PID roughly occupies the same space that logistic regression does in ML. Sure one can use more complicated DNNs when the task really needs it but for many many classification or conditional probability estimation problems a regularized LR goes a long way and its hard to beat it on the axis of simplicity. Of course your resume wont look as shiny.