\documentclass[11pt]{article}

\usepackage{graphicx}
%\usepackage{rotating}

%to kill bld in label for 209:
 \def\descriptionlabel#1{\hspace\labelsep #1}

%\topmargin=-.5in
%\oddsidemargin=0in
%\evensidemargin=0in
%\textwidth=6.5in
%\textheight=9in

%\def\online{\raggedright\textwidth=4.6in\let\bullet=o}
%\online  % use this with dvi2tty to help make online version
          % It will still need manual fixing up, but this is a start

\begin{document}

\title{{\bf {\huge Preface}}}
\author{Selmer Bringsjord\\{\small {\tt selmer@rpi.edu} $\bullet$ {\tt
www.rpi.edu/$\sim$brings}} \and
Dave Ferrucci\\{\tt ferrucci@us.ibm.com}}
\maketitle

\section*{The Marriage of Logic and Creativity}

This book marks the marriage of logic and creativity.

While it may be true that incompatible humans often wed, there are doubtless
unions of a less palpable sort that can never even come to pass.  Such is the
case, by the lights of many, for precisely what we are about herein. 
Creativity and logic?  {\em Married}?  Upon hearing of our plans, seven years
ago, to harness theorem-proving technology in order to create a computer
program able to autonomously generate belletristic fiction, a rather famous
novelist informed us that creativity and logic are as far apart as the east is
from the west (and he proudly quipped that even such a metaphor is beyond
logic, and hence beyond machines).  Just an anecdote, yes, and just the
opinion of one, but the truth of the matter is that this attitude is widely
(and often fiercely) affirmed.  Creativity is generally regarded to involve
breaking the kind of rigid rules standing at the heart of logic; creativity,
at least of the artistic variety, is commonly identified with the emotions and
the ``irrational."  Freud, whose specific claims are today a bit tenuous,
remains a seminal figure for often getting at least the tenor of things
right.  Freud believed that creativity is the link between art and play, and
requires the ``suspension of rational principles."  He wrote that ``The
creative writer does much the same as the child at play.  He creates a world
of phantasy which he takes very seriously --- that is, which he invests with
large amounts of emotion --- while separating it sharply from reality"
(\cite{freud.creativity}, p.~144).  However problematic Freud's rather dark
theories may be today, here he is simply making an observation that cannot be
doubted.  But the issue is whether such sophisticated play can in the end be
reduced to logic.  Is the play of
Joyce and Tolstoy and Updike and Tyler at bottom logic in action?

Many used to ask different question:  namely, Could a computer ever beat the
best human chess-player?  With Kasparov brooding and Deep Blue and his silicon
cousins improving every week, many
are {\em now} asking:  Could a computer beat all human grandmasters {\em
time and time again in normal tournament play}?  To this the both of us
unhesitatingly answer in the affirmative (as should, we daresay, anyone
who knows a thing or two about the dizzying ascension of raw computing
power on this planet --- though by our calculations it will nonetheless
take a decade for machines to achieve such metronomic triumph.)
Then
again, when it comes to AI, we're a rather sanguine duo:  we're inclined
to believe that (e.g.)
\begin{itemize}
\item NASA will run successful missions to Mars and
other planets largely on the strength of ``immobots," HAL9000-like AIs
that will control the ships in question.
\item AI-controlled cars, safer than their human-controlled counterparts,
will be available sooner rather than later.
\item General house-cleaning
robots will arrive --- again, sooner rather than later.
\item Even now, the bulk of medical diagnosis can be carried out by
computers, at an accuracy level surpassing all but a small number of human
diagnosticians; and in the future, machine diagnosis will reach a point
where it is downright {\em irrational} to consult a human M.D.~first.
\end{itemize}
And so on.
But, again, what about creativity?  Robotic drivers may be securely in
our future, but Einstein, G\"{o}del, Tolstoy, Turing, Shakespeare, Plato,
Cantor,
$\ldots$ --- could machines ever reach {\em their} rank?  Could we ever
build a genuinely creative machine?

We seek to answer ``the creativity question" --- not from the comfort of
our armchairs, but from the workbenches in our laboratories.
Specifically, we seek to ascertain whether or not literary creativity is
the sole province of humans by attempting to {\em build} artificial
authors.  The first fruit of our labor, five years in the making (with
another half-decade prior to this one devoted to less ambitious systems),
is Brutus$_1$, a storytelling agent specializing in narrative that involves
betrayal first and foremost, and also self-deception and other literary
themes.  The mind of Brutus$_1$ is revealed in the book you're holding.

