Came here to cite the same thing! :) 

Note that Dustin Moskovitz says he's not a longtermist, and "Holden isn't even much of a longtermist."

So my intuition is that the two main important updates EA has undergone are "it's not that implausible that par-human AI is coming in the next couple of decades" and "the world is in fact dropping the ball on this quite badly, in the sense that maybe alignment isn't super hard, but to a first approximation no one in the field has checked."

(Which is both an effect and a cause of updates like "maybe we can figure stuff out in spaces where the data is more indirect and hard-to-interpret", "EA should be weirder", "EA should focus more on research and intellectual work and technical work", etc.)

But I work in AI x-risk and naturally pay more attention to that stuff, so maybe I'm missing other similarly-deep updates that have occurred. Like, maybe there was a big update at some point about the importance of biosecurity? My uninformed guess is that if we'd surveyed future EA leaders in 2007, they already would have been on board with making biosecurity a top global priority (if there are tractable ways to influence it), whereas I think this is a lot less true for AI alignment.

Can anyone give any outside-view reason to think EA is "unusually good at 'epistemics' / thinking about things" [...]?

Here are some possible outside-view reasons, not saying any of them is necessarily true (though I suspect some probably are):

• Maybe EAs (on average) have higher educational attainment than the population at large, and having higher educational attainment is correlated with better epistemics.
• Maybe EAs write and read more about epistemics and related topics than the population at large, and ...
• Maybe EAs would score better on a battery of forecasting questions than the population at large, and ...
• Maybe EAs are higher earners than the population at large, and ...
• Maybe EAs read more philosophy than the population at large, and ...

Of course it depends which group you compare to, and which thing people are meant to be thinking about.

Can anyone give any outside-view reason to think EA is "unusually good at 'epistemics' / thinking about things", or that "the community isn't corrupted too badly by random founder effects and information cascades"?

I have some evidence that it isn't: a commonly cited argument for the importance of AI research says nothing like what ~20-80% effective altruists think it does.

There might also just have been a natural progression. Toby Ord was always concerned about it, and 80,000 Hours made it a focus from very early on. At one relatively early point I had the impression that they considered shifting someone from almost any career path into AI-related work as their primary metric for success. I couldn't justify that impression now, and suspect it's an unfair one, but I note it mainly as an anecdote that someone was able to form that impression, well before the 'longtermist turn'.