The solutions to problems are represented as symbol structures. A physical symbol system exercises its intelligence in problem solving by search - that is, by generating and progressively modifying symbol structures until it produces a solution structure.
To state a problem is to designate (1) a test for a class of symbol structures (solutions of the problem) and (2) a generator of symbol structures (potential solutions). To solve a problem is to generate a structure, using (2), that satisfies the test of (1).
What makes a problem a problem is not that a large amount of search is required for its solution, but that a large amount would be required if a requisite level of intelligence were not applied. When the symbolic system that is endeavoring to solve a problem knows enough about what to do, it simply proceeds directly towards its goal…
Thus by this procedure [for solving equations], which now exhibits considerable intelligence, the generator produces successive symbol structures, each obtained by modifying the previous one; and the modifications are aimed at reducing the differences between the form of the input structure and the form of the test expression, while maintaining the other conditions for a solution.
For most real-life domains in which we are interested, the domain structure has not proved sufficiently simple to yield (so far) theorems about complexity, or to tell us, other than empirically, how large real-world problems are in relation to the abilities of our symbol system to solve them… It is likely that any system capable of matching [human] performance will have to have access, in its memories, to very large stores of semantic information… A particular, and especially a rare pattern can contain an enormous amount of information, provided that it is closely linked to the structure of the problem space. When that structure is "irregular", and not subject to simple mathematical description, then knowledge of a large number of relevant patterns may be the key to intelligent behavior.
First, each successive expression is not generated independently, but is produced by modifying one produced previously. Second, the modifications are not haphazard, but depend upon two kinds of information. They depend on information that is constant over this whole class of algebra problems, and that is built into the structure of the generator itself: all modifications of expressions must leave the equation's solution unchanged. They also depend on information that changes at each step: detection of the differences in form that remain between the current expression and the desired expression. In effect, the generator incorporates some of the tests the solution must satisfy, so that expressions that don't meet these tests will never be generated. Using the first kind of information guarantees that only a tiny subset of all possible expressions is actually generated, but without losing the solution expression from the subset. Using the second kind of information arrives at the desired solution by a succession of approximations, employing a simple form of means-ends analysis to give direction to the search.
My principal thesis today will be that the input is never into a quiescent or static system, but always into a system which is already actively excited and organized. In the intact organism, behavior is the result of interaction of this background of excitation with input from any designated stimulus. Only when we can state the general characteristics of this background of excitation, can we understand the effects of a given input.
Temporal integration is not found exclusively in language; the coordination of leg movements in insects, the song of birds, the control of trotting and pacing in a gaited horse, the rat running the maze, the architect designing a house, and the carpenter sawing a board present a problem of sequences of action which cannot be explained in terms of successions of external stimuli.
Let us start the analysis of the process with the enunciation of the word. Pronunciation of the word "right" consists first of retraction and elevation of the tongue, expiration of air and activation of the vocal cords; second, depression of the tongue and jaw; third, elevation of the tongue to touch the dental ridge, stopping of vocalization, and forceful expiration of air with depression of the tongue and jaw. These movements have no intrinsic order of association. Pronunciation of the word "tire" involves the same motor elements in reverse order. Such movements occur in all permutations. The order must therefore be imposed upon the motor elements by some organization other than direct associative connections between them.
This is true not only of language, but of all skilled movements or successions of movement. In the gaits of a horse, trotting, pacing, and single footing involve essentially the same pattern of muscular contraction in the individual legs. The gait is imposed by some mechanism in addition to the direct relations of reciprocal innervatioin among the sensory-motor centers of the legs. The order in which the fingers of the musician fall on the keys or fingerboard is determined by the signature of the composition; this gives a set which is not inherent in the association of the individual movements.
The remaining alternative is that the mechanism which determines the serial activation of the motor units is relatively independent, both of the motor units and of the thought structure… It seems to follow that syntax is not inherent in the words employed or in the idea to be expressed. It is a generalized pattern imposed upon the specific acts as they occur.
The problems raised by the organization of language seem to me to be characteristic of almost all other cerebral activity. There is a series of hierarchies of organization; the order of vocal movements in pronouncing the word, the order of words in the sentence, the order of sentences in the paragraph, the rational order of paragraphs in a discourse. Not only speech, but all skilled acts seem to involve the same problems of serial ordering, even down to the temporal coordination of muscular contractions in such a movement as reaching and grasping.
