Re: a simulated utilitarian

From: Anders Sandberg (asa@nada.kth.se)
Date: Tue Aug 05 2003 - 06:18:25 MDT

  • Next message: Alfio Puglisi: "Re: Power from human blood"

    On Tue, Aug 05, 2003 at 04:36:02AM -0400, Wei Dai wrote:
    >
    > So we have one prominent extropian telling us to live for today, and
    > another one telling us to live for the far future. Who is right? It
    > occurs to me that we should take a probability-weighted average of the
    > two positions. If the simulation argument is correct, there is a tiny
    > probability that we live in root reality and what we do today affects
    > an astronomical number of potential future individuals (including all
    > future simulated individuals), and a near 1 probability that we live in
    > a simulation and our actions affect relatively few people. We need to
    > take both possibilities into account when making decisions. When we do,
    > I think the two positions cancel out somewhat and we can live a more
    > "normal" life.

    Assume there are n people living within the average sim (perhaps
    around 10^10 or so) and N people living in the root reality. Root
    reality runs X sims, making the total amount of lives in it N+nX.
    The probability of living in a sim (if you are randomly selected) is
    nX/(N+nX). So if you don't know where you live, the expected number
    of people to be utilitarian about (all people in the "universe"
    you can reach) is n^2 X/(N+nX) + N^2 /(N+nX).

    If X is small the second term will dominate, as well as if
    n sqrt(X) << N. In this case we will likely live in root reality and
    should act that way. If we have reason to believe that n sqrt(X) >>
    N, then we should act more like sims.

    Can we estimate N and X? A basic (and minimal) guess of the future
    superciv would be something on the order of an M-brain. Robert
    calculated that it could do an ancestor sim in 1e-5 cpu seconds,
    which presumably has n=10^11 or so. It could likely sustain around
    1e16 human-level entities, so N=1e16 if all activity is used for
    beings. If the fraction f of resources is used for sims, we get
    N=(1-f)10^16 and X=f 10^5. But these are numbers per second; if we
    assume T repetitions of this we get X = f T 10^5, N=(1-f)10^16
    (since the entities are likely immortal and do not depend on T). T
    is on the order of trillions of years, so we set T=1e19

    We get
    10^27 f T /((1-f)10^16+f T 10^16) + (1-f)^2 10^32 /(10^16(1-f+fT))
    = 10^11 fT/(1-f+fT) + (1-f)^2 10^16/(1-f+fT)
    = 1e30 f/(1-f+f1e19) + (1-f)^2 1e16 /(1-f+f1e19)
    ~= 1e11 + 1e-3 (assuming f >> 1e-19)

    So the answer seems to be that we should behave like sims. But this
    is largely based on the idea that the M-brain inhabitants (N) are
    just immortal - if they have generations shifting, then they also
    get a sizeable T factor and they will be dominant - hence morality
    should be root reality oriented.

    -- 
    -----------------------------------------------------------------------
    Anders Sandberg                                      Towards Ascension!
    asa@nada.kth.se                            http://www.nada.kth.se/~asa/
    GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y
    


    This archive was generated by hypermail 2.1.5 : Tue Aug 05 2003 - 06:23:30 MDT