I note the fake replies are quite limited & repetitive. It would be extremely interesting to see if deep Q learning could be used to develop more realistic repl

I note the fake replies are quite limited & repetitive. It would be extremely interesting to see if deep Q learning could be used to develop more realistic replies. "rewards" would simply be getting more (and faster) responses.