When I said intelligence can only be reasonably defined with respect to goals, that doesn’t mean we have to define everything useful to a goal as intelligent. I would not call a calculator intelligent since it does not learn, that is, it does not incorporate information from outside itself to improve its performance towards its goal.
(If you give it a wide fixed goal, such as “helping humans answer mathematical questions”, then it never absorbs information to get better at it. If you give it a narrow goal, such as “answering the question that was entered into it”, then it arguably absorbs information in the form of keypresses, but those keypresses just define the goal and doesn’t improve its performance on them.)
I would be open to calling a human-with-calculator more intelligent than a human alone, though.
You are right that not many people will use the word as I use it, but that’s on them! People have kept moving the goalposts for intelligence as long as computers have existed. Our use of the word intelligence is vague and woolly, and plagued with what Schopenhauer called “scholasticism”: treating the vague as if it were precise.