Critique: PROMPT: An Innovative Design Tool
Murthy, Addanki
*
Here is another attempt at bridging the gap between weak and strong
problem solving methods. Analyze-Modify, which has been used in many
of the systems we've studied so far, has been identified (in this paper)
as a powerful, general but inefficient solution. the general gist of
Prompt is to provide a level of explicit knowledge that is one step
above first principles, the term the authors use to denote fundamental
domain knowledge, to save the overhead of reasoning with it.
I questioned the claim that Prompt could "invent" the concept of a tube
in the introduction. I was very pleased at how well the example problem
and presentation of the system explained and justified this claim. Most
similar statements in other research work do not achieve this... or
I lack the intellect and insight needed to understand the explanations.
Bravo to Murthy and Addanki for speaking down at my level.
A point is made of the prototypes in Prompt containing structural
descriptions and how this helps analyze the behavior of components
from first principles. They fail to mention that having a description
of structure is necessary for Prompt to be able to change the structure
at all.
An interesting concept is evaluation of a design and attributing
the result to the various characteristics of the design in order
to choose a modification to produce a better design. This is not
trivial task. Prompt does analysis using first principles. A different
approach for the same problem could be using back propagation in
a neural network. Back propagation is (roughly speaking) a method for
propagating error back through a network, adjusting individual nodes
until an acceptable value (or evaluation) is achieved at the output
end. This requires that nodes that effect the outcome be identified
and modified. The nifty part is that the knowledge to do this is not
explicitly stated in the system. How well could this technique be used
in Prompt?
What do the author's mean when they say "...envisioning apparatus within
Prompt has been implemented..."? In what sense does Prompt picture
itself? Also, the paragraph before the Conclusions section seems to
belong in that section. Not that that matters all that much I guess.
*
S. S. Murthy & S. Addanki,
PROMPT: An Innovative Design Tool.
In: Expert Systems in Computer-Aided Design,
(Ed) J. S. Gero, North-Holland, 1987, pp. 323-341.
|