Self-creating games
Dave Gileadi
I recently finished reading the Ender's Game series, by Orson Scott Card
(excellent book, if you haven't read it yet). Anyway, among several really
cool computer games used in the book is one called the Giant's Drink. It is
a strange game in that it was designed to not be winnable (one of those
psychology things). The player was to choose one of two cups the giant
offered. Supposedly there was a right choice and a wrong one, but in
reality you died whichever one you drank. However, after fruitlessly
playing it for a very long time, Ender did something that no-one had ever
done before. He killed the giant (cleverly).
Then comes the interesting part: the game had no precedent for what was to
happen then, so it set aside a bunch of resources and simply invented what
was to follow. The rest of the game (which led through several interesting
places) was completely invented on the fly, by an intellegent supercomputer.
Now, is this possible in a real situation? A computer game that would
create itself, depending on what the player did? Assuming it's possible,
what would it need?
I think it would need:
* A fairly large database of lots of different
objects/characters/terrain, etc.
* Some form of AI techniques, that would be able to respond to the
player's actions by picking out certain objects, etc. from the database and
organizing them into scenes and situations.
It would be even more interesting if it were made into a multiplayer
community, so the outcomes would be affected by more than one person--each
player bends reality to his/her own role. I also think it would need to
have some fantasy to it, so that strange combinations of characters and
environment would seem to fit, rather than being out of place.
Anyway, would this work? Is it possible to bend a game's reality based on
what a player does? How would it play? And would it be fun?
-Dave
Dr. Awkward
Just like the real world, it's more interesting when people create than
machines. Think of computer-generated art (I mean algorithmic, not 3d studio
or whatever human-driven tools), vs. what people do. Then think of game
design as an art, which it is, and you'll begin to get the picture. A
computer can create an infinity of mazes, but a human can create one that
will specifically "trick" you quite easily.
Dave Gileadi wrote in message <35008...@news.itsnet.com>...
James Arendt
Dr. Awkward wrote:
>
> Just like the real world, it's more interesting when people create than
> machines. Think of computer-generated art (I mean algorithmic, not 3d studio
> or whatever human-driven tools), vs. what people do. Then think of game
> design as an art, which it is, and you'll begin to get the picture. A
> computer can create an infinity of mazes, but a human can create one that
> will specifically "trick" you quite easily.
I think the idea proposed could work in the future, but currently we
just don't have powerful enough computers to make the computer
intelligent enough and is intelligence enough? Does intelligence
necessarily mean creative? I don't believe so.
So, not only do computers have to become "smarter" to come up with
alternate situations (sorta like a GM who runs paper based rpgs), but
they need to have the creativity to come up with interesting, fun, and
challenging situations.
However, a limited form of the original idea could be playable, but
how long will it take before the players see a "pattern" in how the
situations are being developed?
James
LEFEBVRE Cedric
Self-creating games ?
Have you ever worked on AI or LISP and even C ?
Self programming languages already exist and these are mainly LISP and C
(because not strong-typed languages).
At this time I am in engineering school and one of our projects at this
time (I finished it last week) was a computer program able to play
chess.
The first time he only knows the rules and randomizes everything, then
game after game he learns and programs new routines (everything in C)
and now I am unable to beat him.
Initially the program was 500Ko and now more than 12Mo.
I know it could have been done with a database but the purpose of the
project was to learn how to make your program reprogramming itself.
At this time such programming is not very much used but it begins to,
since for Seven Kingdoms (I think it was it) they used this kind of
system, letting the computer learn more abd more and program new
reactions.
The problem at this time is the computer learns far less fastly than
human, and to learn him on simple ripost you have to play it ten times
or so (depending on the complexity of it)
If you are interesting there are good books on LISP. (but it is
adaptative to C)
Isaac Kuo
In article <35015A82...@earthling.net>,
James Arendt <grass...@earthling.net> wrote:
>Dr. Awkward wrote:
>> Just like the real world, it's more interesting when people create than
>> machines. Think of computer-generated art (I mean algorithmic, not 3d studio
>> or whatever human-driven tools), vs. what people do. Then think of game
>> design as an art, which it is, and you'll begin to get the picture. A
>> computer can create an infinity of mazes, but a human can create one that
>> will specifically "trick" you quite easily.
>I think the idea proposed could work in the future, but currently we
>just don't have powerful enough computers to make the computer
>intelligent enough and is intelligence enough? Does intelligence
>necessarily mean creative? I don't believe so.
You're right. The problem is that we haven't developed any
algorithms to do something "creative". It might be that
we already have computing power necessary--we just don't
know how to apply it in the right way.
>So, not only do computers have to become "smarter" to come up with
>alternate situations (sorta like a GM who runs paper based rpgs), but
>they need to have the creativity to come up with interesting, fun, and
>challenging situations.
>However, a limited form of the original idea could be playable, but
>how long will it take before the players see a "pattern" in how the
>situations are being developed?
How about modifying the original idea to make the original intent
somewhat possible?
Perhaps there is some way to make a computer game which can be
expanded upon by many users. Somehow, users can create modular
"features" that the computer game can somehow mix and match.
This way, the game evolves through user creativity.
For instance, consider a text adventure game. When a user
plays a game, and notices the computer just spits out a
canned response to "KISS OGRE", he might decide to add
a simple feature that changes this response to "You feel
nauseous." Thus, the next player who tries that will get
the new response.
--
_____ Isaac Kuo k...@bit.csc.lsu.edu http://www.csc.lsu.edu/~kuo
__|_)o(_|__
/___________\ "Mari-san... Yokatta...
\=\)-----(/=/ ...Yokatta go-buji de..." - Karigari Hiroshi
Aldin Larice
On Fri, 6 Mar 1998 16:22:32 -0700, "Dave Gileadi"
<DaveG...@TheOffice.net> wrote thusly:
>I recently finished reading the Ender's Game series, by Orson Scott Card
>(excellent book, if you haven't read it yet). Anyway, among several really
>cool computer games used in the book is one called the Giant's Drink. It is
>a strange game in that it was designed to not be winnable (one of those
>psychology things). The player was to choose one of two cups the giant
>offered. Supposedly there was a right choice and a wrong one, but in
>reality you died whichever one you drank. However, after fruitlessly
>playing it for a very long time, Ender did something that no-one had ever
>done before. He killed the giant (cleverly).
Actually, I think the game *was* winnable; killing the giant was the
only way to win. I imagined it was a test for genius; if somebody
beat the Giant's Drink, Those In Charge new that they had their man
(or seven-year-old boy, in this case).
>Then comes the interesting part: the game had no precedent for what was to
>happen then, so it set aside a bunch of resources and simply invented what
>was to follow. The rest of the game (which led through several interesting
>places) was completely invented on the fly, by an intellegent supercomputer.
>
>Now, is this possible in a real situation? A computer game that would
>create itself, depending on what the player did? Assuming it's possible,
>what would it need?
Yes, it's possible. Think of a computer game that generates random
maps; as soon as the player moves into uncharted territory the game
just generates more territory to explore (within game-defined
constraints; in an RPG, if the player were exploring a forest, have
the forest continue, then maybe create a mountain, and so forth).
It's not that hard.
The only thing about the Giant's Drink and that whole environment is
that it's very fluid. E.g. the player has complete freedom of
movement, which is rare in today's games and, wherever I've seen it,
the player is always moving around a human-created environment. I
don't know how hard it would be to generate random environments like
this, but I imagine extremely difficult.
>I think it would need:
>
> * A fairly large database of lots of different
>objects/characters/terrain, etc.
> * Some form of AI techniques, that would be able to respond to the
>player's actions by picking out certain objects, etc. from the database and
>organizing them into scenes and situations.
This actually depends on what kind of game it is. If we've got a
hack-n-slash RPG, you can just throw a band of orcs at the player
every so often. I've often wondered why side-scrollers don't do this;
respond to the player's ability by just throwing more (or more
powerful) enemies at him/her. I think a fairly recent (shareware?)
computer game did this to some extent.
>It would be even more interesting if it were made into a multiplayer
>community, so the outcomes would be affected by more than one person--each
>player bends reality to his/her own role. I also think it would need to
>have some fantasy to it, so that strange combinations of characters and
>environment would seem to fit, rather than being out of place.
Yes, but then who takes precedence if two players are together and
they both are "bending reality"?
Anyway, this would probably be a lot easier to solve if done as a
single-person game, then studied and applied to a multiplayer
environment.
