Foundation language models through the lens of manufacturing

Abstract

Although recent studies have focused on foundation language models’ architectures, scaling properties, and applications in fields such as healthcare and business, no detailed investigation has addressed their role in manufacturing. This paper fills this gap by examining foundation language models, with a particular focus on large language models (LLMs) as their most prominent instantiation, through the operational manufacturing lens, emphasizing their capabilities and practical applications. In the first part, the core capabilities of LLMs are categorized and analyzed. These capabilities include text understanding and generation, reasoning, multi-modality, interactivity, generalization, and continual learning. The second part examines how these capabilities translate into practical applications across the operational phases of manufacturing. The areas include planning, production, material handling, engineering, quality, maintenance, and warehousing. By aligning LLM functionalities with operational manufacturing phases, the paper shows LLMs’ potential to augment decision-making, enhance efficiency, and increase adaptability in the context of Industry 4.0.