I haven't thought deeply about the various tagging experiments out there [yet], but it strikes me that this idea is approximately on par with taxonomies, which are okay in a small (or closed) enviroment, especially if there's only one user; you. However, as Dave Taylor points out, the bigger the tagging space, the less likely this idea will maintain the benefits experienced on the ground floor.

I predict that these concepts will continue to grow in popularity because people want to organize information. However, they will paint themselves into a corner and the only way out is through the transformation of tagging schemes into ontologies. Don't get me wrong - I think these early tagging systems are useful because they will educate us to the requirements beyond the near horizon and just how low the ceiling is for emergent (taxonomy-based) tagging.

As such, I predict that from a broad and divergent set of terms will emerge groups that create ontological overlays (see Topic Maps, XTM 1.0) of agreed definitions. I suspect these groups will form around a domain of expertise and a desire to eliminate chaos created by ground-up taxonomies. This scenario requires the availability of tagging API's that allow a higher-order architecture to emerge.

Indeed, we are witnessing the errant experiment that will lead to the dawn of ontological awareness. ;-)