All knowledge is functionally Bayesian – it’s a matter of probability of being true, which we can sometimes even formally quantify, but it’s never 100%, with the exception of constructed abstractions (like mathematics or other formalized abstract systems, where things can have definite truth values because we construct them that way) and the existence of at least one ‘mind’ – some system capable of cognition such that I can even be here considering the questions.
This is due, as Ibis3 points out, to the solipsism problem, which can never be resolved – not even if there was a god (or some other outside observer of the universe) and you or I met it, because those experiences would still be subjective and questionable (indeed, many people HAVE met a god as far as they are concerned, and I doubt the veracity of those experiences categorically).
What’s ridiculous is the “therefore god [or other mystical claims]” leap Armstrong is implicitly defending – once you know that (nearly) all knowledge is provisional, you need some basis for whether you provisionally accept any proposition, and with what degree of certainty. Evidence that is validated by multiple subjectivities (assuming that other people exist, of course; and if they don’t, then really, none of this matters because I’m just arguing with myself, on the Internet, which is actually a delusion, so whatever) has been demonstrated over a long time to be our very best means of determining what is most likely to be true. Side note: this is why the scientific method has proved so useful, though that epistemic approach shows up everywhere we have groups of humans, as with cultural norms, shared historical or cultural narratives, etc. – we agree, so it is true (even when it isn’t). So what defenders of faith ignore is that lack of certainty does not mean that anything imaginable is thus equally certain – even without the possibility of perfect knowledge or actual independent verification (everyone verifying what I see could also be my own delusion), not all possibilities are equally likely.