Computer writes a book?
Ted Pedersen
I saw the following in the Books column (by Olin Chism) of the Dallas
Morning News Sunday January 24, 1993. Copied without permission. It
concerns a computer program that allegedly wrote a book in the style
of Jacqueline Susann.
I think this is a stunt. I don't really believe a computer program
wrote this book. However, if anyone has any information about this
book or the author that would give this some credibility I would like
to hear about it. Comment on the general idea would be interesting as
well. I don't think the idea itself is impossible, it just seems like
this case is a little far fetched.
-----------------------------------------------------------------------
"Birch Lane Press of New York says that Scott French, a
free-lance writer and computer consultant from California, invested
eight years and $50,000 to develop a computer program that writes like
Jaqueline Susann. Mr. French's electronic novelist has output Just
This Once, a 256 page sizzler that Birch Lane will publish next
summer."
"The first thought of any good skeptic is that Birch Lane is
pulling a fast one. Remember "Naked Came the Stranger", the potboiler
that Lyle Stuart published back in the 60's? And how Penelope Ashe,
its author, turned out to be a whole committe of jokesters? Well, Lyle
Stuart is a subsidiary of Carol Publishing Group, which also owns
Birch Lane."
"Suspicians are not allayed by the tone of the commentary in
Birch Lane's catalog: "Scott French, a self-professed fan of
mega-selling author Jacqueline Susann, imagined what it would be like
if she were still alive. Secretly he hoped that, following her
premature death, Susann had be frozen, a la Walt Disney. The doctors
could thaw her out by setting a giant microwave to the Low/Defrost
setting, then cure her cancer and put her back in front of the
typewriter where she belonged...""
---
* Ted Pedersen pede...@seas.smu.edu *
* Department of Computer Science and Engineering, *
* Southern Methodist University, Dallas, TX 75275 (214) 768-2126 *
SubGenius
>I think this is a stunt. I don't really believe a computer program
>wrote this book. However, if anyone has any information about this
>book or the author that would give this some credibility I would like
>to hear about it. Comment on the general idea would be interesting as
>well. I don't think the idea itself is impossible, it just seems like
>this case is a little far fetched.
[deletia]
+-------------------------------------SubG---------------------------------+
There's at least one book authored by a computer I can think of--a thinnish
volume that came out six or eight years back called _The Policeman's Beard
is Half Constructed_ and, while most of it was on the ragged edge of
comprehensibility I'd rate it far above any of J.S.'s efforts. Because
although it contained a good deal of repetition due to a conspicuously
small vocabulary, it did deliver occasional jems like, `A black pig is
like a tormented bat,' explained in tortuous logic, and the wild
`inconsistancy is the hobgoblin of little minds,' which leads one to
wonder what trash the programmers had been feeding their machine.
-SubGenius
Eric Behrens
Pedersen) writes:
>
>
> I saw the following in the Books column (by Olin Chism) of the Dallas
> Morning News Sunday January 24, 1993. Copied without permission. It
> concerns a computer program that allegedly wrote a book in the style
> of Jacqueline Susann.
>
> I think this is a stunt. I don't really believe a computer program
> wrote this book. However, if anyone has any information about this
> book or the author that would give this some credibility I would like
> to hear about it. Comment on the general idea would be interesting as
> well. I don't think the idea itself is impossible, it just seems like
> this case is a little far fetched.
>
I would have to guess it's a stunt as well; strikes me like the odds of a
monkey at a typewriter writing Hamlet...
But you never know. I recall an NPR report perhaps two years ago. It was
about a computer scientist who had spent a decade creating a program which
could compose in the style of Bach.
If you provided a simple melody line, the program would orchestrate an
entire opus for you. I must confess that the results were quite
impressive-- not perfect, but impressive.
But music has only a dozen words in its vocabulary (notes of the scale) and
the English language has tens of thousands. I cannot believe that computer
science is capable of replicating the intricate constructions of language.
Maybe it could do e e cummings, but a novel... :Ñ)
Paul Frankenstein
*
* I saw the following in the Books column (by Olin Chism) of the Dallas
* Morning News Sunday January 24, 1993. Copied without permission. It
* concerns a computer program that allegedly wrote a book in the style
* of Jacqueline Susann.
[stuff taken out]
* -----------------------------------------------------------------------
*
* "Birch Lane Press of New York says that Scott French, a
* free-lance writer and computer consultant from California, invested
* eight years and $50,000 to develop a computer program that writes like
* Jaqueline Susann. Mr. French's electronic novelist has output Just
* This Once, a 256 page sizzler that Birch Lane will publish next
* summer."
[rest of article deleted]
* ---
* * Ted Pedersen pede...@seas.smu.edu *
* * Department of Computer Science and Engineering, *
* * Southern Methodist University, Dallas, TX 75275 (214) 768-2126 *
I remember hearing something like this a while back; it might have been in
the SF Chronicle (for some reason, I associate it with one of my vists to the
Bay area). However, the name Scott French does check out (in memory, at
least); he wrote a how-to book back in the mid-seventies called "The Big
Brother Book." It was something along the lines of the Anarchist's Cookbook,
showing you how to pick locks, build bugs, etc, etc. (Although I don't remember
anything about building bombs in it.)
The book might be legit; but then again, it might not...
paul
--
fran...@gas.uug.arizona.edu
"Bagels are great for backpacking."
-- Mike Cottingham
Jack Shirazi <js@biu.icnet.uk>
>Morning News Sunday January 24, 1993. Copied without permission. It
>concerns a computer program that allegedly wrote a book in the style
>of Jacqueline Susann.
>
>I think this is a stunt. I don't really believe a computer program
>wrote this book. However, if anyone has any information about this
>book or the author that would give this some credibility I would like
>to hear about it. Comment on the general idea would be interesting as
>well. I don't think the idea itself is impossible, it just seems like
>this case is a little far fetched.
Markov chaining using a mixture of sentences and words, or possibly
just with words, using a parser to get rid of any unlikely combinations.
Probably added in the use of a style checker.
--
Jack j...@biu.icnet.uk
If you only have a hammer, you tend to see every problem as a nail.
-- Maslow
Erik Max Francis
> I think this is a stunt. I don't really believe a computer program
> wrote this book. However, if anyone has any information about this
> book or the author that would give this some credibility I would like
> to hear about it. Comment on the general idea would be interesting as
> well. I don't think the idea itself is impossible, it just seems like
> this case is a little far fetched.
I don't know about this particular case, but it could in all probability
be real. There are at least two obvious ways to get a computer to
"write" prose; both involve the scanning or analysis of another text.
One involves Markov chains, and the other involves a phonetic dictionary.
Markov chain generators and traverses tend to ramble, but phonetic
dictionaries can make a strange amount of sense at times. In both cases,
it generally requires a human operator to organize the output into
something that makes any sense whatsoever.
I have written a Markov chain generator and traverser, and can run it on
any text that you choose (as long as I have it handy). Email me for more
information.
Erik Max Francis -)(- ..!apple!uuwest!max -)(- m...@west.darkside.com __
USMail: 1070 Oakmont Dr. #1; San Jose, CA 95117 ICBM: 37 20 N 121 53 W / \
Like strategic games of interstellar conquest? . . . Ask about UNIVERSE! \__/
-)(- Omnia quia sunt, lumina sunt; all things that are, are lights. -)(-
Alejandro Rivero
>
->+-------------------------------------SubG---------------------------------+
->There's at least one book authored by a computer I can think of--a thinnish
->volume that came out six or eight years back called _The Policeman's Beard
->is Half Constructed_ and, while most of it was on the ragged edge of
->comprehensibility I'd rate it far above any of J.S.'s efforts. Because
->although it contained a good deal of repetition due to a conspicuously
->small vocabulary, it did deliver occasional jems like, `A black pig is
->like a tormented bat,' explained in tortuous logic, and the wild
->`inconsistancy is the hobgoblin of little minds,' which leads one to
->wonder what trash the programmers had been feeding their machine.
->
that was from RACTER, was it? I read a reference to Racter in Scientific Amer.
time ago, probably in computer games. There was a Racter vs Eliza match, funny,
showing how RACTER routines can make fool a sophisticated parser as Eliza is.
The more important point about racter was the interactive version. He could
speak with a human and take records of conversations for future use. A good
character to include in a game...
