As I predicted recently, AI in predictable future will not be dangerous, literally, in the usual sense. But it will be "dangerous" taking away jobs, especially routine works and requiring low qualifications and rather poor education. I gave earlier information about this trend. Recently I repeated my thesis in many interviews, e.g.: http://biznes.newsweek.pl/miliardy-w-rozumie-czyli-jak-sztuczna-inteligencja-rozpala-biznes,film,384313.html.
Quite recently I've received information about China project concerning project for replacing humans with robots called Du Mi in the fast-food restaurants KFC. This project has been realized - among others - by China' company Baidu. Restaurants network KFC, in cooperation with the Chinese company Baidu, has just opened in Shanghai, the world's first fast food restaurant without any human! Here you are advertissement of Du Mi: https://www.youtube.com/watch?v=JrGD_1haJxY or https://www.youtube.com/watch?v=LL9i085gUH8. It's only beginning... (I told and wrote about it several times). Generally my opinion is quite different than that of Prof. Stephen Hawking. Yes, AI can be dangerous for us, but only under several precise conditions, not without reservation, absolutely ! More here (SOHU): http://learning.sohu.com/20160429/n446815594.shtml