My reading of it seems to be fairly close to Danny Ayers', that is, an appreciation of the effort to understand what's going on mixed with curmudgeonly disliking of the gee-whiz frothiness.
There could indeed be something interesting going on with contemporary communications media, but I don't think it is in terms that this author asserts, nor are the particulars he claims that unique.
Semantics: what knowledge is, how it is processed and represented. Beyond syntax (form) into meaning.
If semantic behaviour, (cough) thinking?, occurs outside of human heads then we could well have a global mind on our hands. Except it would be out of our hands. It could well be that recent techniques have, or will shortly, enable this.
The usual suspects are underneath somewhere: Dennett, Dawkins, Minsky.
Some spins of my own: without explicitly codifiying a general scheme, semantic data and processes already exist and have done for millenia. They are present in systems and behaviour - what people do with the words and texts - and are occasionally represented by gestures, punctuation and the whole continuum of what is sometimes called meta-language. They are augmented by all sorts of social considerations. The fact that a given conference was in London rather than Budapest is "meta-data". But we need to be careful with our terms, many examples given as language meta-data (tone, gesture) are actually explicitly linguistic. Tone in Chinese being a simple example. Even pre-written language had asynchronous storage: well, hello there wandering poet!
As I keep saying whenever someone asks me bitter or lager: Categories Don't Exist. There is no clear division between data and meta-data in nature; genes and memes don't particularly care how they are expressed and transmitted. DNA, RNA, mother's milk -- there's no system definition.
Knowledge seems to be slippery anyway, and it's not reliant on syntax, more on modelling another's internal states. What do I think you mean? That's what I guess you meant, I think. Who are you anyway? What do you want?
The step change in capability, if there is one, is not that TADA! a given "layer" of metadata can now be described, rather that the thinking itself can be let go of, delegated, by virtue of distributed processing. Maybe by eyes and brains, maybe by wires and scripts. This paper illustrates how this could operate very nicely. It could well be that XML is what makes a global neuron fire, or is a step in the waggle dance of the hive mind. But I'm not sure that this, or even the internet, is a necessary part of such a thing.
Want an example of pre-existing emergent semantic processing, on a global scale, with no semantic layer definition? The Economy.
However, semantic layers and their objects are genuinely interesting. If one knowledge object can "mark up" another than some additional evidence can be left behind of process. Aha! Writing! The trackback (the two-ended link, the link as relationship) is seen from this viewpoint as the locus of this paradigm shift, rather than any implementation of XML. But that brings into question the notion of open data fields (blogs with TB, comments) vs closed ones (commercial entities, sealed from graffiti).
Those are some technical issues I have with the paper (I'm also with Ayers on other issues, like the non-applicability of the Turing test argument). But I'd also argue with what it all seems to mean.
If there is such a thing as "the" weblog community, and if it indeed has a mind then it is not a wise one. At present its personality seems to be infantile with disturbances bordering on the psychotic. Impulses related to vanity, currency and other faddishness seems to dominate. Hence my grinching today rather than writing a 3000 word paper of my own, tomorrow.
Also, were I his editor -- gosh! In blogdom, I am! -- I would also recommend that the paper be de-fluffed. Mention "my company" once and once only puhleeze.