Adaptive Training and Education for Adaptive AI Systems
Abstract
Progress in the research and development of AI systems that can adapt to unstructured environments has made the adoption of intelligent autonomous systems (IAS) a priority for the U.S. Department of the Navy. However, developing a large workforce that can effectively reprogram or debug IAS systems is not only difficult through conventional curricular pipelines but also insufficient because many of the operators need expertise in other, role-oriented curricular specializations in addition to effectively using IAS in their missions.The objective of this project is to enable broad segments of the workforce to learn how to re-task and debug adaptive AI systems without having to obtain college degrees in computing. Its research outcomes will impact broad sections of the workforce and support the Navy#s strategy towards IAS by lowering barriers to entry in working with AI systems. They will also have a direct impacton the Naval IAS Vision of integrating IAS as trusted members of the Naval Enterprise by enabling unified, adaptive training platforms that can autonomously adapt to a given AI system as well as to a given trainee and provide appropriate training.Our key contribution is a new paradigm for Adaptive Training and Learning for AI Systems (ATLAS) that automatically adapts to the user and to the IAS. This project aims to develop new AI techniques that will help a trainee learn how to re-task a given AI system to achieve their desired objectives as well as to debug errors in their own instructions and to understand the scope of safe operability of the given AI system.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- May 15, 2023
- Source ID
- N000142312416
Entities
People
- Siddharth Srivastava
Organizations
- Arizona State University
- Office of Naval Research
- United States Navy