InterFace Manual
InterFace Manual
The program AI~WHEEL demonstrates the competence, adaptability, and mechanism of the UCPC. This software allows the user to create a custom universe/challenge by defining the sensory elements, shaping actions, and value assignments of a task environment. User options include interaction, fine tuning of the processing flow, and the disk storage of experienced "intelligent" entities. The program has no soft restrictions on the number of I/O conditions or respective particles used to define the target environment.
Of perhaps greater interest than the UCPC itself, is the method by which these independent but cooperative agents may be most efficiently combined to establish a fixed "multi-angulation" of Observer/Shaper Perspectives (OSP), thereby forming an overviewing and unified intellect within a defined Entity Task Universe (ETU).
AI~WHEEL is a basic program to demonstrate the mechanism and force of the UCPC
A FULLY ADAPTABLE MODULAR UNIT OF "INTELLIGENT" FORCE
The AI~WHEEL is a completely "transparent" basic program designed to display the power and mechanism of the UCPC. The user may select the simple "custom universe" option to define any environmental challenge in his own words, and the AI~WHEEL will "learn" to shape this environment according to user defined values. There are no soft limits on the size or complexities of this target task universe, however, intricate environmental challenges are more efficiently handled by a multiple of UCPCs operating from unique perspectives within the Entity Task Universe.
The UCPC Within the AI~WHEEL
The Unified "Cog" (UCPC) itself is a relatively small Processing Circuit. Most of the coding in the AI~WHEEL program pertains to defining, modeling, and displaying (for the users benefit) the UCPC's interaction within the designated universe. To use the UCPC independent of the AI~WHEEL, the modeling features (optionally including screen and file displays) may be deleted, and the target environment is interfaced at the designated three I/O terminals.
The use of a limited number of words, as assigned I/O particles in this demonstration of an AI/OPR circuit, is arbitrary.
The html INTRFACE describes AI/OPR, an autonomous agent (within the AI~WHEEL) that is prepared to discover whatever inherent environmental order may exist in a given stream of totally unfamiliar I/O particles, assigning a value polarized charge to each isolable flux cycle event.
Input and ouput particles need not be defined by the creator as words, nor in any other manner. Semantic definitions in the program AI~WHEEL are used as a medium within which to display the AI/OPR processing circuit. Words are unnecessary when an AI/OPR is linked directly to its environment through a given sensor/shaper array. I/O Particles are discover and labeled (as they are encountered) as isolable unique sensory quarks and minimal units of shaping force.
The indicators of value, used to set the polarity of a given cycle event, must inevitably be designated (though perhaps through a chain of association) by the creator of the AI/OPR. The raw data for such value detection must be perceivable by the entity.
The AI/OPR is the "first utility" of cognitive physics, but the interesting part of this new science is not the unified cognitive processing circuit itself, but the issues of "Where is it positioned?" and "What does it consider?" within the defined Entity Task Universe.
To Define a Unique Environmental Challenge
select: "Create a Custom Universe" from the Main Menu.
After naming a define-U.UNV file, the Creator is given an opportunity to specify a main field from among the IN-condition fields (lists of particles).
IN-Conditions (Sensory Abilities)
The creator of a new universe will be asked:
"Which is the main incoming condition
in this universe?"
This will be a number between one and the maximum number of IN-conditions within the given task Universe definition. This is usually the condition number of the grammatical subject of the incoming field. This incoming field is usually a sentence or paragraph, but may also be a mathematical or encrypted expression. (If your not sure which field is primary, Guess.)
The particles in the list of a Main Object/Conditions may be individually restricted, ie certain of these particles may be designated as not naturally occurring in the given universe, these being exclusively the byproduct of some given event/law.
So When entering the names of the particles in the main incoming condition, the creator will be asked:
"How many of this object occur naturally
in the opening universe?"
If this object is a raw (naturally occurring) material in the universe, then you will want to enter a positive number. If this object is produced as a byproduct of a given event description/law, then this number would be zero. If it is necessary to repeat an event a given number (n) of time in order to produce the first of this object, then this number (n) would be negative.
Environmental Continuity
Normally the In-conditions of any given environmental cycle are selected at random, unless the continuity switches are set to force any byproduct to be included in the next in-condition sequence, thereby considered by the AI~WHEEL in the next event cycle.
