Large language models do not think and do not feel.
Therefore large language models must not behave as if they are human.