I dont know if racter source is avalaible. In fact, I dont know if the program
is avalaible actually from any place. I have heard there are other projects running
in internet, some of then in .ca areas, but I have not info about. Perhaps someone
there.
-Alejandro Rivero
->
->
->
->
->
->
->-SubGenius
->
->
->
Steve Blinkhorn
*does* write stories and *could* write books. At the moment its major
use is in building disaster simulators for crisis management training,
where it writes scenarios and then generates newspaper articles and
the like. It delights in the name of SCIROCCO, which of course is an
acronym: Source Code Integrator for Report Origination with
C-Compilable Objects. A feature on its use appeared in Chemical &
Engineering News around the third week in September 1992.
No, I'm not going to reveal how it's done, but if there is interest I
will post a sample binary, which uses a limited (and early) version to
write 300-word romantic stories in a send-up of the style known in the
UK as Mills & Boon. It can write literally millions of variations,
took about four hours to write and it occupies all of 30Kbytes.
It takes less than a second to run.
Or if there's an ftp site that would like a copy?
BTW, would someone please advise on preferred encoding method for
MS-DOS binaries? I have a hate-hate relationship with MS-DOS........
--
Steve Blinkhorn <st...@prd.co.uk>
Magnus Olsson
>that was from RACTER, was it? I read a reference to Racter in Scientific Amer.
>time ago, probably in computer games. There was a Racter vs Eliza match, funny,
>showing how RACTER routines can make fool a sophisticated parser as Eliza is.
Sophisticated parser? Eliza? Most versions of Eliza are very
simple-minded indeed and don't actually do any parsing, just pattern
matching.
>I dont know if racter source is avalaible. In fact, I dont know if the program
>is avalaible actually from any place.
Racter is a commercial program, so I'd be very surprised if the source
is available. I've seen (and used) a version for the Macintosh; that
was a few years ago and I don't know if they still sell it. Let me
just say that in real "life", Racter isn't quite as good a
conversationalist as one can be led to believe. The texts quoted in
the articles (and ads) are probably edited, or at least the most
interesting parts have been selected.
Magnus Olsson | \e+ /_
Department of Theoretical Physics | \ Z / q
University of Lund, Sweden | >----<
mag...@thep.lu.se, the...@seldc52.bitnet | / \===== g
PGP key available via finger or on request | /e- \q
Oliver Jakobs
>But music has only a dozen words in its vocabulary (notes of the scale) and
>the English language has tens of thousands. I cannot believe that computer
>science is capable of replicating the intricate constructions of language.
I'd say the analogy is not of the right order. I've heard that scales are
more like phonemes or letters, and that there are groups or patterns of
scales which could be regarded as ``words of music''. There has been done
some work trying to apply linguistic knowledge to musical ``texts''. I don't
remember the scientist's name, but I could look up some references if someone
was interested.
Oliver
--
Oliver Jakobs, Dept. of Computational Linguistics, Trier University, Germany
| Internet: jakobs@ldv01@Uni-Trier.de |
+--------------------------------------------------------------------------+
Alejandro Rivero
>Sophisticated parser? Eliza? Most versions of Eliza are very
>simple-minded indeed and don't actually do any parsing, just pattern
>matching.
It seems that some phrases sond sound in engish as in spanish. .
"sohisticated parser" was meant to de despective, the point being that
you can implement eliza style by using some decent parser program (by example
it could be done in a if-programing language).
Jim Kasprzak
|> wrote this book. However, if anyone has any information about this
|> book or the author that would give this some credibility I would like
|> to hear about it. Comment on the general idea would be interesting as
|> well. I don't think the idea itself is impossible, it just seems like
|> this case is a little far fetched.
computers writing books. Here at Rensselaer Polytutescrew Incorporated,
professor Selmer Bringsjord has been working on several different projects
aimed at getting a computer program to produce meaningful stories. Its
primary goals are really the investigation of the philosophical and
computing problems of getting something like this to work, and they haven't
produced any computer-generated novels yet, but I've seen some rudimentary
output. It looks something like children's stories written by Eliza.
------------------------------------------------------------------
__ Live from Capitaland, heart of the Empire State...
___/ | Jim Kasprzak, computer operator @ RPI, Troy, NY, USA
/____ *| "And the men who hold high places must be the ones who start
\_| To mold a new reality -- closer to the heart."
==== e-mail: kas...@rpi.edu or kasp...@mts.rpi.edu
Elena A Christofides
My husband wrote a program that writes poetry. This program is actually in
basic.
In my modern poetry class a while ago, we had to submit up to three poems that
were our own original work. Both my poems got A-/B+, whereas the computers poem
got an A+!
My instructor was none the wiser at the time - it was a little joke to show
that the computer could write a L=A=N=G=U=A=G=E poem as good if not better than
the language poets themselves!
Elena
Fred Welden
Alan Turing (somebody else can provide the bio, I don't have it)
proposed the now-famous Turing Test of artificial intelligence: put a
person in a booth where he can communicate only by teletype. Have him
communicate with two respondents, one a human, one a machine. If he
can't tell which is the machine, the machine is behaving with artificial
intelligence.
Based on this I propose the Welden Test of natural intelligence: put a
person in a booth where he can communicate only by teletype. Have him
communicate with two respondents, one a human, one a machine. If he
can't tell which is the machine, he is behaving without natural
intelligence.
Congratulations to your instructor--he has failed the Welden Test!
--
--Fred, or another blind 8th-century BC | sas...@dobo.unx.sas.com
Hellenic poet of the same name. |
gbut...@desire.wright.edu
Indeed, I would consider notes to be much more synonymous to the
letters of the musical vocabulary. Patterns of such notes, then, could be
the words of the musical language. Finally, phrases (groups of music, usually
in 4 or 8 bars) are the musical sentances.
Although modern music has very few rules for composition, music has in
the past (especially in the Baroque era) had much definition for what was
'acceptable' and what was not (I am pulling this and my musical analogies from
my memory of a course I took in my freshman year specifically on how Mozart
influenced the development of western music).
Regarding Mozart's style, the rules could fill many pages (about 300,
if I remember). Example: A nota cambiata, a 'word' which consists of going
up one step, down one, down another, up one, and back up to the original tone.
However, the rules govern which beat this may begin on (only the first or
third), how the counterpoint must be written, and that the nc must end in
consonance (major 3rd or major 5th) from the lowest note of the bass voice
(I'm not sure if these rules are exactly correct, as that course is long in
the past; I am using this as an exanmple). This is simply one word out of
several hundred, with rules governing how they may be strung together, too.
Although I think the definition for Mozart's style may be a bit more
straightforward than those for the English language (I believe it would be
slightly easier to know what 'makes sense' in the musical vocabulary), it is
no small feat to copy the style of a composer. Inductively, I don't know how
hard it would be to copy the style of an author (not beyond the realm of
possibility, though).
-Garrett Butulis
Undergraduate EE/music
Wright State University, Dayton, OH.
thomas.p.pschar
A, which on my machine was a 3 1/2 (the disk was a 5 1/4). I never could get it
to run, and even sent the disk back. Does anyone still have it? Did anyone
hack it to work off any drive?
Tom
>
>
>
>
>
>-SubGenius
>
>
>
--
+----------------------------------+------------------------------------------+
|Tom Pschar - AT&T CAOSS Dev. Team | "Fifteen is my limit on Schnitzengruben!"|
| It is very dark. You are likely to be eaten by a grue. |
+-----------------------------------------------------------------------------+
Fred Welden
In article <1k3scj...@montaigne.lif.icnet.uk>, j...@montaigne.lif.icnet.uk (Jack Shirazi <j...@biu.icnet.uk>) writes:
|In article <1993Jan25....@seas.smu.edu> pede...@seas.smu.edu (Ted Pedersen) writes:
|>>I saw the following in the Books column (by Olin Chism) of the Dallas
|>Morning News Sunday January 24, 1993. Copied without permission. It
|>concerns a computer program that allegedly wrote a book in the style
|>of Jacqueline Susann.
|>
|>I think this is a stunt. I don't really believe a computer program
|>wrote this book. However, if anyone has any information about this
|>book or the author that would give this some credibility I would like
|>to hear about it. Comment on the general idea would be interesting as
|>well. I don't think the idea itself is impossible, it just seems like
|>this case is a little far fetched.
|
|Markov chaining using a mixture of sentences and words, or possibly
|just with words, using a parser to get rid of any unlikely combinations.
|Probably added in the use of a style checker.
