Emulation vs. Simulation

From: Lee Corbin (lcorbin@ricochet.net)
Date: Tue Mar 20 2001 - 22:24:58 MST

Eliezer Yudkowsky wrote:

> If you want to explore the configuration space
> of sentience, then read CaTAI and code yourself a new
> citizen. If you want to run historical simulations,
> use zombies. I don't see any kind of important extropic
> exploration that suddenly becomes impossible. It just
> has to happen with volition, in a friendly universe.

Hal Finney and I object to this use of the important
term "zombies". Following Danniel Dennett and others,
we regard zombies as impossible, and I believe that
most people here will agree. A terminological quibble,
perhaps, but a significant one.

I think that none of the people I've seen who have
written at length on this topic will disagree with:

1. A "zombie", defined by Dennett and others as an
    entity that can behave the same way that a
    feeling and conscious creature does, but which
    is not feeling or conscious, is impossible.

2. the term "simulated", when applied to an entity,
    is ambiguous; better is "emulated", to refer to
    entities whose mental processing is on the order
    of a human being's or greater, and "portrayed",
    to refer to a very shallow implementation which
    is not much more than the image of a real entity.
    Thus, for example, please be kind to emulated
    beings, but it doesn't matter how you treat
To bolster my argument, Webster's New Twentieth Century
Unabridged gives

simulate: 1. to give a false indication or appearance of;
           to pretend; to feign; as *simulate* an interest.
           2. to have the external characteristics of; to
           look or act like; as, the insect simulated a twig.

emulate: 1. to strive to equal or excel
          2. to rival successfully

Thus when Eliezer wrote, "If you want to run historical
simulations, use zombies", I suspect that he's asking
the impossible. History is so complicated, I maintain,
that only emulations can prove anything.

Lee Corbin

This archive was generated by hypermail 2b30 : Mon May 28 2001 - 09:59:42 MDT