EXPERIMENTS WITH A THEOREM-UTILIZING PROGRAM,

Abstract

A discussion is presented of the issue of the importance for machines of acquiring skill as human beings do, by learning theorems, and developing sensitivity to the occasions on which it is appropriate to apply them. It is reported that a theorem-learning-and-utilizing program has been written and tested in a simple but paradigmatic problem domains. Also that although the program does solve problems, in some cases rather well, each version of it reached a threshold beyond which increase and modification of its theorem store did little to improve its problem-solving power. Experiments with the program suggest that there may be no practical way to achieve machines very skillful at solving difficult sequence-synthesizing problems unless they can be empowered with something like an effective theorem-utilizing ability. Further, the experiments point to many difficulties which stand in the way of empowering machines with such an ability in interesting problem domains, especially difficulties associated with pattern recognition and with integrating isolated previously proved theorems into general solution schemata. (Author)

Document Details

Document Type
Technical Report
Publication Date
Feb 05, 1964
Accession Number
AD0431056

Entities

People

  • Larry E. Travis

Organizations

  • System Development Corporation

Tags

DTIC Thesaurus Topics

  • Identification
  • Learning
  • Pattern Recognition
  • Recognition
  • Sensitivity
  • Sequences

Readers

  • Artificial Intelligence
  • Systems Analysis and Design

Technology Areas

  • AI & ML
  • AI & ML - Machine Learning Algorithms