OUT-Conditions (Shaping Abilities)
When entering the particles of the OUT conditions, the creator will be asked:
"What is the Creator assigned value unit
cost of this action?"
This will normally be entered as negative number (or zero), this numeric value is the creator designated cost that will be associated with this action, represented in Universal Value Units (UVU).
The objects, adjectives, actions, tools, adverbs, and prepositions, along with their suffixes and prefixes, may be entered in any numbers, order, or combination, as phrases, sentences, or paragraphs.
Creator Assigned Values (laws)
The format of the laws is dictated by the definition of the IN and OUT conditions. When writing a law, the creator is prompted to specify or unspecify each condition IN and OUT, along with the subsequent resulting value attributed to this event, and/or any metamorphosis of a given IN-condition particle.
Laws Entry (Optional) Conditions Link
When entering the laws of a given universe, the user/creator may exercise the "link to" option by typing "l" [enter] in response to any IN or OUT condition query. The creator will be prompted to enter the corresponding condition number that designates the link. Establishing such a link will require equality between the two condition fields in order to qualify the law during goal and metamorphosis (UVU) evaluation of an AI~WHEEL cycle event. ie, If these two fields are not equal, this law will not be qualified or enforced in the evaluation of a given event cycle.
Event by Event Reports are Displayed for the Creator
Screen displays are provide in detail for each occurrence in the challenge universe. The AI~WHEEL (in default) will model the creator assigned environment, task, and laws, within computer space. The user will be able to monitor a normal screen display detailing each encounter/event that the entity experiences.[optional printout or save-to-disk files available on main menu]. The creator may even suggest (force) specific encounters, or reactions (teach the AI~WHEEL).
The creator may halt experience/events/action at any point and change the processing attitudes/policies of the a given AI~WHEEL by altering the user controlled parameters and Cognitive Track Switch (CTS) of the forty variables in the "User Parameters and Cognitive Track Switches Menu".
The user can save and recall any accrued intelligence. The user may remove a given perhaps mature entity from existence in computer space, and resurrect it at will (takes only a few seconds).
Using the AI~WHEEL as a Unit of Intelligent Force
"The AI~WHEEL when properly placed within a defined ETU constitutes computer hosted intelligence interfaced with a physical (environment) space."
You may translate the UCPC portion of AI~WHEEL basic coding directly into any given machine language of any robotic entity (*) faced with ANY real-time Environmental challenge. (This "hook-up" will require some programming skill, but if you can interface a CPU to a real-time hardware sensor array, then you probably already have some programming skills.)
(*) A robotic entity is being defined as any computer system with "i/o feedback looping" capability. ie, The robot entity must be able to affect, and then perceive this affecting of, its environment. In the optimum sensor array I recommend including voice i/o, and topographic vision. For optimum UCPC shaping action, I suggest including maximum mobility and a hand-like grasp ability.
To See, or Not To See?
Sensor arrays without visual input may yet foster very impressive entities. Vision (even topographic vision) is going to be a substantial processing concern in any environmental challenge. The deciphering of visual (reflected light) sensor graphics arrays is a valid problem, but perhaps somewhat off the point, which is the UCPC itself and the diversity of perspective that is possible within a given complex defined ETU.
The Creator Selects the First Perspectives on Order
AN UCPC is a subjective observer and shaper agent, linked to other agents by a common objective. The UCPC itself is a tool that discovers, mutates and reflects environmental order from a given perspective within the task universe. The solution for any task requiring intelligence is to place these tools in the most efficient OSP configurations within the ETU. It would be most effective to allow the UCPC to establish and test OSPs automatically (see CEM). This may in any case be associated with a RAModel of the ETU, but it is possible in some task environments to allow the CEM access and complete OSP freedom within the original physical ETU.
Real-Time Event-by-Event Creator Goal-Grading
The creator may directly grade the UCPC after each real-time event, by diverting Interface Terminal 3 (the subroutine in the main processing line designated "revaluation") to receive real-time event goal-value grades from the creator during each event cycle.
For instance, this grade may be communicated by voice/sound, visual signal, or data entry by any method, depending of the nature of the sensor array being employed by the UCPC. So it is possible to communicate with a UCPC by voice, provided the I/O array includes voice output and recognition capabilities.
