Small and medium-sized manufacturers (SMMs) are falling behind larger firms in efficiency and productivity, and this gap is widening [1, 2]. This is particularly concerning because SMMs, manufacturers employing 500 or fewer workers, account for 43 percent of all US manufacturing employment and about half of the nation’s manufacturing services and products [3]. It is not clear whether the advance of AI will cause this gap to widen. Becoming more efficient, productive, and resilient is largely a matter of learning how to use new tools and techniques. However, small manufacturers are challenged to connect with supporting institutions and find guidance in technical areas [1].
It is widely believed that AI is changing the nature of work. The aim of the Human/Machine Teaming for Manufacturing Digital Twins project is to establish techniques and measurement science by which some of the sophisticated tools of manufacturing can be used effectively through productive, incremental exposure to their capabilities. Generative AI and domain-specific languages for manufacturing tasks may make it possible to accelerate learning and narrow the gap between large and small manufacturers in the use of sophisticated tools.
Objective
To remove barriers to the use of sophisticated tools in manufacturing, such as digital twins, by developing the foundation for integrating generative AI, declarative domain-specific languages, and human/AI teaming.
Technical Idea
The mission of NIST’s Engineering Laboratory includes researching methods to make US manufacturing more resilient. In manufacturing, resilience is largely the ability to learn new capabilities. Our focus in the first year of this project is to focus on how people can best team with AI to learn a popular language for constraint-based optimization, MiniZinc. We pair generative AI with AI planning in a chat-based environment to interview people about their production scheduling problem and, in the process, formulate a solution in the MiniZinc domain-specific language. This research will serve as a springboard for out-year development of best practices for designing domain-specific languages for manufacturing tasks and representations of explanations.
Research Areas
The Teaming project seeks to develop methods and solutions for human/AI teaming on select cognitive tasks of manufacturing via the following tasks:
Highlights
References
[1] Ben Armstrong, Suzanne Berger, and Bill Bonvillian. Advanced Technology, Advanced Training: A New Policy Agenda for U.S. Manufacturing. MIT Initiative for Knowledge and Innovation in Manufacturing, May 2021.
[2] Testimony of William B. Bonvillian to the House Defense Appropriations Subcommittee. Hearing on Workforce Development and the Department of Defense, House Defense Appropriations Subcommittee, October 2021.
[3] National Science and Technology Council. National strategy for Advanced Manufacturing. United States Government, October 2021.