The finger strokes of a musician may reach sixteen per second in passages which call for a definite and changing order of successive feature movements. The succession of movements is too quick even for visual reaction time. Sensory control seems to be ruled out in such acts. They require the postulation of some central nervous mechanism which fires with predetermined intensity and duration or activates different muscles in predetermined order.
Consideration of rhythmic activity and of spatial organization forces the conclusion, I believe, that there exist in the nervous organization, elaborate systems of interrelated neurons capable of imposing certain types of integration upon a large number of widely spaced effector elements; in the one case transmitting temporally spaced waves of facilitative excitation to all effector elements; in the other imparting a directional polarization to both receptor and effector elements. These systems are in constant action. They form a sort of substratum upon which other activity is built. They contribute to every perception and to every integrated movement.
A quote from Pylyshyn, Computation and Cognition, page 79-80.
Historically, a major breakthrough in understanding the nature of control was the articulation of the idea of feedback through the environment. With that understanding, a certain balance of control was restored between a device or an organism and its environment. Although only the device is credited with having a goal, responsibility for its behavior is shared. At times, when the environment is passive, the initiative appears to come primarily from the device; whereas, at other times, the environment appears to intervene, and the initiative seems to go in the opposite direction. This notion of the responsibility for initiation of different actions is fundamental for understanding control. In most computer programs the most common idea has been that of control moving from point to point, or from instruction to instruction, in a largely predetermined way. This sequencing of instructions makes the notion of flow of control quite natural; branch instructions make it equally natural to think of passing or sending control to some other "place". When control passing is combined with a primitive message-passing facility - at minimum, a reminder of where the control came from, so it can be returned there later - subroutines become possible. And, because subroutines can be nested - that is, can send control to still lower subroutines, and so on, with the assurance that control will eventually find its way back - the notion of a hierarchy of control emerges. Miller, Galanter, and Pribram (1960), who saw the psychological importance of the idea of hierarchical subroutines, called them test-operate-test-exit or TOTE units and suggested that they be viewed as the basic theoretical unit of psychology, replacing the ubiquitous reflex arc. This idea has been highly influential in shaping psychologists' thinking about cognition.
Concepts and representation
If covariation is sufficient for representation then Brooks too accepts the need for representations. It is clear that by representation, however, he means symbolic, probably conceptual representation. Let us define a symbolic representation as one which can be combined and manipulated. This condition adds the notion of syntax to representation. To get systematic generation of representations it is necessary to have a notation that is sufficiently modular that individual elements of the notation can be combined to make molecular expressions. In this way, ever more complex structures can be constructed and used by a finite system.
There are many ways of thinking that do not presuppose use of an articulated world model, in any interesting sense, but which clearly rely on concepts. Recall of cases, analogical reasoning, taking advance, posting reminders, thoughtful preparation, mental simulation, imagination, and second guessing are a few. I do not think those mental activities are scarce, or confined to a fraction of our lives.
Nor do I think they are slow. When a person composes a sentence, he is making a subliminal choice among dozens of words in hundreds of milliseconds. There can be no doubt that conceptual representations of some sort art involved, although how this is done remains a total mystery.
Situationally determined tasks
For instance, humans, when putting together jig saw puzzles, may be said to be situationally determined if there is enough joint constraint in the tiles and assembled layout to ensure that they can complete the puzzle without wasted placement. No behaviorist theory can explain jig saw performance, however, because there is no readily definable set of structural properties - ie. stimulus conditions - that are the causes of jig saw placements - ie. responses. The agent is too active in perceptually questioning the world. On two confrontations with the same world the same agent might perceive different situations as present because it asked a different set of perceptual questions. These questions are a function of the state of the agent and its most recent interactions with the world.
We can say that jig saw puzzles are perceptually hard but intellectually simple. There is enough local constraint in the world to "determine" successful placement despite there being several tiles that can be successfully played at any moment… the situation contains enough information to pre-empt the need for lookahead… The perceptual problem is tractable in the sense that only a fraction of the visible world state must be canvassed to determine where to move… And what is most salient in the environment is usually discernable and economically detectable from the agent's perspective…
From this it follows that if a task requires knowledge about the world that must be obtained by reasoning, or by recall, rather than by perception, it cannot be classified as situation determined: