Embodied AI refers to artificial intelligence systems that can interact with and learn from their environments using technologies such as sensors, motors, machine learning and natural language processing, according to TechTarget. For example, think of an autonomous vehicle, or a robot or drone.
Rather than relying on data that people provide, embodied AI interacts directly with its environment, gaining more accurate insights that it uses to inform its decisions and actions, according to Built In.
Embodied AI can respond to different kinds of sensory input, like detecting ultraviolet light and magnetic fields, using GPS to know where things are, or understanding inventory levels in a supply chain. Many embodied AI systems can move, like robots, but movement is not a requirement to make something qualify as embodied AI. For example, an autonomous security system might learn from the physical interactions of agents running on networking or storage infrastructure.
Deeper Dive
Elements of embodied AI systems include features such as:
- Sensors: to collect information about an environment that helps the AI build and update its model.
- Actuators: devices in the AI system that convert energy into motion, enabling an embodied AI system to act on or change the environment or its mobility in that environment.
- World model: the model that the AI system creates of its environment and itself, enabling it to make decisions as it pursues a goal.
In health care, examples of embodied AI are surgical robots that can use AI to guide them in making precise incisions, applying the right amount of pressure and performing careful movements during operations. Robots and exoskeletons can gather feedback from sensors to provide the proper support to patients during physical rehabilitation, the Built In blog post notes.