What is KitBit?
The outline of the model fundamentals presented here is only intended to provide a quick look at some of the ideas and concepts which have led to its conception and design. More information about the model will soon be available under licence.
- Model Fundamentals
- Philosophical Bases
- Mathematical Bases
- Conceptual Bases of Intelligent Knowledge
- Conceptual bases of Intelligent Behaviour
- Creative intelligence
Model Fundamentals
- It is a model which attempts to emulate some of the principal abilities of the natural intelligence.
- It considers the human intelligence as an advanced stage of the natural intelligence of animals, plants, and even nature itself.
- It registers experiences and uses them to make intelligent assumptions.
- It continually verifies the validity of these assumptions and stores the verified or rejected conjectures as being useful for later predictions.
- It creates a dynamic structure of data which is modified based on accumulated experience.
- It attempts to emulate the algorithms and internal heuristics which the human brain uses, based on the paradigm we propose, as well as the way in which it internally constructs knowledge.
- It uses time as an independent variable, creating a dynamic system with self-learning capacities.
- Its general structure allows it to create abstract concepts and establish categories.
- It uses and distinguishes between two independent yet interrelated sub-models as emulators of natural intelligence:
- Intelligent knowledge generator KIT (Knowledge Imprint Tracer)
- Intelligent behavior generator BIT (Behavior Imprint Tracer)
Philosophical Bases
The model is based on the following axiom: the human system of perception and reasoning, both external (senses) and internal (cognitive cerebral processes), is indifferent to the truth: its raison dŽetre and functional criteria are its utility for the subject.
The model rests on the naturalist conception of epistemology: ideas do not preexist things, but the inverse is true.
The systems of categories must take into account the time factor in order to concur with the evolutionist assumption.
Phylogenetic knowledge inherited by each subject in its genes is what permits the development of ontogenetic knowledge through experimentation and communication with its surrounding environment.
Both use the same mechanism of trial and error to create knowledge.
Ontogenetic knowledge is capable of experimenting virtually (imagine) before carrying out any real tests. The imagination permits the elimination of erroneous tests before experimentation. For this reason, its learning process is much more efficient than the phylogenetic process which does not possess this capacity.
Mathematical Bases
We define the complexity of a system as the length (in bits) of the smallest program which is able to reproduce it.
We define the intelligibility of a system as the logarithm of the quotient of the length (in bits) of its complete description and its complexity.
The intelligibility of a completely random event is zero. Such an event is unintelligible (incomprehensible, incompressible, algorithmically random). Any other type of event with intelligibility greater than zero will be intelligible (comprehensible, compressible, algorithmically reducible).
Axiom: The existence of natural laws (universal constants or restrictions on possible events) implies that nature always produces intelligible events.
If we know with sufficient precision the initial conditions of a system with an intelligibility greater than zero, then an algorithm capable of predicting the evolution of this system over time exists.
The precision with which it is necessary to know the initial conditions of the system to correctly predict its evolution is inversely proportional to its intelligibility.
In any formal axiomatic system (FAS), there are true maxims which are indemonstrable (incompressible). Logical reasoning does not lead to all theorems. There are "theorems" which consist only in mere description of facts. Since it is proved to be so, such "theorems" must be assumed as new axioms of the FAS. This way the set of axioms of the FAS keeps growing on and on.
The model registers experiences, selects them and classifies them according to their complexity, compresses them (comprehends) and stores them in a dynamic memory system which includes the very algorithms and heuristics used for comprehension.
The configuration of the dynamic memory structure allows it to be used continually in new tasks of registration, selection, classification and compression (comprehension) of experiences.
Conceptual Bases of Intelligent Knowledge
The model is founded on the following axiom: all knowledge is hypothetical or conjectural.
The analysis of registered experiences leads to rational predictions.
The growth of knowledge comes from learning, which in turn is a consequence of the elimination of errors.
We understand an error as being a theory or conjecture which produces a prediction that is refuted by a provable reality.
The theories constructed upon experimental results lead to the choice of new experiments. In other words, they determine that it is useful to experiment with the objective of obtaining new theories or to improve upon and augment the existing theories.
Formulations which are experimentally impossible to refute do not pertain to what we consider as the field of intelligent knowledge.
Conceptual bases of Intelligent Behaviour
The model is founded on the following axiom: a living organism is a dissipative structure far from equilibrium, bound to the Second Law of Thermodynamics.
Part of the entropy generated in the living organism can be dissipated in the environment. It is even possible for the entropy of the system to decrease without violating the Second Law. This allows the spontaneous formation of states of equilibrium of a system which has not reached its maximum entropy.
Complex systems, once some previous conditions are given (multiplicity, diversity, connectivity and autocatalysis) spontaneously tend to evolve towards higher levels of organization. At some point in this process, the system could be considered as a living organism.
Living organisms differ from inert systems in that they contain, as a fundamental part of their internal structure, "operational information" composed of instructions carried by their genes. These instructions constitute a program which allows the organism to react to its environment.
Living organisms with complex brains also contain a program which this "operational information" prints upon them. This program can modify itself in accordance with the interaction of the subject with its environment during his life time.
Intelligent behaviour is that which allows the genes carried by the subject, and consequently the subject itself, to survive. That is, it maintains them in equilibrium far from their state of maximum entropy.
Intelligent beings are able to make predictions or conjectures and can "travel forward in time" or "foresee". This capacity allows them to act in a way that diminishes their own entropy and that of their vital surroundings, with the result of increasing the entropy outside its system in greater proportion. From this perspective, it could be said that the principal aim of intelligent behaviour is to "slow down time".
Creative intelligence: Integration of intelligent knowledge and behavior
We call creative intelligence or simply intelligence, that which integrates intelligent knowledge and intelligent behavior.
Intelligent knowledge
- Evaluates the intelligibility of the systems it attempts to analyze.
- Discards the analysis of those systems with intelligibilities which are incompatible with the calculative, temporal and algorithmic capacity available to the subject.
- Predicts, for the selected systems, future events using patterns borne out of experience.
- Evaluates the certainty of the realized predictions by constructing a dynamic memory structure with intelligent knowledge.
- This dynamic memory structure permits it to learn intelligent knowledge. In this manner, it can improve the efficiency of future predictions.
Intelligent behavior
- Determines actions, for each system analyzed, which modify the future events predicted by intelligent knowledge.
- Discards those actions which are beyond the reach or possibilities of the subject.
- Chooses, among the selected actions, those which minimize the increase of entropy in each system and, when appropriate, maximize its reduction.
- Evaluates its success in choosing actions by constructing a dynamic memory structure with intelligent behavior.
- This dynamic memory structure permits it to learn intelligent behavior. In this manner, it can improve the efficiency in choosing future actions.