It occurs to me to wonder why any publisher would do this. Certainly
there's an incentive for researchers in linguistics, artificial
intelligence, literary criticism, and any number of other academic
disciplines to try to create a program that could write worthwhile prose
in the style of a given author--but why would a publisher do it?
There are scads of humans out there who can write a respectable story in
the style of any given commercially successful author, and whose
manuscripts can be had cheaply. It is not the cost of acquiring
manuscripts that drives up the price of books, it's all the other stuff
that you'd have to do whether the book was written by a human or a
machine.
Jim Kasprzak
|> Racter! I bought that program a long time ago, but it had to run off of drive
|> A, which on my machine was a 3 1/2 (the disk was a 5 1/4). I never could get it
|> to run, and even sent the disk back. Does anyone still have it? Did anyone
|> hack it to work off any drive?
A roommate of a friend of mine had a copy of Racter on his Amiga. One night
in 1989 when I was trying very hard to ignore school, we spent six hours
trying to get it to say that "cats and dogs should get along". We went through
a lot of interesting sentences in the interim, and I think managed to see all
of Racter's vocabulary.
In "Return to Club Kibo", I was going to include a conversation between
Eliza, Racter and Springhead on the subject of whether Spot is allowed. I
really should get around to writing that.
To bring this back to books: Racter was where I first heard of Oblomov.
Who was the author who used Oblomov as a character, and in what novel? Is
it still in print at all?
Charlie J Church
Please post to the network, I would like to see for myself.
Charlie
Paul Christopher Workman
>Sophisticated parser? Eliza? Most versions of Eliza are very
>simple-minded indeed and don't actually do any parsing, just pattern
>matching.
Speaking of which, is there some public-domain English-language
parsing code out there I could ftp or such? I want to write a
(really weird) game generator which would need an English language
parser.
thanks,
--paul
Barbara Hlavin
>In article <1993Jan25....@seas.smu.edu>, pede...@seas.smu.edu (Ted Pedersen) writes:
>|>
>|> I think this is a stunt. I don't really believe a computer program
>|> wrote this book. However, if anyone has any information about this
>|> book or the author that would give this some credibility I would like
>|> to hear about it. Comment on the general idea would be interesting
> There is, in fact, some serious research going on in the area of
>computers writing books. Here at Rensselaer Polytutescrew Incorporated,
B
What I want to know is, why does anyone *want* a computer to write a book?
(OK; why does anyone want a computer to do anything?)
Consider the scarifying number of us out here who are willing to do
it without programming, task forces, bodies of researchers supported
by grants and salaries, or indeed, the least little bit of encouragement.
>professor Selmer Bringsjord has been working on several different projects
>aimed at getting a computer program to produce meaningful stories. Its
Professor Bringsjord should be working on a project aimed at getting
a warm-blooded human being to produce meaningful stories.
--Barbara
Alan Mead
> There are at least two obvious ways to get a computer to
>"write" prose; both involve the scanning or analysis of another text.
>One involves Markov chains, and the other involves a phonetic dictionary.
>[...]
Not too long ago (perhaps in an AI book by Peat), I read an account of
research attempting to get a program to understand stories. The best
example was a program that read incoming UPI or UPS stories and
summarized them. As I recall, it worked fairly well.
This would seem to be the neccessary leap before a computer were to
write a story. I think the above two methods are really only good for
generating sentences.
For example, take the Doug Adams _Hitchhiker_ series. His emphasis on
coincidence might lead you to think that his stories are somehow more
"random" and would be easier to generate by computer. But I don't
think anything could be farther from the truth; much of his humor is in
providing a new fact that sheds just the right light on some old bit so
that you see it in some new and funny way. That kind of thing takes
careful planning and intelligence. I think Hemmingway's short stories
would be similarly difficult. You could generate all kinds of
Hemmingway-esqe sentences, but one of his strengths was in telling 95%
of the story (and knowing which 5% to leave out).
I suspect that the original poster's book was, if not entirely written
by a human, written with the mere aid of software. I could imagine a
program that changed the wording and maybe even the grammer of a
sentence to reflect another writer's style.
Of course, if a computer was involved at all, that would imply that
they had a large amount of Susann's work in electronic form (in order
to anaylize it). Is that likely?
-alan
Ted Pedersen
The computer really isn't writing the book. It's the programmer who writes
the program that writes the book. Despite all appearances computing is
a human enterprise and such a book would reflect the sensibilities of
the programmer and not the computer.
It's a different way to think about writing a book. Don't write the
book, write the program that writes the book. Is it a good approach?
Maybe. Maybe not. We'll see.
Clearly there are authors who use very formulaic (is that a word?)
approaches. The open question is "Is it possible to render such a formula
in a form advantageous to computing?" The even bigger question is "Is
it possible to write a program that does more than follow a formula?
(ie exhibit signs of Artificial Intelligence)". Worth investing? From
a Computer Science point of view? Yes. From a literary point of view?
Maybe.
dave budd
>In <74AX...@cc.swarthmore.edu> beh...@cc.swarthmore.edu (Eric Behrens) writes:
>>But music has only a dozen words in its vocabulary (notes of the scale) and
>>the English language has tens of thousands. I cannot believe that computer
>>science is capable of replicating the intricate constructions of language.
>I'd say the analogy is not of the right order. I've heard that scales are
>more like phonemes or letters, and that there are groups or patterns of
>scales which could be regarded as ``words of music''. There has been done
>some work trying to apply linguistic knowledge to musical ``texts''. I don't
>remember the scientist's name, but I could look up some references if someone
>was interested.
>Oliver
I'd also say that a musical note is more akin to a letter or phoneme, and to
continue the analogy, a chord would be a word, one or two bars would be a
phrase, and a few bars would be a sentence. Then you go on to paragraphs,
which would be like a verse in a song, and chapters, like one movement of a
symphony, etc etc etc
And of course there's intonation/instrumentation to cope with
Actually, Occam wore a beard: philosophers kept borrowing his razor
Dave Budd, MCC, Oxford Rd, Manchester, England (44|0)61-275-6033
William B Dwinnell
Alan: I'm not sure I understand your assertion that "This [attempting
to get a program to understand stories] would seem to be the neccessary
leap before a computer were to write a story." Programming a sentence
generator which produces output which makes sense is a simple matter, yet
such systems do not have to (and generally can't) understand natural
language sentences, not even their own output. Why couldn't this be
extnded to complete stories?
Scott Turner
>What I want to know is, why does anyone *want* a computer to write a book?
>Consider the scarifying number of us out here who are willing to do
>it without programming, task forces, bodies of researchers supported
>by grants and salaries, or indeed, the least little bit of encouragement.
There are many interesting reasons to look at computer storytelling.
For one thing, a computer that can tell stories will likely have a lot
of the communication skills necessary to interact in more meaningful
ways with its users than current computers. Certainly one could see
the usefulness of a storytelling in a domain like CAI.
At a more fundamental level, building a computer program that can tell
stories is a way towards understanding and modelling the cognitive
processes used in storytelling. One purpose of my research in
computer storytelling is as a vehicle for studying issues in
creativity. I trust the value of creative computers is apparent.
-- Scott T.
Scott Turner
>The computer really isn't writing the book. It's the programmer who writes
>the program that writes the book.
The human author isn't writing the book. It's the parents who birthed
and raised him that write the book.
-- Scott T.
Roger Squires
Those interested in the topic, even casually, might
wish to look for a wonderfully entertaining series
of lectures on vinyl records by Leonard Bernstein,
entitled The Norton Lectures, given in '72 at Harvard.
He talks about (and demonstrates on the piano) musical
syntax and semantics, referring to Chomsky, using for
examples Mozart's 40th and that famous piece by Ives
who's name I forget that has competing melodic and chaotic
themes, and asking the question "whither now music?"
It's about 12-14 records, and if you are really, really
lucky, your library may have a copy.
Not that any of this has squat to do with rec.arts.books.
rms
Dr. Disorder
>In article <jakobs.728043779@utrurt>, jak...@utrurt.uni-trier.de (Oliver Jakobs) writes:
>> In <74AX...@cc.swarthmore.edu> beh...@cc.swarthmore.edu (Eric Behrens) writes:
>>
>>>But music has only a dozen words in its vocabulary (notes of the scale) and
>>>the English language has tens of thousands. I cannot believe that computer
>>>science is capable of replicating the intricate constructions of language.
