DoCatsHaveMinds

When you get involved in ArtificialIntelligence, sooner or later you will come up against the question of: but is the machine really intelligent?

And if you are lucky, that will send you down the rabbit hole of PhilosophyOfMind and eventually you'll bottom out at the realisation that the question is just a special case of TheProblemOfOtherMinds. And while philosophy has a lot of good and interesting thinking about that, the one thing it can't offer is a nice convenient and satisfying answer to the question of whether it's like something to be a Large LanguageModel.

But anyway, as a way to distract yourself from that conundrum, I invite you to consider the fate of the question : do cats have minds?

The thing about this question is that there are millions of cat lovers in the world who already have a perfectly good answer to it.

And it won't matter how rigorous your arguments or your experiments are, or that your paper got published in Nature because the reviewers really loved it.

The cat lovers of the world will not give a flying fuck.

Of course their Tiddles has a mind and a rich personal relationship with them.

Cat lovers are literally the limits of science when it comes to the question of cat minds. They define a frontier beyond which science can go no further.

Ultimately, you may harrumph, but science is a social phenomenon. And the cat lovers of the world can and will hold that line. Any story to the effect that cats don't have minds is going to get firewalled to a few cognitive research geeks, while the rest of the world is protected from it. And goes on happily assuming that cats do, indeed, have minds.

One reason I bring this up is because I don't think that it's "like something" to be ChatGPT. Nor do I think that Intrepida, the cat I shared a house with for 10 years, really felt any affection for me or had much of what I'd call a mind capable of social relations with me. As opposed to an evolved instinct towards parasitism on humans.

BUT ... I fully recognise that the status of both these beliefs is equally a kind of folk psychology. It's a set of heuristics that get me around the world, let me make a couple of moral deductions when needed, and which I feel generally comfortable with holding.

But if I were to go into rigorous philosophical circles and announce that it was a cast-iron certainty that neither Treps nor ChatGPT have a mind, I would hope I would get laughed out of the room. What would be valuable in such an environment, either for the case of ChatGPT or the cat, is not a bald assertion one way or the other but a rigorous argument, which considered and answered obvious counterarguments.

Much of the AI debate I see online is people starting with such bald assertions as inputs and claiming that, for example, anthropomorphism is "bad" because obviously LLMs don't have this or that mental facility. Most such arguments make me want to scream "how do you KNOW that the LLM doesn't have that?" And I AGREE that the LLM doesn't have that (sometimes). It's just that the lack of argument or even awareness that the question hasn't been definitively answered, is grating.

Compare :

• MirrorTest