I have read like 1/3 of the comments but its interesting that the most interesting problem of everything we are talking about was only noticed by a non-programmer (so far). Everything that was talked about in the comments all hinge on the state of the art (SOTA) of ai in natural language processing (nlp). John Carmack recommended https://www.youtube.com/watch?v=OQQ-W_63UgQ stanford's nlp class (i didnt know who he was until i saw that post on hn). That class is good. it should take a long time to grok. you'll be solid up to LSTMs a general current standard.