changeset 134:b6cbdd5a9547

moar ideas.
author Robert McIntyre <rlm@mit.edu>
date Thu, 04 Sep 2014 15:58:12 -0700
parents 35eb4c1a7bf7
children 04394e3857e2
files org/ideas.org
diffstat 1 files changed, 55 insertions(+), 0 deletions(-) [+]
line wrap: on
line diff
     1.1 --- a/org/ideas.org	Thu Sep 04 15:33:15 2014 -0700
     1.2 +++ b/org/ideas.org	Thu Sep 04 15:58:12 2014 -0700
     1.3 @@ -31,6 +31,61 @@
     1.4  getting credit.
     1.5  #+end_quote
     1.6  
     1.7 +- problem with Aubrey de Grey's ideas :: Aubrey de Grey says that we
     1.8 +     might be able to live forever by continually repairing our bodies
     1.9 +     at the cellular level -- he details 7 different mechanisms of
    1.10 +     damage and says that if all of them are dealt with /together/
    1.11 +     that it would stop aging. (You can't miss even one because
    1.12 +     they're all fatal.)  However, it doesn't take into account that
    1.13 +     we are also beings of information and that there is a very real
    1.14 +     software component to our existence. Even if our biological
    1.15 +     chassies can be maintained forever, I think it is unlikely that
    1.16 +     our minds will operate well far outside of the design constraints
    1.17 +     that we've evolved to handle. Say I programmed a webserver with
    1.18 +     the express goal of it being able to serve webpages for month on
    1.19 +     some stock server. I'll do fairly rigorous testing to make sure
    1.20 +     that it can handle the expected load then then some. Now say that
    1.21 +     you want to keep a particular instance of this webserver running
    1.22 +     indefinitely. (The program instance is like your mind and the
    1.23 +     computer it's running on is like your body). You might very well
    1.24 +     be able to keep the physical computer infrastructure running for
    1.25 +     forever by replacing hard drives / ram / CPUs, etc. However,
    1.26 +     since I designed the webserver to work for a month, it probably
    1.27 +     has memory leaks, rare stochastic bugs, or other build in limits
    1.28 +     / constraints (think log files or some date rollover shenanigans)
    1.29 +     that will ultimately kill the webserver server even with eternally
    1.30 +     perfect hardware. Do you really expect that a webserver
    1.31 +     engineered to work for 1 month will run for 10 years? In fact, if
    1.32 +     I put in the extreme effort to make it that robust, I've wasted
    1.33 +     time that I could have spent on other projects by pursuing an
    1.34 +     unnecessary engineering goal. Likewise, human minds have only
    1.35 +     ever run for at most 122 years before they are destroyed due to
    1.36 +     hardware degradation. Fixing the hardware doesn't change any
    1.37 +     software bugs that are almost certainly present in the human
    1.38 +     mind. Think of all the pathological things that can go wrong with
    1.39 +     a webserver, multiply it by a million, and that likely how
    1.40 +     evolution has designed our minds. For example, consider memory :
    1.41 +     why should you expect that we have evolved the ability to
    1.42 +     coherently organize memories past say 150 years? There's been
    1.43 +     absolutely no selective pressure for this ability, so you can bet
    1.44 +     that if there's any fitness to be gained from not having
    1.45 +     unlimited memory potential (such as better metabolic efficiency),
    1.46 +     we have it! You might think that maybe we would just forget
    1.47 +     things the same way that we sort of forget things that happen
    1.48 +     earlier in our lives, but complicated information processing
    1.49 +     systems don't have to fail gracefully when they're pushed far
    1.50 +     past their design constraints. A 150 year old person is just as
    1.51 +     likely to suffer a catastrophic psychosis due to software
    1.52 +     limitations associated with memory as he is to do something with
    1.53 +     all those memories we might consider reasonable. More likely, in
    1.54 +     fact, since there are so very many ways for a complicated
    1.55 +     software system to break and so few ways for it to run
    1.56 +     successfully. Therefore, I think Aubrey de Grey's "hardware-only"
    1.57 +     approach is missing a very important component of longevity
    1.58 +     science, and any successful effort to make people live orders of
    1.59 +     magnitude longer than they do naturally will need to deal with
    1.60 +     people's software as well as their hardware.
    1.61 +
    1.62  - validating neurocryopreservation :: Problem : you want to test
    1.63       whether a brain is functionally preserved through vitrification,
    1.64       but you don't want to figure out how to preserve all the other