Let’s just throw all this overly political AI stuff to the side. It sounds like what many people have been doing with good old fashioned linear regression for…d

Let’s just throw all this overly political AI stuff to the side. It sounds like what many people have been doing with good old fashioned linear regression for…decades…would severely upset you. When you’re doing prediction for the sake of prediction, the results speak for themselves. There is no inherent necessity for the indicators to be intuitive. There are certainly contexts in which it does matter, though. When accountability matters. It’s, as you say, “basically just” doing exactly what it’s doing. It works exactly as well as it works. To argue that it’s good or bad based on some implementation detail is … kind of beside the point.