>>
>> I'd say the analogy is not of the right order. I've heard that scales are
>> more like phonemes or letters, and that there are groups or patterns of
>> scales which could be regarded as ``words of music''. There has been done
>> some work trying to apply linguistic knowledge to musical ``texts''. I don't
>> remember the scientist's name, but I could look up some references if someone
>> was interested.
>>
>> Oliver
>>
>>
>> --
>> Oliver Jakobs, Dept. of Computational Linguistics, Trier University, Germany
>> | Internet: jakobs@ldv01@Uni-Trier.de |
>> +--------------------------------------------------------------------------+
>
> Indeed, I would consider notes to be much more synonymous to the
>letters of the musical vocabulary. Patterns of such notes, then, could be
>the words of the musical language. Finally, phrases (groups of music, usually
>in 4 or 8 bars) are the musical sentances.
[Stuff on Mozart's style deleted for space]
> Although I think the definition for Mozart's style may be a bit more
>straightforward than those for the English language (I believe it would be
>slightly easier to know what 'makes sense' in the musical vocabulary), it is
>no small feat to copy the style of a composer. Inductively, I don't know how
>hard it would be to copy the style of an author (not beyond the realm of
>possibility, though).
Perhaps this is the wrong place to take up this thread, but as a composer,
I just can't let this go by. IMHO, the second poster made a much more plausible
analogy (i.e, notes are closer to letters rather than words - although, in
certain styles, a single note can take on many different levels of meaning,
from the letter to the paragraphand beyond). However, I strongly disagree with
the statement that it would be easier to know what "makes sense" in musical
vocabulary. As with language, the basic rules for making combinations out
of primitive expressions can often contradict each other, with some kind of
difficult-to-precisely-characterize meta-system required to resolve the
lower-level conflict. It is my intuitive belief (based on a life-long deep
immersion in music, but with an admittedly paltry understanding of
natural-langauge modeling) that the rules for creating "proper" expressions
in the style of a specific composer (especially one as rich as Mozart) are
even more resistant to explicit definition than the rules for what
"makes sense" in the English language. I think that there are even more
decisions at every level of the musical process that involve aesthetic
judgements (of a variety that would be imposssible to describe unambigously).
As easy to trip up as it is, even ELIZA does a better job of
acheiving its goal than any computer-generated piece in the style of
Mozart that I've ever heard.
--
Practice random acts of randomness.........
theodore...@cs.yale.edu
Jonathan R. Ferro
> I'd say the analogy is not of the right order. I've heard that scales are
> more like phonemes or letters, and that there are groups or patterns of
> scales which could be regarded as ``words of music''. There has been done
> some work trying to apply linguistic knowledge to musical ``texts''. I don't
> remember the scientist's name, but I could look up some references if someone
> was interested.
A well-known sourcebook for this area is _A Generative Theory of Tonal
Music_, by Fred Lerdahl and Ray Jackendoff, a part of the MIT Press
series on cognitive theory and mental representations.
-- Jon Ferro Einsprachigkeit ist heilbar
Jim Kasprzak
|> What I want to know is, why does anyone *want* a computer to write a book?
I think you've missed my point here. As I mentioned in a line of my post
which you deleted, Prof. Bringsjord is primarily concerned with the problems
in philosophy and computing which one faces when trying to get a computer to
write a book. Questions like:
- What sort of data base does one need to work from to write a story?
- What is the difference between a story and a narration?
- Is it possible for a computer to use the subtleties of metaphor and
imagery inherent in language?
- How does the "programming" of an author, whether human or computer,
affect what the author writes?
This is just the beginning. I am not involved in Prof. Bringsjord's work;
I only learned about it through a talk he gave to the local chapter of the
ACM. Perhaps I should tell him that the good folks on Usenet have taken an
interest in his work, and let him speak for himself.
|> Professor Bringsjord should be working on a project aimed at getting
|> a warm-blooded human being to produce meaningful stories.
He is. He's written a novel titled _Soft Wars_, and is currently trying to
get working on a second novel.
Doug Quarnstrom
The universe writes the book through the tool of evolution.
This seems a rather absurd thing to bicker over.
doug
Tim Poston
: Maybe it could do e e cummings, but a novel... :Ñ)
1) e e cummings did write a novel (The Enormous Room),
based on his prisoner of war experiences.
2) he was one of the most `poetic' of poets,
using the lateral, more analogue than digital,
possibilities of English to the limit.
It is not hard to get a computer to produce
correct English syntax (though the _content_ is
liable to seem odd); it is very hard to get
it to break the rules in a way that communicates
emotional force to the reader.
A computer will write like Ms Susanne a long while before one
writes like e e cummings.
Tim Poston
: Based on this I propose the Welden Test of natural intelligence: put a
: person in a booth where he can communicate only by teletype. Have him
: communicate with two respondents, one a human, one a machine. If he
: can't tell which is the machine, he is behaving without natural
: intelligence.
:
: Congratulations to your instructor--he has failed the Welden Test!
Be fair on the poor guy; he was grading a _Modern_Poetry_Class_.
None of the usual tests like coherence, levels of organization, etc.
would allow more than half the class a natural intelligence certificate.
Besides, the A+ submission was written by the poster's husband,
using BASIC to mechanize some selection from words and phrases _he_
installed --- not just a full dictionary and a set of language rules ---
and he and she selected the most successful poem, in their view,
from the output. He was facing mostly natural intelligence.
Finally, he was not able (even by teletype) to interrogate the machine,
which was thus not a `respondent' in the Turing/Welden sense.
So he was not even _subjected_ to the test, let alone failing it.
Tim
________________________________________________________________________
"She was not one of the two North American poets worth more than
a warm mouthful of industrial cherry cola, and it rankled."
Ibid.
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Kathy Johnson
Fred Welden
In article <1993Jan28.0...@nuscc.nus.sg>, t...@iss.nus.sg (Tim Poston) writes:
|sas...@dobo.unx.sas.com (Fred Welden) writes:
|
|: Based on this I propose the Welden Test of natural intelligence: put a
|: person in a booth where he can communicate only by teletype. Have him
|: communicate with two respondents, one a human, one a machine. If he
|: can't tell which is the machine, he is behaving without natural
|: intelligence.
|:
|: Congratulations to your instructor--he has failed the Welden Test!
|
|Be fair on the poor guy; he was grading a _Modern_Poetry_Class_.
|None of the usual tests like coherence, levels of organization, etc.
|would allow more than half the class a natural intelligence certificate.
|
|Besides, the A+ submission was written by the poster's husband,
|using BASIC to mechanize some selection from words and phrases _he_
|installed --- not just a full dictionary and a set of language rules ---
|and he and she selected the most successful poem, in their view,
|from the output. He was facing mostly natural intelligence.
|
|Finally, he was not able (even by teletype) to interrogate the machine,
|which was thus not a `respondent' in the Turing/Welden sense.
|So he was not even _subjected_ to the test, let alone failing it.
All points well taken. Having seen an opportunity to toss in the
Welden Test I wasn't going to let a few nit-picky little details like
this stop me--who knows when another occasion would arise?
Fred Welden
In article <1993Jan27....@seas.smu.edu>, pede...@seas.smu.edu (Ted Pedersen) writes:
|In regards to why we would want a computer to write a book...
|
|The computer really isn't writing the book. It's the programmer who writes
|the program that writes the book. Despite all appearances computing is
|a human enterprise and such a book would reflect the sensibilities of
|the programmer and not the computer.
|
|It's a different way to think about writing a book. Don't write the
|book, write the program that writes the book. Is it a good approach?
|Maybe. Maybe not. We'll see.
|
|Clearly there are authors who use very formulaic (is that a word?)
|approaches. The open question is "Is it possible to render such a formula
|in a form advantageous to computing?" The even bigger question is "Is
|it possible to write a program that does more than follow a formula?
|(ie exhibit signs of Artificial Intelligence)". Worth investing? From
|a Computer Science point of view? Yes. From a literary point of view?
|Maybe.
So you've added authors to the list of folks who might have good reasons
to create a program that writes books. The question I asked (and I
can't tell from your headers whether you're responding to me or not) was
why would PUBLISHERS want such a thing?
