When I read this parable, I was already looking for a reason to understand why Friendly AI necessarily meant "friendly to human interests or with respect to human moral systems". Hence, my conclusion from this parable was that Eliezer was trying to show how, from the perspective of AGI, human goals and ambitions are little more than trying to find a good way to pile up our pebbles. It probably doesn't matter that the pattern we're currently on to is "bigger and bigger piles of primes", since pebble-sorting isn't certain at all to be the right mountain to be climbing. An FAI might be able to convince us that 108301 is a good pile from within our own paradigm, but how can it ever convince us that we have the wrong paradigm altogether, especially if that appears counter to our own interests?

What if Eliezer were to suddenly find himself alone among neanderthals? Knowing, with his advanced knowledge and intelligence, that neanderthals were doomed to extinction, would he be immoral or unfriendly to continue to devote his efforts to developing greater and greater intelligences, instead of trying to find a way to sustain the neanderthal paradigm for its own sake? Similarly, why should we try to restrain future AGI so that it maintains the human paradigm?

The obvious answer is that we want to stay alive, and we don't want our atoms used for other things. But why does it matter what we want, if we aren't ever able to know if what we want is correct for the universe at large? What if our only purpose is to simply enable the next stage of intelligence, then to disappear into the past? It seems more rational to me to abandon focus specifically on FAI, and just build AGI as quickly as possible before humanity destroys itself.

Isn't the true mark of rationality the ability to reach a correct conclusion even if you don't like the answer?