 |
 |
 |
|
Re: Tuesday, February 28, 2006
On "Ending the Tyranny of the Link" and "hyperwords." What's right - links are still useful, but the author is correct in noting their limitations. What's wrong - author's suggested 'hyperwords' is not an answer - it slices salami with a battle-axe. The response list is less restricting, but still very canned, limited, and has only the context that the author thought to give it - or alternatively, by the context that the reader chooses to give it. It saves the reader only the mechanical searching for data via Google or some other mechanism, it does not address the underlying problem. The web turned data to information by organizing it, but subject to the limititations imposed by each website author's technical abilities, linguistic skills, pre-existing prejudices, and worldview, etc. Hyperwords would turn information, shallow and single-dimensional as it is, back to data; in this direction is chaos. Correct (read: useful) interpretation of a line of text requires appropriate heuristics, for which one can substitute that bugaboo of the 1980's, 'Artifical Intelligence' - or equally scarybadbuzzwordiness, 'Expert Systems'. We have touches of this already - in places like Amazon, Tivo. You liked X, you will probably like Y, the system infers. Often wrong, but a noble effort and frequently correct. Even the condescending and irritating Microsoft Bob, Clippy, and Agent made moves in that direction. Humans can no longer be expected to polyglot the Internet, is too much. But the rules humans apply to reason, query, and search can be applied to hueristic systems, which can apply the links in an appropriate manner on-the-fly. This should properly be applied at the client side, not the server side. Why? Because context is everything, and my information is not your information, what I care about is not what you care about - the important bits of a sentence for me are not the important bits for you. In addition, applying server-side personal heuristics requires gathering them, and you can't have mine. If my dataprivacy is important - facts which are historical data about me, then my personal heuristics are even more important - with them, you can predict, and that is not to be encouraged on the storage side. Furthermore - the author identified that links are single-ended, and this would remain the case even with the use of hyperwords. A protocol to allow an overlay of links to be applied to OBJECTS LINKED TO back to whence they came, without having to modify or update the original objects, would be useful. This would function like a clear acetate overlay on an overhead projector (for those who remember those ancient devices) - the reader would see the original object and the overlay (or overlays, or a selection of appropriate overlays) which make links back to from where they were linked. Now we're getting somewhere. Back to Prolog, we've got to get going on with canning human brains. Embrace the horror, don't be afraid to say AI. Smooches, Wiggy
|
