Building safe physical AI: Open questions, risks, and a call for collaboration by Benjamin Alt

Ben­jamin Alt, Lau­rie Burns-Mill, and An­drew NelsonWe thank Finn Metz, Anna Mag­pie, Jan Wehner, Lee Sharkey, Chris Pang, Sig­mund Hen­num Høeg, and Marta Krzem­inska for their feed­back on early ver­sions of this post. All views and er­rors in this post are our own.Ex­ec­u­tive summaryPhys­i­cal AI—in­tel­li­gent robots, drones, and other sys­tems di­rectly act­ing on the phys­i­cal world—could re­move the limits placed by the availa­bil­ity and cost of hu­man la­bor on what hu­man­ity can bu...