I'm going to repost the same comment I made on Reddit:<p>This is quite interesting. It certainly sounds like this does dissolve the Fermi paradox, as they say. However, I think the key idea in this paper is actually not what the authors say it is. They say the key idea is taking account of all our uncertainty rather than using point estimates. I think the key idea is actually realizing that the Drake equation and the Fermi observation don't conflict because they're answering different questions.<p>That is to say: Where does this use of point estimates come from? Well, the Drake equation gives (under the assumption that certain things are uncorrelated) the expected number of civilizations we should expect to detect. Here's the thing -- if we grant the uncorrelatedness assumption (as the authors do), the use of point estimates is entirely valid for that purpose; summarizing one's uncertainty into point estimates will not alter the result.<p>The thing is that the authors here have realized, it seems to me, that the expected value is fundamentally the wrong calculation for purposes of considering the Fermi observation. Sure, maybe the expected value is high -- but why would that conflict with our seeing nothing? The right question to ask, in terms of the Fermi observation, is not, what is the expected number of civilizations we would see, but rather, what is the probability we would see any number more than zero?<p>They then note that -- taking into account all our uncertainty, as they say -- while the expected number may be high, this probability is actually quite low, and therefore does not conflict with the Fermi observation. But to my mind the key idea here isn't taking into account all our uncertainty, but asking about P(N>0) rather than E(N) in the first place, realizing that it's really P(N>0) and not E(N) that's the relevant question. It's only that switch from E(N) to P(N>0) that necessitates the taking into account of all our uncertainty, after all!<p>[Note afterward: Over on Reddit, hxka points out that that should be P(N>1), not P(N>0). Or really it should be P(N>1|N>0)...]