>Algorithm changes are always going to produce winners and losers. The implementers have control of the algorithm. If you choose to go a certain way, then that

>Algorithm changes are always going to produce winners and losers. The implementers have control of the algorithm. If you choose to go a certain way, then that means you are accepting the output. Why should anyone be able to hide behind "it's just an algorithm change." Now I understand that many of these machine learning algorithms lead to unpredictable and unexpected results. But again, they have found these risks acceptable by continuing to use them anyway. There should never be any hiding behind "its the algorithm, not us." You are responsible for your algorithms, whether you understand it or not. The thing I don't understand is why they continue trying these stupid algorithms. What value is there in them? It's very rare for me to encounter one of these things and go "yes, this is exactly what I wanted." More likely, they get in the way of what I want to do. Maybe I'm an outlier, but I really wish as a profession, we'd be much more frank regarding the actual results of these things.