Frontier Systems for the Physical World: The Emerging Paradigm of Physical AI

Oliver Hsu examines three frontier domains—robot learning, autonomous science, and novel human-machine interfaces—as instances of an emerging paradigm extending AI beyond language and code into the physical world. These areas share common technical primitives including learned physical dynamics representations, embodied action architectures, and simulation infrastructure, creating a structural flywheel for physical AI development. The article argues these fields are positioned to enter a scaling regime similar to large language models due to their proximity to the incumbent AI paradigm while requiring non-trivial additional work.