The Marvin Minsky Institute for Society of Mind Theory

Abstract

Building on Genesis's self-reflection capability we will perform experiments where agents learn how solve problems in new and creative ways, learning by example and instruction, and distilling general procedures and recipes. Such systems will require reflective capabilities of various kinds in order to recognize, abstract, and align what they learn. To better understand the interface between language and other senses, we will experiment with biologically plausible models of vision, including architectures that can integrate top-down and bottom-up information. With a similar integrative approach, we will pursue human-scale models of language that can use knowledge and multi-modal senses to constrain meaning. We will integrate a new level of hypothetical reasoning capabilities into our cognitive models, enabling a range of more complex behaviors including forms of imagination, moral reasoning, evaluating, and planning. We believe all of this will be an important step toward understanding the competences necessary for human intelligence.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Dec 01, 2022
Accession Number
AD1190041

Entities

People

  • Davis Randall

Organizations

  • Massachusetts Institute of Technology

Tags

DTIC Thesaurus Topics

  • Air Force
  • Artificial Intelligence
  • Artificial Intelligence Software
  • Automata Theory
  • Cognitive Science
  • Computational Linguistics
  • Computational Neuroscience
  • Computational Science
  • Computer Languages
  • Computer Programming
  • Computer Science
  • Computer Vision
  • Computers
  • Information Processing
  • Language
  • Linguistics
  • Machine Learning
  • Military Research
  • Natural Language Understanding
  • Natural Languages
  • Reasoning

Readers

  • Computational Linguistics
  • Neural Network Machine Learning.
  • Systems Analysis and Design