Physical AI
Readiness
TRL 6 · SRL 5.5 · gap +0.5
Gap = technical readiness − societal readiness. Positive: the technology is ready and society is not. Negative: society wants it more than the technology can deliver.
Not scored. Watchlist entries do not enter any gradient.
Physical AI is the label for machine intelligence that has to act in the world rather than on a screen — perception, reasoning and action under gravity, friction, wear and the fact that mistakes break things.
The distinction is worth having because the two domains have different constraints. A language model that is wrong produces a bad sentence. A machine that is wrong drops something or hits somebody, and the tolerance for tail-case failure is therefore near zero in exactly the environments that are least predictable. That is why Chapter 5 finds robotics mature in structured settings and early in unstructured ones. The gap between a factory floor and a hospital ward is not a capability gap. It is a specification gap.
Readiness sits near the middle on both axes and close together, which reads correctly: this is moving, and neither the engineering nor the acceptance is holding it.
Common misconception: that it is robotics with better software. The hard part is not the intelligence or the actuator. It is the sensing that has to be right when the room is not what the model expected.
What it does
mechanism — not yet written. Published as missing rather than hidden.
Candidate megatrends — unreviewed
- Automation of Cognitive Work
- Digital Sovereignty
- Synthetic Media and the Verification Collapse
- The Application Gap