Carnegie Mellon University’s Software Engineering Institute has released white papers on three pillars of artificial intelligence engineering: human centered, robust and secure and scalable.

The human-centered pillar of AI engineering seeks to ensure that AI platforms are developed in accordance with the ethical principles of the Department of Defense and other agencies, SEI said Wednesday.

The institute’s white paper on this AI engineering pillar has three focus areas: the need for designers and systems to understand context of use and sense changes over time; development of tools, processes and practices to scope and facilitate human-machine teaming; and mechanisms, methods and mindsets to engage in critical oversight.

The document on robust and secure AI focuses on the development of processes and tools for testing, analyzing and evaluating AI systems, improvement of robustness of AI systems and components and the need to design for security challenges in modern AI platforms.

The institute highlights three focus areas in its white paper on scalable AI and those are scalable management of data and models, scalable infrastructure and algorithms and enterprise scalability of AI development and deployment.

"These papers state the open questions we see in the field and identify gaps where work is needed. If we want to drive progress in the field, we need to start taking steps towards defining and answering these hard questions,” said Rachel Dzombak, digital transformation lead at SEI's Emerging Technology Center.

"By putting these pillars in place as AI system design and development starts, you're more likely to build systems that achieve mission outcomes,” noted Dzombak, who also leads the institute’s work on AI engineering.

These papers on the pillars of AI engineering stem from an SEI-led national initiative sponsored by the Office of the Director of National Intelligence aimed at establishing the discipline of AI engineering for national security and defense.

The Government Accountability Office (GAO) used input from government, industry and nonprofit experts to develop a framework on the responsible use of artificial intelligence technology.

GAO said Wednesday that it held a forum to gather input on AI oversight, in addition to interviews with AI experts and federal inspector generals.  The framework featured key practices that tackle the data, performance, monitoring and governance aspects of AI accountability.

Auditors, third-party examiners and other entities may use the framework to answer questions and adopt procedures that support responsible AI use.

GAO aimed to determine AI best practices with respect to federal agency users and organizations that design, develop, monitor and deploy AI platforms.

The U.S. Army has developed a set of software designed to help autonomous, robotic systems navigate and move through complex environments. The Scalable, Adaptive and Resilient Autonomy (SARA) program aims to develop algorithms, tools and methods that support autonomous navigation, the Army said Tuesday.

SARA has been running for a year, in which the program produced advancements in local planning, uncertainty models and adaptive sensor calibration. The effort supports the larger AI for Maneuver and Mobility Essential Research Program.

"What the SARA model allows us to do is drive the performers to integrate their solutions into a government-owned software stack that is compatible with research across the Army," said Brett Piekarski, chief scientist of Army Research Laboratory's (ARL) Computational and Information Sciences Directorate.

Eric Spero, collaborative alliance manager for SARA, said the model allows industry and academia to directly take part in ARL's autonomy pursuits.

SARA has also produced algorithms that reduce errors in long-distance travel, facilitate obstacle navigation and allow for high-speed movement in narrow areas.