Your entire rebuttal depends on treating extortion as a softmax function, which is a very settler way to think about machine learning engineers.