Raw

As reported by Jamie and Harry, and now the subject of a CNET piece and Slashdot thread, Tim Berners-Lee was challenged by Peter Norvig of Google. The subject of Berners-Lee's keynote was putting data on the Web using Semantic Web technologies.

Norvig responded:

... from Google's point of view, there are a few things you need to overcome, incompetence [of the general user] being the first

(my emphasis)

Presumably Google aren't actually against putting data on the Web per se, in fact their GData project is all about doing just that. Their Google Base project isn't far away either, that's about content metadata.

General users are pretty capable of creating good data when provided with good tools. When I ordered some gadgets online (from Germany) the other week, everything worked perfectly. People deal with electronic data all the time, the introduction of the HTTP protocol hasn't rendered it garbage. There's nothing to suggest the use of Semantic Web languages for representing that data would cause any more corruption than say shifting from pre-Web EDI to WS-*.

Sure, legible RDF/XML is much harder to write than HTML. But right now I'm using a WYSIWYG editor. That looks after the HTML. The CMS also takes care of the RDF associated with this post. Things like the date this was written will appear as explicit statements in the feed (and as RDF/XML, if you care to put a .rdf after the URI of this post). Look ma, no markup.

The other two points quoted from Norvig:

The second problem is competition. Some commercial providers say, 'I'm the leader. Why should I standardize?'

...

The third problem is one of deception. ... With less human oversight with the Semantic Web, we are worried about it being easier to be deceptive

Competition and standardization - yes, certainly issues for the Web. But the companies that thrive in this environment tend to be the ones that embrace open standards. The fact is that the rest of the world is likely to be bigger than any leader. Respect the long tail.

Is deception an issue on the current Web? Yes, people game search engines. Also countries (like China) with the aid of commercial organizations (like Google) filter material to produce a false picture of the information that's out there. Lies, damn lies and search engines.

But anyhow, is Norvig really suggesting that Google are currently solving deception issues via human oversight? Whatever, proof and trust are key considerations for Semantic Web systems. The foundations are built into the theory, logic provides one route to addressing the issues. Even if that formalism is ignored and statistical approaches taken, the use of common languages makes data more amenable to analysis. Probabilities and uncertainties can be expressed as logical statements on top of the base languages. However you approach it, the answer is mo' better data, not less.

The other issues here are those of ownership and distribution of the data. Google are comfortable with the notion of general users putting data on the Web, at least as long as it's Google's Web. They are a few steps beyond the usual walled garden org or pre-Web monolith, but maybe their current business model still depends on their ability to control the data - or rather the interfaces to the data. I use Google Search all the time, get the ads in my face. But Google don't own the data I'm looking for.

I guess what I'm trying to say here is that Google appear to see data being distributed on the Web and under the control of individual users as being a threat to their own aims. They see problems unless Web data is in their silos. But this doesn't have to be the case, they can still be leaders in the interfaces to the data. i.e. don't build big DBs, make good mashups.

In one sense at least there is potentially a big problem for them: the more directly addressable, first-class data there is on the public Web, the less people will be dependent on free-text search engines. No doubt they're aware of that already.

PS. The CNET piece made it into techmeme, also Valleywag has a piece entitled Google suit disagrees with inventor of the Web, loses. Can't disagree with its awesomeness point, but Peter Norvig deserves a lot better than "suit". His AI book is excellent (I think it's the 1995 edition I've got, should probably update) covers many of the key ideas behind the Semantic Web. Er, if you forget the Web bit.

@en