An angel comes down to you with two arks, ark A and ark B. You may choose to take either ark A or both arks A and B.
Ark A has either the Mona Lisa, worth millions of dollars, or a worthless spider, and ark B has a goblet worth 1000.
The obvious choice would be to take both, right? But here's where it gets difficult:
40 days ago, the angels made a prediction about what you would choose. If they thought you would take both, the put the spider in ark A. if they thought you would just take ark A they put the Mona Lisa in it. Remember that they are almost certainly right.
I see two ways to look at this:
(1) You take both; what they predicted 40 days ago barely makes an impact on your choice now, why waste the $1000 cup?
(2) Taking both would only get you the Mona Lisa under the assumption that the angels were wrong, so that is obviously the wrong choice; the angles are almost certainly right, in which case taking ark A would get you the Mona Lisa.
Which would you take?
I think mr fantastic might be referring to the St. Petersburg paradox. If we have a probability that the angels are right (or two probabilities, one for false positive and one for false negative), then we can calculate expected value, but the St. Petersburg paradox highlights why a reasonable person might not decide based just on expected value.
Also, I got this from a book called "The Math Book" and they didn't give the probability of the angels being right. I just assumed it was somewhere about 95-99%, but "almost certainly" is all the information the angels give you. I don't think the paradox applies to this anyways. For one thing, there is no fee to play. You just have to meet some overly-generous angels.
Given the looseness of the connection, I suppose the paradox is probably not what mr fantastic was referring to, though.
(By the way for the purposes of problems like this there is a technical meaning to "almost surely" which I will assume means the same as "almost certainly", which is in effect with probability 1, which for most purposes is indistinguishable for certainty. Though of course the CIA takes "almost certainly" to mean with probability of about 93+/-6%, but this we will ignore as the CIA is not on the side of the angels)
So if we know what is mean by "almost certainly" there is no paradox just a common decision problem, the confusion here is in treating "almost certain" as the same as certain.