Untitled

\documentclass{article}
\title{The Nature of Thought}
\author{Matthew Brace}

\begin{document}
\maketitle
\tableofcontents

\section{Introduction}
Many philosphers, scientists, students and doctors have spent
countless hours, months and years working to divine the true nature of
thought.  While there has been growth and learning, we still have very
little understanding of how and why we think.  I will not be solving
the mystery, nor will I be providing any radical new insight.
Instead, I will explore the most basic interactions that can be
considered thought and present a system to model them within a
programmable computer.
\subsection{Purpose}
There are numerous artifical intelligence systems today, from the
very specific to as generic as possible.  This is not a way to replace
them.  What is explored here is a base system for providing learned
behavior from external stimuli.  While overly broad for traditional
machine learning applications, and perhaps any directly beneficial
uses of artificial intelligence, it should provide a sound basis to
begin exploration into developing a system capable of higher level thinking.
\subsection{Goals}
At conclusion, we will have fully detailed a system capable of
mimicking traditional learned behaviors.  It should be sufficiently
generic that any task that can be learned through input, response and
feedback can be implemented using this system.  It should require the
minimum definition of input, output and measurable values to produce a
viable system for use.
\subsection{Learned behaviors}
The focus of this paper is on simulating learned behaviors.  The
capability to apply previous results to subsequent situations is a key
component of forming higher level thought.  Additionally, it has the
most direct applicable benefits for common uses of artificial
intelligence today.
\subsection{Foundation for AI systems}
The system covered is designed to be usable once completed.  It will
not fulfill the hopes and dreams of a strong Artificial Intelligence,
capable of unbounded learning.  Instead, it forms the necessary base
to allow advancement in other areas of simulated thought.

\section{Instinct As A Base}
According to Merriam-Webster, instinct is defined as
\begin{quote}
1 : a natural or inherent aptitude, impulse, or capacity ``had an
instinct for the right word''

2 a : a largely inheritable and unalterable tendency of an organism to
make a complex and specific response to environmental stimuli without
involving reason

2 b : behavior that is mediated by reactions below the
conscious level
\end{quote}
For our purposes, definition 2a is the most suitable and is what I
will use when referring to instinct.  It is the responses to stimuli
that we contain at birth.  It is the responses that are exhibited even
if the stimulus had never been encountered before.  It is the
recognition that hunger relates to the need to feed, thirst is a need
to drink and pain is injury to the body.
\subsection{What is instinct?}
From this definition of instinct, the actions of an organism are
guided by the responses their local system has to stimuli.  It is a
physiological response to stimuli.  As an unlearned, untrained and
innate reaction, instinct could be qualified as a physiological
response.  However, as can be seen in humans, instinct can be
overridden by conscious thought.  Religious fasting is one such
example of using conscious thought to perform actions counter to
instinct.  This means that the reaction is not one of specific action,
merely overwhelming desire to perform a specific action.  I will not
explore the mechanism by which instinct works within organisms.  I
will make an assumption that instincs exist for the purposes of
survival of the organism and survival of the organism's species.
\subsection{Instinct as a preprogrammed response}
When something is said to act on instinct, it is an immediate response
that has no signs of conscious thought.  A ``short-circuit'' response
that requires significant effort to not follow.  It provides the base
reactions to stimuli that affect the continued existence of the
organism and it's species.  Due to the way in which instinct
functions, it appears to not be based on some form of intelligence.
If you place your hand on a hot stove, immediately you will move your
hand away without the appearance of any decisions.  Without instinct
to remove your hand from the source of pain, the process would be
fast but not nearly as immediate.  Instinct, then, is simply a special
case of thought, one with immediate responses.
\subsection{Other behaviors built off of instinct}
An organism's first experiences are based only on instinct.  If the
system is to be unique to each organism, and the experiences of the
organism define the system, there must be a base ``bare-bones'' system
that each organism begins with.  This is the framework for all learned
behaviors.  With instinct as this base system, it helps elucidate why
it is a special case of thought and why each organism within a species
appears to have the same instincts.
\subsection{Learned behaviors are a result of stimuli}
With instinct as merely a special case in thought process, we can
naively assume that all thought derives from stimuli.  This is
obviously a naive assumption since we are capable of thought
independent of stimuli.  However, as a defining, base system, this
assumption allows us to accurately model learned behaviors.  Further
systems that work in concert with the stimuli, analyze, response
system would provide the higher level thought.
\section{Thought as analysis of memories}
\subsection{Actions determined by previous similar experiences}
\subsection{Previous experience retained as memories}
\subsection{Each memory includes the thought process}
\section{Memories as a weighted graph}
\subsection{Memories all contain similar components}
\subsection{Stimuli can evoke a memory by familiarity}
\subsection{Weighting on edges relates the similarity}
\section{Traversing the weighted graph as analysis}
\subsection{Entry point}
\subsection{Edge determination}
\subsection{Nodes saved for processing}
\subsection{Processing the nodes}
\section{Disambiguation of potential actions}
\subsection{Rank against current stimulus}
\subsubsection{Similarity}
\subsubsection{Favorableness of outcome}
\subsection{Highest rank action is executed}
\subsection{If there's a tie, select most recent}
\section{Memory summarization and memory exclusion}
\subsection{Summarization based on similarity}
\subsubsection{Creatures don't traverse every memory naturally}
\subsubsection{Near-perfect matches reinforce actions}
\subsubsection{Near-perfect matches should be deduplicated}
\subsubsection{Weighting for deduplicated nodes represent the number summarized}
\subsection{Memory exclusion}
\subsubsection{Nothing ever truly forgotten}
\subsubsection{Use relative threshold of similarity to not follow an edge}
\section{System Stimuli}
\subsection{Categories of input types}
\subsection{Each input is distinct}
\subsection{Each stimulus can have different expressive values}
\subsection{Each stimulus has a preprogrammed reaction}
\subsection{Preprogrammed reactions are strong weighted memories}
\subsection{Actions for desirable, undesirable and indifferent qualities of stimulus}
\section{System Responses}
\subsection{Methods of response preprogrammed}
\subsection{Each method of response is a function of the input}
\subsection{Methods of response can use any output path}
\subsection{Once response is sent, immediate outcome is measured}
\section{Arbitrary valuation of outcome}
\subsection{Must rank outcome for analysis to have meaning}
\subsection{Ranking determines behavior}
\subsection{Needs/wants can be programmed}
\subsection{``Un-needs'' (conditions to avoid) can be programmed}
\subsection{To truly approximate organic thought, likes/dislikes must be mutable}
\end{document}