Ensured Reliability for Artificial Intelligent Systems
Abstract
Artificial intelligence (AI)/assistive automation applications are becoming more common in the Army's developmental technologies and systems. Ensuring these systems perform reliably is critical to the success of the Army's evolving mission. In addition to traditional hardware and software aspects, development will also need to account for human and environmental interactions within the systems. New areas regarding the data pipeline and AI algorithm training become instrumental in the reliability of AI systems. Previous reports have examined AI-specific design-for-reliability activities and identified potential failure modes in AI systems. This report expands on those efforts and discusses AI system elements that are necessary to mitigate reliability risks. These elements will form the basis for future reliability risk assessment methodology.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 26, 2023
- Accession Number
- AD1217511
Entities
People
- Alexander Marchioni
- Barry Hum
- Nathan Herbert
- Patiana Theragene