Ed Stastny
>|> The human author isn't writing the book. It's the parents who birthed
>|> and raised him that write the book.
The CHICKEN wrote the book (or was that the egg?)...
My $.02 is that computer programs writing books is substancially
disassociated with "random word pulling" in that the program (if well
written and meticulous) has an inset series of associations and tangents
of which some are "woven into" it's fabric and others are merely happy
biproducts of the program...dynamics and phenomenon.
There's no doubt the programmer influences the writing of the book, just
as environment and nature influence the writing of all books, but I
think it's safe to say that the computer IS writing books.
I guess this type of thinking hinges on the belief that humans are,
themselves, computers.
...e
--
Ed Stastny...ed@cwis.unomaha.edu END PROCESS
PO BX 31104, Omaha,NE 68132 USA SOUND N & A
vmessageline:(402)3934207 VISIONTRIBE
attn artists: send scans for project (gif or jpg) STIMULATION
bi...@waikato.ac.nz
>>In article <1993Jan25....@seas.smu.edu>, pede...@seas.smu.edu (Ted Pedersen) writes:
>>|>
>>|> I think this is a stunt. I don't really believe a computer program
>>|> wrote this book. However, if anyone has any information about this
>>|> book or the author that would give this some credibility I would like
>>|> to hear about it. Comment on the general idea would be interesting
>
>
>> There is, in fact, some serious research going on in the area of
>>computers writing books. Here at Rensselaer Polytutescrew Incorporated,
>
> B
> What I want to know is, why does anyone *want* a computer to write a book?
Well one reason would be that once computers can churn out mass-market books
such as Mills and Boon romance novels, humans would no longer be able to
compete because the computers would be far cheaper. Humans would then be
forced to turn their creative talents to other kinds of writing, be more
unpredictable and perhaps invent many more new styles of writing as well.
I look forward to the day. Especially for TV scripts. I'm convinced there
is a computer out there already churning out scripts for American TV
programs. (Hollywood sub-plot 2.1 : The hero always marries the heroine).
Kathy Johnson
|>
|> So you've added authors to the list of folks who might have good reasons
|> to create a program that writes books. The question I asked (and I
|> can't tell from your headers whether you're responding to me or not) was
|> why would PUBLISHERS want such a thing?
|>
|> --
|> --Fred, or another blind 8th-century BC | sas...@dobo.unx.sas.com
|> Hellenic poet of the same name. |
So they could replace their staff writers with it? 8-)
Graeme Hirst
>> But you never know. I recall an NPR report perhaps two years ago. It was
>> about a computer scientist who had spent a decade creating a program which
>> could compose in the style of Bach.
>>
>
>I would be very interested in hearing more about this program from anyone.
>I've always wanted to try something like that.
Sounds like a reference to the work of David Cope, who applies AI-ish
techniques such as ATNs to represent the style of a composer. His book
on the topic (details below) is reviewed in the forthcoming :-( December
1992 issue of "Computational Linguistics" journal.
Computers and musical style
David Cope
(University of California, Santa Cruz)
Madison, Wisconsin: A-R Editions Inc (The computer music and digital
audio series, edited by John Strawn, volume 6), 1991, xvii+246 pp;
hardbound, ISBN 0-89579-256-7, $45.95
--
\\\\ Graeme Hirst University of Toronto Computer Science Department
//// g...@cs.utoronto.ca / g...@cs.toronto.edu / 416-978-8747
Don Zirilli
>In article <1993Jan25....@seas.smu.edu>, pede...@seas.smu.edu (Ted Pedersen) writes:
>|>
>|> I think this is a stunt. I don't really believe a computer program
>|> wrote this book. However, if anyone has any information about this
>|> book or the author that would give this some credibility I would like
>|> well. I don't think the idea itself is impossible, it just seems like
>|> this case is a little far fetched.
>
>computers writing books. Here at Rensselaer Polytutescrew Incorporated,
>aimed at getting a computer program to produce meaningful stories. Its
>computing problems of getting something like this to work, and they haven't
>produced any computer-generated novels yet, but I've seen some rudimentary
>output. It looks something like children's stories written by Eliza.
>------------------------------------------------------------------
You call that "serious research"? Even I have accomplished more. I wrote
my own c-program to write novels. The program is _finished_. Unfortunately,
though I am a brilliant theorist, I am not a terribly efficient programmer
and the c-program, which I leave running, is only producing approximately
one word per hour. Here is what I have so far, after days of waiting:
Elsie was tired of being human. It just didn't seem right. What
can humans do, really? They can pick daisies, but who cares? It all seemed
so pointless.
A truck hit Elsie at the corner at a speed exceeding thirty five
miles per hour. This is only sad if you consider human life worthwhile.
The truck driver soon came to the conclusion that he should quit his job.
After all, what good is a truck? It has no intelligence to speak of, and
only does what its driver directs it to do. And what does he accomplish in
his truck? He brings people food and other material items which more or
less perpetuate a society of fleshy creatures of dubious importance.
The former truck driver was named Jojo, and after paying proper
respects at Elsie's funeral, he went back to his apartment to contemplate
his life now that he was no longer driving trucks. It all seemed so
pointless. He called a woman, Margarita, who would often come over and sit
on his
This is as far as its gotten. I anxiously await the next word. Anyway,
in spite of the repetition of "It all seemed so pointless", I'm sure this
prose far outshines that RPI crap. After all, if they could write novels
they'd be in Columbia, and if they could program computers they would be
in MIT! (just a joke, dammit)
Yours,
Posthumous O'Toole
--
Chucklehead awaits the return of his .sig
Aaron Radomski, where are you???
dave budd
>As easy to trip up as it is, even ELIZA does a better job of
>acheiving its goal than any computer-generated piece in the style of
>Mozart that I've ever heard.
>--
It must be remembered that Mozart himself once wrote a piece by assembling
chunks chosen by rolling a die. But the chunks were all 'properly' written
first.
jsterw...@acs.harding.edu
>In article <1993Jan25....@seas.smu.edu>, pede...@seas.smu.edu (Ted
>Pedersen) writes:
>>
>>
>> Morning News Sunday January 24, 1993. Copied without permission. It
>> concerns a computer program that allegedly wrote a book in the style
>> of Jacqueline Susann.
>> I think this is a stunt. I don't really believe a computer program
>> wrote this book. However, if anyone has any information about this
>> book or the author that would give this some credibility I would like
>> to hear about it. Comment on the general idea would be interesting as
>> well. I don't think the idea itself is impossible, it just seems like
>> this case is a little far fetched.
>>
>
>monkey at a typewriter writing Hamlet...
>
>But you never know. I recall an NPR report perhaps two years ago. It was
>about a computer scientist who had spent a decade creating a program which
>could compose in the style of Bach.
>
>entire opus for you. I must confess that the results were quite
>impressive-- not perfect, but impressive.
>
>But music has only a dozen words in its vocabulary (notes of the scale) and
>the English language has tens of thousands. I cannot believe that computer
>science is capable of replicating the intricate constructions of language.
>
>
are so many things to consider in music: Dynamics, key signatures, time signa-
tures, accents- I could go on and on. A better comparision would be the 26
letters of the alphabet to the twelve notes of the scale, but even then it's
not very accurate. Thank you for allowing me to speak my peace.
Jeffrey>
jsterw...@acs.harding.edu
>: >The computer really isn't writing the book. It's the programmer who writes
>: >the program that writes the book.
>:
>: The human author isn't writing the book. It's the parents who birthed
>: and raised him that write the book.
>
>The universe writes the book through the tool of evolution.
>
>This seems a rather absurd thing to bicker over.
>
>doug
jsterw...@acs.harding.edu
>|> >The computer really isn't writing the book. It's the programmer who writes
>|> >the program that writes the book.
>|>
>|> The human author isn't writing the book. It's the parents who birthed
>|> and raised him that write the book.
>|>
>|> -- Scott T.
>It's the all powerful being who created the first human that writes the book.
Graham Higgins
+ Well one reason would be that once computers can churn out mass-market books
+ such as Mills and Boon romance novels.
Mass-market = many sales = satisfied customers (else no re-purchase).
--
Graham Higgins
<< g...@hplb.hpl.hp.com || g...@hplb.hp.co.uk || g...@hplb.uucp >>
<< My opinions above are exactly that: my own, and opinions. >>
Fred Welden
A fine example of so condensing an argument that it ceases to have any
content whatsoever. Or are you attempting only to observe the obvious?
