Deanthropomorphising NLP: Can a Language Model Be Conscious?