\section*{From Chess to Literary Creativity}

In our experience,
the public is quite comfortable with the notion that a machine can play
invincible chess --- because even those who know nothing of the niceties
of search algorithms intuitively grasp the
mathematical fact that chess, at bottom, is utterly mechanical, that if
one can ``look far enough ahead" the game becomes trivial.
%Likewise, while the general
%public may be reluctant to bring their symptoms to a machine, there
%this same public seems to grasp the mechanical, rule-based nature of
%diagnosis (as evidenced by the fact that lots of people purchase simple
%computerized medical diagnosis systems for their personal computers).
%(Sometimes those who have toiled mightily to make
%first-rate chess-playing machines find this belief more than a tad
%disturbing.)
On the other hand, given the reaction of the public to
Brutus$_1$'s prowess as reported in the media (as evidenced by
a persistent stream of rather emotional communication we receive),
we think it's safe to say
that while we (and many other AIniks, e.g., Douglas Hofstadter
\cite{fluid}) merrily press ahead in the attempt to reduce creativity to
computation, the lay mind is fundamentally disturbed by the prospect of
creative machines. In presenting the anatomy of Brutus$_1$'s brain herein, we
will sooth the souls of those who, hearing about his 
%(we use `his' rather than `its' in order to remain
%sensitive to {\sc Brutus}$_1$'s intimate relationship to the late, 
%corporeal
%Brutus, who was of course male)
exploits, fear that humans will soon have nothing over machines.  It will
become crystal clear below that Brutus$_1$ should give his human creators
rather a lot of credit.  Put in terms of our terminology, we say that
Brutus$_1$ has weak, rather than strong, creativity.

What we call ``strong creativity" is what might be called
``raw origination."  Raw origination is akin to creation {\em ex nihilo},
and though this form of creativity may well be impossible, the fact of the
matter is that the {\em concept} of creating something from nothing is
very real not only to monotheists, but also to many hardheaded scientists
who have pondered creativity.  The paradigmatic example is Margaret
Boden, arguably the world's leading authority on computational
creativity.  Boden \cite{boden.dartnall} distinguishes between a brand of
creativity associated with the novel combinations of old ideas (she gives
the example of the Lennon-McCartney arrangement of {\em Yesterday}, marked
by the unprecedented combination of a cello with music of this type), and
a type of creativity in which something utterly and completely new is
produced (e.g., non-Euclidean geometry).  Computers, of course, have no
trouble with the former type of creativity.  The latter type is somewhat
more difficult for them.  It's exceedingly hard to see how a computer
could, say, autonomously discover a new class of numbers through new
proof techniques, which was one of Cantor's novel achievements.

The distinction between strong and weak creativity isn't a new one.
When Alan Turing, one of the grandfathers of computer science and AI, proposed
that if a machine could pass his famous ``imitation game" (in which a
computer passes if it's linguistically indistinguishable
from a human; the game is now known as the ``Turing Test"), we humans should
immediately conclude that such a machine can genuinely think
he considered an objection from Lady Lovelace that was given
on the strength of raw origination.  She argued: ``Computers will
never be creative, for creativity requires {\em originating} something, and
this is something computers just don't do.  Computers do what they are
programmed to do, nothing more."  (Turing presents his imitation game, and
discusses the Lovelace objection, in his
\cite{tt}.)

Suppose for the sake of argument that Lovelace is correct. 
Even so, the other sense of
creativity, ``weak creativity," remains intact.  Weak creativity has
its roots in the ``operational" notion of creativity devised by
psychologists.  For example, E.~Paul Torrance, who more than any other
psychologist has probed the nature and concrete signs of
creativity, holds that $x$ is to be deemed creative just in case
$x$ scores well on the dominant test for creativity in children and
adults: The Torrance Tests of Creative Thinking.\footnote{See
\cite{torrance.test} for the test itself.  For reviews of the test, see
\cite{chase},
\cite{swartz},
\cite{treffinger}.}
This test comes in both ``visual" and
``verbal" forms.  In the visual form, test-takers are asked to draw
pictures (often by enriching existing sketches); in the
verbal form, test-takers are asked to write --- creatively.
For example, one of the activities subjects engage in in the verbal test
is the following.

\begin{small}
\begin{quote}
Most people throw their tin cans away, but they have thousands of
interesting and unusual uses.  In the spaces below and on the next page,
list as many of these interesting and unusual uses as you can think of.
Do not limit yourself to any one size of can.  You may use as many cans
as you like.  Do not limit yourself to the uses you have seen or heard
about; think about as many possible new uses as you can.  (From the
verbal version of \cite{torrance.test}.)
\end{quote}
\end{small}

\begin{figure}
\centering
%\centerline{\psfig{figure=s3g.xfig.ps,height=2in}}
\includegraphics[width=3in]{s3g.xfig.ps}
\caption{\label{s3g}S$^3$G}
\end{figure}

