Re: Bye Bye HyperText: The End of the World (Wide Web) As We Know It!

Michael De La Rue (mikedlr@indy.unipress.waw.pl)
Thu, 03 Oct 1996 14:27:08 +0200


In message <199610011710.KAA11233@webcrawler.com>, "Gerry McKiernan" writes:
> Bye Bye Hypertext:
> The End of the World (Wide Web) As We Know It!
>
> With the potential application of text extraction and
>summarization, data mining, intelligent software agents, and
>neural networks and other forms of AI, it has occurred to me
>that we may be on the brink of the End of the World (Wide Web)
>as We Know It.

>etc. Basically meaning will provide links.

This is all very well. In fact it's a wonderful idea. There are many
other people who agree with you. We only have one problem to solve to
get it to work. We call it `Artificial Intelligence'. All you have
to do is build a machine that thinks and works like a man and you've
nobody needs to do the linking. The machines just do it themselves.

seriously.

Or put it another way. When I provide a link in the Scottish climbing
archive, it doesn't mean `here is some other related document that has
similar content' it means `here is the document that Michael thought
was the best to cover that topic in the present context of _all_ the
documents that he is aware of on the entire WWW'.

I agree there are many problems which are being solved with links that
are better solved by search engines etc. I agree that sometimes you
are looking for `all the documents Michael wanted to hide from me'.
But this all misses the whole value of links. They have been very
specifically selected by another human, and until infinitely clonable,
absolutely obedient, totally reliable artificial intelligence is
available, this won't be something that can be replaced. Even then,
unless that artificial intelligence can work nearly infinitely fast,
all it will be replaced by is lists of links that are built up by the
AI entities and replaced on a regular basis...

Incidentally, even most workers in the AI field think that this kind
of AI is somewhat unlikely. Rodger Penrose says it's all impossible,
but he seems to have gone a little space cadet over this one (I'm not
denying that it is (or isn't) impossible, just that Rodger Penrose has
reasonable reason to think so).

You have to think of the links as something like a good index in a
book. A bad index is built up by saying where the different words and
concepts are mentioned; a good one is built up by hand by the author
by thinking about which bit of the book is the bit which covers which
topic that who would be looking for in what way. It's not a trivial
skill. Look at the Latex book (Lamport) or the Chicago Manual of
style (is that the right title) for more info on this kind of thin

Michael

<http://www.tardis.ed.ac.uk/~mikedlr/biography.html>
Scottish Climbing Archive: <http://www.tardis.ed.ac.uk/~mikedlr/climbing/>
Linux/Unix clone@ftp://src.doc.ic.ac.uk/packages/linux/sunsite.unc-mirror/docs/