I really appreciate that the discussion around the issues behind the Semantic Web (or GGG -- Giant Global Graph) are starting to mature significantly. Still, I find it very interesting and relevant that I stumble upon these discussions more frequently than having them dished up for me by my mechanical servants. This perennial frustration of relying on luck points to a spot very close to the heart of the problem. The gestation period of knowledge seems to be retarded by eloquence rather than accelerated by it. My navigation within the WWW or GGG is still much more luck or brute force than inspired intentionality. If one could map the related concerns, irrespective of the relative eloquence of the missive, we might start seeing a greater acceleration in quality contributions which do more than reiterate the same oft-repeated vollies.
This post is inspired by a great case study; brought about because a friend sent me a link to Nova Spivack's 2003 post extolling the wonders of RSS. In the side navbar, I happened to notice that Nova's latest post from today was a response to Anne Zelenka's post comparing the aims of the Semantic Web to making machines more like the autistic Rain Man, able to recount facts in extraordinary precision and relations, but unable to use that data to further a real human connection. (I read Anne's article after I read Nova's... I found Anne's perspective extremely refreshing).
I like Nova Spivack. I don't know him personally, but I really like the quality of his thinking. It is particularly because he can be so eloquent that I feel compelled to reply. For me, the most cogent sounding and practical speakers on this particular subject are unfortunately arguing for a system that will never get us to our destination, even though the technologies they are working on are the best that we have to date. It is abundantly clear to me that we are going to have to discover a radical paradigm shifting solution to actually get to where we're pointing. And it seems that these eloquent arguments for tinker-toy elevators for space exploration are just giving just fueling the fire for skeptics who claim that the goal itself is absurd, and that we should just resign ourselves to the conclusion that algorithms will never amount to more than an interesting deterministic diversion from the real work of human powered thinking.
The position that I take falls in neither of the popular camps. I'm not looking to technology to pull some Deus ex machina rabbit out of its binary hat, nor am I inclined to believe that the forward march of technology can only enslave us. (Here is where I wish I were more eloquent myself.) I think we can pull ourself up by our bootstraps with the assistance of our technology, but it is going to require a radical new perspective on the way we approach problem solving. Simply jacking up the horse-power on existing technologies isn't sufficient in my opinion. And while Semantic Web (RDF/OWL) is a spiffy new sparqly toy, it doesn't have the quantum level nano-tube like strength to overcome the gravity of our nuvo-cognitive situation.
The crux of the problem, as I see it, relates very closely to David Weinberger's thesis in Everything is Miscellaneous; I don't trust man or machine to get the meaning of a piece of data right in an absolute and general way, because the meaning has everything to do with the context with which I want it in this very moment, and not so much to do with the intent of the person who wrote it down. Imagine if every poet encoded the semantics of each poem along with the poem. There are two obvious problems, one is the inflexibility of deciding the meaning at one point in time from one perspective, and the second is the problem of using that metadata to find a specific instance when the tags are so universal and common as: 'Love', 'Loss', 'Beauty', etc.
My wife contributed significantly to a successful solution that defined an interchane language/protocol for the oil industry. It was a massively complex but useful tool, and significantly simplified that the task of exchanging data stored in a wide variety of formats. But trying to envision a general purpose approach to extending that elegant solution to all of human knowledge is futile to the point of inducing despair. The reason that the solution she worked on is so successful is because the solution is closely tied with the intent for which the data was being used. Petroleum engineers know the data is there, know very much what the 'shape' of the data will be like, and have specific questions they are asking about the data. The intent of the how the data will be used only appears to be bound to the data itself because of the shared concern with others who would use that data. Thinking that the intent can actually be bound up with the data in a way that the 'concern' can be discerned by looking only at the contextual-less data and metadata misses the crucial point about data that Anne makes so well: meaning isn't simply constituted from facts, no matter how flawlessly the facts are memorized and cataloged.
Just to make sure that my point isn't lost in my ineloquence, I not only think that computerized-assistance with meaning-discernment is a good idea, I think it is inevitable, positive, and empowering. The catagorization scheme that I would like to see used would allow me to compare and contrast various perspectives on a given problem, not by finding the "right," "true," or most popular perspectives, but rather by giving me the ability to map the various perspectives and navigate among them irrespective of what any given authority thought about the correctness or popularity of any given perspective. Even, and perhaps especially, when that expert happened to be the author of the said perspective. Just because I disagree with Nova and Tim Berners Lee about the method of storing, retrieving or constructing meaning, doesn't mean I don't wish them all success. I genuinely hope I'm wrong. Just not so much that I can sit by contentedly and wait to see what they come up with. I would love to be able to easily find all the other people who look to where we're going and see other facets of the discussion that need to be had. Unfortunately, right now, I can only rely on the A-listers for a particular topic to link to them, or hope I stumble upon them in the comments, or through forwarded links from friends.
Post Script: An interesting aside. In discussing this idea with my wife, another whole perspective emerged. She indicated that she wasn't only uninterested in how anyone went about solving this particular set of problems (whether be GGG or some other new paradigm), she was concerned that such a solution may interfere with the expression of our humanity. In trying to discern why exactly that was, I think I narrowed the concern down to one of limitation enforced by algorithmic determinism imposed by machines. There are also issues and concerns around privacy, autonomy, freedom, etc., but to try to simplify the discussion, I asked that we think of an analogy: what if you had a faithful and skilled human servant who knew your patterns, could anticipate your concerns, and presented a simplified menu of choices, with full capability of expanding that menu of choices if you indicated that you wanted such. Trying to imagine a future in which computers could perform these tasks as well as a human servant (or assistant if you prefer), in her mind the human assistant is acceptable and desirable, but the computer assistant borders on repulsiveness. This response has nothing to do with the efficacy or beneficiality of the solution, but of the ethical sensability of the situation. It's a position that absolutely must be taken into consideration, even if the major participants in the discussion don't see it as particularly important.
Comments