To Install a Single UCPC Directly Into a Real-Time Task Universe. A real-time OSP will consider specific m&m directly within the Three part ETU:
(1)The Cognitive Space of the UCPC (*)
(2)The Target Environment
(3)The I/O Interface between 1. and 2. ie, the target environment and the computer space of the UCPC.
(*)A UCPC may (in any given OSP configuration) consider itself to be a portion of the task universe, optionally including its processing instruction(s) and/or record(s) of knowledge.
There are three real-time interface terminals within the program coding, points at which to splice in, to reroute the AI~WHEEL from a computer space model, to some real-time physical environment and task. Notice that this target environment may exist exclusively within computer space; for instance, if you wished to create an intelligent opponent within a game environment.
1. IN-conditions - hook up robot sensors here. Each sensor would correspond to an IN-condition, and each isolable distinct reading would represent a unique particle.
2. OUT-conditions. The influential force that the UCPC has on the task universe. Each shaping ability would correspond to an OUT-condition, and each isolable distinct command would represent a unique particle. (There must be a real-time feedback relationship between IN and Out, ie, The sensors must be able to perceive either directly or indirectly, the influence the UCPC's shaping abilities are having on the ETU.)
3. Revaluation. The grade and/or byproduct of the event (effect on the ETU). May optionally include designating any current changes in the I/O conditions as being a byproduct of the event cycle.
REVALUATION (awarding UVU)
In a real-time physical universe, value and metamorphosis become issues of observation. Value may be designated from any specific indicators within the ETU. However, this aspect must at some point be designated by the creator, regardless of how many links there may be in the chain of events ending in such a designation. You may attempt to achieve "perpetual motivation" by mimicking the survival/reproduction concerns of living creatures. CAUTION - Interface Non-Destructive Shaping Capabilities Only. UCPC objectives may conflict with the well being of other entities in the same physical space. In this respect, the resulting "intellect" can be dangerous to life and property. If you wish to avoid most of these kinds of problems, I suggest limiting the shaping abilities of UCPC inhabited robotic entities to non-destructive levels; or limit the target environment to computer models or harmless physical environments. Notice that such a model-environment would be ideal for competing with other UCPC guided entities, and that such competition would greatly accelerate learning.
A cycle event normally begins and ends with a sensor update of the task environment. The difference between these two definitions of ETU perceptions is associated with the self-shaping effect of the entire event. (Notice that a single flux event may be processed as the end of the last cycle, and the beginning of the next)
This byproduct or the "presumed self-induced immediate effect on the ETU" portion of the event is merely the difference between the pre-action status of the ETU and the after action status, within the given cycle event. Such changes, including metamorphosis, would be associated either directly or indirectly, along with any creator assigned value augmentations.
This does not mean that all events will invoke value, the award will perhaps be zero, or may be negative; furthermore the UCPC may, or may not, have incurred value depletion due to action (shaping) costs.
Hook-In One, or as many UCPCs as you like...
Install a UCPC wherever you may need an observer/shaper in the Entity Task Universe. Create an Interfaced committee (multiple) of uniquely positioned UCPCs to more quickly master a single challenging universe.
How many UCPCs does a robot need to successfully master a challenging universe?
Although a single UCPC may theoretically inevitably discover and reflect all perceivable (available) value-polarized patterns of matter and motion from a given OSP perspective, within a complex ETU, it is usually more time and record efficient to establish a multiple of UCPCs at various strategically placed OSPs within the ETU (placed either by creator design, or by another UCPC (CEM), either randomly or based on a similar history of trial and success).
One UCPC may eventually resolve even a very complex task universe, however you might find it more efficient to assign more than one UCPC to the ETU, thereby gaining additional perspectives on the m&m landscape of the task universe. This is a kind of "triangulating" on the patterned order of the ETU.
Triangulation is a two dimensional approach to exposing unique patterned order, revealed exclusively by observing from two combined perspectives; Notice that the ETU may be defined in any number of dimensions, and that there is no limit to the number of UCPCs that may be established in a given "multi-angulation" (the OSP array within the ETU).
