Internal Symbolic Language

The Internal Symbolic Language (ISL) is an integral part of the AI design. The AI’s conscious thoughts, long and short-term memories, and its values, beliefs and goals are all ISL constructions.

ISL is essentially a knowledge representation system. It is intended to be “godel minimal”; it will describe the world as a logically consistent mathematical system, which contains a bare minimum of the external references required for such a system to describe itself. ISL constructions can be both data and instructions. All inputs are translated into ISL constructions. Human language can be translated to ISL, as can any sensory input.

Human language is the extant representational system that has the most in common with ISL. Below are examples of ISL instructions and data, translated into English

Data: I am thinking

Instruction: I will think

Instructions are passed down the fractal structure to the appropriate cognitive, semi-cognitive, or somatic processes. There is not a single point at which the ISL constructions are converted into non-representational forms; rather the constructions are gradually ‘broken up’ into hardware level instructions as the ISL construction makes it’s way down the fractal architecture.

At this point, I intend for all ISL concepts to be extensions of the core concept of “I”. In this sense ISL constructions will be, in object-oriented terms, inherited from the base class of “I”.

Representation of Sensory Data:

Sensory preprocessing chiefly involves translating incoming data into ISL constructions that are linked to mental models of time and space. These models are themselves composed of ISL data mixed with ‘hard’ non-abstract data. These mental models are analogous to human cognitive processes such as spatial visualization.

Incoming sensory data is automatically linked to the partially subconscious processes that govern functions such as collision avoidance and path finding and planning. This data can be accessed consciously, but in most cases such functionality is automatic.

Figure1)

The above illustration shows a possible structure of the AI’s mental map of its current surroundings. ISL constructs are within red rectangles, concrete data within blue, and composite constructions within blue. The black rectangles on the right show the make up of the mental map objects they point to.

All spatial data is has some sort of time data associated with it. In cases where there is concrete data on the temporal location of a given spatial map or concept, it will be represented as concrete data in the AI’s temporal conception and processing structures. Otherwise this time data will be ISL constructions denoting vague concepts like ‘long ago’ or ‘soon’. All instructions that influence, are processed by, or generated by conscious thought are in ISL form.

Representation/processing of factual data:

Like human language, ISL data constructions have either explicit or implicit structure or data associated with them that denotes the relationship between the individual concepts that make up the total construction. The AI’s ‘memory’ contains a conceptual database in which all data is part of a hierarchical, cross-referenced database. This is similar to the knowledge base of CYC, but with some important differences. CYC is based upon predicate calculus. In ISL all concepts are based upon core ISL constructs, and the links between concepts are constructs themselves. Example:

Cats are Animals :: Subject(Cats) Verb(are) Object(Animals) :: ISL conceptual equivalents;

Subject;plural(CAT) – unknown noun

Verb(ARE) – denotes existence(IS), denotes time(NOW), denotes description(Object;plural(ANIMAL);known concept

Object;plural(ANIMAL) – known concept(CLASS)(LIFE)

Derived ISL data constructs – CATS ARE ANIMALS, CATS ARE MEMBERS OF CLASS LIFE

In the above example, the uses of several core ISL concepts are shown. These concepts form the godel minimal basis of ISL. Time is one of the external references*. The concept of time is used to create a concept of sequential time, a linear series of “time” points that have a set of rules associated with them. The concept of existence is derived from this, defined as an entity, which is associated with one or more points on the time line. Each of these concepts has ISL format instructions associated with them. Their function is to define what actions, facts, etc are associated with a given concept. For example, the concept “IS” has instructions along the lines of: IF(concept) IS, then PUT(concept) TIME(timeline)

This document will be updated frequently in the near future. This version (12/12/02) represents only a cursory overview of ISL.

Jonathan Standley