I agree and it also seems like the truth is probably the opposite of what people are likely to take away. Loosely, when you train a neural net repetitively in

I agree and it also seems like the truth is probably the opposite of what people are likely to take away. Loosely, when you train a neural net repetitively in one way it adapts in order to be better at solving that specific case. While there may be some variation in initial conditions it seems likely that a lot of the variance is a result of the training itself.