Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

lost early in Curses

6 views
Skip to first unread message

Matthew J. MacKenzie

unread,
May 25, 1994, 9:26:01 PM5/25/94
to

I need a gentle nudge here, I think -- not ready to bow my head and
take a walkthrough just yet...

I've got 84 points and am just getting into the rod-collection
business (two down, one more located, some ideas what to do after
that). But I don't know how to charge the silly things, much less
use 'em. Could someone give me just a subtle clue? Post or email --
Thanks.

--
_____________________________________________________________________________
"'round here we stay up very very very very late" -- Counting Crows
Placement of smilies is left as an exercise for the reader.
ma...@glue.umd.edu Matthew MacKenzie

Matthew J. MacKenzie

unread,
May 25, 1994, 10:39:51 PM5/25/94
to

I see I put that in the wrong group. I'd still appreciate a pointer
in Curses, but while I'm here I may as well join a discussion. Let's
see what Michael Booth has to say here (chopping liberally from the
quotes)...

Sean Barrett <se...@stat.tamu.edu> wrote:
>Greg Ewing <gr...@huia.canterbury.ac.nz> wrote:
>>The real problem is to describe the results to the player in
>>a literary manner.
>>
>>It's not hard to arrange for all objects that should be
>>burnable to be burnable. But if all you see when you
>>burn most things is "The x burns" then burning things
>>gets boring very quickly.
>
>> GET STICK
>You get the stick.
>
>> BURN STICK
>The stick burns.
>
>> BURN MATCH
>The match burns.
>
>Sure, it's a little stupid sounding, but so
>is "get" and "drop".
>
>Of course, most games replace "You get the stick"
>with "Ok."
>
>The idea is that getting/dropping are fundamental
>actions in the universe, with one very standard
>"outcome" for the normal cases.
>
>You make burning things that fundamental, it's
>acceptable to replace it with "Ok." as well.

This is precisely where Mr. Barrett and I differ. I agree completely
with Greg Ewing, in that the most difficult part of the simulation
game is generating decent prose.

It is not neccessarily acceptable to just have the game say 'OK' when
something is dropped, taken, burnt, and so on. That is your opinion,
quite obviously, but it presents a rather bleak atmosphere. I would
rather see something like:

> BURN MATCH
The match sputters to life with a sharp hiss, the smell of sulfur
filling your nostrils.

[snipper-snack...]

>Certainly this is true, in some sense. But I don't think people
>mind the "boringness" of the core simulation, especially if
>their primary goal is to achieve some particular effect;
>then they may want "just the facts, ma'am".

This is true also. No need to overstate everything with flowery
language. However, there is a happy medium.

The biggest problem then becomes not the simulation system
(which is formidable in itself), but the natural language generation
system, which is as tangled as the english language. Just spitting
out canned phrases is not what I'm referring to here, but rather having
the system generate interesting prose describing the environment, such
as a portion of a room description, or the description of an object.
Something that is NOT a series of disjoint sentances like 'You are a strong
human male. You are tired. You are carrying a burned torch, a sword,
and a gem. You have a scar on your left cheek. You are not feeling very
well.'

I have a slightly different view of this part, and I think I can
connect it to this discussion. You're describing the story-telling
system in terms of modules: here's the simulation part, here's the
prose generation part. To me, the goal of IF is uncomfortably close
to the general AI problem (which both makes it very interesting and
forces us to scale back our expectations when the time comes to
actually implement something): to something like a dungeon master.
This program both runs the simulation and tells you (or shows you)
what's going on, and it can change both the world and the prose to
whatever effect it's after. It should guide both events and language
to a certain effect. Let's call it Sam, in honor of Mr. Clemens.

Now suppose you, the player, have just picked up a piece of paper and
a match. You're standing in a graveyard, under a new moon on an
October night in Salem. The witch-hunters are closing in. You decide
to start a brush fire in the dry autumn grass to give the hunters
something to deal with, since their homes are nearby. Sam's trying to
heighten the tension, and bring you at least close to being caught
because his arsenal of preprogrammed pieces of story lines, which he's
trying to pick a dramatic arc out of, says that you're at the height
of tension in the plot.
_______________________________________
---> strike match on headstone

The match is very bright in the darkness. It probably can be seen a
mile away.

---> burn paper with match

The paper flares like a torch. You hear shouts close by.

---> burn bush with paper

The flame crawls along one branch, then spreads to another. You see
old Preacher Helms in the light, not fifty feet away. He's got his
Bible in one hand and his whipping cane in the other, and is
momentarily frozen, aghast, as the dry bush flares like a bonfire in
front of him.

Preacher Helms finds his voice as four of the others come into view.
"Burn the witch!" he rumbles.
_______________________________________

Now Sam's storytelling here is hard to separate into segments. He's
allowed the bush to take fire, and taken advantage of the dramatic
turn of events in describing what you see. In my opinion the prose
and the higher-level story line could be chosen by the same system, as
higher and lower levels of the same process. A certain story line can
increase the tension, and so can a certain phrase.

The trick would be to catalog all things Sam knows about -- events,
verbs, characters, juxtapositions, idioms -- in a uniform system,
designed to say what effect each thing could have on a story (that is,
on the player). A garden spider, say, might have a death association
of .8, association with spider webs of 1.0, association with bugs of
.4, association with trees of .3, zero on all other catagories.
So the characters and events are not simulated as separate entities,
but are parts of Sam.

In this story, fire has strong implications of danger, maybe .9.
Sam's goal is to combine elements which are strongly matched, using
some quick comparison function -- the abstract object called witch
burning has a catalog entry which matches up nicely with the available
elements (fire, tension, Preacher Helms). Once witch burning is
selected as a decent match, it's gonna have at least a temporary
effect on events because it's there, just as real to Sam as the paper
and the headstone. If you'd tried to escape across a lake instead
Helms might have remembered to check whether you float...

I'm not trying to describe a heavy-duty Cyc-style AI system here, as
much as one more trick to try to get along without such a thing, by
the way. Sentence structures would have to be canned, the same way
plots would.

Anyway, I think a system like this, with a database of plot and
language effects, could reinforce the sense of genre and setting, and
add variety to the narration. It'd be a pain to create though, and
you'd have to be careful to keep in running in some low-bounded
polynomial time; if I get to something like this it'll start with a
very small micro world where I can play with the comparison function,
etc.

If people have any comment on this post, aside from giving me a hint
in Curses ;-), it'll probably be about treating characters as just
plot elements; but we'll see if any of this makes sense in the
morning.

0 new messages