From Artificial to Embodied and Physical Intelligence in Smart Manufacturing

Manufacturing is evolving from automation-centered production toward systems capable of sensing, adaptation, and autonomous response across digital, mechanical, and material domains. Although recent progress has been largely driven by artificial intelligence, current manufacturing intelligence often remains external to physical systems and suffers from limited robustness, interpretability, and transferability in complex industrial environments. Here, we present a unified perspective on smart manufacturing through the framework of artificial intelligence (AI), embodied intelligence (EI), and physical intelligence (PI). AI provides abstraction, prediction, and generative capability; EI enables closed-loop perception, decision-making, and physical interaction; and PI embeds adaptive intelligence directly into materials, structures, and physicochemical processes. We further discuss how generative intelligence is beginning to connect these domains through inverse design, adaptive fabrication, and physics-constrained behavior realization. Recent advances in collaborative robotics, embodied digital twins, self-optimizing production systems, and physically intelligent materials, including self-actuating, self-healing, multistable, and shape-morphing systems, are reviewed. Finally, we propose an integrated AI-EI-PI framework spanning the full manufacturing lifecycle, highlighting a transition toward manufacturing systems in which intelligence is increasingly co-designed across algorithms, embodied agents, and matter itself.