Mind

Mind is whatever it is that perceives, thinks, feels and acts. The philosophy of mind asks what that is, how it relates to the body, and what else might have one. For most of its history the debate between dualists, physicalists and functionalists was conducted over a single case: us. Artificial systems add a second case, and second cases are where concepts show their seams. Which features of mind are essential, and which are accidents of being an animal?

Two cautions shape how I approach the question. The first concerns description. Shanahan has urged care with the mentalistic vocabulary we reach for when talking about language models, since words like “believes” and “knows” carry implications we may not intend. The deflationary description of the systems as mere mimics is one my arguments neither need nor assert either.

The second concerns where the weight of an argument should rest. Freiman and others have charged that the received epistemology of testimony is anthropocentric, ruling machines out because they lack minds, intentions or a distinctively human capacity for responsibility. I think the charge is often right. An argument that a system cannot inform us because it has no mind concedes that diagnosis while denying its conclusion, and it purchases a verdict that will not survive its premises: if the systems come to be describable in mental terms, the argument lapses without anything having been learned about what they do. My arguments about what language models can transmit are therefore built so that nothing turns on mentality.

That is not indifference to the question of machine minds. It matters for consciousness and moral status, and for what understanding is. It is a decision to keep the questions apart, so that progress on one is not held hostage to the hardest of the others.