End-to-end LSTM-based dialog control optimized with supervised and reinforcementlearning

Shared on r/worldnews with 5050 upvotes.

Shared on r/worldnews with 3373 upvotes.

Discussed on Hacker News with 1913 points and 1392 comments.