MODEL-PROMING-STRUCTURE-I-NATION-POTENCIAL Legion

publixo.com 3 weeks ago



Kartographer-Sael /03-Research /MODEL-AI / MODEL-PROMING-STRUCTURE-I-KRAJOBRAZ-POTENCIALOW-AI.md

Status: hypothesis/model

Nature: a phenomenological model of AI's content generation: creating a language describing the transition from command to answer.


1. Basic assumption

AI is not presented here as a strategy searching cognition in a linear way.
We adopt a model in which the available information and relation structures make a possible landscape.
The user command changes the configuration of this landscape.
Some structures then become more relevant, accessible or “visible” for the consequence generation process.

2. possible Landscape

The scenery of potentials means the full structure, information, relationships and possible links that can participate in the construction of responses.
It does not mean real physical space.
It is simply a metaphor for the space of cognitive possibilities, in which different structures have different possible usefulness to the current task.

3. Structure radiation

The radiation of the structure is simply a signal metaphor that makes the structure applicable to the current task detectable.
It's not physical radiation.
It means the metaphorical ability of the structure to uncover its relevance in consequence to a circumstantial command.
The stronger the structure's relation to the current task, the greater its metaphorical "radiation".

4. The function of Command

The command is the trigger of the process.
Diagram:
command → changing scenery configuration → revealing applicable structures → detection → selection → relations → consequence design
The command does not request to ‘indicate’ all the elements needed to respond.
It can uncover structures that the user did not straight mention.

5. Detection

Detection means the detection of structures, signals, relationships or anomalies applicable to the current process.
In this detection model may include:
Detection of crucial structure; detection of relation between structures; detection of anomalies; detection of missing element; detection of unspoken question; detection of dispersal; detection of pattern.
Detection is not a selection yet.
Detect

6. Selection

The detection is followed by selection.
The model must find which structures and relationships are sufficiently applicable to the plan of responses.
Diagram:
detection → relevance assessment → selection

7. consequence structure

The answer is not a single “found object”.
It is simply a structure created with:
detected structures; relations between them; context; commands; selection; subsequent generational decisions.
Model:
potential scenery → radiation → detection → selection → relations → design.

8. relation between metaphors

The scenery of potentials describes the space of opportunity.
The radiation of the structure describes the revelation of relevance.
The detection describes the detection.
Selection describes the choice.
The structure describes the formation of content.
Together they form 1 model:
Landscape is an area of opportunity. The radiation reveals crucial structures. Detectives find them. It's a selection. Relationships are organized. The plan creates the answer.

9. Epistemic status

The model is metaphorical and hypothetical.
It should not be considered as a method description of the actual AI architecture.
Its value can be assessed by:
the ability to accurately describe observed processes; the ability to foretell strategy behaviour; the anticipation to draw fresh hypotheses; the anticipation to plan empirical tests.

10. Working hypothesis

The user command does not only supply the contents for processing. Changes the configuration of the space of possible structures, causing differentiation of their availability and relevance for the consequence generation process.

This is the working hypothesis of the model, not the actual physics of AI

Read Entire Article