From: Anders Sandberg (asa@nada.kth.se)
Date: Sun Dec 15 2002 - 16:36:16 MST
On Sun, Dec 15, 2002 at 12:36:31PM -0800, Robert J. Bradbury wrote:
>
> On Thu, 12 Dec 2002, Anders Sandberg wrote:
>
> > What Robert was asking about wasn't whether some future technology
> > could abolish death of some future super-technology could resurrect
> > the dead, but how to keep the vision of dead transhumanists alive
> > today. [snip]
>
> I missed Anders' comments and Lee's response. Catching up...
>
> I was *not* thinking about a Tipler-ian Omega Point kind of
> scenario (with lots of math involved). I do think the idea
> of keeping the ideas of transhumanists alive and extending
> them is a very good one (it could help in the scenario I
> was actually thinking of). What I was really going for was
> whether the recreation of a resonable facsimile could be
> managed?
OK, I stand corrected. My apologies to Lee and Avatar (and Robert)! I
still think it is better to look at the here-and-now, since we will have
to do without Sasha for many many years to come even if supertech comes
around. If we just have to base our emotional handling of loss by the
invocation of hoped for technology, we are no better off than someone
invoking divinities to do it.
> The point is that as we go forward during this century we are going to
> have an increasing understanding of the fundamental processes for how
> human minds work and a much greater capacity for simulating a human mind.
> I was trying to get to the question of what more might be done to have a
> sufficient record that it becomes feasible to really "recreate" someone?
> At least to the extent that they walk, talk, act and may actually "think"
> like the original.
All the information we leave behind us can in some sense constrain a
model of ourselves, it is just that some of it is awfully hard to use.
My messy desktop tells a complex story of my various projects, habits
and occasional actions, but it only tells part of the whole story.
Similarly for my emails - they tell a bit about my mental state and what
I may think and believe, but not the whole story.
A whole mind is around 10^16 bits or more, and may change over the
passage of a lifetime. If we assume that one second differences are the
same "moment" we get around 10^9 such slices, or a total information
content of 10^25 bits to encode a lifetime. That is roughly the number
of degrees of freedom we need to constrain. A picture of a person taken
at a certain point would constrain the mind in that slice to some extent
- given the orientation of the eyes and the place, we could deduce the
input to the visual system to some extent, lets say around 10^6 bits at
best. An email like this would be the result of a few minutes mental
effort, likely constraining my frontal lobes and speech centers over 600
slices (it would be great here with an edit history) with a few thousand
degrees of freedom, also ending up with around 10^6 bits. So maybe we
could assume (in this happy dance of handwaving and picking numbers from
the air :-) that each piece of information on average produces 10^6
constraints (MRI scans provide a lot more, a messy desk a lot less). So
to get a total constraint we need 10^19 pieces of evidence. How many can
we actually get? I imagine that an ordinary person leaves one piece of
evidence every hour or so (government records, surveillance camera
images, work, photos, email, surf logs). That would produce around
600,000 pieces of evidence over a lifetime; totally it would be 6e11
bits, just 10^-13 of the degrees of freedom. That is unfortunately not
even enough to constrain one mental slice (our real goal). If we assume
one piece of evidence every second, then we get roughly enough evidence
to constrain a slice at the end. That sounds like a highly documented
life.
I guess a "good enough" simulation is far easier. After all, I would
estimate that a simulation of me more like the real me than any other
existing person would be "good enough". That means it belongs to the
10^-9 fraction of state space (2^10^16 states or so) closest to me right
now. That is nearly 2^10^16 states (the division hardly changes
anything), and the 6e11 bits of information I might leave behind is
enough to constrain the model to be within this cluster.
-- ----------------------------------------------------------------------- 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 : Wed Jan 15 2003 - 17:58:45 MST