↓ Skip to main content
tomrochette.com
  1. Artificial General Intelligence/

Problem solving algorithm

Author
Tom Rochette
History / Edit / PDF / EPUB / MD
Created: / Updated: / Status: in progress / Readability: 3 / 2 minutes read (~317 words)
artificial-general-intelligence 244
Table of Contents

Context
#

Learned in this study
#

Things to explore
#

  • How do “non-intelligent” agents solve their problems?
  • Can one solve problems without being aware of them?
  • Can one solve problems if they don’t understand them?
  • Let say you are an agent in a given environment and you want to solve problems, what components would you need to do so?

Overview
#

  • Two types of problems:

    • Ill-defined: No clear goals, solutions path, or expected solution
    • Well-defined: Specific goals, clearly defined solution paths, and clear expected solutions

  • Observe a problem

  • Determine the causes of the problem

  • Generate potential solutions

  • Test solutions

  • Evaluate and record results

  • Iterate

IDEAL
#

  • Identify the problem
  • Define the context of the problem
  • Explore possible strategies
  • Act on best solution
  • Look back and learn

PDCA/PDSA
#

  • Plan
  • Do
  • Check/Study
  • Act

Alternative
#

  • Determine micro tasks
    • How does one recognize which micro tasks are necessary? (seems to be based on prior experience)
  • Determine constraints (what needs to be done before something else can be done or what can’t be done)

Requirements
#

  • Language: A way to temporarily represent objects/entities and states in order to manipulate them
  • Causality decorrelation: The ability to extract the “true” causes and effects relations (what implies what).
  • Memory: To store prior experiments and their results. Memories are then reused in order to predict actions-effects (causality) when mentally manipulating models, which should be more energy efficient (and reproducible) than executing the action again.
  • The ability to simulate a sequence of action-effect

Evolution of problem solving with age
#

  • Reflexes
  • Random discovery (trial and error)
  • Understanding of causality
  • Pattern recognition
  • Reuse of developed algorithms
  • Learning from others
  • Fine-tuning of acquired skills

Level of abstraction by age
#

  • Equation with numbers and missing total
  • Variable manipulation
  • Spatial reasoning
  • Problem recognition and application of the proper tools to solve them problem

See also
#

References
#