Apprentice Agents

Abstract

To become full-fledged members of human-robot teams, artificial intelligent agents must be capable of learning. In human teams the best learning method is apprenticeship. Robotic learning today does not address human-inspired methods like apprenticeship. We propose to apply the apprenticeship style of learning to agents. Agent apprenticeship will allow human team members to gradually develop trust in the apprentice~s ability to autonomously perform a designated set of tasks.We propose to develop an approach to one- or few-shot learning through mixed-initiative natural language dialog in which the agent, while processing language inputs from human interlocutors, recognizes that it has a need to learn, carries out the learning process and augments the knowledge resources in its long-term memory accordingly. The agent then connects the newly learned material to the content of its long-term memory to make it available for reasoning and action.The proposed research will combine theoretical, algorithmic and descriptive work with computer modeling and validation of theoretical claims. The proposed research is synergistic with two ongoing efforts supported by ONR: a project (N00014-16-1-2212) devoted to communication for enhancing human-robot collaboration (Brian Scassellati, PI) and a project (N00014-16-1-211) devoted to naturalistic dialog understanding by agents (Marjorie McShane, PI). The former will provide an excellent environment for experimentation and evaluation of the results of the proposed research. The latter will contribute enhanced component technologies for the proposed here.The results of the proposed research will advance the stated goals of the Third Offset strategy and specifically contribute to ONR research thrusts such as Language in Action, Autonomy and Proactive Decision Support.

Document Details

Document Type
DoD Grant Award
Publication Date
Feb 03, 2017
Source ID
N000141712218

Entities

People

  • Sergei Nirenburg

Organizations

  • Office of Naval Research
  • Rensselaer Polytechnic Institute
  • United States Navy

Tags

Fields of Study

  • Computer science

Readers

  • Agent-Based Social Robotics and Mobile-Assisted Learning in Virtual Environments.
  • Research Science/Academic Research

Technology Areas

  • AI & ML
  • Autonomy
  • Autonomy - Human-Robot Interaction