
Overview
By Richard Weyhrauch
This note is called:
[OVERVIEW]
The notations used in this note is described in
[NOTATION]
Under development
My goal is to build an artifact that can think. This goal drives everything
I am currently working on. This bote discusses my understanding of what this
entails.
To repeat myself:
GOAL: to build an artifact that can think
I use the slogan 'an artifact that think' to engender a mental image of a
thing that exists in our world and, at the very least, whose interact with us
is sufficiently rich that we will be willing to call it both knowlegdgeable
ans even intelligent. This is a strong requirement because it includes at
least as good an understanding of its world as a young human child (let's say
a five year old). In fact, at a more adult level we might ask if we could
imagine ourselve hireing it as an assistant: could it fill out an application;
could I send it to the library to get a book; or maybe ask it to read a paper
and deliver to me a summary.
I will call the part of it that thinks its head. Thus I will say
things like 'I wanty to build a head'. or 'What is in its head'. There are
three ideas that drive my thoughts.
Structural Similarity
Structural similarity is the belief that the organization of
the mind is essentially the same throughout our species.
Accordingly, an [agent]s model of other [agent]s have the same
structure, but certainly not the same content, as an [agent]s
model of itself. This principle is employed at all levels,
and hence the model that the user is believed to have about
an [agent] has the same structure as [agent]s model of the user.
However, an [agent]s model of itself need not be the entirety of
the [agent] -- for instance, its model of its "mental" processes
may be different from the set of actual processes that comprise it.
Solipsism
Heads are finite
As an artifact in the world it must have finite extent. After thinking
about it this may seem like a trivial observation, but in fact it has profound
consequences.
With this in mind, the system was designed from an
"egocentric" standpoint -- from Fred's view of the world.
This approach is appropriate when one's goal is to construct
an [agent] that can perform in real situations and can use
perception. The opposite approach, one I
attribute to
logic, is to describe people from the "outside". In this
vein, we see some researchers in Artificial Intelligence,
following philosophers of logic, attempt to characterize
"knowledge" by incorporating both concepts in the mind and
objects in the real world into their formalisms.
Properties of Heads
Our goal requires us to understand what might be used to build a head and
what things the head knows about.
Heads contain mental images.
Heads cannot be modeled by theories
- heads are finite
- heads are updatable
§ x.The development of ideas
- just mental moidels of things - nou=ns
- Classification
- new types of classification
- numbers/counting
- naming/language
- classes of objects (as an abject)
- theories
- relations
- parts of speach (eg verbs)
- attributes/adjectives
- attributes of verbs (adverbs)
Is there a history of parts of speach?
- abstract notions
eg, love pain, racissm, ...
words fo these?
meta theory - Meet Fred
- new things and view
new things and new meta theory
- scienticic method
- practical computation
- sets
is this analysis new> - places to look
- Child developmewnt
- History of language
- history of ideas
The purpose of education
- To transmit mental images of society
- shortcut the need for reinvention
- the invention of education
- writing - books - computers
§ x.
What is language?
|