Could OpenAI's 'too dangerous to release' language model be used to mimic you on online? Yes, says this chap: I built a bot to prove it https://www.theregister.co.uk/2019/03/20/openai_language_model/ via @theregister