According to the symbol-systems approach, every item in the world is mapped onto a symbol. For example, the image of a chair is mapped onto the label `chair'. After having labeled all items, every subsequent task is solved by manipulating the labels. For example, the `chair' is in the `middle' of the room, I am standing at the `door'. Therefore, given the known geometry of the room, I need to `go forward' before I can `sit down'.
This strategy has been followed by the so-called `Good Old Fashioned Artificial Intelligence' or short GOFAI (Haugeland, 1986). GOFAI was successful to cope with tasks that are difficult for humans, like playing chess, but GOFAI did not succeed on tasks that humans do effortlessly, like reaching for a pen on a cluttered desk. This difficulty suggests that the brain has a different strategy for solving these tasks (Pfeifer and Scheier, 1999). Moreover, following the symbolic approach, the mappings from real objects onto symbols were usually done by a human, leaving only the symbol manipulation to the machine. Therefore, the approach distracts from the real problem of object manipulation (Brooks, 1986b). As it turned out, this gap to the real world could never be closed (Pfeifer and Scheier, 1999).