[Up: AI] [Robot Wisdom home page]
The premise(s) of this webpage:
- Every undergrad, encountering Artifical Intelligence for the first time, imagines they can solve its central problem... until they actually try it.
- The XML community has been delaying this day of reckoning by blaming their tools.
- The AI community has been remiss in not warning them.
- Their time has run out: Let's watch the fun!
The central problem of AI is to find a finite vocabulary that can be used to express any idea. [more]
Because we spend our lives adroitly juggling thousands of ideas every day, we intuitively believe we are the master of these ideas, and that expressing them concisely-- with a finite vocabulary-- is just a matter of sitting down and rattling them off.
But just try it! Ambiguities and overlaps are everywhere. Any idea worth mentioning will involve some aspect of the human psyche, so expressing it will require a finite model of the human psyche-- and nobody is anywhere close to offering that.
A trivial example, from the domain of XML webpages:
Say you have a fanpage devoted to Drew Barrymore.
It's easy to label the basic subsections with a standard vocabulary: pictures, biography, filmography, links, etc. (Think Yahoo.)
But as a Drewfan, you probably also realise that what people admire about her is her relentlessly upbeat attitude, despite her pretty-appalling Hollywood childhood. So maybe you want to create a page about her outlook on life-- her philosophy.
Granted, this is not a standard subsection for a typical fansite, but neither is it outlandish. And if you're instead (also!) a fan of James Joyce, then a subsection about his philosophy would be absolutely logical... so any system of XML meta-information needs to have a finite way of expressing this concept: 'philosophy of person X'.
But now you've opened the floodgates of the 'AI problem'. My Joyce site includes resources on his eyesight, his writer's craft, his education, his travels in Europe, the books he read, his view of Shakespeare, his style of punctuation, errors he made in Ulysses, ...etc etc etc. [more]
So I predict that in the coming year we'll get to relish the spectacle of dozens of prominent XML-advocates running up against this AI-barrier, and having to back down from their extravagant claims for XML. Soon after this, I predict, XML will be quietly abandoned as a successor to HTML for webpage markup.
So this webpage will be accumulating links that track this evolution, starting with this index page of Tim Berners-Lee's meditations on XML and AI. In one particular example here, Tim tries to represent the concept "Ralph said Ora wrote the book."
(Tim has the most ego on the line, in this battle, so I'm not surprised to see the first signs of a nervous breakdown in his writings here.) [anti-W3C expose]
Nervous-breakdown update
What XML-advocates have to do now is propose 'ontologies' that arrange the finite vocabulary into a sorted hierarchy. These are always pretty funny, because they have to turn such somersaults to achieve even miniscule results. Most people are ashamed to publish them, but here's some I've found: Jeff Heflin; KA2
AI AWOL?
I have to wonder where the AI community has been while all this has been going on.
There's one prominent example of an ambitious universal ontology, called 'Cyc' [more], but if Cyc wants to represent 'buy a car' then 'buy' becomes:
"Thing: Individual: TemporalThing: Event: Action: PurposefulAction: Transaction: CommercialActivity: SalesActivity"and 'car' becomes:
"Thing: Individual: TemporalThing: SomethingExisting: Artifact: Conveyance: TransportationDevice: TransportationDevice-Vehicle: RoadVehicle: Automobile". [cite] [thread]Hopefully this isn't the future of webpage markup!
One of the founders of the Cyc project, R.V. Guha, [Epinions page] has in fact been deeply involved in the evolution of XML, and should have been able to warn people that Cyc is nowhere close to offering universal indexing, even after 15+ years of effort.
Alternatives
I have a series of rants: against structural markup [1], in favor of browsers that parse [2], and proposing a better startingplace [3].
SuggestionsYou can submit a new URL or any other suggestion for this page by typing it into the box below. It will instantly become visible to anyone at this comments page. I should get around to checking it out and updating it above within a week or three, at which point I'll delete it from the comments page.
If you want credit, include your name and email (otherwise it's anonymous). You can use HTML but you don't have to.
[Up: AI] [Site map] [Robot Wisdom home page] (Feedback)