> the robot has a model of itself and the world around it, and when it "looks at the world" it's actually looking at that model. That would be the classical "s

> the robot has a model of itself and the world around it, and when it "looks at the world" it's actually looking at that model. That would be the classical "sense, plan, act" approach to AI. There is also a whole array of more "reactive" solutions, that only relies on sensor inputs without building any world model. Brooks' "Intelligence without reason" [^1] is probably one of the best papers out there in describing how this way is better suited to build complex robotic/AI systems and apply them to real-world scenarios. [^1]: http://people.csail.mit.edu/brooks/papers/AIM-1293.pdf http://people.csail.mit.edu/brooks/papers/AIM-1293.pdf