This was on purpose, as "brain" is a fuzzy, complex thing whereas taking a license to use NN as a (limited) model of learning and reinforcement echoes to some much more precise concept including its behaviour and corner cases that many around here know about when playing with CUDA but fail to make the leap that our brain/mind/psychology (see? fuzzy!) works quite the same way.