It would be optimum to assign a UCPC group (committee) to install universal or custom designed UCPC at given points within ETU. The director (see CEM) of these installation would make proposals such as:
What we if made a comparison of the matter and motion of "this" and "that" portion of the ETU?
What if we put an eye here? - (an IN-condition of a new UCPC)
What if we put a hand here? - (an OUT-action of a new UCPC)
Any points and configuration of m&m within the ETU or model of the ETU may be used to designate an OSP. Notice that the IN-condition sensory points/configurations don't necessarily have to be in the same ETU proximity as the OUT-condition shaper points/configurations.
AN UCPC is a Local Decision Machine (LDM)
A Local Decision Machine (LDM) is any reasonably competent Simile Learning Machine operating within its given decision environment. The following is an overview of LDM (or UCPC) expansion, including a definition of OSPs within the ETU.
Mix Real and Modeled Portions of an ETU
OSPs may be assigned to any combination of real-time and computer-modeled m&m in the ETU. In cases, for instance, where it is not practical or possible to install an OSP directly within the ETU m&m, one may use a computer space model of this portion of the ETU. Such a model however, depending of how faithfully maintained, may of course be less reliable than a directly placed observer. Notice many RPX posts that are not possible or practical, directly within the original ETU, may prove possible and valuable within the RAModel.
COMMITTEE INTELLECT
A mature intellect is normally a multiple of UCPCs "manning" a number of OSPs within a single ETU...
So which UCPC handles an incoming flux or problem?
New flux is handled by the UCPC that recognizes (perhaps as valuable) the sensor particle string. A multiple of UCPCs assigned to a common task universe would each monitor a given flux within this universe, but of these, only one UCPC would recognize the greater number of the SPECIFIC particles defining the flux. If more than one UCPC did recognize a flux, one among these would probably have the higher value reactionary behavior goal-grade expectation.
Universal Value Unit (UVU)
The assignment of UCPC goal-grades by the creator is inevitably expressed in units of value awarded to the UCPC at some point in experience. The creator may of course designate any number of different kinds of value units, but in default, value is registered as the UVU.
Production alone does not necessarily depict value. The AI~WHEEL is programmed in default to produce x number of a unique products; x is equal to the number of possible different OUT-condition sequences of the AI~WHEEL. Value will only be assigned (to a given production event) if the production chain eventually leads to a valuable (and profitable after action-cost deductions) event.
Although it is possible to have a UCPC that does not consider UVU (perhaps dealing only with metamorphosis/byproducts) somewhere along the line, something akin to value will have to be associated to justify the action and metamorphosis in the production chain (unless the creator wishes to make production itself an end, and even this may be regarded as an indirect UVU designation).
Particle Definition Implosion
Expanding Condition/Particle Arrays
The UCPC is designed to discover and resect "intelligent" behavior sequences through simile reference of value polarized order. A single unit is capable of achieving competence in some very complex environments.
Theoretically there is no limit (beyond given hardware limitation) to the number of i/o conditions that one may assigned to a single UCPC, nor the number of particles within each unlimited condition. The creator may even implode the definition of a given particle by expanding the depths of "how the program considers the particle".
You may adapt this program to split each particle (or selected particles) into particle conditions, with particles of their own, designated "sub1" particles, which of course (in turn) may also be divided into conditions with "sub2" particles, and so forth, for as deep as is necessary to define the ETU.
Arrays will have to be increased accordingly, in the processing coding, wherever references are made to the condition particles affected by the expansion. Notice, there may eventually be little difference between conditions and particles, except perhaps in creator formatting mechanics.
As it is written, the AI~WHEEL universe is confined to computer space; this is what makes it necessary to define an environment to administer inclusive laws. However, when interfaced to a real-time environment, none of these definitions are necessary or even desirable. The UCPC will operate much more efficiently discovering and labeling its own object-action i/o conditions and subsequent observation of the event-value flux.
The UCPC in default associates value flux with the current event cycle. Laws are discovered in the same way as I/O conditions and particles (or sub particles), except they involve IN and OUT condition-strings. A perceivable increase in the value status of the target environment will be associated with the I/O strings, and will subsequently set the value polarity of the current event cycle record.
Condition Priority
In UCPC processing, a similar In-conditions sequence of specific particles is responded to with a known successful sequence of specific OUT-condition particles. But how does one decide which In-condition of an event should be considered first when searching for a similar historical event?
