Definition
Embodied AI refers to artificial agents — typically robots or simulated bodies — that learn and act by perceiving and physically interacting with an environment, rather than only processing static datasets. The embodiment forces the agent to deal with real-world constraints like noisy sensors, delayed feedback, and the physics of motion, which proponents argue is necessary for developing grounded, generalizable intelligence rather than pattern-matching on text or images alone.