The buzz around Web 2.0 may have only started in the last year or so but already industry commentators are putting their opinions in the hat for what will constitute Web 3.0? Such talk strikes me as more than a little premature and what is being discussed appears to be a regurgitation of the technologies proposed during the dot-com boom of the mid-nineties rather than original ideas on how to take what we have learned from the previous two incarnations of the Web.
Discussing Web 3.0 is premature because no one has come to grips with what exactly what the concept of Web 2.0 is right now. There are loose ideas of community, interaction and the writeable Web but no simple, easy to understand description has yet crystallised. Until this occurs its hard to tell where one set of conceptual ideas finishes and another begins. The bursting of the dot-com bubble signaled the end of one distinct period of Web development much like the K-T boundary marked the end of the dinosaurs (mostly). This intense moment of destruction followed by relative calm gave those on the Web time to pause, disseminate what came before and evaluate the best way forward.
To make matters worse discussion about what Web 3.0 could be appears to be centered around the relatively old concepts of the Semantic Web. Whilst a nice idea such arguments ignore the fact that Semantic Web ideas existed well before Web 2.0 concepts and in terms of realising these grand ideas not a great deal has changed. From a technical perspective the enabling technologies are still overly complicated and at a practical level no clear upgrade path exists from our current dumb Web to this idealised space (apart from millions of hours of painful, manual classification). Of greatest significance the Semantic Web relies on our ability to generate classification systems for many different forms of data. Given that a single office document standard cannot be agreed to and development of in-depth, domain specific semantic languages such as Industry Foundation Classes are stalled such a proposition seems far off.
Rather than a complicated semantic model I believe a more suitable candidate for the Web 3.0 crown would be the concept of the Learning Web. A Learning Web would successfully marry human input (gestures) to digital processes and would be an interesting proposition because unlike the Semantic Web it builds on social groups, can be understood by the casual observer and most importantly offers a clear monetisation pathway in the form of the Gesture Economy (something Steve Gillmor has been talking about for a long time). Emerging tools like Riya are beginning to embody such concepts. Riya is a visual search tool that uses human input to train search algorithms (similar in the way speech recognition works). Without human input searching is no better than contemporary technologies, but once real-time user feedback enters the fray things start to get really interesting. The beauty of such an approach is that whilst the results are smart the underlying data-set is still inherently dumb which means simple underlying technologies and an easy uptake for developers and users alike. Such a place would not be nearly as intelligent as that proposed by Semantic Web advocates but from a practical perspective it would work just as well for many tasks. But perhaps of more significance to us, the end-users and inhabitants of this space, it would operate at a human level well below the Matrix-like autocracy of that envisaged by the semantic model.