people believe the semantic web should implement logic (including syllogisms), others don't, and that's another big split in opinions. It would be nice to see a summary of how all these opinions group/correlate, and which are the most popular. My personal interest is in A.I., and to get a feel for the semantic web I've been looking at how logic might be implemented on the semantic web, which is largely why the article appealed to me.
A former member
Post #: 2
I am a beginner in terms of learning about semantic web, but here is my take on that article, which isn't making much sense to me. It looks like an attempt at a "mashup" of AI critique wth semantic web critique and it fails miserably. It seems to be saying, because semantic web does not solve all the world's problems by some magic its useless. Well the news I have for you is semantic web isn't supposed to solve "all" the world's problems, just a subset of the problems in computing/information science! Its just like any other technology in the computing field from Operating systems to databases, to TCP/IP, HTTP etc. All of these had doubters, but these are the most ubiquitous technologies in use today.
Semantic web so far as I have understood it is about building an infrastructure for the web. Its building upon already existing infrastructure technologies such as TCP/IP, HTTP, HTML etc. Its main idea is Linked Data instead of Linked Documents, or as Tim B-Lee calls it "Web of Data".
There are already very useful applications out there using the Semantic web (e.g. Zemanta, yahoo search monkey etc.) and there are many more coming. There are even frameworks being built to create semantic oriented applications.

The other point is that the idea of semantics is not new. I recently heard at a conference someone from NYTimes saying "we have been doing semantics since 1913". What he meant was they have been organizing data similar to semantic concepts like RDF/Triples, OWL, but it was on paper and they had a whole building and a slew of staff dedicated to it. To quote Chis Welty/IBM, "In the Semantic Web, it is not the Semantic which is new, it is the Web which is new".

Next equating AI with semantic web. I don't know where that comes from. Maybe this presentation will help clear some misconceptions.

http://www.w3.org/200...­

Anyways, there are always doubters and some of it can also be attributed to hype which are generated not by proponents of technology itself, but by others with little understanding of the technology or the problems its trying to solve. So people developing robotics might not have been trying to replace humans, but that was what got hyped. So guess the disappointment of people when robots haven't taken over the world yet and so they criticize robotics???

Anyways, just my 2 cents.

See another useful article:
http://cacm.acm.org/m...­
Graham M.
user 7585556
Newbury Park, CA
Post #: 1
The semantic web is not just about machine understanding of text. Sure people use semantic web technology in text mining and NLP, but that is only part of the picture. Semantic technologies also have relevance in information modeling, model driven architectures, data integration, etc, none of which have anything to do with text.

So I don't think the semantic web is "merely an A.I application" ...
A former member
Post #: 4
Graham, I'm willing to accept that the semantic web might have more applications than just understanding text, and data mining, but I'm afraid I don't understand your examples. Maybe links or references would help. I suppose you have seen the DIKW (data-information-knowledge-wisdom) pyramid, with the peak at the wisdom level, so when you say "information modeling" I envision either the next higher level of the pyramid, which brings us back to machine understanding and knowledge, or merely the structuring of information, which to me just means data structures, which are old hat to computer science people. The same with the other terms you mentioned. I'd be interested in knowing more about what exactly you think other good semantic web applications would be.

P.S.--I'm still formulating a response to the previous post with the A.I. vs. semantic web slides. Also, if somebody will invite me to this group's Yahoo group, I'll post there, too.
A former member
Post #: 5
Eqbal,

It might take me 2-3 posts to cover all your main points and the points of those links you provided, but here's one such post.

I appreciated reading those slides I haven't seen before, but personally I really didn't care for Berners-Lee's opinions of AI/NLP vs semantic web. He starts off with a lot of playing around with language that seems useless to me, by saying AI is a field whereas the semweb is a project. So if you study it it's a field, but if you work on it it's a project? The difference to me is so trivial, meaningless, and interchangable that I don't know why he would go to the trouble to emphasize something like that. Then he never defines AI, or intelligence. A lot of his style and details (a grammatical error, sterile academic examples, seemingly irrelevant technical details of the semweb) bothered me, too, though those could be considered only aesthetic issues. The justification for inconsistent ontologies seems really illogical and strange to me. Maybe it's just because I missed his lecture and because I'm looking only at skeletal slides that I didn't get a feel for his main points, but I felt I should've gotten a much better understanding of how the various fields of AI, NLP, and semweb relate.

I think probably Berners-Lee and I have very different visions of what the semantic web should be. Maybe we all do. I really want to start some polls here to see what people's vision of the semantic web is, but ideally I'd first like to find some good references (book or online) that clearly state there's a difference of opinion on some fundamental topic within the semantic web community (e.g., logic, knowledge representation, universality of definitions).

