To be fair, due to the nature of neural networks, there are a great number of domains where they can be applied in some way without that fact actually explainin

To be fair, due to the nature of neural networks, there are a great number of domains where they can be applied in some way without that fact actually explaining anything about how the domain works. For example, TD-Gammon plays a mean game of Backgammon, but this does not mean that Backgammon reduces to connectionism. TD-Gammon applies neural networks in a specific way with other algorithms involved, and then there are still underlying facts about what it takes to win at Backgammon which do not necessarily mandate anything that resembles connectionism.