After the Torrance Test is administered, one can send it out to be
professionally judged.  Our aim on the problem of
literary creativity is to build an artificial agent capable of producing
stories that would be scored as highly creative by human judges in the
dark as to whether or not the stories they receive are from humans or
machines.  One of us (Bringsjord) has refined this scenario into what he
calls the ``short short story game," or just S$^3$G for short.  The idea
is simple; it is summed up in Figure \ref{s3g}.
A human and a computer compete against each other.  Both receive
one relatively simple sentence, say:  ``Barnes kept the image to
himself, kept the horror locked away as best he could."  (For
a much better one, see the ``loaded" sentence shown in Figure
\ref{s3g}.)
Both mind and machine must now fashion a short
short story ($\approx$ 500 words) designed to be truly interesting; the
more literary virtue, the better.
Our goal, then, is to build an artificial author able to compete with
first-rate human authors in S$^3$G, much as Deep Blue went
head to head with Kasparov.

Unfortunately, this goal is too tough to
reach, at least for the foreseeable future; it may even be a goal that
is forever beyond the reach of a machine.
Our more immediate goal is
therefore to build a machine capable of passing a less demanding
Torrance-like test:  that is, a silicon author able to generate stories
that would be regarded creative, even if these stories are well below
what a muse-inspired member of {\em homo sapiens} can muster.

\section*{How?}

How does one go about building such a silicon author?  Our answer comes in the
following pages.  At this point we'll mention only one property we believe
a good
story generator must have:  {\bf wide variability}.

There a many dimensions
over which a story can vary.  Plot is only one of them. 
Characters, settings, literary themes, writing style,
imagery, etc.~--- these are other dimensions, and there are many more. 
Generally speaking, belletristic fiction has very wide variability
across these dimensions.  Mark Helprin's latest novel is likely to have a
rather unpredictable plot traversed by rather unpredictable characters in
rather unpredictable settings tossed by unpredictable mixtures of love,
revenge, jealousy, betrayal, and so on, as reported in prose with a
cadence and clarity rarely seen; and one of the chief effects of it all
is to conjure unforgettable images in the reader's mind. (One of us  is
haunted weekly by the image of the lost gold in Helprin's {\em Memoirs
From The Antproof Case}.)  At the other end of the
spectrum fall formulaic fiction and film; here the variability is
narrow.  Some romance novels, for example, fail to offer wide variability
of plot and characterization:  it's the same character types time and
time again, dancing hot and heavy to the same choreography.
(If
Brutus$_n$, some refined descendant of Brutus$_1$, is to soon find
employment at the expense of a human writer, in all likelihood it will be
as an author of formulaic romance and mystery.)

Whether or not a story generator can be implemented to achieve
wide variability hinges on what we call {\bf architectural
differentiation}.
%and robust knowledge-bases.
A story generation system has
architectural differentiation if for each substantive
aspect of the story that can
vary, there is a corresponding distinct component of the technical
architecture that can be parameterized to achieve different results. 
While we owe many debts to the pioneers who have come before us in the
field of story generation, it's safe to say that their systems failed to
enable wide variability via architectural
differentiation.

From the start, our approach has been to
bestow Brutus$_1$ with a counterpart to {\em every} substantive aspect of
human literary genius.  One of the ways to encapsulate this approach is
to say that Brutus$_1$ is designed to satisfy what we call `the seven
magic desiderata' for a successful story generator, namely,

\begin{small}
\begin{description}
\item[D1] {\em Give proposed rigorous accounts of strong creativity a
run for their money.}  An impressive
storytelling AI is one that satisfies, or at least comes close to
satisfying, proposed sophisticated accounts of {\em strong} creativity. 
{\sc Brutus}$_1$ does this:~as we show in our forthcoming book, the system
qualifies as capable of raw origination on Margaret Boden's definition of
this concept.
\item[D2] {\em Generate imagery in reader's mind.}  An artificial agent 
aspiring to be counted among the literat\'{\i} must be able to spark
significant
readerly imaging.  (Sometimes even literary fiction can earn
classification as such despite displaying ordinary prose.  Victor Hugo's
{\em Le Miserables} is a case in point:  The writing is simple (relative
to other immortals, anyway), but what readers
can forget the scenes set in the sewers beneath Paris?)
\item[D3] {\em Situate story in ``landscape of consciousness."}  A good 
storytelling AI 
must produce stories having not only a landscape of action, but also a 
landscape of consciousness, that is, a landscape defined by the mental 
states of characters.
\item[D4] {\em Mathematize concepts at core of belletristic fiction.}  No
artificial agent will lay claim to being counted literarily creative
unless it processes the immemorial themes (e.g., betrayal) at the heart of
literature; and such processing can presumably come only if the themes in
question have been formalized.
\item[D5] {\em Generate genuinely interesting stories.}  A true artificial
storyteller must produce genuinely interesting stories.
Among
the things that readers find interesting are particular topics like
sex and money and death (as the well-known cognitive scientist Roger
Schank has explained:~\cite{schank.interestingness}), and also classic
themes like betrayal, ruthless ambition, and unrequited love.
\item[D6] {\em Tap into the deep, abiding structures of stories.}  Any 
truly 
impressive artificial author must be in command of story structures that 
give its output an immediate standing amongst its human audience. 
For Brutus$_1$, these structures take the form of what are called `story
grammars.'
\item[D7] {\em Avoid ``mechanical" prose.}  Last but not least, there is
a point we have yet to face up to:  an artificial author must produce
compelling literary prose.
\end{description}
\end{small}

