This document summarizes a conference paper published at ICLR 2020 that proposes a method called Plug and Play Language Models (PPLM) for controlled text generation using pretrained language models. PPLM allows controlling attributes of generated text like topic or sentiment without retraining the language model by combining it with simple attribute classifiers that guide the text generation process. The paper presents PPLM as a simple alternative to retraining language models that is more efficient and practical for controlled text generation.