So, the New York Times has finally given some legitimacy to the Semantic Web. Woo-hoo!
Or has it? This article is about Web 3.0; it isn't until paragraph 7 that it equates Web 3.0 = Semantic Web. I find it a bit odd that this article came out on the weekend following the fifth International Semantic Web Conference, but no mention of that conference was made in the article. In fact, none of the people mentioned in the article were involved in the organization of that conference, nor in the W3C effort that has made all the standards that support the Semantic Web (except the oblique reference to Sir Tim in the second page, but he isn't quoted, just mentioned). I guess that the W3C and ISWC just aren't relevant to the semantic web.
Or maybe they are. The article states, for instance, that "The classic example of the Web 2.0 era is the “mash-up” ", and goes on to describe a mash-up as taking information from one site and using it on Google maps. While in reality, the basic Semantic Web language, RDF, is the ultimate in mash-up data languages. I find it odd that the word "mash-up" has come to mean things like "taking info from a real-estate site and mapping it all on Googlemaps". Mash-up used to mean, well, mashing two or more sources together. You know, like taking the real estate map and mixing it with a list of gyms and organic supermarkets, to give a real, integrated picture of a neighborhood. With RDF, this isn't any more difficult than a "unary" mash-up (Piggybank demoed this over a year ago using RDF). Oddly, mash-ups like this aren't very common in Web 2.0 sites. The Semantic Web is really about doing mash-ups right; as such, it is a key enabling technology for mature Web 2.0 sites.
The article says that critics of Web 3.0 say that it is unattainable; I wonder what critics like that are left. It is pretty common nowadays to find web applications that do things that were considered AI just a few short years ago (those lovely route-planning algorithms at Google and Mapquest are a good example. Just check out Google Scholar to see all the work in the 90's by highly respected AIers on this problem, including the current director of research at Google). The intelligent web is here today.
The real issue is whether we can bring the web together so that the whole functions even as much as the sum of its parts, not to mention something greater. Route planning doesn't seem very smart, when you have to re-type addresses from other pages. Mash-ups don't seem very smart when you have to turn your head fast and blink your eyes to see everything together. It doesn't take an intelligent web infrastructure, or the massive common senes of Cyc, to do this; just a bit of sensible glue.
The tail end of the article interleaves reports of very good work with web information mining done at IBM; this is a tale of some very standard and well-founded analytic techniques bringing new value because they can access more data than ever before. They did this with the current web infrastructure, and a lot of merging of information. Did they use RDF to merge this data? I don't know - a lot of the students in my classes have invented their own ways to do this sort of thing. But one thing all the veterans of these projects agree upon; they wish they had used a standard, 3rd-party supplied system to do the merging, so that they could have focused their development efforts on their high-value analytics.
I haven't asked the IBM folks about this, but I would be surprised if their experience was tremendously different.