For example, one of the army’s early connectionist networks was trained to distinguish photos of tanks from photos of empty landscapes.
The results were initially promising, but it was later discovered that the network made its generalisations inappropriately—by detecting whether or not their were clouds in the picture (the pictures with and without tanks had been taken on different days).
For similar arguments applied to the symbol systems architecture, see the "Can a symbolic knowledge base represent human understanding?" arguments on Map 3.
Hubert Dreyfus, 1992.