No one needs computers to churn out mass-market romances. The masses
churn them out quite cheaply by themselves. Now, somebody might want
to market the software itself, so a formula romance fan could push a
button and get a new romance every time, but you can bet the mass-market
publishers don't want that.
Eduard Tulp
>In article <Bob_Hearn-2...@192.35.50.165> Bob_...@qm.claris.com (Robert Hearn) writes:
>>> But you never know. I recall an NPR report perhaps two years ago. It was
>>> about a computer scientist who had spent a decade creating a program which
>>> could compose in the style of Bach.
>>>
>>
>>I would be very interested in hearing more about this program from anyone.
>>I've always wanted to try something like that.
>
>Sounds like a reference to the work of David Cope, who applies AI-ish
>techniques such as ATNs to represent the style of a composer.
To me it sounds like the work of Kemal Ebcioglu. If I remember correctly,
he wrote an expert system which harmonizes chorales in the style of Johann
Sebastian. I have his thesis somewhere at home. I can look it up if anyone
is interested. I am in the middle of moving, so it should be in a box
somewhere :-( I seem to remember that he did his research at SUNY (State
University New York, Buffalo). Last I heard he works for IBM somewhere.
Eduard T.
Jim Kasprzak
|> to create a program that writes books. The question I asked (and I
|> can't tell from your headers whether you're responding to me or not) was
|> why would PUBLISHERS want such a thing?
I think publishers would want an authorial program if the software was
cheap, ran on a Macintosh, and cranked out bestsellers at the click of a
mouse button.
The current research going on in this area is not something that publishers
would have much interest in.
B. Van de Wetering
>
>I vaguely recall something about a poet - think he may have been called Dada,
>from whence Dadaism, but this is probably just my deranged memory - who wrote
>poems, then tore up the paper into words/sentences and shook them up in a hat.
>Then he pulled them out again, writing them down. I think the comparisson to
>current poetry/fiction writing programs is valid. Few people would claim that
>Dada's hat wrote poetry, after all.
>
>Okay, fire when ready :-)
>
>-- Olly Betts --
I know for certain that the term Dadaism didn't come from a poet's name, but
rather was coined by a poet whose name escapes me at the moment. For some
reason Tristan Tzara comes to mind, but I'm probably mistaken. This
technique of using cutups to write poetry or prose was also used by W.S.
Burroughs and Brion Gysin. It can be very fun and amusing especially with a
group of people. Several years ago I spent some time cutting up bible tracts
and rearranging them into "poetry." One of my favorites came from a tract
detailing the evils of Rock & Roll.
her big fantasy is to "seduce
a priest ignorant of this
his own chaos, and he found it to be a
kind of music
whose goal in life is to be more like
world, the flesh and the devil
tune and tempo of
big beat matches the
trouble with his conscience
he knew this music
every time he entered
Another would
carry this beat with them
overthrowing the
erotic politicians.
Yet Christians are
pressed by Satan about this thing.
He confessed to
anything about revolt, disorder,
especially activity that
has no meaning.
"It is all sex."
--
Brian Van de Wetering INTERNET:wete...@sea.com
UUCP:...!ucsd!soledad!wetering
VOICE:619/581-6181 x307
Systems Engineering Associates, 2204 Garnet Ave Suite 303, San Diego CA 92109
Dawn Myfanwy Cohen
seen/programmed computer generated stories/poetry/text...Could anyone
who actually has access to some of this text (like if you have a
working program) please post some of it? I would be extremely curious
to see anything that is (1) poetry or (2) text that is anywhere near
coherent over more than two sentences.
--
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Good Morning!
--Dawn
Send fan mail to:
uucp: {ames,bosgd,harvard,moss}!rutgers!paul.rutgers.edu!dcohen
arpa: dco...@paul.rutgers.edu
Daniel Barnes
I have been following the topic of composing computers with interest. I
am working a program myself to compose music, although I have yet to find a
good way to represent the knowledge it needs. I plan to have it look for
patterns in types of music by a composer (for example, Beethoven sonatas).
From this, it will write a sonata in Beethoven's style. My question is this:
has anyone attempted something like this, and if so, how was the music database
represented and what pattern recognition algorithm would work best? I would
appreciate any help on this. Thanks in advance!
Mark Israel
> In article <93Jan28.12...@smoke.cs.toronto.edu>, g...@cs.toronto.edu (Graeme Hirst) writes:
>>>> But you never know. I recall an NPR report perhaps two years ago. It was
>>>> about a computer scientist who had spent a decade creating a program which
>>>> could compose in the style of Bach.
>>>>
>>> I would be very interested in hearing more about this program from anyone.
>>> I've always wanted to try something like that.
Do you have an academic interest in doing that? (If so, please e-mail me;
there's someone here whom you should correspond with.)
>> Sounds like a reference to the work of David Cope, who applies AI-ish
>> techniques such as ATNs to represent the style of a composer.
>
> To me it sounds like the work of Kemal Ebcioglu. If I remember correctly,
> he wrote an expert system which harmonizes chorales in the style of Johann
> Sebastian.
You remember correctly.
> I have his thesis somewhere at home. I can look it up if anyone
> is interested. I am in the middle of moving, so it should be in a box
> somewhere :-( I seem to remember that he did his research at SUNY (State
> University New York, Buffalo).
Right again! His Ph.D. dissertation was _An Expert System for Harmonizing
Chorales in the Style of J. S. Bach_, SUNY, 1986.
But here's some more recent work:
"HARMONET: A Neural Net for Harmonizing Chorales in the Style of J. S. Bach",
by Johannes Feulner, Hermann Hild, and Wolfram Menzel (all of Universit"at
Karlsruhe, Germany), in _Advances in Neural Information Processing 3_ (NIPS 3),
Kaufmann 1991.
I have e-mail addresses for the authors.
mis...@csi.uottawa.ca Mark Israel
Sunbeams brightly play, where Fancy's fair pavilion once is pight.
Ted Pedersen
>someone who has to read pages of copy every day, seems like an exercise
>in academic autoeroticism. Teach more human beings to be able to "use
>the subtleties of metaphor and imagery inherent in language" and you've
>accomplished something. Teach a human being the difference between a
>story and a narration. Give a human being the database to work from.
>Then, put a computer to work solving the HIV problem, or guiding
>satelites.
>Until every manuscript that crosses my desk is meet for publication, I
>see no need to start teaching computers who to WRITE. Teaching them how
>to read, so that they can, as tools, tell me what is worth giving a
>second, more personal look, now *THERE* is something worth the federal
>funding that must pour into the universities that produce the academic
>minds that come up with these futile experiments.
>It may seem harsh, and I'm sure Bringsjord is a brilliant man, but there
>are quite enough writers in this world. Adding (programming?) another
>is redundant. What this world needs are solutions to problems, not
>solutions to non-problems.
I would also suggest that there are "quite enough" literary agents. What
great problems of the world do literary agents address?
There are two levels on which to judge the idea of computers writing a
book. You must judge the book and the program that wrote the book. It
might be that the book is trash. But the odds are that if you have a
program that can write even a trashy book you've got an amazing
program. That might not be important to literary agents but it's a big
deal for computer scientists.
What we've arrived at is a restatement of the "basic research vs.
applied technology" debate. Having a computer write a book (even writing a
paragraph for that matter) is basic research in Artificial
Intelligence. There are always those that say let's not worry about
basic research, let's do some practical stuff. Fine. We need to do
practical stuff be we also need people stumbling around in uncharted
areas maybe accomplishing nothing on the chance that some fundamental
breakthough might result.
Ted
---
* Ted Pedersen pede...@seas.smu.edu *
* Department of Computer Science and Engineering, *
* Southern Methodist University, Dallas, TX 75275 (214) 768-2126 *
Terje Johansen,o90b
I would love to see such a machine. Reason: the shelves of the bookstores
today already have endless heaps of worthless paper written by people who
could not care less about literary qualities. They write by the help of
company-made scripts complete with storylines and names. They just supply
the adjectives...it seems. So, when machines can show just how
worthless those tons of paper are, we might get some consciousness
as to what we read and what it make us become.
--
Terje Johansen at Trondheim College of Engineering, Norway.
