but if you replace AC with a weaker axiom, Axiom of Dependent Choice, you can construct a model in which all subsets of reals are measurable (called the Solovay model), which precludes the Banach-Tarski paradox (which requires non-measurable sets).
Could BT be constructed without AC but with other way of ensuring there exist non-measurable sets? No idea.
The problem here is not that there does not exist a procedure to describe which number to choose, the problem is that there does not exist any way to describe which number you chose. Those are not the same thing.
https://www.solipsys.co.uk/new/APointAgainstTheAxiomOfChoice...
Wait, I think you just made me realise that my intuition for AoC is wrong and actually matches AoCC instead, and that I'd never stopped to consider that the collection of sets you're choosing from could be uncountable. If that's correct, that explains so much.
I have a masters in math. Concrete numbers are irrelevant to me, and their explicit representation is even less relevant. Any number can easily be described by x.
Btw this is just the standard reason for the existence of the Axiom of Choice. If there exists a procedure to describe which number to choose (i.e. an algorithm/formula), you don't need AC at all.
IMO the only "intuitive" anti-AC argument is, that you can get most of math with Axiom of Countable Choice, while you avoid paradoxes like Banach-Tarski.