>Anyway, would this work? Is it possible to bend a game's reality based on
>what a player does? How would it play? And would it be fun?
I think the main problems would be the interface and environment. We
simply don't have the technology to build a game that's as free as the
Giant's Drink (as described in Ender's Game), though I think you could
build a genre game (hack-n-slash, action) along these lines.
Aldin Larice
------------------------------------------------------------------------
If doctors are so smart, why do they call it a practice?
Karl Low
Ender's Game series Spoilers below.. read at your own risk.
Aldin Larice wrote in message <350190b8...@news.gmu.edu>...
>><snip basic desc of Giant's Drink in Ender's Game>
>Actually, I think the game *was* winnable; killing the giant was the
>only way to win. I imagined it was a test for genius; if somebody
>beat the Giant's Drink, Those In Charge new that they had their man
>(or seven-year-old boy, in this case).
Nope. Reading further into the series the game wasn't intended to be
winnable. It was to be a psychological test to see how "suicidal" a kid was.
>>Then comes the interesting part: the game had no precedent for what was
to
>>happen then, so it set aside a bunch of resources and simply invented what
>>was to follow. The rest of the game (which led through several
interesting
>>places) was completely invented on the fly, by an intellegent
supercomputer.
>>
Actually... an completely artificial intelligence.. the point of whether the
intelligence counts as alive or not is looked at more closely in the third
book.
>>Now, is this possible in a real situation? A computer game that would
>>create itself, depending on what the player did? Assuming it's possible,
>>what would it need?
>
>Yes, it's possible. Think of a computer game that generates random
>maps; as soon as the player moves into uncharted territory the game
>just generates more territory to explore (within game-defined
>constraints; in an RPG, if the player were exploring a forest, have
>the forest continue, then maybe create a mountain, and so forth).
>It's not that hard.
The kicker is though.. is it Fun? What made Giant's Drink so interesting is
that the computer was taking stuff from Ender's files, psychological
studies, previous work, family history, etc. and putting that all together
to make a game specifically tied to Ender.
Sure you can generate random maps.. trying to develop a coherent game
experience though.. that's a lot tougher.
>
>The only thing about the Giant's Drink and that whole environment is
>that it's very fluid. E.g. the player has complete freedom of
>movement, which is rare in today's games and, wherever I've seen it,
>the player is always moving around a human-created environment. I
>don't know how hard it would be to generate random environments like
>this, but I imagine extremely difficult.
As you just said previously, random maps/environments are the easy part.
>>I think it would need:
>>
>> * A fairly large database of lots of different
>>objects/characters/terrain, etc.
>> * Some form of AI techniques, that would be able to respond to the
>>player's actions by picking out certain objects, etc. from the database
and
>>organizing them into scenes and situations.
You'd also need a very hefty intelligent system that could look at what the
player was doing and create *appropriate* responses on the fly. For that, I
don't think we have the computing power yet.
>This actually depends on what kind of game it is. If we've got a
>hack-n-slash RPG, you can just throw a band of orcs at the player
>every so often. I've often wondered why side-scrollers don't do this;
>respond to the player's ability by just throwing more (or more
>powerful) enemies at him/her. I think a fairly recent (shareware?)
>computer game did this to some extent.
This isn't actually expanding the game though, this is just upping the
difficulty level.
>>It would be even more interesting if it were made into a multiplayer
>>community, so the outcomes would be affected by more than one person--each
>>player bends reality to his/her own role. I also think it would need to
>>have some fantasy to it, so that strange combinations of characters and
>>environment would seem to fit, rather than being out of place.
>
>Yes, but then who takes precedence if two players are together and
>they both are "bending reality"?
>
>Anyway, this would probably be a lot easier to solve if done as a
>single-person game, then studied and applied to a multiplayer
>environment.
Personally, I think this is a bad idea, and part of the problems that most
multi-player games out there have. They're designed as single player games,
then just "ported over" to multi-player, without any thought into how
creative, vindictive, petty, mean, generous, helpful or just plain different
people are. As such, when somebody isn't playing "the right way" for you,
the game loses its fun.
>>Anyway, would this work? Is it possible to bend a game's reality based on
>>what a player does? How would it play? And would it be fun?
Technically, yes it's possible, it would play as slow as molasses, and as
for it being fun.. without an actual intelligence doing the designing..
doubtful in my opinion.
There's a world of difference between making a game and making a GOOD game.
The easiest example of this is to check out various Doom WADs.. some of
them are great, some of them are really no "fun" at all, if even people
can't get it right all the time, what hope is a computer going to have?
-Karl
Isaac Kuo
In article <350190b8...@news.gmu.edu>,
Aldin Larice <jem...@geocities.com> wrote:
>On Fri, 6 Mar 1998 16:22:32 -0700, "Dave Gileadi"
><DaveG...@TheOffice.net> wrote thusly:
>>Then comes the interesting part: the game had no precedent for what was to
>>happen then, so it set aside a bunch of resources and simply invented what
>>was to follow. The rest of the game (which led through several interesting
>>places) was completely invented on the fly, by an intellegent supercomputer.
>>Now, is this possible in a real situation? A computer game that would
>>create itself, depending on what the player did? Assuming it's possible,
>>what would it need?
>Yes, it's possible. Think of a computer game that generates random
>maps; as soon as the player moves into uncharted territory the game
>just generates more territory to explore (within game-defined
>constraints; in an RPG, if the player were exploring a forest, have
>the forest continue, then maybe create a mountain, and so forth).
>It's not that hard.
That's also not what Dave is getting at. A random terrain
generator was purposefully already programmed into the game
you're thinking of. While it might appear like something
new the first time a player tries to explore uncharted
territory, after that the precedent has been set.
Basically, the question is whether or not the game can be
programmed to contend with options the programmer never
anticipated in an interesting way. (Any program can
contend with unanticipated player actions in an
uninteresting way.)
>>I think it would need:
>> * A fairly large database of lots of different
>>objects/characters/terrain, etc.
>> * Some form of AI techniques, that would be able to respond to the
>>player's actions by picking out certain objects, etc. from the database and
>>organizing them into scenes and situations.
>This actually depends on what kind of game it is. If we've got a
>hack-n-slash RPG, you can just throw a band of orcs at the player
>every so often. I've often wondered why side-scrollers don't do this;
>respond to the player's ability by just throwing more (or more
>powerful) enemies at him/her. I think a fairly recent (shareware?)
>computer game did this to some extent.
Back in the early boom in arcade games, there wasn't enough
ROM to store the locations of all enemies and maps. Back
then, most video games used randomly placed enemies, which
got more numerous and more dangerous with time.
At some point, enough ROM was affordable to put predesigned
enemy locations on them. For whatever reasons, people
_prefer_ having the enemies pop up at the same place in the
same time. It seems that most players like being able to
memorize what to expect on a level so they can learn how
to beat it. Indeed, a game like Sonic the Hedgehog would
be absolutely unplayable if the "maze" were randomly
generated each time it's played.
I like the idea of extra random enemies thrown in to keep
players on their toes, but unfortunately I don't think it
would be very popular.
John Q. Public
Dave Gileadi wrote in message <35008...@news.itsnet.com>...
>Then comes the interesting part: the game had no precedent for what was to
>happen then, so it set aside a bunch of resources and simply invented what
>was to follow. The rest of the game (which led through several interesting
>places) was completely invented on the fly, by an intellegent
supercomputer.
>
>Now, is this possible in a real situation? A computer game that would
>create itself, depending on what the player did? Assuming it's possible,
>what would it need?
>
>I think it would need:
>
> * A fairly large database of lots of different
>objects/characters/terrain, etc.
> * Some form of AI techniques, that would be able to respond to the
>player's actions by picking out certain objects, etc. from the database and
>organizing them into scenes and situations.
It would also need a scenario generator.
Such as : Rescue the princess from the dragon (simple example) or more
complex such as: Rescue the princess from teh shapshifting dragon that
happens to be the king's advisor and wishes to coerce the king into giving
up the gem of Alazdar which the advisor has discovered grants immortality to
the person that crushes the gem and drinks it mixed with royal blood.
Such scenarios were generated by a wheel we had in g&t and provided dozens
of story ideas. By drawing on a larger base and tagging the objects to
exclude silly quests (such as rescuing a pig held for ransom)
one could create a world of endless adventure.
Gerry Quinn
>It would also need a scenario generator.