War isn't fought to decide who is right
but who is left.
Hovig Heghinian
(Terje Johansen,o90b) writes, with lotsa stuff deleted:
>>In article <1k56lg...@shelley.u.washington.edu>, tw...@milton.u.washington.edu (Barbara Hlavin) writes:
>>> What I want to know is, why does anyone *want* a computer to write a book?
>I would love to see such a machine. Reason: the shelves of the bookstores
>today already have endless heaps of worthless paper written by people who
>could not care less about literary qualities. They write by the help of
>company-made scripts complete with storylines and names. They just supply
>the adjectives...it seems. So, when machines can show just how
>worthless those tons of paper are, we might get some consciousness
>as to what we read and what it make us become.
A thought: any computer that could write a book would be in the
charge of programmers that ARE in fact concerned with literary quality,
or they wouldn't engage in the experiment. Of course, when the art is
``perfected'' we'll definitely see the mass-production of books;
I wholeheartedly agree.
I do not advocate the line that a book produced by a computer will
constitute ``literary quality,'' nor will it do so any time soon.
I simply think that for now, the thought experiment of getting a computer
to write a book is no less than an effort to describe that which makes
literary quality, or at least to describe what we read and what it makes
us become, n'est-ce pas?
If ``book processors'' become commonplace any time soon, the event
might have exactly the opposite effect than what Terje describes (in my
pitiful opinion, that is). It may cause people to think twice about what
they read, and the glut of reading material (which can't be any worse
than it is now!) might actually start favoring those more thoughtful
pieces of writing, as we'd all like. As much as we worry, a computer
will never, I repeat, never be an Asimov ... prolific, lightly readable,
enjoyable, yet utterly brilliant in relating everyday story-telling to
__humanity__. I.e.: DON'T PANIC! =)
Hovig
Mark 'Mark' Sachs
In article <s4p...@rpi.edu>, kas...@isaac.its.rpi.edu (Jim Kasprzak) says:
> A roommate of a friend of mine had a copy of Racter on his Amiga. One night
>in 1989 when I was trying very hard to ignore school, we spent six hours
>trying to get it to say that "cats and dogs should get along". We went through
>a lot of interesting sentences in the interim, and I think managed to see all
>of Racter's vocabulary.
You can still -- at least I could last time I was there, which was a few years
ago -- use Racter on an ancient IBM PC at the Computer Museum in Boston.
Quite frankly, I was impressed -- I know it's basically Eliza with a long-term
memory, but it was fascinating to "converse" with anyway.
Of course, the thing was set up improperly, and after a few minutes of
conversation it would fill up the disk, crash, and start over... but still...
Geez, what I wouldn't do to get a copy of Racter for my own self... anyone
have an Amiga version they wanna part with? Or even an IBM version?
"If you can't dazzle them with brilliance, riddle them with bullets."
[Your blood pressure just went up.] Mark Sachs IS: mbs...@psuvm.psu.edu
DISCLAIMER: Penn State University cares about my money, not my opinions.
Terje Johansen,o90b
Hehe. Got to add a few more words, I see. I agree with you, Hovig,
eons will pass before a machine write something like Caves of Steel.
But other writers have less imagination. How many storylines are the
Barbara Cartland books based on? 10? 5? 1? Or your cowboy books, agent
series, war series?
I would like to draw your attention to the bookwriting machine in
Wilbur Orwell's '1984'. This machine didnt write books word for word,
but took several story lines and mixed them together.
Sounds easy to implement, eh?
Like, a mayor thread would be: country invades neighbour
or: man hunts girl
of which each thread could have side-threads like:
- hero is hurt
- hero falls in love
- hero's best friend dies
feels familiar, eh? I bet you 10 to one any programmer could make such
a program in a week, supplemented only with a base of a few hundred mayor
and minor threeds. Then, for each thread there could be a few dosen
descriptions, with variables set for kind of area, time of year, etc.
Well...would this be a GOOD book? It certainly wouldnt be innovative
or actually very interesting, even with all the usual clichees.
So where is my point?
Merely that it would indeed be possible, but not literally gaining, to
make a machine writing or aiding majorly in writing a book in one of
the mayor paperback-genres.
It is my hope, that when humanity realizes we waste our time on reading
drivel even a machine could write, maybe more people would start searching
for books of better quality, and hopefully there might be a decline in
the sale of trash-literature.
But, seeing how we already embrace those accursed machines known
as television, I accept that this is a dream.
But hope never must die. Let Asimov, Verne, Buchanan and Tolkien
carry the torches, through this dark age of tired writers.
They can fight the storytellers, but never beat them...
--
Terje Johansen at Trondheim College of Engineering, Norway.
Man will survive into eternity
he's too stupid to know when he's beaten.
Stanley Shursky
In article <1993Feb2.18...@viper.edb.tih.no>, ter...@edb.tih.no (Terje Johansen,o90b) writes...
>It is my hope, that when humanity realizes we waste our time on reading
>drivel even a machine could write, maybe more people would start searching
>for books of better quality, and hopefully there might be a decline in
>the sale of trash-literature.
>
>But, seeing how we already embrace those accursed machines known
>as television, I accept that this is a dream.
>But hope never must die. Let Asimov, Verne, Buchanan and Tolkien
>carry the torches, through this dark age of tired writers.
>They can fight the storytellers, but never beat them...
>
>--
>Terje Johansen at Trondheim College of Engineering, Norway.
>
I hate to break the news to you, but the high sales of trash-literature
indicate only that people like and will pay for trash. Personally, I
believe you will make more money writing trash than writing literature.
(Doesn't mean I don't want to write literature. But if trash is easier
and pays more...)
Stan
Who lives in the country where they made three (3) (count'em 3) TV movies
about the Amy Fischer/Joey Buttafuco story (gag, wretch, cough). Makes
me want to change countries.
Jorn Barger
same boast, that they could whip something out in a matter of weeks that
would be able to rival Barbara Cartland ;^)
It's hard to say simply why these efforts are doomed, but a hint of the
problem might be visible if you realize that you have to simulate
characters-making-choices, which is the AI-planning-problem, undiluted,
and that if you succeeded, you ought to be able to easily switch the
text-output into a 'verbose' mode that *portrayed the characters'
Joycean stream of consciousness* in microscopic detail.
Every simple act we 'execute' has to have be pre-molded to *avoid* an
*infinity* of ways it might go sour...
jorn
Tim Poston
:
: For an interesting short story concerning the computer-writes-book
: theme, there's one by Roald Dahl who's name escapes me. Has anyone
: read it?
Not me. But if we're getting into human fiction about machine fiction,
Fritz Leiber's novel "The Silver Eggheads" is an absolute must.
A future where the entire bookwriting process has been mechanized
for a century, and then the machines break down.
(You also get some insight into the economics of a robot brothel.)
A very funny book, and very wise.
Much of what has been said in this thread was anticipated in that book,
over two decades ago.
Tim
Fraser Wilson
For an interesting short story concerning the computer-writes-book
theme, there's one by Roald Dahl who's name escapes me. Has anyone
Fraser.
discoco = discoco
To the one fella who is trying to write a program to compose
music "in the style of [XX]..." READ COPE, ET AL. He's WAY
ahead of you. Listen to recordings he has made using his EMI
LISP-based system; you'll learn something.
clsmith
T...@ndsuvm1.bitnet
>Scott Turner (s...@aero.org) wrote:
>>pede...@seas.smu.edu (Ted Pedersen) writes:
>>> the computer really isn't writing the book. It's the
>>> programmer who writes the program that writes the book.
>> The human author isn't writing the book. It's the
>> parents who birthed and raised him that write the book.
> This seems a rather absurd thing to bicker over.
IMHO, it might be a silly distinction today, but if we ever create sentient
machines in the future, giving too much credit to the creators/programmers
would be tantamount to slavery.
--Marcus Tye, t...@vm1.nodak.edu
Andy Barnhart
Computers are not intelligent; they are just fast idiots. This thread would
be more suited to comp.ai or something, though.
If I were ever to write such a program, I would be sure to exclude knowledge
of copyrights and litigation from the computer's database. Hmmm, there
might be a story idea there...
--
-Andy
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Andy Barnhart sas...@unx.sas.com
"A mind that is stretched to a new idea never returns to
its original dimensions" - Oliver Wendell Holmes
Gustav C. Rettke
>
> I would also suggest that there are "quite enough" literary agents. What
> great problems of the world do literary agents address?
