Information Technology Reference
In-Depth Information
(i) Omit perception: the system is incapable of representing internally
environmental regularities.
(ii) Omit memory: the system has only throughput.
(iii) Omit prediction, i.e., the faculty of drawing inferences: perception
degenerates to sensation, and memory to recording.
Second Example: If the conceptual linkages of memory with the other two
faculties are removed one by one, nolens, volens “memory” degenerates first
to a storage and retrieval system and, ultimately, to an inaccessible storage
bin that is void of any content.
After these reductiones ad absurdum I shall now turn to a more con-
structive enterprise, namely, to the development of a crude and—alas—as
yet incomplete skeleton of cognitive processes.
III. Cognitive Elements and Complexes
I shall now develop my thesis in several steps of ascending complexity of
quality, rather than of quantity, beginning with the most elementary case of
apparent functional isolation of memory but of zero inferential powers,
concluding with the most elementary case of functionally unidentifiable
memory, but of considerable inferential powers. Throughout this discussion
I shall use examples of minimal structural complexity for the sake of clarity
in presenting the argument. I am well aware of many of the fascinating
results that can be derived from a rigorous extension of these minimal cases,
but in this context I feel that these findings may divert us from the central
issue of my thesis.
My first case deals with the computation of concomitance. The detection
of concomitances in the outside world is of considerable economic signifi-
cance for an organism immersed in this world, for the larger an equivalence
class of events becomes the fewer specific response patterns have to be
developed by the organism. The power of inductive inference rests on the
ability to detect concomitance of properties, and—as was believed until not
long ago—the efficacy of the conditioned reflex rests on the ability to detect
concomitance of events.
The principle of inductive inference is essentially a principle of general-
ization. It says that, since all things examined that exhibited property P 1 also
exhibited property P 2 , all as yet unexamined things that have property P 1
will likewise exhibit property P 2 . In other words, inductive inference gen-
eralizes the concomitance of properties P 1 and P 2 . In its naive formulation
the “conditioned reflex” can be put into a similar logical schema which I
will call “Elementary Conditioned Reflex” (ECR) in order to establish a
clear distinction between this model and the complex processes that regu-
late conditioned reflexive behavior in mammals and other higher verte-
brates. However, I shall return to these in a moment.
Search WWH ::




Custom Search