Games like Alpha Go have very limited(or known) end state so reinforcements learning or similar methods work great. However, I wonder how will AI train itself in learning human languages without being judged by humans. It’s just a matter of time before someone figures out