AI Engineering: Building the Discipline and Growing an Ecosystem

2021 Year in Review

AI Engineering: Building the Discipline and Growing an Ecosystem

Most artificial intelligence (AI) applications fail, sometimes spectacularly. The drive to achieve one-off capabilities precludes a broader, disciplined approach that would enable the rapid uptake of AI demanded by the Department of Defense (DoD).

To mature AI practices and help national defense and security agencies adopt AI, the SEI has begun formalizing the field of AI engineering, much as it did for software engineering in the 1980s. AI engineering is an emerging field of research and practice that combines the principles of systems engineering, software engineering, computer science, and human-centered design to create AI systems in accordance with human needs for mission outcomes.

In October 2020, the Office of the Director of National Intelligence sponsored the SEI to lead an AI engineering initiative to guide the development of a multiyear research and development roadmap and develop capabilities based on partners’ core competencies. Bolstered by its recently formed AI Division, the SEI leverages its researchers’ expertise in AI; its deep knowledge of the government technology space; its status as a trusted federally funded research and development center; and its relationships with government, the armed services, industry, and academia. In 2021, partners from these spheres collaborated with the SEI to develop three initial pillars of AI engineering: AI systems should be scalable, robust and secure, and human centered.

By putting these pillars in place as AI system design and development starts, you’re more likely to build systems that achieve mission outcomes.

Rachel Dzombak
Digital Transformation Lead, SEI AI Division Photo of Rachel Dzombak

The SEI’s government partners cited scalability challenges in the private sector, amplified by government-sector barriers, as particularly worrisome. Often, AI projects fail to move past the prototype phase. The scalability pillar of AI engineering includes three areas of focus:

scalable management of data and models
enterprise scalability of AI development and deployment
scalable algorithms and infrastructure

Even highly scalable systems will not fulfill mission outcomes if they are not robust and secure. AI systems must be robust against real-world variations—those that the systems can reason about and those that they cannot. The pillar of robust and secure AI calls out three focus areas:

improving the robustness of AI components and systems, including going beyond measuring accuracy to measuring mission outcome achievements
development of processes and tools for testing, evaluating, and analyzing AI systems
designing for security challenges in modern AI systems

While security is a must for AI implementations in the DoD, so is keeping humans at the center. The human-centered pillar of AI engineering is intended to ensure that AI systems are built in alignment with the ethical principles of the DoD and other government agencies. The pillar of human-centered AI engineering highlights these areas:

the need for designers and systems to understand the context of use and sense changes over time
developing tools, processes, and practices to scope and facilitate human-machine teaming
methods, mechanisms, and mindsets to engage in critical oversight

The three pillars would lend AI systems more than just their namesake qualities, according to Rachel Dzombak, digital transformation lead at the SEI and a leader of the SEI’s work in AI engineering. “By putting these pillars in place as AI system design and development starts,” she said, “you’re more likely to build systems that achieve mission outcomes.”

The AI engineering initiative invites collaboration on research projects to advance the discipline and build a community. It is also developing symposia for 2022 to further evolve the state of the art; gather lessons learned, best practices, and workforce development needs; and foster critical relationships. “By creating an ecosystem around the discipline,” said Dzombak, “we can coalesce insights and establish best practices around how we design, deploy, and maintain AI capabilities.”