With the elimination of the supercomputer simulation and connectionist models as candidates for the model architecture of the Organon Sutra, it was with keen disappointment that the design effort began to examine the least attractive approach, a knowledge engineered systematic modeled by some form of symbolic representation scheme.
In a symbolic representation, each independent element of a knowledge domain, together with the monikers for their interactions with other elements, and the rules with which new knowledge is acquired are all represented symbolically.
To be sure, symbolic representations have many strengths, mainly in their various utilities for argumentation. There are well developed constructs to form true deductive inferencing, with abilities to form possible constructs for inductive as well as abductive inferencing.
There are structures that can form axiomatic proof systems, with a defined semantics that deals with meanings assigned to symbols, and systems of syntax which establish how symbols stand in relation to each other.
In symbolic logic systems, there are competing notions of provability and notions of truth. Semantics generate meaning and truth, syntactics demonstrate truth preservation across inference.
And unlike the neural network models, it is easier to convey multi-dimensionality in a symbolic systematic, which is the connectionist models’ chief weakness, and the principle reason why ANNs have great difficulty with structured representations. Although it would be even better if this expressiveness were an implicit virtue of all symbolic representations, conveying multi-dimensionality is one of the more appealing factors of the symbolic language of mathematics.
However, symbolic representations suffer from several fundamental limitations when considered as the basis for a knowledge engineering systematic, the very least of which is the inherent difficulty in expressing concepts of time and change when implementing symbolic logics.
In the symbolic worlds of logic and mathematics, we start out with the rules with which those worlds behave, and then we imagine only those objects that obey the given rules. Because of this, these symbolic worlds are closed domains. The concept of an object that operates beyond the given rules is a non sequitur (Wiley Miller really did get it right!), and there cannot be exceptions to the given rules.
In the real world, on the other hand, we first learn of objects that inhabit that world, and only then do we begin to imagine the rules that they obey. Not only is the knowledge process in reverse, but for the real world, the knowledge domain is open. There is fundamentally no limit to the rules that can be imagined, and exceptions abound.
By itself, symbolic logic can only serve to move us from one truth to another. But because symbols possess zero dimensionality, there is no underlying finality of truth. Ultimate truth in the sanitary world of pure symbolic logic is a sophist ruse, and we must be wary of the seductive confusion of the ‘concept of proof’ with the ‘concept of truth’.
And when it comes to describing the ever-present apparent chaos of the real world, mathematics can be a difficult horse to ride. In closing his imaginative book on Emergence: From Chaos to Order, John Holland remarked that at the formal level of mathematics, there is only a limited body of forms that deal with nonlinearity. Almost all of the well-established tools of mathematics are built upon assumptions of linearity and additivity, and those parts that take nonlinearity as a subject matter usually depend upon linear approximations. This creates a lack of fidelity when a systematic relies upon mathematics as the principle language of definition in objectifying the real world. Much like logic, mathematics can only lead us from one proof to another, and this lack of fidelity to real world experience, where nonlinearities, singularities and apparent chaos is the norm, creates a disconnect in the expressiveness of mathematic models for the real world.
And it can be said that there is an even more fundamental limitation in the logics and mathematics when being used as a representation basis for any real world model building, in that there is no operator in either language that conceptualizes ‘memory’, or state. As soon as we introduce the concept of state into the lexicon of symbolics, we untether ourselves from axiomatic proof, at which time we clearly see where the lawless frontier border of artificial intelligence lies with respect to the gentrified neighborhoods of formal symbolics.
During the final analysis, however, symbolic systems did reflect one attribute which kept it in consideration. It is unknown how a purely symbolic representation of the Organon Sutra would exhibit gestalt abstraction, but the argument was made that if the process could define abstraction literally, then certainly it should be able to model it symbolically.
So with all of its detractors, the design effort did not leap to discard symbolic representations as an architectural basis for the Organon Sutra, even as it was recognized that as a standalone foundation, it too was insufficient to provide the functionality needed.
It was at this point that the design process for the Organon Sutra “went rogue” and began to draw analogies to hemispheric specialization in mammalian cerebral organs, which led to speculations that perhaps a symbolic representation system could have utility in conjunction with a complementary form of specialized processor.
Left-right hemispheric asymmetries in cerebral specialization have likely evolved to make optimal use of bilateral nervous systems. Where there is a convergence of purpose in evolving separate left and right nervous systems to control the motor and sensory functions of bilateral organisms, Nature takes advantage of every opportunity to create something better.
In the case of this convergence of purpose in creating bilateral nervous systems, Nature adds a divergence of functionality between the otherwise identical, lateral hemispheres of mammalian brains.
It is generally recognized that in human cerebrums the left hemisphere is typically dominant in language activities, logic processing, and symbolic manipulation, whereas the right hemisphere is typically dominant in spatial abilities, holistic thinking and non-language specializations such as musical comprehension.
The consequence of having two diametrically opposed dispositions in behavior living in a single head becomes a resultant (would some say reluctant?) synthesis of the two, an emergent behavior we characterize as intelligence.
Extending this analogy of hemispheric architecture to the design that was desired for the Organon Sutra, it seems evident that the symbolic manipulation systems that we have been discussing would be perfectly suited to model what is ascribed to as “left-brain thinking”.
But what form would a systematic take to model the enigmatic “right-brain thinking”? Since language has evolved predominantly in the left hemisphere of humans, our words are perfectly adequate to describe that hemispheres’ disposition, but they are wholly uncomfortable in the world of right-brain activities.
To gain a better understanding of the teasingly veiled processes behind the generalized compartment of the human central nervous system we call the right hemisphere, we have to look beyond the dichotomy between left-brain and right-brain, and consider what we can learn by examining the currency of exchange between the two, the currency of the human mind.
And that currency is what we call knowledge.
Just as one can tell a lot about a society merely by observing when money changes hands, we can tell something about the human mind by studying when knowledge is formed and exchanged.
GO TO NEXT DISCUSSION
GO TO TOP OF DISCUSSION
Copyright © 2019 All rights reserved