Quote (AbDuCt @ Jun 9 2016 04:26pm)
I've come to grasps with myself and decided I can not make an AI that is good enough to my standards. Lack of massive amounts of treated data, and the hardware to use real RNN//sequence to sequence networks really hold me back. My closest attempt was using a text ranking algorithm to try to extract keywords from text and reply based on those keywords. It should work well in theory, but in practice... not so much. It was likely due to my tainted data I harvested, but I don't have the time or resources to manually train the database. I would need people to help via inputting statements/responses and/or talking to it and rating the responses. Although there are too many trolls for this to work.
With that being said, someone tell Microsoft to hurry up and port Xiaoice over to English.
Damn, I was psyched about that too.