Aaaagh!! How about the connection of literary sources (i.e. authors) with
the labyrinth that is the literary industry (i.e. publishers, distributers)
so that great ideas and insists can be shared. And thereby at least offer
a chance that some remnant of humantiy might gain the benefit of anothers
experiences without having to duplicate the causes, which could be traoumatic.
Gus Rettke | Programming - the key to understanding and
| utilizing the computer.
>
>
>
>
>
>
>
The Great Grendel-Khan
whatever I create. I figure if I'm to have some god status, then I should
be able to enjoy it. Anyway, what would you propose if computers gained
sentience? Give them citizenship, and let 'em vote? What kind of life
would a computer have sitting on a desktop?
--
The Great Grendel-Khan, | Vote jo...@iastate.edu
Earth Rim Roamer. | and arg...@iastate.edu
arg...@iastate.edu | for your friendly internet Presidents!
David Whitten
University (Bob Hearn was in the same class...) and realized there is a
different kind of programming needed to write a program that is interactive
and a program that tells a story.
Perhaps if I give some examples, you'll see why.
One of the structures you create when writing a story-writer contains links
about what needs a 'character' has. In Bob's example, it was a bear that
needs to eat to live.
You set up a situation where the bear can move, eat, and get, and then you
place the bear in an arbitrary part of your map, you put honey in another
part of your map, and you generate a story (actually closer to a travelogue)
of how the bear needs the honey, how he goes here, and there looking for the
honey, how he finds the honey, gets the honey, eats the honey, and then his
hunger is gone.
(I can go into more detail if anyone is interested ...)
Anyway, If you are writing Interactive Fiction, you don't really have a
story 'going' on, so you don't have to worry about things like changing the
full name of a character into a pronoun when generating the next sentence
about the character, but you do have a problem that when the programmed
character has done a search for some needed object (like honey) that it
may not have been in the kitchen when the bear was in the kitchen, but now
it is, since the player put it there after the bear left.
Assuming for a moment that you have a dynamic plot built up on the idea that
the computer maintained characters have desires or needs that are met partially
through the actions of a player, there must be some kind of signalling
mechanism that says some details that the character 'knows' to be true
are no longer true.
I think this is called the 'frame problem' in the artificial intelligence
literature.
Any thoughts about this ??
Dave (whi...@fwva.saic.com) US:(619)535-7764 [I don't speak as a company rep.]
Rujith S DeSilva
Grendel-Khan) writes:
>MHO on the other hand says that I should have complete control over whatever
>I create. I figure if I'm to have some god status, then I should be able to
>enjoy it.
Do you create your children?
>Anyway, what would you propose if computers gained sentience? Give them
>citizenship, and let 'em vote? What kind of life would a computer have
>sitting on a desktop?
Ask Stephen Hawkins. He should have a pretty good idea what it's like.
Anyway, why not fit a few manipulators on the computer? Give it some wheels?
Even incorporate it into an autonomous robot?
Later,
Rujith de Silva.
Carnegie Mellon.
P.S. If someone can remember that verse by Kahlil Gibran about our being the
bows from which we shoot our children like arrows, I'd appreciate being sent
it.
Tim Poston
: IMHO, it might be a silly distinction today, but if we ever create sentient
: machines in the future, giving too much credit to the creators/programmers
: would be tantamount to slavery.
More like Confucianism, in its popular form: `ancestor worship'.
Those who respect their ancestors and honour them as creators,
even pray to them, do not thereby become their slaves.
Mind you, I think buying mobile phones made out of paper,
and burning them so your ancestors will be able to use their ghosts
in the spirit world, is taking respect for one's progenitors a bit far.
I've no kids, so that won't arise for me with them;
but if ever a program I wrote starts burning paper gifts for me post-mortem,
my shade will definitely class that as a bug.
Tim
Leonard Newnham
> I saw the following in the Books column (by Olin Chism) of the Dallas
> Morning News Sunday January 24, 1993. Copied without permission. It
> concerns a computer program that allegedly wrote a book in the style
> of Jacqueline Susann.
Found in alt.atheism in thread Re: Atheism and Death:
------------------------------start-------------------------------
Also worth investigating is "The Policeman's Beard is Half-Constructed", a
book of short stories and poems written by RACTER. It was published by
Warner books, and is unfortunately out of print. (I found a copy in the
Cambridge University Library.) The RACTER program itself is available for PC
and Macintosh; it was written by William Chamberlain and Thomas Etter of the
Inrac Corporation, and published by Mindscape. I think you can still order
copies from Nickers International in New York. I have the Mac version; it
works under System 7 on my PowerBook 100, as long as you turn off the speech
synthesis.
-------------------------------end--------------------------------
--
Leonard e-mail: L.Ne...@bradford.ac.uk
Ice
>I was thinking about the 'novel writing' program I wrote when I was at Rice
>University (Bob Hearn was in the same class...) and realized there is a
>different kind of programming needed to write a program that is interactive
>and a program that tells a story.
I think that a nontrivial story-telling program must necessarily be a
superset of an interactive fiction program. You not only need all the
data structures of an IF system, but then the program has to use them
to create a plausible narrative.
>of how the bear needs the honey, how he goes here, and there looking for the
>honey, how he finds the honey, gets the honey, eats the honey, and then his
>hunger is gone.
>(I can go into more detail if anyone is interested ...)
Please do. Use e-mail if you think it more appropriate. Here it seems
to be doable with standard STRIPS-style planning and a data structure
for modelling the bear with a variable that decreases with time and
is only increased by food. The planning algorithm would use this and
other variables.
>Assuming for a moment that you have a dynamic plot built up on the idea that
>the computer maintained characters have desires or needs that are met partially
>through the actions of a player, there must be some kind of signalling
>mechanism that says some details that the character 'knows' to be true
>are no longer true.
>
>I think this is called the 'frame problem' in the artificial intelligence
>literature.
>
>Any thoughts about this ??
Perhaps object-oriented programming can help where semi-sentient objects
in the program update the variables in the game for the synthetic character's
benefit. It's consistent with at least a few views on how the mind works
to solve the frame problem.
>Dave (whi...@fwva.saic.com) US:(619)535-7764 [I don't speak as a company rep.]
>
Ice() ice@skynet.{uucp|tdkcs.waterloo.on.ca}
Phil Goetz
The frame problem isn't really concerned with whether object B is
still in location C. That's easy. Harder problems involve propositions,
i.e. if yesterday "John is hungry" was true, is he still hungry today?
If yesterday "John is tall" was true, is he still tall today?
Object-oriented programming is just a method. It is unreasonable to
expect a new method of programming to solve any hard AI problem.
A first reaction might be, "I'll have the food that John eats
send a message that John is not hungry. I'll have John keep track of
when he gets hungry again." Problem is, complete forward chaining
DOESN'T WORK. When your world gets big enough and complex enough to be
inferencing, you would spend all your time forward-chaining.
My theorem: The mind uses lazy evaluation. We don't try to determine
the truth-value of a proposition until we want to know it.
I imagine there is some forward chaining also, but less than backward
chaining. Partly this is because we are not deductive, and we would
be foolish to continue inferencing further along an ever less-and-less
likely path unless we already knew the unlikely consequent were in fact
true.
Gavin Inglis
>>I was thinking about the 'novel writing' program I wrote when I was at Rice
>>University (Bob Hearn was in the same class...) and realized there is a
>>different kind of programming needed to write a program that is interactive
>>and a program that tells a story.
>
>I think that a nontrivial story-telling program must necessarily be a
>superset of an interactive fiction program. You not only need all the
>data structures of an IF system, but then the program has to use them
>to create a plausible narrative.
Now that's interesting. I was thinking down opposite lines; that to
write a good interactive fiction program you need a storyteller at the
core. I reasoned: you need to be able to generate stories. The art of
interactive fiction is in chopping up stories and incorporating changes
of direction at appropriate points. (gross simplification, I know; and
I'm not sure I really believe that definition. Nevertheless...) So you
need to take your storyteller and interrupt it with directions
sometimes.
Therefore your interactive fiction program must be a superset of the
storyteller, not the other way round.
Any more thoughts either way?
--
Gav | I hate clever signatures.
<ai...@uk.ac.ed.castle> |