RE: Cryonics and information theory

From: Rafal Smigrodzki (rafal@smigrodzki.org)
Date: Wed Jul 09 2003 - 13:04:05 MDT

  • Next message: Extropian Agroforestry Ventures Inc.: "Re: Sol-like system discovered...SETI new directions?"

    Eliezer wrote:
    > Rafal Smigrodzki wrote:
    >> Eliezer wrote:
    >>
    >>> One problem I have, though, is that it still looks to me like it
    >>> would be better to just chop off the head and drop it into a bucket
    >>> of liquid nitrogen as fast as possible.
    >>
    >> ### Probably a bad idea. The freezing damage (microstructural) is
    >> extensive, maybe reversible with some as yet not-invented technique,
    >> but quite possibly too severe after all. Vitrification, and the,
    >> recently proposed by Alcor, intermediate temperature storage at just
    >> below the glass transition phase, should greatly reduce the damage,
    >> even including the gross fractures.
    >
    > on't tell me about "freezing damage". Tell me about information
    > loss, in the information-theoretical sense of that term. I don't
    > care about "fractures" or "damage" - tell me what happens to the
    > configuration space.

    ### We cannot at present adequately describe the configuration space of the
    system in question, but we can make educated guesses. We know that spatial
    proximity is important for synaptic function. We also know that many types
    of mechanical damage result in the loss of positional information, that is,
    without being able to trace the trajectories of objects as the damage
    occurs, it is impossible to reliably back-compute the original
    configuration. Therefore we can make the assumption that a piece of tissue
    without mechanical disruption will be, all else being equal, closer to the
    original configuration than a visibly damaged one, and at the very least,
    recovering function will require fewer computational steps.

    -----------------------------------
    >
    >> ### You are correct to look at the problem from the information
    >> theory point of view, but the specific application of your reasoning
    >> is IMO inappropriate - freezing does much more damage in terms of
    >> information loss (just a guess, prompted by looking at fresh-frozen
    >> brain sections)
    >
    > How can you guess this by looking at slides? You can sever a neuron
    > and half and what matters is not that it has been torn in half, but
    > whether the jagged edges uniquely identify the other piece of the
    > puzzle. This I would expect them to do.

    ### Well, this is just an intuition, but I expect the puzzle to be
    difficult. Looking at the flash-frozen tissue you have more than one
    neuron - you have literally thousands of synapses and hundreds of processes
    interrupted by every single ice crystal (it looks a bit like the holy-est
    part of Swiss cheese). Minute differences in the diffusion of water during
    freezing will shift the loose ends by microns, and to compute back to
    pre-freezing state you might need to model the diffusion during every ice
    crystal formation. To do that, you need to have a very good picture of
    concentrations of ions and proteins in the frozen tissue, *and* in the
    microenvironment before freezing. Maybe it is computationally tractable, but
    I would prefer not have my reanimation depend on this issue, especially
    since vitrification really doesn't add that much time to the procedure,
    compared to what we know the tissue can last after death.

    ------------------------------------

    >
    >>> If dendrites and axons retract into the cell body within half an
    >>> hour after the neuron has been starved of oxygen (!!!),
    >>
    >> ### This is not true.
    >
    > Maybe I've got it wrong. What's the correct time?

    ### On autopsy tissues you can see nice undamaged neurons even if fixation
    was performed 24 hours after death. There is rather quick loss of synaptic
    boutons during warm ischemia, but once the tissue cools down to ambient you
    should have no less than 12 hours and perhaps as much as 24 hours for
    getting good slides. You will see some mitochondrial swelling within 20
    minutes after death, but this is irrelevant to reconstruction of neural
    networks.

    Rafal



    This archive was generated by hypermail 2.1.5 : Wed Jul 09 2003 - 10:13:50 MDT