What Is Physical AI?
Physical AI is artificial intelligence that acts in the physical world: systems that perceive an environment through sensors, decide, and move matter — robots, autonomous machines, embodied agents — closing the loop between computation and consequence. It is the discipline where AI's errors stop being wrong words and start being dropped totes or worse, which is why evidence standards matter more here than anywhere in AI. This site tracks its most demanding test bed: general-purpose humanoid robots.
The stack that makes AI physical
Physical AI layers perception (cameras, depth, force, proprioception), world modeling, task planning, and control that survives contact with real physics — plus the unglamorous scaffolding of safety cases and fleet operations. Progress in any single layer photographs well; products require the whole stack to hold across shifts, which is what deployment evidence uniquely measures.
Why humanoids are the stress test
A humanoid commits to the hardest version of every layer at once: bipedal balance, whole-body manipulation, human-space perception, and tasks defined by human workflows. That makes the humanoid record — tracked on this site profile by profile — the clearest public signal of where physical AI genuinely stands versus where it merely performs.
How to evaluate physical-AI claims
Three questions cut through most hype: was the behavior independently observed or self-published; does it run in a customer's environment or the maker's; and is the operating mode disclosed (autonomous, supervised, teleoperated)? The methodology here turns those questions into scored, sourced dimensions — and refuses to blend them into one number.
Frequently asked questions
- Is physical AI the same as robotics?
- Robotics is the machinery discipline; physical AI is the intelligence layer that makes machines general. Every physical-AI system is a robot; not every robot runs physical AI — a fixed-program arm does not.
- What are examples of physical AI?
- Humanoid robots are the flagship example this site tracks; the class also spans autonomous mobile robots, self-driving systems and embodied agents. The company directory maps who builds what.
- Why does everyone say physical AI now?
- The term rose as chip and AI leaders framed embodiment as AI's next frontier; the substance behind the term is exactly what evidence-first tracking exists to measure.