Another of Rodney’s robots, named COG, may well be the most advanced robot on the planet. Rodney’s aim is to get COG to learn about the world as a child does, to eventually reach the intelligence, dexterity and understanding of a real human, and to be able to operate like a human in the real world. COG’s brain mimics the human in that it utilizes distributed multiprocessors.

For example, there are computers for the eyes, others for the neck and still others to communicate between eye and neck to enable turning the head in the direction of observation. For each of a pair of eyes, COG has 2 cameras, one Wide Angle for motion detection and one Close-up "Fovea" camera for object detection. COG’s arm has an elbow, shoulder and wrist, as well as pressure-sensitive strips between joints to enable COG to vary the resistance depending on the type of load.

Wrapped around one of the fingers of COG’s hand is a touch-sensitive material that detects touch-pressure. When touched by an object, the hand closes and tries to grasp the object, as directed by an on-board computer in the hand. To demonstrate COG’s overall learning and functioning ability, Rodney reset COG’s memory to 0 and started the machine. At this point COG didn’t even know how to operate its arm.

Rodney started waving a hockey puck in front of COG, with the intention that COG should eventually recognize the puck and reach for it. COG’s Wide Angle Motion Detection cameras detected the motion which in turn triggered the Close-up cameras to work on detecting the object. Once detected, COG started a trial-and-error process of repeatedly launching his arm in the general direction of the puck, measuring the shortfall or overshoot, and modifying the behavior in the direction of best results, until finally grasping the puck. Though human babies take several months to learn this skill, COG mastered it in a few hours, in another striking demonstration of the Genetic Algorithm.