connectionism (idea)

The theory of mind is a long debated topic throughout philosophy, many individuals have sought an answer and this has led down many paths. An enduring characteristic of philosophy is that it is an originator of thought, it provides a springboard for ideas. A basic structure for philosophical progress paralleled with scientific advance could be represented as: an individual makes some claim A, another makes claim B, and still another may produce a claim C; as progress builds and scientific advances begin to support one over the others, then those less well evidenced theories will be “pruned” away. By this method you can trace the lineage of the theory of mind, through platonic dualism and Aristotelian structure, through to Galilean mathematics and Cartesian dualism. A unique branching of the philosophical tree happens here in the split of Berkely’s idealism and Hobbesian/Lockian early materialism. This is of particular interest simply due to the fact that the split remains, with the mind lying in the balance, and the scientific method playing the role of the swords of the duel.

Thesis Section

A primary concept of the thesis of this paper is the idea of a “threshold neural net complexity”. Robots can be constructed which exhibit base animal behavior, similar to bees, simple insect behavior is of a certain “neural net complexity”; it is representative of a level of intelligence. Another premise of the thesis of this paper is that it must be assumed that there is no mind, in the classical idealist sense, we are material objects, nothing more. Given that we are material and are not in any way animated by an illusory force, it can be assumed that we are on the same scale as insects or rocks for that matter, regarding intelligence as we are all composed matter whose actions are guided by a neural network, we are all at some level of “neural net complexity” (although the lack of a brain/CNS limits the ability of the rock to be determined as intelligent). The central thesis of this paper is that there is a certain level of neural net complexity at which wonderful things happen; a structure which we will call a mind emerges. In the scientific method, specifically in psychological research when categorizing a certain characteristic you must operationally define what observations can be detected and recorded; this approach is what should be used in the question of mind. When you look at a piece of matter, there are certain characteristics which denote a level of intelligence: social interaction (this is a large category as many of what we would operationally define as intelligent behaviors lie in this domain), tool usage, and complexity of communication abilities are a few examples of these characteristics. By observing those characteristics as compared to the structure of their neural network (in regards to neural net complexity/size, as well as the complexity of the CNS), a parallel between perceived intelligence and neural complexity becomes evident. By this method we can assign a certain level of neural net complexity as a “threshold neural net complexity” dictating the presence of "mind", so to speak.

Implications

The implications of this are twofold, in that it separates the concept of mind from our concept of humans while allowing us to keep our “cozy ideas“ about us possessing a mind, and it creates an intriguing paradigm in regards to the development of intelligence. By designating a certain level of neural net complexity as a “threshold neural net complexity” it allows for other living animals to be regarded as being in possession of a mind. The truly interesting conclusion led to by this line of thought is that; given that insects are governed by the same system as we are, simplified, and that robots can exhibit equivalent observable behaviors as paralleled to neural net complexity, denoting comparable intelligence to their insect counterparts, then robots can have minds; having crossed the threshold level in neural net complexity as paralleled to observable, intelligent behaviors. The beauty of this is in the parsimony of this concept: inanimate objects can have a mind, its simply a matter of possessing a neural network which exceeds the threshold value; and exhibiting behaviors which denote intelligence as can be seen in the swarm intelligence based robots.

The central idea follows from the precepts of implementational connectionism, in that it seeks an accommodation between connectionism and classicism. The identifying feature is the concept of a “threshold neural complexity” at which one could consider the neural network (living or non-living) as possessing a “mind”.

Objections

Honestly, I feel I’m unqualified to create an argument of this scope, as this specific area of philosophical work (that of consciousness, intelligence, and the mind in general) is of an overwhelming nature; that being said I find myself resolved to discover a solution which will allow the best of both worlds, as it were - it depends solely on the ability to differentiate between “living” and “intelligent”, as concepts relating to the aforementioned “presence of mind” (having crossed a certain threshold neural complexity). Therein lies the primary difficulty with the argument; the evidence we have greatly shows that in order for a neural network to have sufficient complexity to denote a degree of intelligence comparable to the threshold value, it must be biological in nature. This is an undeniable argument against this paper’s thesis, as there is very little evidence as of yet regarding computer neural networks which could be considered to have crossed the “threshold complexity” and exhibited observable intelligent behaviors. Simple neural nets based on weighted connections and symbolic representations are capable of facial recognition tasks, as well as numerous basic linguistic communication tasks. These are indicators of intelligence, and assuming we are no more unique than bees in regards to what animates us (our neural net), and given that robotic neural nets can simulate behavior more complicated than those of insects, it would seem to follow logically that inanimate groupings of circuits, set up in a neural network fashion with sufficient complexity, could think, learn, create, rationalize, theorize, communicate, remember, etc…Basically, exhibit all the characteristics we view as indicators of intelligence, as well as having the corresponding parallel of a complex neural structure.

The issue is a redefinition of “mind” as an arbitrary but consistent measurement of intelligence, rather than a symbolic reference to “That which thinks”; the terminology in the English language with regards to the mind as related to human intelligence is too ingrained to remove entirely, so a compromise must be made - the redefinition of it’s meaning.

Another key objection lies in the inherent complexity of the brain and the difficulty in replicating it artificially. A brain is made up of several billion neurons suspended in glial support cells, each neuron has one receptor site at which any number of synaptic connections from other neurons can be taken. This alone represents an enormous number of potential brain states, however, there are 12 identifiable neurotransmitters which act through the synapses whose function we are aware of, and there are several more whose purpose we are unaware, and it is suspected there are yet more that are entirely undiscovered, this expands the number of potential brain-states from hundreds of billions of combinations to trillions of potential brain-states. The complexity inherent in a biological neural system makes replication via artificial means seemingly insurmountable, and therefore serves as an effective argument against this line of thought. The response to this lies in our developing technology, in that as time goes on and we discover more about biological neural nets, and as advances in artificial neural network creation become more and more sophisticated - time will tell in regards to this argument.

The argument is heavily dependant on other arguments, it is heavily reliant on the assumption of matter consisting of one substance (monadic) and behavior being governed by the same rules that apply to a rock, a flower, or a monkey. Also, this thesis is heavily dependant on the assumption that living matter is no different from non-living matter, in that it is a relation of atoms, regardless of animated characteristics, all matter is matter; the defining factor is the presence or lack of a central nervous system or “neural network”.

Having recognized the numerous flaws in this theory, I still feel it creates an interesting paradigm for how one could look at the mind. The implications of the theory’s principles are far reaching, in that having differentiated between “living” and “intelligent”, as well as differentiating between possession of, or lack of, a mind (in the sense of the threshold neural complexity), it allows an objectively defined view of what dictates intelligence and discernable qualities dictate the presence of “mind”. The elegance of this theory is that it allows us to keep our terminology regarding the mind intact with a slight redefinition of what a “mind” is.

In the end, the ever growing and expanding tree of philosophical knowledge will be the judge; as time passes, and evidence compiles, some theories lose credence and are pruned away and forgotten, while others are supported and grow further, addressing new problems in their respective disciplines. This is my branch, prune away.