>Such as : Rescue the princess from the dragon (simple example) or more
>complex such as: Rescue the princess from teh shapshifting dragon that
>happens to be the king's advisor and wishes to coerce the king into giving
>up the gem of Alazdar which the advisor has discovered grants immortality to
>the person that crushes the gem and drinks it mixed with royal blood.
>Such scenarios were generated by a wheel we had in g&t and provided dozens
>of story ideas. By drawing on a larger base and tagging the objects to
>exclude silly quests (such as rescuing a pig held for ransom)
>one could create a world of endless adventure.
Rescuing a pig held for ransom sounds endearingly wacky!
Could we do more with current techniques. Suppose we took a random
map, maze and monster generator, a quest generator like the above, and
a puzzle-maker like Moriarty, Minesweeper (or my own Detective Chess),
and let them mate (the use of a genetic algorithm is metaphorical
here).
How close could we get to the desired scenario? The puzzle generator
is the hardest - we can make puzzles of a chosen degree of difficulty,
but how can we integrate them so they appear part of a quest?
Moriarty-style puzzles seem the most promising - bits of knowledge
could be spread around the scenario to form sub-quests, and collated
to give essential info for the main quest. Door and switch puzzles
could be generated in other ways.
- Gerry
===========================================================
ger...@indigo.ie (Gerry Quinn)
http://indigo.ie/~gerryq
Original puzzlers for PC, Amiga, and Java
===========================================================
Richard Hobson
A list of books or even web sites would be very cool, I am really
interested but don't know where to start.
Richard
-ZZ
On Fri, 6 Mar 1998 16:22:32 -0700, "Dave Gileadi"
>Now, is this possible in a real situation? A computer game that would
>create itself, depending on what the player did? Assuming it's possible,
>what would it need?
>Anyway, would this work? Is it possible to bend a game's reality based on
>what a player does? How would it play? And would it be fun?
To work properly, the game-world would have to be defined from a root
level upwards. IE the definition would need to incorporate the
fundamental laws of the game universe from which new development could
be based.
Now that's a lot of work.
I've seen it done in an rpg (superbly), but on the other hand a human
Ref is still far, far better at that kind of thing that a computer.
-ZZ
Torben AEgidius Mogensen
k...@mouse.csc.lsu.edu (Isaac Kuo) writes:
>In article <35015A82...@earthling.net>,
>James Arendt <grass...@earthling.net> wrote:
>>Dr. Awkward wrote:
>
>>I think the idea proposed could work in the future, but currently we
>>just don't have powerful enough computers to make the computer
>>intelligent enough and is intelligence enough? Does intelligence
>>necessarily mean creative? I don't believe so.
>You're right. The problem is that we haven't developed any
>algorithms to do something "creative". It might be that
>we already have computing power necessary--we just don't
>know how to apply it in the right way.
I guess most people would (possibly unconsciously) define things that
you can program a computer to do as "not creative", just like you
often hear that "intelligence is what we have not (yet) taught a
computer to do well". Hence, while chess was considered a primarily
intelligence-driven game around the turn of the century, this is no
longer considred the case now that computers regularly beat even the
best human players.
>How about modifying the original idea to make the original intent
>somewhat possible?
>Perhaps there is some way to make a computer game which can be
>expanded upon by many users. Somehow, users can create modular
>"features" that the computer game can somehow mix and match.
>This way, the game evolves through user creativity.
>For instance, consider a text adventure game. When a user
>plays a game, and notices the computer just spits out a
>canned response to "KISS OGRE", he might decide to add
>a simple feature that changes this response to "You feel
>nauseous." Thus, the next player who tries that will get
>the new response.
I think this is far too low level. I don;t think it is very difficult
to let a computer create a "world" suing algorithms for terrain,
climate, cities, trade routes and even composition of people in
cities. However, individual people responses are less likely to be
so. A simple idea, related to the original GM suggestion is to require
the players of the game to handle all but trivial NPC responses. When
something gets beyound the game AI, it will pick a random PC who is
logged on (preferably not someone too close to the player interacting
with the NPC) and let him play the part of the NPC. Players can be
required to handle these respones in order to get permission for
continued play.
If the AI can be made to teach itself according to examples, it can
use the responses that the players generate for the NPC (and the
responses of PC's to each other and to NPC's)to extend itself. This
is, however, not going to be easy to do.
Torben Mogensen (tor...@diku.dk)
Dave Gileadi
[snip lots of interesting discussion]
Isaac said:
>
>That's also not what Dave is getting at. A random terrain
>generator was purposefully already programmed into the game
>you're thinking of. While it might appear like something
>new the first time a player tries to explore uncharted
>territory, after that the precedent has been set.
>
>Basically, the question is whether or not the game can be
>programmed to contend with options the programmer never
>anticipated in an interesting way. (Any program can
>contend with unanticipated player actions in an
>uninteresting way.)
Yes! The way the computer did it in Ender's game required two things:
1) it happened to also be the computer that controlled all the happenings on
the base where they lived, as well as linked to the computer that handled
all the affairs on Earth (again, the nature of the computer was pointed to
by Karl Low). The point is that it had knowledge about everything--ex. ask
it what a shoe was, and it could tell you what they were used for, what they
were made of, all the different varieties available, how much they cost,
etc. It could even simulate what might happen if a shoe were, say, to be
dropped onto a roof from a thousand-foot cliff. In other words, the
computer had access to data for just about anything needed for the game.
2) it was a truly intelligent computer, and contained some kind of program
to analyze its player's personalities. Thus it was able to decide
intelligently what might affect the player certain ways.
Now, it is clearly impossible for any computer nowdays to accomplish either
one of the above tasks; but as we know, computer games are simulations of
reality, and why not simulate the above?
We could have a limited database of things that the player might encounter
in this particular world. That is one reason why in the original post I
said the game ought to be some sort of fantasy game--it limits the set of
needed data, as well as makes strange combinations less odd.
The challenge would be to come up with a creative AI that could interpret
what a player was trying to do or say, and come up with some solution using
the possible data. The difficulty would be making the solutions plausible,
or realistic.
I like the idea of letting players add to the game--that idea is worth
following. It also would allow the computer to learn from their additions.
On the other hand, it could lead to some pretty messed up playing, if people
just used garbage for their additions...
Aldin Larice
On Tue, 10 Mar 1998 11:00:45 -0700, "Dave Gileadi"
<DaveG...@TheOffice.net> wrote thusly:
>
>[snip lots of interesting discussion]
>
>Isaac said:
>
>>
>>That's also not what Dave is getting at. A random terrain
>>generator was purposefully already programmed into the game
>>you're thinking of. While it might appear like something
>>new the first time a player tries to explore uncharted
>>territory, after that the precedent has been set.
>>
>>Basically, the question is whether or not the game can be
>>programmed to contend with options the programmer never
>>anticipated in an interesting way. (Any program can
>>contend with unanticipated player actions in an
>>uninteresting way.)
>
>Yes! The way the computer did it in Ender's game required two things:
>
[snip: The computer controlled the entire world, with full physics
built-in, AND was a truly intelligent program, able to analyze its
players' personalities]
>Now, it is clearly impossible for any computer nowdays to accomplish either
>one of the above tasks; but as we know, computer games are simulations of
>reality, and why not simulate the above?
>
>We could have a limited database of things that the player might encounter
>in this particular world. That is one reason why in the original post I
>said the game ought to be some sort of fantasy game--it limits the set of
>needed data, as well as makes strange combinations less odd.
Indeed, I don't see why not. Basically, you totally model a shoe
(physical view, weight, color, magical properties...). Then you
totally model a sword. And so forth, adding to the world step by
step.
>The challenge would be to come up with a creative AI that could interpret
>what a player was trying to do or say, and come up with some solution using
>the possible data. The difficulty would be making the solutions plausible,
>or realistic.
Well, I think the real problem would be this bit. The Giant's Drink
was able to model the person's personality, and thus present
challenges based on that.
Though it could be possible to do something along these lines.
Present the user with a small beginning world that has a set of
personality-revealing situations.
The actual Giant's Drink is a good example: The only way to win is to
kill the giant in a vicious way. If the player does that, notch up
the aggressiveness level in their profile.
>I like the idea of letting players add to the game--that idea is worth
>following. It also would allow the computer to learn from their additions.
>On the other hand, it could lead to some pretty messed up playing, if people
>just used garbage for their additions...
Indeed, though it's difficult to do that in traditional games, since
it's generally easy to upset the delicate balance. OTOH, in this game
it shouldn't be a problem, since there really is no balance....
