You think language models are alive/aware and have feelings about token inputs?