From: Eugen Leitl (eugen@leitl.org)
Date: Sun Jun 16 2002 - 08:19:12 MDT
On Wed, 12 Jun 2002, Smigrodzki, Rafal wrote:
> Hal Finney wrote:
>
> To me, this casts doubt on the intuitive sense that there is "one
> person" associated with the first run, and "another person" associated
> with the second run, with the first person getting snuffed out when he
> pushes the button while the second person continues to run. That
> model just doesn't work.
It should be pointed out that we're allocating an undue amount of mental
processes and list bandwidth to what is basically a pathological case.
Really just a gedanken.
In numerics applications frequently each run produces very different
results for different architectures. Even changing the compiler
optimization options (in absence of infinite accuracy rearrangement of
instructions results in different results) results in a different outcome.
Whenever people rerun a job they usually want to validate that the changed
code still produces sane results, or that the system suffered an uncaught
hardware glitch (since these are rare, and manifest in a stochastical
manner even a single repetition of a run is sufficient to validate the
result).
Mental processes are a classical instance of nonlinear numerics. Here
differences in state amplify exponentially, so above certainly applies.
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:14:49 MST