Bjoern Guenzel
k...@mouse.csc.lsu.edu (Isaac Kuo) wrote:
>You're right. The problem is that we haven't developed any
>algorithms to do something "creative". It might be that
>we already have computing power necessary--we just don't
>know how to apply it in the right way.
What about genetic algorithms?
Bjoern
jay...@imaginestudios.com
In article <890334767.14429.0...@news.demon.co.uk>,
Good point, that's fairly close to 'creative', at least as creative as we've
gotten so far. The only difficulty with genetic algorithms is you must
carefully set up the range of 'paths' they can choose in their creativity and
you must create a good way of testing their success. One you've done all this
it seems a little less creative since you've defined so much.
For example, using the title as an example, you must program in a good number
of general rules to use to build games and then you must define a way to test
how successful or good the game is that is creates. Not an easy task...
jayson
-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/ Now offering spam-free web-based newsreading
David H. Bolton
In article <6dte55$nlo$1...@its1.ocs.lsu.edu>, Isaac Kuo
<k...@mouse.csc.lsu.edu> writes
>I like the idea of extra random enemies thrown in to keep
>players on their toes, but unfortunately I don't think it
>would be very popular.
It depends on game type.The Ultima series had NPCs and certain key
baddies at fixed locations on a predesigned map but there were loads of
randomly generated monsters on this map and I also think in the
underground dungeons.
--
David H. Bolton
Bjoern Guenzel
jay...@imaginestudios.com wrote:
>In article <890334767.14429.0...@news.demon.co.uk>,
> guenz...@usa.net (Bjoern Guenzel) wrote:
>>
>> k...@mouse.csc.lsu.edu (Isaac Kuo) wrote:
>>
>> >You're right. The problem is that we haven't developed any
>> >algorithms to do something "creative". It might be that
>> >we already have computing power necessary--we just don't
>> >know how to apply it in the right way.
>>
>> What about genetic algorithms?
>>
>> Bjoern
>>
>Good point, that's fairly close to 'creative', at least as creative as we've
>gotten so far. The only difficulty with genetic algorithms is you must
>carefully set up the range of 'paths' they can choose in their creativity and
>you must create a good way of testing their success. One you've done all this
>it seems a little less creative since you've defined so much.
>For example, using the title as an example, you must program in a good number
>of general rules to use to build games and then you must define a way to test
>how successful or good the game is that is creates. Not an easy task...
In the game like the one from the title, maybe the players actions
could be made to set the path. For example, if the genetic algorithms
generates a blue room, a red room and a green room for the player to
enter, and whatever room he enters gets evolved next.
Aren't games like Creatures (which I haven't played or seen, only
heard about) in a way self-creating games, with the player being the
directing force? Of course there are hardwired limits to the possible
number of games, but that is just a matter of detail. Once the genetic
code reaches down to machine code, everything should be possible:-)
All this of course only hippothetical, I think we are a long way from
getting to self-creating games. However, with the growing demand for
detail in virtual worlds, I could imagine that sooner or later games
producers won't have a choice but have their worlds somehow
automatically Computer-created (I don't mean random, I think if there
is an interesting logic behind such a world, it could be interesting.
Like I saw in another post - start with a landscape, then see where
the rivers are and put settlements there and so on).
Funny, maybe there is a factor that describes how much percent
'simulation' of real life a programmer can code into a games - just
because it takes longer for example to describe somebody walking down
a street than to simply walk down a street, which limits the how much
of your own experience of reality you can encode - producing a virtual
would be the equivalent of writing... then different media could have
different factors... Of course one could just mount some people with
live cams (and other sensors?) and send them down the street and let
the Computer do the rest. Reminds me of the Gargoyles from Snow Crash.
Or imagine little Info-Roboters crowding our real world streets,
collecting data in the name of game companies (unfortunately, this
would be too much of a Big Brother situation):-)
Bjoern
Bjoern Guenzel
"Karl Low" <kw...@deletethis.cadvision.com> wrote:
[...]
>The kicker is though.. is it Fun? What made Giant's Drink so interesting is
>that the computer was taking stuff from Ender's files, psychological
>studies, previous work, family history, etc. and putting that all together
>to make a game specifically tied to Ender.
I read an article about the israelic game 'virus', which would
apparently scan the conten of your Computer and then use the items it
found later in the game. Apparently the game was a flop, but I think
the idea sounded interesting. And after all, a personal Computer in
most cases does contain a lot of data about it's user.
Yes, I know - the problem is still that the Computer doesn't
understand the data and therefore can hardly use it in interesting
ways. Although I guess the Computer could determine how important
certain snippets of data are for you (depending on how often you use a
certain program or read a certain document or whatever). Maybe it
could even come up with a simple psychological profile by determining
how forceful or hectically you move the mouse-pointer....:-) Seems as
if todays Computer COULD get loads of psychological info abbout their
user if they wanted to.
In fact, now I seem to remember that Microsoft is already working on
this, programs detecting when the user gets angry or nervous and
reacting with more help messages etc.
I suppose the whole data gets also send to the Microsoft Marketing
section at the same time...:-) In fact, the Computer could probably
easily detect how intelligent you are, measuring how long it takes you
to learn to use a program (identifying a user by his typing behaviour
has already been done), and send the data to your employer.
Have things like that ever been used in games?
I think the character creation of the Ultimas where really
interesting, when you went to the gypsy and had to answer certain
questions to determine which character type you are. Any other
examples?
Bjoern
[...]
Geoff Howland
> I read an article about the israelic game 'virus', which would
> apparently scan the conten of your Computer and then use the items it
> found later in the game. Apparently the game was a flop, but I think
> the idea sounded interesting. And after all, a personal Computer in
> most cases does contain a lot of data about it's user.
> Yes, I know - the problem is still that the Computer doesn't
> understand the data and therefore can hardly use it in interesting
> ways. Although I guess the Computer could determine how important
> certain snippets of data are for you (depending on how often you use a
> certain program or read a certain document or whatever).
This is not possible, this is not how computers work, they dont log howoften
you use any program or access any files. At the MOST you can
find out the last time a file was SAVED, and possibly the last time an
application was launched (such as a looking in the document folder
on Win95 and seeing the assoicated programs to the files, or on the
mac I think it actaully has a last used applications deal), but still
this is trivial knowledge, and someone could realistically just have
used a program they use only 6 months and then you make it a
major feature?
Also, AI is not an option now, the only working AI's I know of have
VERY specific topics of "thought" such as playing chess, to even
imagine throwing any sorts of random data at a program is simply
trying to make science fiction a reality, for now. The best and only
way to do this would be for the programmer to go through and make
up routines on things he knows of. Such as, he can scan through
data looking for addresses or phone number and try to attach them
to names. Thats relevant and can look like the program is acting
intelligently...
> Maybe it
> could even come up with a simple psychological profile by determining
> how forceful or hectically you move the mouse-pointer....:-) Seems as
> if todays Computer COULD get loads of psychological info abbout their
> user if they wanted to.
> In fact, now I seem to remember that Microsoft is already working on
> this, programs detecting when the user gets angry or nervous and
> reacting with more help messages etc.
This is totally ridiculous, they MAY have something like this, whichI doubt,
but their not going to be getting it on anyones computer in
the next 50 years, IMO.
> I suppose the whole data gets also send to the Microsoft Marketing
> section at the same time...:-) In fact, the Computer could probably
> easily detect how intelligent you are, measuring how long it takes you
> to learn to use a program (identifying a user by his typing behaviour
> has already been done), and send the data to your employer.
> Have things like that ever been used in games?
This is illegal and immoral, and also non sensical. Even if they couldtry to
come up with a formula for how fast your reaction time to
messages are, how do they know your making an intelligent choice?
Maybe you just answer yes to everything. Once again this would
take more AI than the best right now has, and probably more processing
power than exists for those AI's as well. This is very subtle stuff, IQ
tests are pretty bogus as well, and if your trying to determine someones
IQ not even from a list of questions but just from day to day reactions to
messages the system produces your going to be making it that much
exponentially harder on yourself.
> I think the character creation of the Ultimas where really
> interesting, when you went to the gypsy and had to answer certain
> questions to determine which character type you are. Any other
> examples?
>
> Bjoern
This is nothing like a self-creating game, thats simply an interface for:
I am a thief, I am a fighter, I am a very brave person, I like the color blue.
Nothing else, alos, randomly craeted monsters, does not make a
self-creating game. The rules have already been set up, its just
throwing things at random.
This is like saying a randomly generated maze is a self creating game.
