The Semantic Web is the infrastructure for the coming age of "Ambient Information".
Imagine that you are in a group touring Spain and have all decided to visit the Gaudi-designed cathedral in Barcelona. One of your troupe is interested in Spanish history, one in architecture, another in religion. With enough homework, Web-use training, and patience each would be able to have prepared for the visit so that much of the experience would be pertinent to particular interests. As ambient information facilities become widespread it will be simple to know where one is and what one is looking at - the accompanying subtext is presently impossible to get in a form appropriate to individual interests.
There will be machines to know your position in space/time and what the Semantic Web will do is enable your machine to furnish you with the viewpoint you've chosen to present individualized presentations of what it all means. The architect could know the technical details of the structure, even view its plans, etc. as the historian was hearing about how Gaudi was regarded when the project began and for the traveler interested in religion there could be a discussion of the symbolisms inherent in the sculptural features of the building.
Instead of one tour guide droning on with some "least common denominator" canned speech that only skimmed the event, there would be the sort of in-depth resource derived from the near-limitless treasure trove of the Web. As you left the area of the cathedral you would be told how to get to the restaurant with the special saffron rice dishes you'd been anticipating and even be told that you had a reservation and how to get there - it could speak the address to the cab driver in Catalan!
What underlies the informational revolution that this promises- what we're calling the "Semantic Web", is a system for handling huge databases of information by machines which slavishly follow your preferences for retrieval and presentation. The W3C technologies used for this include CC/PP, RDF, XML, and a host of techniques like XLink, XForms, VoiceXML, and others which permit you to control machines' performance for your own information purposes.
The fundamental notion is that everything on the Web will be "indexed" for automated retrieval because of the Resource Description Framework (the aforementioned "RDF"). The changes that must be made to insure machine-readability are closely akin to those proposed in WAI's various guidelines because much of what must be dealt with is so like what must be handled by assistive technology. In particular authoring tools that conform to ATAG (or even by hand-editing in conformance with WCAG) will attend to "writing what you mean, rather than what you want done with it". This is a principle tenet of the Device Independence Activity as well as the Semantic Web Activity.
In summary, the Semantic Web, based on RDF, provides the ability for machines to access the information stored on the Web without human intervention, but at human whim/command. It must be combined with a "Web of Trust" involving "digital signatures" to be fully effective and requires careful consideration in authoring/storing of content.