< Web Thinks

Transcript

Friday, January 05, 2007

BOB GARFIELD:
This is On the Media from NPR.

And I'm Bob Garfield. For a couple of years now, we've been living in a Web 2.0 world. Web 2.0 is the nickname given to the second generation of the Web in which the static pages of the early Internet gave way to something more fluid, where users can more easily find and share information, music and video.

We were just settling into Web 2.0 when suddenly, back in November, The New York Times ran a page-one story about the future of a new Semantic Web, a Web that thinks. Web 3.0? Huh?

Nigel Shadbolt is a professor of Artificial Intelligence, or AI, at the School of Electronics and Computer Science at Southampton University in England. He says that Web 3.0 will not only give us the facts we seek, it will choose among them, organize them – essentially do our work for us.

NIGEL SHADBOLT:
So I might ask of Web 3.0, if it was actually there and running at the moment, book me a holiday in Vienna in a nice resort near Innsbruck for my family next week. Oh, and by the way, I've got a 12-year-old.

Now, that kind of expression of a complex question is nothing that we can precisely enter into the Web as it currently exists. People's hopes and aspirations for this more complete Web are that it will be able to deal with these kinds of complex queries.

BOB GARFIELD:
Under Web 3.0, you're talking about the computer actually making some judgments for me. How do it know?

NIGEL SHADBOLT:
Well, when you look at the Web page of a holiday provider and look at all of those features on those page, a particular location, a particular hotel type, how long it'll take to get there, you are extracting all of that information using your wetware and your eyeballs and four billion years of evolution. That's what human beings are exquisitely good at.

What we're developing in this next generation of Web are markup languages which are able to take the content on pages and documents like that and describe the bits of that content that are about the hotel and its location, about the expected amount of sunshine at a particular month in the year. It's called metadata, and the metadata is now allowing us to have information about the information available. And that's how the machines know.

Essentially, one machine will be using a particular vocabulary to describe its markup, and it will go and tell another machine that this is the place to look for to find out how it's using those definitions.

BOB GARFIELD:
It sounds to me what you're describing is Artificial Intelligence.

NIGEL SHADBOLT:
I kind of think of the Web as almost like a natural ecology. And what we're seeing start to pop up in various parts of the Web are simple sorts of micro-intelligences. Now, this isn't full-blown AI, but in just the way that we now have shopping assistants, shop-bots that can go out there and get a good price for us from particular stores, or an auction agent that can sit around and make a bid for us on eBay, these very limited sorts of adaptive behavior are going to start to spring up across the Web.

So I think of the Web starting to illustrate lots of micro-intelligences rather than being some kind of super-sentient information fabric. I think that really does remain in the realm of science fiction.

BOB GARFIELD:
In The New York Times, there was a piece about Web 3.0, talking about a system called – forgive me, I don't know the pronunciation, but it's CYC.

NIGEL SHADBOLT:
"Sike."

BOB GARFIELD:
Okay.

NIGEL SHADBOLT:
The CYC Project was an attempt, some years ago now, to try and use AI techniques to construct encyclopedic knowledge that humans have about the world. Now, one of the very hardest things to get our computers to understand is all the common-sense knowledge we have about how the world works and is structured.

The fact that if I have a can of water above a cup and I pour the can of water under normal gravity, that water will just flow into the cup – all that common-sense knowledge that we have packed in over years of development is hard to reproduce in computers.

Now, what CYC has been about is trying to regiment large amounts of common-sense knowledge and put that to work in programs that reason about various aspects of the world.

Now, CYC has been seen as a potential help to Web 3.0. I don't think we're going to be able to kind of plot all of that encyclopedic knowledge into the Web and just have good things happen. It's much more likely that we'll take facets and aspects of CYC's knowledge base and use them for particular tasks.

BOB GARFIELD:
This is all fascinating. Is there any reason to think that this is going to materialize in the next year, the next five years, the next thirty years? How close are we to a Web 3.0 Artificial Intelligence online world?

NIGEL SHADBOLT:
I don't see that just around the corner. I think what we'll see is increasing amounts of these limited intelligent services. I also see a Web that will be still very much driven by people. People are the real intelligence in the Web.

And what we see happening in very interesting developments like Wikipedia, where you get this collective wisdom, large numbers of individuals collaborating to reveal and put down their experience, their intelligence, what we see there is this key facet of the Web - that people are extremely good at being the general intelligence on the planet.

And what the Web machines deliver up to us are particular ways of holding vast amounts of detailed information and serving and filtering that information in accurate ways.

BOB GARFIELD:
Nigel, you're kind to talk with us. Thank you very much.

NIGEL SHADBOLT:
Thank you.

BOB GARFIELD:
Nigel Shadbolt is Professor of Artificial Intelligence in the School of Electronics and Computer Science at Southampton University.