On the Dual-Use Dilemma in Physical Reasoning and Force

William Xie, Enora Rice, Nikolaus Correll

Keywords: contact-rich manipulation, dual use, safeguarding, visual prompting

TL;DR: applying safeguards to VLMs employed for physical reasoning & contact-rich manipulation reduces both harm and help

Abstract: Humans learn how and when to apply forces in the world via a complex, lifelong physiological and psychological learning process. Attempting to replicate such a process in vision-language models (VLMs) presents two challenges: VLMs can produce aggressively harmful behavior, which is particularly dangerous for VLM-controlled robots which interact with the world, but imposing behavioral safeguards can limit their functional and ethical extents. We conduct two case studies on safeguarding VLMs which generate forceful robotic motion, finding that safeguards reduce both harmful and helpful behavior involving contact-rich manipulation of human body parts. Then, we discuss the key implication of this result--that value alignment may impede desirable robot capabilities--for model evaluation and robot learning.

Submission Number: 3