I think GP means in a quasi-statistical/ML sense: if you build your model for an upbringing for success on data of yesteryear, it doesn't necessarily work (give good predictions for) today.