The argument is so strong that there's hardly a way around it actually.
I would have much difficulty to explain why, but I have a strong belief that the kind of simulation he's talking about can not run faster than reality. Therefore, It is not possible for evolved conscious beings to simulate their evolutionary history up to their current level of consciousness, as it would take way too much time. So they would not do that, either because it would be pointless (who wants to wait a billion years for the result of a computation?), or because they would just have no time to do it in the universe they live in (considering their universe might have a finite lifespan for cosmological reasons).
So I think amongst his three hypothesis, it's the second which is true:
« any posthuman civilization is extremely unlikely to run a significant number of simulations of their evolutionary history (or variations thereof); »
It doesn't mean they can not gather enough computing power. It just means that having an infinite computing power does not mean you can simulate anything as fast as you want.
It's a very interesting paper though. Thanks. I'll read it more thoroughly some day.
With Moore's law nothing takes a billion years.
A billion years of computing today is 1 million in 10 years time, 1000 in 20 years, 1 year in 30 years, 6 hours in 40 years, 22 seconds in 50 years, 0.022 in 60 years.
ie in 60 years we should be able to do a billion years worth of today's computing in substantially less than 1 second.
(I'm using Kurzweil's price performance model of doubling every year- perhaps this is somewhat optimistic)