The seven magic desiderata are cashed out in
Brutus, a rich and highly differentiated system architecture for story 
generation.  Brutus$_1$ is the current implementation of the Brutus
architecture.

\section*{Why?}

Finally, a question interviewers and member of the audience have asked us
time and time again through the years:  Why do it?  There are three general
reasons, two theoretical, one practical.  The first theoretical reason for
investing time, money, and talent in the quest for a creative machine is to
work toward an answer to the question of whether we ourselves are machines. 
If the creative side of human cognition can be captured by computation, then
it's surely likely that we are at bottom computers.  (The more quotidian side
of human cognition can presumably be mechanized.)
The second theoretical reason for our work is to
silence those who believe that logic is forever closed off from the
emotional world of creativity.  Brutus$_1$ is Vulcan through and through,
utterly devoid of emotion, but he nonetheless seems to have within his
reach things that touch not only our minds, but our hearts.  The practical
rationale for our endeavor is simply that machines able to work alongside
humans in arenas calling for creativity would have incalculable worth.  A
machine able to write a feature-length film, or create and manage the
unfolding story in an on-line game, would be, we suspect, pure gold.

\vspace{1in}

\noindent S.B.~Troy NY / D.F.~Yorktown Heights NY --- July 1998


\newpage

\begin{thebibliography}{99}

\bibitem{boden.dartnall} Boden, M.~(1994) ``Creativity and Computers," in
{\em Artificial Intelligence and Creativity} (Dordrecht, The
Netherlands:~Kluwer), pp.~3-26.

\bibitem{chase} Chase, Clinton I.~(1985) ``Review of the Torrance Tests
of  Creative
Thinking," in Mitchell, James V., ed., {\em 9th Mental Measurements
Yearbook, vol. II} (Lincoln, NB:~Buros Institute of Mental Measurement),
1631-1632.

\bibitem{freud.creativity}
Freud, S.~(1959)
{\em Creative Writers and Daydreaming}
(London, UK:~Hogarth Press and the Institute of Psychoanalysis).

\bibitem{antproof} Helprin, M.~(1995) {\em Memoir from the Antproof Case}
(New York, NY:~Harcourt Brace).

\bibitem{fluid} Hofstadter, D.~(1995) {\em Fluid Concepts and Creative
Analogies} (New York, NY:~Basic Books).

\bibitem{schank.interestingness} Schank, R.~(1979) ``Interestingness: 
Controlling Inferences," {\em Artificial Intelligence} {\bf 12}:~273-297.

\bibitem{swartz} Swartz, Jon D.~(1988) ``Torrance Tests of Creative
Thinking," in Keyser, D.J. \& Sweetland, R.C., eds., {\em Test Critiques,
vol. VII}, 619-662.

\bibitem{torrance.test} Torrance, E.P.~(1966) {\em The Torrance Tests of
Creative thinking:~Technical-Norms Manual} (Princeton, NJ:~Personnel Press).

\bibitem{trabasso} Trabasso, T.~(1996) ``Review of {\em Knowledge
and Memory:~The Real Story}," Robert S.~Wyer, ed., Lawrence Erlbaum,
1995, {\em Minds \& Machines} {\bf 6}:~ 399-403.

\bibitem{treffinger} Treffinger, Donald J.~(1985) ``Review of the Torrance
Tests of Creative Thinking," in Mitchell, James V., ed., {\em 9th Mental
Measurements Yearbook, vol. II} (Lincoln, NB:~Buros Institute of Mental
Measurement), 1632-1634.

\bibitem{tt} Turing, A.~(1964) ``Computing Machinery and Intelligence,"
in Anderson, A.~{\em Minds and Machines} (Englewood Cliffs,
NJ:~Prentice-Hall), pp.~4-30.

\end{thebibliography}

\end{document}