Newsgroups: rec.arts.int-fiction
Path: gmd.de!xlink.net!howland.reston.ans.net!vixen.cso.uiuc.edu!sdd.hp.com!decwrl!decwrl!waikato!canterbury.ac.nz!huia!greg
From: greg@huia.canterbury.ac.nz (Greg Ewing)
Subject: Re: More rambling 
Message-ID: <CGnsoA.JGD@cantua.canterbury.ac.nz>
Nntp-Posting-Host: huia.canterbury.ac.nz
Reply-To: greg@huia.canterbury.ac.nz (Greg Ewing)
Organization: University of Canterbury, Christchurch, New Zealand
References: <2caanp$dfc@agate.berkeley.edu> <753468129snz@frootbat.demon.co.uk>
Date: Wed, 17 Nov 1993 23:16:57 GMT
Lines: 67

In article <753468129snz@frootbat.demon.co.uk>,
zobbo@frootbat.demon.co.uk (Ian Cottee) writes:
|> for it any maybe you'd have a game where you could do many more things
|> if they are logical and possible. Virtual Realitity promises to do the same 
|> thing graphically. Surely - before you can do that - you can do the whole
|> thing textually. 

No, you can't!

Consider this: There is a river that you want to cross, too deep and
swift to swim. You find some pieces of wood, a hammer and some nails
and decide to build a boat. In a decent reality simulator, you should
be able to do this even if the authors never explicitly included the
concept of a boat.

By simulating physics at a low enough level, it's probably not too
difficult. If the system keeps track of the shapes and spatial
relationships between the parts, and knows about things like density
and buoyancy, it will be able to figure out that the thing will
float on water, how much weight it will carry, and so forth.

But what happens when the player looks at what he has built?

In a graphical VR, all the system needs to do is show you a picture
of the object, which, given that it already has quite a detailed
spatial model of it, is almost trivial (or at least is a well-understood
problem).

In a textual VR, somehow the the object must be described in words.
At the very least it should say something like "There is a boat here".
Ideally it should contribute some atmosphere by saying something like
"There's a small, rather crudely-constructed boat here, made from
some pieces of scrap wood. It's pretty amateurish, but it looks like
it might float."

But without any built-in concept of a "boat", generating even the
simplest textual description that's remotely reasonable, starting
with a raw physical model, seems nearly impossible.

The problem is that the English language (or any natural language)
contains a huge vocabulary of words like "boat" that stand for
a complicated set of ideas regarding both the physical form and
intended use of an object.

To solve the problem would require a large dictionary of English
words and phrases together with a very, very sophisticated pattern
matching algorithm to map the objects to the descriptions. 
It would be at least as difficult as solving many of the hardest
problems in machine vision and reasoning which currently occupy
the AI community.

And that's just to get from the object to the word "boat", without
even considering how to make the description interesting to read!

The moral is that it's easy to overlook how much the processing
power of the human brain contributes to the experience of good IF.

|>   Ian Cottee - 15, Gerards Close, Varcoe Road, S Bermondsey, London,
SE16 3DF
|>     zobbo@frootbat.demon.co.uk  ***************  zobbo@cix.compulink.co.uk
|> * Don't go hanging onto rainbows, the colours just slip and you start
to slide*

Greg Ewing, Computer Science Dept, +--------------------------------------+
University of Canterbury,	   | A citizen of NewZealandCorp, a	  |
Christchurch, New Zealand	   | wholly-owned subsidiary of Japan Inc.|
greg@cosc.canterbury.ac.nz	   +--------------------------------------+
