(you may click the
subfile number to be viewed, or scroll down)
This
file contains the following subfiles:
4 - testing in a variety of ways - models should make predictions
5 -
Why this paper?
(subfile 4: models should make predictions)
When the behavior involved is linguistic, the model will predict, for
example, the existence of words that might actually be found in the
dictionary, or the need for new parameters that must be added to the
definition-space. There is also a mechanism for optimizing the topology
of the space in which words’ definitions exist (this is equivalent to
saying that the algorithm can make certain predictions about the axes
in that space), and the program will produce some sentences that are
“legal” and some that are “correct” for the grammatical and reasoning
environment (this is equivalent to saying that it
makes predictions about combinations of symbols that are legal in a
grammar, or are sensible in a context).
When the behavior is musical, the form of the “prediction” would be a
finished melody: it predicts that some particular series symbols could
be mistaken for a human-created musical composition. When this product
is compared with
examples from the
training set, in performances heard by human observers, part of
the Turing Test for the presence of artificial
intelligence is completed. When in the robot-navigational realm, a
"prediction" would consist of series
of commands and a stated goal - evaluation would involve the distance
of the actual arrival from the stated goal.
(subfile 5: why this paper)
It is unpleasant to write about a project that has, in a sense,
produced no results as yet, but I can see that the road ahead is years
long, and experience has shown that isolation is fundamentally
counter-productive. Although I am working on a couple of related
programs for a software company, my “day job” in an unrelated field
prevents the type of collegial interactions most researchers enjoy, and
thus I am presenting my work in this form.