How might language influence how an AI "thinks"? — LessWrong