Say the predominant computing paradigm was not based on electricity, but rather something like mechanics. You would still say the AI built on these computers is concious, right? And how about any other physical process which "computes" this particular function (i.e., the one corresponding to the neural net), whatever that means?
I don't see any reason why you would think a neural net on a computer is concious, but not this other physical implementation (except perhaps that the brain also uses electricity, like a computer). In that sense, you really are claiming that "math is concious" (Compare this with a physical theory, which might be, for example, positing a "conciousness particle").
I find it reasonable to object to that. At the very least, it raises some difficult questions. First, what counts as a computation of our function? The universe doesn't have a well defined input and output.
You mention that writing the equation on paper doesn't count. But what if a human computed the function numerically using pen and paper, without understanding the underlying meaning (as in the Chinese room thought experiment)? The computational theory seems to imply that it would give rise to conciousness (other than that of the human). This is very unintuitive.
Secondly, what is special about the functions obtained from training neural nets? Why are their physical instantiations concious, but not those of some other function obtained by randomly choosing the weights of a neural net? It is unclear how to differentiate between the two cases (though I can't confidently rule it out as impossible).