Interpretability And Controllability In The Age Of Language Models