The ACM article had some of the same quality as the Berners-Lee article, too, I thought. Most of it was a fascination of macro scale / collective emergent phenomena, which reminds me of people who wave their hands and claim that collections of agents are the future to true AI, which I don't believe. I guess my personal interest is mostly in knowledge representation, which I believe is the key underlying technical issue in both AI and the semantic web, and neither article addressed that point that I felt was of prime importance. But at least I got to see other points of view and read material I hadn't seen before. Thanks for posting those links.
A former member
Post #: 3
Dobalena,

What I understood from "AI is a field vs SW is a project" is that AI is a field that deals with developing theory whereas SW deals with trying to solve some practical problems and sometimes it applies the theories developed in various fields and sometimes its just organic in that it just creates/fits parts that just seem to fit. So for example representing data in form of RDF didn't really come from any base theory, its just something people decided to use as a standard, but after that I am sure there could be things learnt in graph theory that could be applied to a graphs represented in from of RDF. Implementations of SPARQL probably do that and we could apply theory learnt in AI "field" to reasoning and learning things from the graph to gather "intelligence". So if Clay Shirky and others are saying SW is just an application of AI, in my personal opinion they are not 180 degrees from reality, but perhaps their understanding is at 120 degrees from it! Am I making sense?

Can you give specific examples from the links I provided where you had issues with the how SW was presented?
A former member
Post #: 6
Eqbal,
I'm not so sure RDF/triples aren't based on theory. One link in a semantic net, which is probably the simplest formal way to represent semantics, obviously has 3 components, so it's pretty clear the semantic web is founded on successive abstractions of that (simple) theoretical observation. And since true AI doesn't yet exist, using "theory learnt in AI" is mostly pointless in my opinion, which goes back to why I liked Clay Shirky's point of the semantic web as basically reigniting AI under a new name. I see what you & Berners-Lee mean now, though, concerning ad hoc engineering vs an extension of theory. Ad hoc engineering is fine with me if it works, but there's no way ad hoc engineering is going to make a machine that understands natural language, so designing the semantic web that way seems like a horrendous waste of time to me that will eventually require rewriting the entire thing correctly later on.

Some of my specific complaints about the Berners-Lee slides...
slide 4: Asks what AI is, then "answers" the question with 100% questions, omits known useful definitions.
slide 5: Grammatical error ("A architecture...")
slide 7: What symbols is he talking about? And is he answering the question, or asking more questions?
slide 9: What is a "foaf"? What meaning of "card" is he using? What is he even talking about?!
slide 10: Why use the old meaningless foo & bar? Why not give a real-world example?
slide 12: What information is being left? Queries? Data? Inferences? What is a "delegated query"?
slide 13: What does OWL have to do with description logic?
slide 17: Many capitalization and punctuation errors. When do local ontologies become global ontologies?
slide 18: More punctuation errors. Why not just include a diagram of each type?
slide 19: What does he mean by "logic"? Which type of logic?
slide 21: What does this slide have to do with anything? Seems unrelated to all slides preceding it.
slides 31+: Are these an appendix or what? Isn't the presentation over now?

It's painful for me to go through those slides. They're all over the place, the intended ideas aren't coming through to me, and the minor text errors are distracting. Sorry to be so critical, but every time I see those slides I get the same impression.
A former member
Post #: 4
Hi Dobalena,

I have very little or no background in AI and I still can understand a lot of the pieces of Semantic Web. I think you're probably right as to people are "seeing" many different things into semantic web, like blind men trying to feel an elephant. The community probably needs to communicate this idea a lot better. As far as making machines understand natural language, I never thought that was the purpose of semantic web (or its sole/main purpose anyway).

My understanding of SW so far is that it is evolving some sort of structure out of the web so machines can make sense of the data available and in turn can allow humans/applications to gather/deduce much more useful information from the web than what is possible today. So instead of linking of documents ("web 1.0") or linking applications ("web 2.0") the attempt is to link data. Obviously data in relational databases isn't so conducive to linking and suffers from many other deficiency in terms of dealing with semi-structured or un-structured data. That is why there is a need to create an infrastrucutre around which this linking of data can be made possible (was that a grammatical mistake? ;-)). To that end (building infrastructure) is the the development of RDF, SPARQL, OWL and such.

To answer some of your issues on the presentation:
slide 4: You maybe right, he could've used more formal/academic definitions, but I am not sure who his audience was
slide 5: Good catch!
slide 7: I am not clear myself :-)
slide 9: For more information on foaf see http://www.foaf-proje...­
slide 10: Valid point
slide 12: Delegated query is a query made on behalf of some other agent
Slide 13: There is a flavor of OWL that supports description logic called OWL DL
Slide 17: I don't see anything related to local ontologies becoming global on this slide. One of the points is about interconnected ontologies.
Slide 18: You're right and to think he is British!!!
Slide 19: I don't know, he could be more specific
Slide 21: Yeah I am scratching my head too
Slides 31+: Yeah, looks like he doesn't want us to see beyond slide 29!

Anyways, here are some more presentations I found which might explain what people take SW to be:
http://www.w3.org/Peo...­

I am also surprised as to why no one else is stepping up in this discussion. I am a newbie/amateur in this area, so I was hoping other people can do a lot better job of explaining things!
Powered by mvnForum

Our Sponsors

People in this
Meetup are also in:

Sign up

Meetup members, Log in

By clicking "Sign up" or "Sign up using Facebook", you confirm that you accept our Terms of Service & Privacy Policy