There's one significant difference between posing the problem in terms of AIs and posing it in terms of angels: if AIs are going to exist we'll have to design them[1] and whoever's designing them will be trying to ensure that their behaviour fits (something like) our values; whereas if angels exist, they were designed by someone else whose values may be quite different from our own and it's no business of ours to decide how they should behave.

[1] Perhaps indirectly.

Yeah, I like concise writing too. Concise and clear is even better. Concise, clear and funny, better still. Yudkowsky doesn't do too well on conciseness, but I think he does just fine on clarity and humour. (You might want to bear in mind that the article linked from here is part of a lengthy ongoing series (perhaps I should say: series of series) that EY was writing at the time; it's doubtless clearer when read in the context of the rest of it.)

Although "epistemic" would have been better, I really don't think "epistemological" need have been such a roadblock. If someone refers to an organism's "biological makeup" or "physiological condition", I hope it would be clear that they mean the kind of makeup/condition with which biology/physiology is concerned, rather than the organism's pet theories about biology and physiology. So also with "epistemological state".

What do other users of such terminology mean by it? I just asked Google for <<<"epistemological state">>> and of the first page results I reckon: first one is this discussion; second one is EY's meaning; third is ambiguous; fourth is yours; fifth is EY's; sixth is EY's (and says in so many words: 'Philosophers tend to suppose that one's "epistemological state" is constituted by beliefs'; the authors are philosophers); seventh is yours; eighth is ambiguous but I think nearer EY's; ninth is a sort of hybrid, nearer to yours; for the tenth (of which I can see only the snippet Google provides, the rest being behind a paywall) I can't tell. Some of those hits are from people whose use of philosophical language I wouldn't trust for an instant, but at least four seem reputable. (I am not sure whether to be relieved or alarmed that the ones that look reputable to me on other grounds are also the ones that favour EY's usage; perhaps I'm suffering from some bias or something.) It seems like EY's usage is pretty reasonable. I still think "epistemic" is better; as you may have noticed, he's now changed it.