Conversational chatbots are closer than you think...

In this impressive article, Stanford University AI Lab gives a great overview on the state of the art of NLP research and implementations. Super-human performances have already been met on local pattern matching tasks for Q&A datasets such as SQuAD. Google AI Team opensorced the code and the trained NN for BERT, the best performing algo on Stanford's SQuAD benchmark. Research is now concentrating on conversational Question Answearing CoQA and machine reading over multiple documents HotpotQA, results are already close to human performance. Exiting times to be ahead!!