... if you factor in all the scientific knowledge so far accumulated and use the universal prior.
I agree with you in principle, a good Bayesian indeed should use all his prior informations and adopt a universal prior, but I think that putting God at 0dB is still beneficial in the context of the OP debate, and wouldn't change the outcome.
I strongly doubt that the OP's father would accept a starting point where the plausibility of God is already very low, while the probability of God at 1/2 is much more palatable, and usually the accepted starting point in traditional rationality.
The fact is that the OP can share an uninformative prior and then have the father make God a complex hypothesis, by presenting scientific knowledge directly in opposition with naive opinion on God existence. Then he just can rely on his preference of simple explanations to accept God non-existence.
This procedure still make sense from a Bayesian POV but produces much less friction when used within a heated debate.