This is interesting, a couple off the cuff thoughts.

Denying premise 1 doesn’t seem that crazy. Suppose you have a small Rasberry-Pi that implements a sorting algorithm S. Suppose that the Rasberry-Pi doesn’t have enough computational juice to handle lists of length 10, but can do everything less just fine. Well, it implementing S or Junky-S (S but it fails after 10); we answer that question by consultion a counterfactual: if the rasberry-pi had sufficient computing power, would it give the right answer for a list of length 11? So then you just deny that there is no fact of the matter on physicalism. There is, it just depends on the counterfactuals.

There’s also a bit of an ambiguity going on I think. It’s obviously true that when I add two numbers together I engage in some sort of mental activity; for the sake of illustration, suppose I picture two groups of sticks equal to the value of each term to be add, then bring them together in my mind, and count the number of sticks. It’s true that I’m engaging in that picturing activity, and that there is ipso facto a fact of the matter concerning that. Is it really that crazy to say there is no fact of the matter about whether that mental process implements this or that function? This doesn’t strike me as uniquely crazier than the calculator intuition, and so my guess is that if you bite the bullet there you should just bite it here, and any answer to that problem will also be an answer to this one.