Its not,. its a maze game EVERY time, the rules never change, simply
which direction you need to go.
The BEST way to simulate a self creating game is to build flexible
rules, rules that can be put together at different times. Such as
when youll be able to do certain things. Or having rules such as
2 out of 3 things have to be accomplished.
This means you dont ALWAYS have to do each thing, and can theorhetically
take different paths of actions.
Plus if you throw in events that are pre-created in a random order, and give
them different levels of merit, meaning one time the whole point of the game
may be to slay the dragon, the next time you play, you can still slay the
dragon
but its only a stepping stone to recovering a clue for recovering a lost
artifact.
This presents different ways of playing, in different orders, but the rules are
still the same in each context. This is the best that can be done now...
-G. Howland
jem...@geocities.com
On Tue, 24 Mar 1998 11:34:23 GMT, guenz...@usa.net (Bjoern Guenzel)
wrote:
>"Karl Low" <kw...@deletethis.cadvision.com> wrote:
>[...]
>>The kicker is though.. is it Fun? What made Giant's Drink so interesting is
>>that the computer was taking stuff from Ender's files, psychological
>>studies, previous work, family history, etc. and putting that all together
>>to make a game specifically tied to Ender.
>
>I read an article about the israelic game 'virus', which would
>apparently scan the conten of your Computer and then use the items it
>found later in the game. Apparently the game was a flop, but I think
>the idea sounded interesting. And after all, a personal Computer in
>most cases does contain a lot of data about it's user.
>Yes, I know - the problem is still that the Computer doesn't
>understand the data and therefore can hardly use it in interesting
>ways. Although I guess the Computer could determine how important
>certain snippets of data are for you (depending on how often you use a
>certain program or read a certain document or whatever). Maybe it
>could even come up with a simple psychological profile by determining
>how forceful or hectically you move the mouse-pointer....:-) Seems as
>if todays Computer COULD get loads of psychological info abbout their
>user if they wanted to.
[snip]
>In fact, the Computer could probably
>easily detect how intelligent you are, measuring how long it takes you
>to learn to use a program (identifying a user by his typing behaviour
>has already been done), and send the data to your employer.
The problem with this is that you'd have to keep track of computer use
over a *very* long time, so the game wouldn't be playable until this
program had sat in the background, psychoanalzying your computer usage
for three months.
I think it's possible for a game to construct some sort of a
psychological profile, though, as I mentioned in a previous message:
have an initial set of challenges that build your psychological score
based on how you react. If you try to kill if possible, pump up the
aggressiveness. If you try to talk your way out, increase the
"diplomaticness" score. And so forth.
Paul 'Ozymandias' Harman
Geoff Howland wrote in message <3517F307...@ntr.net>...
>This is not possible, this is not how computers work, they dont log
howoften
>you use any program or access any files. At the MOST you can
>find out the last time a file was SAVED, and possibly the last time an
>application was launched (such as a looking in the document folder
>on Win95 and seeing the assoicated programs to the files, or on the
>mac I think it actaully has a last used applications deal), but still
>this is trivial knowledge, and someone could realistically just have
>used a program they use only 6 months and then you make it a
>major feature?
I think that NT *does* log this information, actually.
>> In fact, now I seem to remember that Microsoft is already working on
>> this, programs detecting when the user gets angry or nervous and
>> reacting with more help messages etc.
>
>This is totally ridiculous, they MAY have something like this, whichI
doubt,
>but their not going to be getting it on anyones computer in
>the next 50 years, IMO.
But it's easy to do these sort of checks. Somebody who has smooth mouse
movements is calm; somebody who has fast, jerky mouse movements is irate.
Ever know anyone who slams the keyboard when they are annoyed? If the
computer detects you have just hit a large number of keys extremely quickly
(faster than you'd normally type), it can assume you've punched the
keyboard. If the user is spending a lot of time browsing a program's menus
and/or dialog boxes without actually *doing* anything, the application can
assume that the user is "lost", and can offer to help.
Dead easy.
>> I suppose the whole data gets also send to the Microsoft Marketing
>> section at the same time...:-) In fact, the Computer could probably
>> easily detect how intelligent you are, measuring how long it takes you
>> to learn to use a program (identifying a user by his typing behaviour
>> has already been done), and send the data to your employer.
>> Have things like that ever been used in games?
>
>This is illegal and immoral, and also non sensical. Even if they couldtry
to
>come up with a formula for how fast your reaction time to
>messages are, how do they know your making an intelligent choice?
Sending the data to the employer is immoral (but probably not illegal),
yes... but what is to stop the program judging the "skill" of the user by
their behaviour, and tempering it's help prompts accordingly? To try and
drag this back on-topic, there are several games that I've heard of that
tailor the game difficulty according to how well the player is doing. In
what way is it different for a WP application to give me more help if I
appear 'lost'?
>Maybe you just answer yes to everything. Once again this would
>take more AI than the best right now has, and probably more processing
>power than exists for those AI's as well. This is very subtle stuff, IQ
>tests are pretty bogus as well, and if your trying to determine someones
>IQ not even from a list of questions but just from day to day reactions to
>messages the system produces your going to be making it that much
>exponentially harder on yourself.
The sort of scheme I have outlined is not beyond the wit of man. I studied
A.I at university and got within a hare's whisker of doing a PhD on
*precisely* what we are talking about here (well, not the IQ test stuff). So
this sort of thing (if it doesn't actually exist already) is not rocket
science.
>The BEST way to simulate a self creating game is to build flexible
>rules, rules that can be put together at different times. Such as
>when youll be able to do certain things. Or having rules such as
>2 out of 3 things have to be accomplished.
>
>This means you dont ALWAYS have to do each thing, and can theorhetically
>take different paths of actions.
>
>Plus if you throw in events that are pre-created in a random order, and
give
>them different levels of merit, meaning one time the whole point of the
game
>may be to slay the dragon, the next time you play, you can still slay the
>dragon
>but its only a stepping stone to recovering a clue for recovering a lost
>artifact.
>
>This presents different ways of playing, in different orders, but the rules
are
>still the same in each context. This is the best that can be done now...
I have sympathy for this point of view, it must be said. Certainly this is
the best that has been done at the moment. The trouble with scanning the
hard disk of the user for information is that so much of it is probably
irrelevant - even the few 'personal' files that the user has (if the game
can find them) will probably be outside the 'context' of the game (in what
way would Quake use my best friend's phone number, for example?).
I have a feeling that "self-creating" games, if we went about it this way,
would feel like those personalised books you used to get when you were a
kid. You remember - your parents wrote off for the book telling the
publisher what your name and your best friend's name was, and these names
were used as the lead characters in the story.
Ozzy.
--
+-+ Paul Ian Harman +-+-+-+-+-+-+- Ozzy +-+ Games Guru & Sci-Fi Admirer +-+
-+- oz...@kasterborus.demon.co.uk +-+ http://www.kasterborus.demon.co.uk -+-
Geoff Howland
Paul 'Ozymandias' Harman wrote:
> Geoff Howland wrote in message <3517F307...@ntr.net>...
> >This is not possible, this is not how computers work, they dont log
> howoften
> >you use any program or access any files. At the MOST you can
> >find out the last time a file was SAVED, and possibly the last time an
> >application was launched (such as a looking in the document folder
> >on Win95 and seeing the assoicated programs to the files, or on the
> >mac I think it actaully has a last used applications deal), but still
> >this is trivial knowledge, and someone could realistically just have
> >used a program they use only 6 months and then you make it a
> >major feature?
>
> I think that NT *does* log this information, actually.
NT *MAY* log the last programs run, or the programs you runmost on an everage,
but when you access files?
Absolutely not. How many times youve run each program? No way.
They would show that information somewhere, and Ive never seen it.
> >> In fact, now I seem to remember that Microsoft is already working on
> >> this, programs detecting when the user gets angry or nervous and
> >> reacting with more help messages etc.
> >
> >This is totally ridiculous, they MAY have something like this, whichI
> doubt,
> >but their not going to be getting it on anyones computer in
> >the next 50 years, IMO.
>
> But it's easy to do these sort of checks. Somebody who has smooth mouse
> movements is calm; somebody who has fast, jerky mouse movements is irate.
> Ever know anyone who slams the keyboard when they are annoyed? If the
> computer detects you have just hit a large number of keys extremely quickly
> (faster than you'd normally type), it can assume you've punched the
> keyboard. If the user is spending a lot of time browsing a program's menus
> and/or dialog boxes without actually *doing* anything, the application can
> assume that the user is "lost", and can offer to help.
