Logical or Connectionist AI? — LessWrong