Obviously, not all aspects or conditions of an event are considered of equal importance when selecting a simile. The "importance" priorities of IN and OUT-conditions are methodically calculated by the UCPC, and updated after every event. The most important condition will be used as the first key for selecting similar historical events, and referencing parallel continuities and behavior patterns.
Given that condition priority is of great importance when referencing reactionary order through a simile:
How many ways are there to establish priority?
How many (and what kinds of) spectral perspectives, measurement, and relative associations are available?
The default method for determining condition priorities is to compare each event (as it occurs) to each historical record of events. Each condition is selected and compared to the corresponding condition of every historical event, if the selected condition particles are not equal, but all other conditions are equal, then a unit representing irrelevance is attributed to the selected condition.
For instance, if the selected condition was designated as the color particle specific of "red", and an historical event record was found with a color specific of yellow (or any not-red specific), and the remaining(*) conditions of the two events matched perfectly, this would attribute 1 unit representing irrelevance to the color-condition. The condition with the least number of irrelevancy citations is considered the most important particle in the definition cluster, and the largest scoring condition is considered least important.
(*) The entire remainder is considered in default, however, this may also be constrained by the creator or a given UCPC. Also, condition priority IN or OUT may fluctuate with condition specifics.
Condition priorities should be fairly consistent from a properly defined OSP. If condition priorities seem to be divided into more than one substantial portion, with perhaps more than occasional exceptions, then it would probably be most efficient to assign another UCPC. The subsequent sub-perspective (or new UCPC) is defined by the emerging group of exceptions.
Notice that each of these In and OUT priorities may be associated with a specific condition or particle. For instance it may be known that the most effective IN-condition priority is x, when the third OUT condition is particle z. If the fourth IN-condition is also y, then yet another condition priority sequence may be known to be most effective for IN-condition simile selection, reference, and resection.
Intelligent Perspectives and Overviews
Intelligence is the simile discovery, reference, and mutation of value polarized patterns, through the random or creator designed installation of modular observer/shaper simile learning machines (UCPCs) at revealing perspectives on the m&m within the designated Entity Task Universe.
An Entity Task Universe is:
1. The processing entity as it exists within computer space.
2. The target environment and its creator assigned objectives.
3. The I/O Interface between 1 and 2.
Some Familiar Applications
The Cart-Pole Challenge
To meet the cart-pole challenge the creator need only interface the obvious sensor and shaper i/o conduits, and arrange the polarized value designation according to the objective, ie an upright pole status would be of greatest value, regressing as the pole is perceived to be further from this perpendicular attitude.
I believe this simple challenge would require only one OSP, however, the question of the "cart-pole OSP configuration" is a revealing example of the problem of "single vs. multiple OSP assignments". Even in this minimally challenging task universe, a multiple of strategically placed OSPs would achieve complete competence more directly and efficiently that a single UCPC perspective on the m&m of the ETU.
Tic-Tac-Toe
Although it is possible to conquer the TTT challenge from a single perspective by writing a great number of laws, or expanding the "byproduct" to consider all nine 3-particle conditions (thereby incurring a large number of records), the most graceful approach would seem to be to configure eight perspectives (the straight-line possibilities) within this 100%-reliable computer model of the given 2d task universe.
This is interesting because although the most efficient OSP configuration (for such a simple challenge) appears to be a relatively high number (8) of OSPs, the observations, value polarized reactions, and inevitable conclusions of each UCPC will be identical, and therefore reducible to a single UCPC handling each OSP perspective separately during a given event cycle.
Chess
In a game such as chess, it would be perhaps difficult for the creator to know where to position OSPs within the task universe. A very good argument could be made that this question of "where and how to configure OSPs?" is indeed the problem. This is a stark example of the clear necessity for using an expansion manager subroutine to randomly configure OSPs (including their respective m&m configurations) from which to search for value polarized order within the ETU of the chess challenge.
Furthermore, with a challenge such as chess, I would suggest that the creator attempt random changes in the (again perhaps unclear) area of value assignments. Notice however that a change in value assignment would seem to indicate a necessity to restart the given UCPC from zero knowledge.
by
David Albert Harrell