Predicting innovation is something of a self-canceling exercise: the most probable innovations are probably the least innovative.