diff org/good-ideas.org @ 162:bdeaad2b1507

cryonics links.
author Robert McIntyre <rlm@mit.edu>
date Sat, 23 Apr 2016 18:01:05 -0700
parents ba80a6a67b55
children 7a6b855cfb99
line wrap: on
line diff
     1.1 --- a/org/good-ideas.org	Mon Mar 14 07:21:07 2016 -0700
     1.2 +++ b/org/good-ideas.org	Sat Apr 23 18:01:05 2016 -0700
     1.3 @@ -40,6 +40,63 @@
     1.4    <hr/>
     1.5    #+END_HTML
     1.6  
     1.7 +
     1.8 +
     1.9 +** A simple merge procedure for uploads
    1.10 +   Putting aside the philosophical problems around mind-uploading for
    1.11 +   a moment, imagine that you were /already/ a computer program
    1.12 +   running on a suitable robotic body. Let's say that the program is
    1.13 +   based off a detailed emulation of your brain -- no one really
    1.14 +   understands exactly /how/ the program is doing its computations,
    1.15 +   only that it faithfully recreates the original biological
    1.16 +   computations. At this point, you might want to be able to exist in
    1.17 +   multiple places at once while still maintining a coherent unity of
    1.18 +   identity, but how do you do it? You /can't/ modifiy your mind to
    1.19 +   handle multiple bodies at once, because this would require
    1.20 +   extensive understanding of exactly how your program is doing its
    1.21 +   computations in order to scale it to multiple bodies. But you can't
    1.22 +   just copy yourself either, because the copies will eventually
    1.23 +   diverge leading to a loss of unity of identity. You can't easily
    1.24 +   merge copies together after they've "diverged" because again, you'd
    1.25 +   have to come up with a coherent theory of mind to merge the
    1.26 +   datastructures. One simple way to do merges is this: First, you
    1.27 +   need to acchitect the robot bodies to record every single bit
    1.28 +   that's passing through all sensory nerves throughout a small time
    1.29 +   interval that's insufficient to create significant divergence. For
    1.30 +   sake of argument I'll assume that you do this each day. This stream
    1.31 +   of data represents the ultimate "life recording" and can be easily
    1.32 +   accomplished either in simulation of physical reality by using
    1.33 +   custom sensory organs, like in [[http://aurellem.org/#CORTEX][=CORTEX=]]. This set of information,
    1.34 +   if played back to a copy of you in the exact mental state right
    1.35 +   before the start of recording, would perfectaly match with the
    1.36 +   choices of the individual and serve as an adequate replacement for
    1.37 +   the world, even though it contains almost no information compared
    1.38 +   to the world! Once you can do the "full life records," then the
    1.39 +   procedure works like this: At the morning of Day 1 you "checkout"
    1.40 +   the one copy of you from yesterday and make around 10 copies which
    1.41 +   each live out their day. Each copy records a full "life recording"
    1.42 +   for the day. Then one copy is chosen as the "trunk" and the rest as
    1.43 +   branches. The trunk replays the life-recordings of the branches in
    1.44 +   accelerated time over the night, and the branches are then deleted
    1.45 +   leaving only one individual that remembers 10 consecutive days
    1.46 +   spent in different contexts. Then you repeat the process for Day 2
    1.47 +   and so on. I'll call this the "small delta-T approximation" method
    1.48 +   for mind-merging. It works as long as the timeframe is not too long
    1.49 +   an is limited by how fast you can faithfully replay life
    1.50 +   recordings. You don't have to be "offline" for any amount of time
    1.51 +   while doing this: You can also alternate two sets of 10, one for
    1.52 +   the day and one for the night, and have the night and day trunks be
    1.53 +   generated from the last day / night group's trunk,
    1.54 +   respectively. This trades having to "sleep" for the day/night crews
    1.55 +   not being able to remember what happened last night/day
    1.56 +   respectively. The more you know about how brains work, the faster
    1.57 +   you can integrate previous experience and the more copies you can
    1.58 +   sustain. This methods means that you never lose any experience, but
    1.59 +   if you're willing to lose some nonessential experiences, a more
    1.60 +   extreme version of this might be to make a copy of yourself that
    1.61 +   accomplishes a task and then reports anything of note in a written
    1.62 +   report.
    1.63 +
    1.64  ** Earth, Air, Water 
    1.65  
    1.66     Probably all intelligent species name their planet "dirt" in their
    1.67 @@ -335,7 +392,7 @@
    1.68     expressed in what I call a "hybrid base", a combination of base 10
    1.69     and base 60. You can get absurd things like 100 < 61, and 120 ==
    1.70     80! I wonder if these hybrid base systems could be very useful for
    1.71 -   some purposes!
    1.72 +   some purposes! Dylan [[http://logical.ai/microwave/org/sawtooth.html][wrote a blog post on this subject!]]
    1.73  
    1.74  ** Three Eyes
    1.75     If you had three eyes, would you still draw cubes like we currently