>
> Dead easy.
That whole statement is COMPLETELY ridiculous. If you give ANYpsychoatrist
only a keyboard and mouse to determine what the mood
of ANY individual is theyll be totally lost. Saying someone may move
the mouse shaky when their upset is an ENORMOUS over simplification
and really pretty out there. Dont people use things at different speeds
for other reasons then being angry or upset? Like they have to go somewhere?
Or their sick of sitting at the computer? How many thousands of reasons
could there be for that? They drank to much jolt and now their shaky? They
havnt slept enough and their tired? How are you going to calculate all of
these details from ONLY a keyboard and a mouse? Or were you just planning
on ignoring these?
What if theres a problem with the mouse, if its dirty? What if they spilled
something on the keys and their stuck? Your going to account for that?
The last time I checked there is no way to tell the difference between fast
typing and POUNDING on the keyboard, no PSI info is passed along in
any input routine Ive seen. So really, you have no idea whats going on.
As I said, the BEST you can do is track their response time. What if
the users 5 year old is persistantly asking about every move they are
making? How does your magical AI account for this?
I can tell you that any program that INTERRUPTED me while I was looking
for something or working "too slowly" would be the first one to hit the
trash bin. That would be INCREDIBLY annoying. How about you just
design a GOOD interface, so that the user doesnt have to search like a
blind man.
So as far as Ive stated does this still sound, "dead easy" to you?
It looks a little more like "dead impossible" to me.
> > I suppose the whole data gets also send to the Microsoft Marketing
> >> section at the same time...:-) In fact, the Computer could probably
> >> easily detect how intelligent you are, measuring how long it takes you
> >> to learn to use a program (identifying a user by his typing behaviour
> >> has already been done), and send the data to your employer.
> >> Have things like that ever been used in games?
> >
> >This is illegal and immoral, and also non sensical. Even if they couldtry
> to
> >come up with a formula for how fast your reaction time to
> >messages are, how do they know your making an intelligent choice?
>
> Sending the data to the employer is immoral (but probably not illegal),
> yes... but what is to stop the program judging the "skill" of the user by
> their behaviour, and tempering it's help prompts accordingly? To try and
> drag this back on-topic, there are several games that I've heard of that
> tailor the game difficulty according to how well the player is doing. In
> what way is it different for a WP application to give me more help if I
> appear 'lost'?
Its illegal where I live, its called invasion of privacy. They could get
yourcredit cards like this too, couldnt they?
Help prompts dont need to pop-up if the interface is designed correctly.
Games like this have been around for YEARS, but this is NOT a self
creating game, and this is NOT AI. I dont know HOW you came to
this conclusion, but this is simply:
Is the player losing? Make it easier.
Is the player winning? Make it harder.
> >Maybe you just answer yes to everything. Once again this would
> >take more AI than the best right now has, and probably more processing
> >power than exists for those AI's as well. This is very subtle stuff, IQ
> >tests are pretty bogus as well, and if your trying to determine someones
> >IQ not even from a list of questions but just from day to day reactions to
> >messages the system produces your going to be making it that much
> >exponentially harder on yourself.
>
> The sort of scheme I have outlined is not beyond the wit of man. I studied
> A.I at university and got within a hare's whisker of doing a PhD on
> *precisely* what we are talking about here (well, not the IQ test stuff). So
> this sort of thing (if it doesn't actually exist already) is not rocket
> science.
I honestly dont believe you came ANYwhere NEAR *EARNING* a PhDin this subject.
The examples you have given are way out there and you
dont seem to have a real grasp on things any more than someone who
just picked up a science fiction book without ever seeing a computer.
I dont mean to be derrogatory to you, but I cant see any other way to say
this, and your arrogant demeanor at how simple this is makes it even
more impossible for me not to be critical. You speak like you know what
your doing, but you havnt made one even decent argument for why this
is so.
This type of thing, in reality IS a rocket science. I am really blown away
by you not understanding this. You are attempting to do a complex
psychological examation to determine a users FEELINGS and STATE
OF MIND at any given time, given 2 input devices, which are SO
LIMITED in giving information that they really cant even be USED
for this purpose. How do you even know if its the SAME PERSON
whos at the computer? You dont.
I really am blown away by HOW OUT THERE this is.
> >The BEST way to simulate a self creating game is to build flexible
> >rules, rules that can be put together at different times. Such as
> >when youll be able to do certain things. Or having rules such as
> >2 out of 3 things have to be accomplished.
> >
> >This means you dont ALWAYS have to do each thing, and can theorhetically
> >take different paths of actions.
> >
> >Plus if you throw in events that are pre-created in a random order, and
> give
> >them different levels of merit, meaning one time the whole point of the
> game
> >may be to slay the dragon, the next time you play, you can still slay the
> >dragon
> >but its only a stepping stone to recovering a clue for recovering a lost
> >artifact.
> >
> >This presents different ways of playing, in different orders, but the rules
> are
> >still the same in each context. This is the best that can be done now...
>
> I have sympathy for this point of view, it must be said. Certainly this is
> the best that has been done at the moment. The trouble with scanning the
> hard disk of the user for information is that so much of it is probably
> irrelevant - even the few 'personal' files that the user has (if the game
> can find them) will probably be outside the 'context' of the game (in what
> way would Quake use my best friend's phone number, for example?).
Obviously something that is going to grab your info will not be a FPS, itwould
be a game that has to do with peoples interactions in some way
and would build a story around this info. Personally I think this is stupid
I just gave how it could and couldnt be done.
> I have a feeling that "self-creating" games, if we went about it this way,
> would feel like those personalised books you used to get when you were a
> kid. You remember - your parents wrote off for the book telling the
> publisher what your name and your best friend's name was, and these names
> were used as the lead characters in the story.
That wouldnt be a self creating game, that would be a game with your namein it,
like MOST games that let you chose your characters name.
Those books arnt personalized books, they dont have ANYTHING about you
except your name, its crap.
> Ozzy.
>
> --
> +-+ Paul Ian Harman +-+-+-+-+-+-+- Ozzy +-+ Games Guru & Sci-Fi Admirer +-+
> -+- oz...@kasterborus.demon.co.uk +-+ http://www.kasterborus.demon.co.uk -+-
I came down on you pretty hard, but I dont have much simpathy for things
that dont make sense. I would have had more if you had been less
arrogant and acting like you know what your doing. But as far as I can see you
really dont. I dont know what you learned in school, and I dont know what you
do, and really. It doesnt matter, the points youve given are invalid, and
the ideas you claim to be simple are impossible.
If you can prove me wrong, terrific, but Id bet my liver that you cant.
-G. Howland
Dave Gileadi
Bjoern Guenzel wrote in message
<890739017.4730.3...@news.demon.co.uk>...
>"Karl Low" <kw...@deletethis.cadvision.com> wrote:
>[...]
>>The kicker is though.. is it Fun? What made Giant's Drink so interesting
is
>>that the computer was taking stuff from Ender's files, psychological
>>studies, previous work, family history, etc. and putting that all together
>>to make a game specifically tied to Ender.
>
>I read an article about the israelic game 'virus', which would
>apparently scan the conten of your Computer and then use the items it
>found later in the game. Apparently the game was a flop, but I think
>the idea sounded interesting. And after all, a personal Computer in
>most cases does contain a lot of data about it's user.
>Yes, I know - the problem is still that the Computer doesn't
>understand the data and therefore can hardly use it in interesting
>ways. Although I guess the Computer could determine how important
>certain snippets of data are for you (depending on how often you use a
>certain program or read a certain document or whatever). Maybe it
>could even come up with a simple psychological profile by determining
>how forceful or hectically you move the mouse-pointer....:-) Seems as
>if todays Computer COULD get loads of psychological info abbout their
>user if they wanted to.
[snip]
I don't think I'd call this a self-creating game, though it has potential to
adapt to your habits and thus "create itself" in a sense.
The thing that gets me, though, is an idea I'd been thinking about for a
while...
What if you start playing a game, a fairly realistic game that pulled you
into the plot and had lots of tension, tons of intrigue, etc. You would
play some sort of hero, maybe a detective, and there would be a chilling
villian, evil henchmen, the lot.
Suppose you began to "fear" the characters (not really, but like you do in a
movie). Then, a few days into the game you suddenly got an e-mail from one
of the henchmen, telling you to lay off the case, you're life might depend
on it. Say it is followed by a terse e-mail from the boss himself, asking
you to cut a deal--strongly advising you to take a check and go retire in
the bahamas...
