Interdisciplinary Workshop on Mechanical Intelligence: Summary Report

A major interdisciplinary workshop convened by the National Science Foundation (NSF) has formally defined 'Mechanical Intelligence' (MI) as a distinct and promising field of research. The workshop, held in May 2024 and involving 38 academic researchers and 8 NSF program officers, resulted in a summary report authored by Victoria Webster-Wood and 39 collaborators. The core thesis posits that intelligence—encompassing responsiveness, adaptivity, memory, and learning—can be encoded directly into the physical and material structure of a system. This stands in direct contrast to the dominant paradigm of computational intelligence, where smart behaviors are generated solely through software algorithms and electrical signaling in a central processor.

The report outlines MI's potential applications across robotics, materials science, and bio-inspired engineering. By designing structures that can compute, sense, and react through their mechanical form, researchers aim to create more robust, energy-efficient, and adaptive machines. Examples include robots that navigate through shape-changing limbs without complex central control, or materials that 'remember' and adapt to previous stresses. The workshop's intensive discussions were structured to identify key challenges and opportunities, aiming to establish a foundational roadmap for future NSF-funded research in this emerging domain.

The formalization of MI as a concept signals a strategic shift in how intelligence for physical systems is engineered. Instead of solely focusing on making software smarter to control 'dumb' hardware, the MI approach seeks to distribute intelligence into the hardware itself. This could lead to breakthroughs in soft robotics, prosthetics, and adaptive infrastructure, creating systems that are inherently safer and more capable in unpredictable real-world environments. The NSF's backing indicates this is becoming a prioritized area for foundational research funding.