Anyway, you probably get the drift. The point is, a simple mechanism like
this (e-mail coming from your own computer), as well as maybe finding other
pertainent data from the hard drive, could really blur the lines between
fact and fiction.
Kind of scary, huh?
-Dave
Aldin Larice
>> I read an article about the israelic game 'virus', which would
>> apparently scan the conten of your Computer and then use the items it
>> found later in the game. Apparently the game was a flop, but I think
>> the idea sounded interesting. And after all, a personal Computer in
>> most cases does contain a lot of data about it's user.
>> Yes, I know - the problem is still that the Computer doesn't
>> understand the data and therefore can hardly use it in interesting
>> ways. Although I guess the Computer could determine how important
>> certain snippets of data are for you (depending on how often you use a
>> certain program or read a certain document or whatever).
>
>This is not possible, this is not how computers work, they dont log howoften
>you use any program or access any files. At the MOST you can
>find out the last time a file was SAVED, and possibly the last time an
>application was launched (such as a looking in the document folder
>on Win95 and seeing the assoicated programs to the files, or on the
>mac I think it actaully has a last used applications deal), but still
>this is trivial knowledge, and someone could realistically just have
>used a program they use only 6 months and then you make it a
>major feature?
Sure it's possible, if you had a program tracking computer usage in
the background over a period of time.
>Also, AI is not an option now, the only working AI's I know of have
>VERY specific topics of "thought" such as playing chess, to even
>imagine throwing any sorts of random data at a program is simply
>trying to make science fiction a reality, for now. The best and only
>way to do this would be for the programmer to go through and make
>up routines on things he knows of. Such as, he can scan through
>data looking for addresses or phone number and try to attach them
>to names. Thats relevant and can look like the program is acting
>intelligently...
*BZZT* wrong, sorry. Not only are there plenty of routines for
generalized pattern matching, simple learning algorithms could do what
we're talking about here, no problem.
>> Maybe it
>> could even come up with a simple psychological profile by determining
>> how forceful or hectically you move the mouse-pointer....:-) Seems as
>> if todays Computer COULD get loads of psychological info abbout their
>> user if they wanted to.
>> In fact, now I seem to remember that Microsoft is already working on
>> this, programs detecting when the user gets angry or nervous and
>> reacting with more help messages etc.
>
>This is totally ridiculous, they MAY have something like this, whichI doubt,
>but their not going to be getting it on anyones computer in
>the next 50 years, IMO.
I've heard of this too.
Look at it this way: if someone is trying a number of options,
successively, quickly, the program can infer that the user has a real
problem and needs comprehensive help. If, on the other hand, the user
never uses certain basic routines (Find, CTRL-HOME, etc.), the program
can infer that the user is probably "afraid" of the computer, and
extra tips might be well-received.
Not "totally ridiculous".
>> I suppose the whole data gets also send to the Microsoft Marketing
>> section at the same time...:-) In fact, the Computer could probably
>> easily detect how intelligent you are, measuring how long it takes you
>> to learn to use a program (identifying a user by his typing behaviour
>> has already been done), and send the data to your employer.
>> Have things like that ever been used in games?
>
>This is illegal and immoral, and also non sensical. Even if they couldtry to
>come up with a formula for how fast your reaction time to
>messages are, how do they know your making an intelligent choice?
What's illegal about keeping track of how a person uses a computer?
It's what that stupid "Clippy" in Office 97 does.
Why would someone make a lot of rapid, non-intelligent choices? If
someone starts only using very simple pieces of an application, then
within a few weeks is using very sophisticated pieces, it can be
inferred that that person is intelligent.
>Maybe you just answer yes to everything. Once again this would
>take more AI than the best right now has, and probably more processing
>power than exists for those AI's as well. This is very subtle stuff, IQ
>tests are pretty bogus as well, and if your trying to determine someones
>IQ not even from a list of questions but just from day to day reactions to
>messages the system produces your going to be making it that much
>exponentially harder on yourself.
This would be very hard to program, yes.
>> I think the character creation of the Ultimas where really
>> interesting, when you went to the gypsy and had to answer certain
>> questions to determine which character type you are. Any other
>> examples?
>>
>> Bjoern
>
>This is nothing like a self-creating game, thats simply an interface for:
>
>I am a thief, I am a fighter, I am a very brave person, I like the color blue.
>
>Nothing else, alos, randomly craeted monsters, does not make a
>self-creating game. The rules have already been set up, its just
>throwing things at random.
>
>This is like saying a randomly generated maze is a self creating game.
>Its not,. its a maze game EVERY time, the rules never change, simply
>which direction you need to go.
>
>The BEST way to simulate a self creating game is to build flexible
>rules, rules that can be put together at different times. Such as
>when youll be able to do certain things. Or having rules such as
>2 out of 3 things have to be accomplished.
>
>This means you dont ALWAYS have to do each thing, and can theorhetically
>take different paths of actions.
But this isn't a self-creating game. It's simply a game with
different possible ways to win.
This is *part* of having a self-creating game. But not "The BEST way
to simulate" one.
>Plus if you throw in events that are pre-created in a random order, and give
>them different levels of merit, meaning one time the whole point of the game
>may be to slay the dragon, the next time you play, you can still slay the
>dragon
>but its only a stepping stone to recovering a clue for recovering a lost
>artifact.
>
>This presents different ways of playing, in different orders, but the rules are
>
>still the same in each context. This is the best that can be done now...
>
>-G. Howland
>
>
Aldin Larice
Geoff Howland
> I don't think I'd call this a self-creating game, though it has potential to
> adapt to your habits and thus "create itself" in a sense.
>
> The thing that gets me, though, is an idea I'd been thinking about for a
> while...
>
> What if you start playing a game, a fairly realistic game that pulled you
> into the plot and had lots of tension, tons of intrigue, etc. You would
> play some sort of hero, maybe a detective, and there would be a chilling
> villian, evil henchmen, the lot.
>
> Suppose you began to "fear" the characters (not really, but like you do in a
> movie). Then, a few days into the game you suddenly got an e-mail from one
> of the henchmen, telling you to lay off the case, you're life might depend
> on it. Say it is followed by a terse e-mail from the boss himself, asking
> you to cut a deal--strongly advising you to take a check and go retire in
> the bahamas...
>
> Anyway, you probably get the drift. The point is, a simple mechanism like
> this (e-mail coming from your own computer), as well as maybe finding other
> pertainent data from the hard drive, could really blur the lines between
> fact and fiction.
>
> Kind of scary, huh?
>
> -Dave
I dont think its scary, it would be an interesting feature to the game
though,
giving it more importance in your life, because you may get a message through
email. This can also be annoying and should be used sparingly.
-G. Howland
Geoff Howland
> >This is not possible, this is not how computers work, they dont log howoften
> >you use any program or access any files. At the MOST you can
> >find out the last time a file was SAVED, and possibly the last time an
> >application was launched (such as a looking in the document folder
> >on Win95 and seeing the assoicated programs to the files, or on the
> >mac I think it actaully has a last used applications deal), but still
> >this is trivial knowledge, and someone could realistically just have
> >used a program they use only 6 months and then you make it a
> >major feature?
>
> Sure it's possible, if you had a program tracking computer usage in
> the background over a period of time.
This is the second time Ive been told I was wrong with NO back up.
What program is going to track this? What are they going to record?
Your going to run your game that needs this info in the background
for 6 months, and record thousans of file transations and their
dates for this game? Is this in any sense a realistic idea?
> >Also, AI is not an option now, the only working AI's I know of have
> >VERY specific topics of "thought" such as playing chess, to even
> >imagine throwing any sorts of random data at a program is simply
> >trying to make science fiction a reality, for now. The best and only
> >way to do this would be for the programmer to go through and make
> >up routines on things he knows of. Such as, he can scan through
> >data looking for addresses or phone number and try to attach them
> >to names. Thats relevant and can look like the program is acting
> >intelligently...
>
> *BZZT* wrong, sorry. Not only are there plenty of routines for
> generalized pattern matching, simple learning algorithms could do what
> we're talking about here, no problem.
Once again Id like to see this backed up. Point out what kind of algorithmcould
look through any number of files on your comptuer and match up cases.
I maintain that there isnt enough information given the normal user to MATCH
any patterns.
Say they have a phone number list on there computer, and they have a few
documents for work, and they have their browser favorites and they have
some financial data and maybe a bunch of emails which contain names
of people they work with.
This is going to be just a HANDFUL of data, this isnt enough for any program
that isnt SPECIFICALLY designed to just look for these specific things,
which is what I was saying is possible.
Simple learning algorithsm dont do this type of thing, ive seen them, ive
used them in demonstation of learning programs, genetic evolution, expert
system, they dont work off a few fragments of data.
> >> Maybe it
> >> could even come up with a simple psychological profile by determining
> >> how forceful or hectically you move the mouse-pointer....:-) Seems as
> >> if todays Computer COULD get loads of psychological info abbout their
> >> user if they wanted to.
> >> In fact, now I seem to remember that Microsoft is already working on
> >> this, programs detecting when the user gets angry or nervous and
> >> reacting with more help messages etc.
> >
> >This is totally ridiculous, they MAY have something like this, whichI doubt,
> >but their not going to be getting it on anyones computer in
> >the next 50 years, IMO.
>
> I've heard of this too.
>
> Look at it this way: if someone is trying a number of options,
> successively, quickly, the program can infer that the user has a real
> problem and needs comprehensive help. If, on the other hand, the user
> never uses certain basic routines (Find, CTRL-HOME, etc.), the program
> can infer that the user is probably "afraid" of the computer, and
> extra tips might be well-received.
>
> Not "totally ridiculous".
Let me give you a few scenarios and you tell me how your mystery algorithmis going
to handle it.
A user likes to use the menus, hes comfortable with the mouse, and so
never uses CNTRL-HOME. Should this user always be considered a
novice user by the program?
Expanding on this, the user is well accustomed to his menus, so can move
through them quickly, does this mean hes nervous? No.
I gave examples of hardware failures (ie the mouse ball is sticky), distractions
with other things in the room, someone being sick. All of these things
are going to influence your data and cant be accounted for.
How are you accounting for them, how am I wrong about this?
> >This is illegal and immoral, and also non sensical. Even if they couldtry to
> >come up with a formula for how fast your reaction time to
> >messages are, how do they know your making an intelligent choice?
>
> What's illegal about keeping track of how a person uses a computer?
> It's what that stupid "Clippy" in Office 97 does.
>
> Why would someone make a lot of rapid, non-intelligent choices? If
> someone starts only using very simple pieces of an application, then
> within a few weeks is using very sophisticated pieces, it can be
> inferred that that person is intelligent.
Intelligence has NOTHING to do with knowing how to use a more"sophisticated"
function. This is ridiculous. Thats just knowledge, and
any idiot can pick up a manual. Suppose the user only needed the program
to DO this ONE thing, so they learned that in the first few days. Later
the user needs to do others, but is totally lost on what to do, are
they now to be considered a "intelligent" and "procifient" use because
they used a sophisticated function a few times?
Everything your presenting is just simplified conjecture. I have always
turned that little paper clip off, but from what I can tell about it, it
pretty much gives you any information about whatever your doing.
This isnt intelligent, this isnt AI.
> >Maybe you just answer yes to everything. Once again this would
> >take more AI than the best right now has, and probably more processing
> >power than exists for those AI's as well. This is very subtle stuff, IQ
> >tests are pretty bogus as well, and if your trying to determine someones
> >IQ not even from a list of questions but just from day to day reactions to
> >messages the system produces your going to be making it that much
> >exponentially harder on yourself.
>
> This would be very hard to program, yes.
This just is NOT feasible now, which is what Ive been saying. You donthave the
tools to collect the input, you dont have a controlled environment,
you dont have ANY of the things necessary to complete any kinds of
real judgement, the best your doing is guess work, and its the programmers
guess work, not a learning machine.
> >This is nothing like a self-creating game, thats simply an interface for:
> >
> >I am a thief, I am a fighter, I am a very brave person, I like the color blue.
> >
> >Nothing else, alos, randomly craeted monsters, does not make a
> >self-creating game. The rules have already been set up, its just
> >throwing things at random.
> >
> >This is like saying a randomly generated maze is a self creating game.
> >Its not,. its a maze game EVERY time, the rules never change, simply
> >which direction you need to go.
> >
> >The BEST way to simulate a self creating game is to build flexible
> >rules, rules that can be put together at different times. Such as
> >when youll be able to do certain things. Or having rules such as
> >2 out of 3 things have to be accomplished.
> >
> >This means you dont ALWAYS have to do each thing, and can theorhetically
> >take different paths of actions.
>
> But this isn't a self-creating game. It's simply a game with
> different possible ways to win.
>
> This is *part* of having a self-creating game. But not "The BEST way
> to simulate" one.
I didnt say this is a self creating game, I said its the best you can do withthat
now and a rough example at that. Its not detailed, its not precise, so
the IDEA of it is the best way to simulate it, thats true, the programmer
makes the game look like its creating itself each time, when its really
not, its just mixing things together randomly with a clever gluing to make
it seemless.
Truly self creating games, as I said, make their own rules, and this just
isnt possible now in any kind of intelligent fashion.
> Aldin Larice
-G. Howland
Jlun...@jlphoto.com
>This is not possible, this is not how computers work, they dont log howoften
>you use any program or access any files. At the MOST you can
>find out the last time a file was SAVED, and possibly the last time an
>application was launched (such as a looking in the document folder
>on Win95 and seeing the assoicated programs to the files, or on the
>mac I think it actaully has a last used applications deal), but still
>this is trivial knowledge, and someone could realistically just have
>used a program they use only 6 months and then you make it a
>major feature?
This isn't a problem. Assuming this game isn't something that you use
once and are done with. (If it is, perhaps useing recently accessed
files ~would~ be a good idea. They're fresh in the player's mind.)
Every time the game starts check the "Last Accessed" time/date of the
files.(I'm talking Win95 here.) And find out how long ago it was used.
Then keep a running average time-since-last-accessed and use that. Of
coarse you'd have to restrict what types of files so system files
didn't come out on top of the list every time. (And the game's own
files.)
Of coarse if the computer has more than one user then there are
privicy problems that would have to be delt with.
-Andy
Geoff Howland
> This isn't a problem. Assuming this game isn't something that you use
> once and are done with. (If it is, perhaps useing recently accessed
> files ~would~ be a good idea. They're fresh in the player's mind.)
> Every time the game starts check the "Last Accessed" time/date of the
> files.(I'm talking Win95 here.) And find out how long ago it was used.
> Then keep a running average time-since-last-accessed and use that. Of
> coarse you'd have to restrict what types of files so system files
> didn't come out on top of the list every time. (And the game's own
> files.)
>
> Of coarse if the computer has more than one user then there are
> privicy problems that would have to be delt with.
>
> -Andy
You also wouldnt know if it is the persons file or not. And unless, like
I said, the programmer specifically designed the program to look for say,
phone numbers, names, email addresses, things of this SPECIFIC
nature, your not going to get a learning algorithm, like has been suggested,
to figure out on its own what this data means, at all.
This is the point, I obviously knew you could scan the whole hard drive looking
for freshly WRITTEN files, but this doesnt give you a good sample of data
to do any sort of learning system on. It only does any good if the program
searches for specific data.
-G. Howland
Jlun...@jlphoto.com
>This is the point, I obviously knew you could scan the whole hard drive looking
>for freshly WRITTEN files, but this doesnt give you a good sample of data
Windows95 keeps track of recently ACCESSED files as well.
-Andy
Michael Kissin
Geoff Howland wrote:
> > I read an article about the israelic game 'virus', which would
> > apparently scan the conten of your Computer and then use the items it
> > found later in the game. Apparently the game was a flop, but I think
> > the idea sounded interesting. And after all, a personal Computer in
> > most cases does contain a lot of data about it's user.
> > Yes, I know - the problem is still that the Computer doesn't
> > understand the data and therefore can hardly use it in interesting
> > ways. Although I guess the Computer could determine how important
> > certain snippets of data are for you (depending on how often you use a
> > certain program or read a certain document or whatever).
>
Hmmm, I hope I am thinking about the same game here...
That is _almost_ what the game did. It scans your HDD and then you fight a computer
"virus" through your own file system, trying to eradicate it and most importantly,
trying to stop it from infecting the key files in your root directory and you
windows directory.
The game would use the bitmaps/gifs that it found on your HDD as textures in the
game and your MIDI and WAV files as background sound/music.
That's is all I know except that I don't actually think that it has been realeased
yet. Or at least, not in New Zealand.
Maybe I am thinking of a newer version.
Michael