X-Message-Number: 7853
Date: Fri, 14 Mar 1997 09:29:57 +0100 (MET)
From: Eugene Leitl <>
Subject: Re: timescale of cryonics & uploading

On Thu, 13 Mar 1997, CryoNet wrote:
> ----------------------------------------------------------------------
> 
> Message #7840
> From:  (Thomas Donaldson)
> Subject: Re: CryoNet #7828 - #7833
> Date: Wed, 12 Mar 1997 11:27:59 -0800 (PST)
> 
> [...]
> About replacement or uploading of brain structures: while it is quite true
> that the whole idea of neural nets came from study of how our brains work,
> the electronic ones do not attempt to imitate brains except in the very

> broadest sense. Perhaps the name has become misleading. And certainly, there's

The concept is still quite new, and we hardly even started to learn. The 
90's are sure a decade of the brain, and just now the Japanese are 
embarking on a 15-20 year long-term project. Masai Ito, president of the 
japanese Science Council, and Shun-ichi Amari (a leading neuroscientist) 
proposed such a program in "The Age of Brain Science", Rieken study. 10 
G$ are alotted for the next two decades, already 100 M$ for 1997 alone.
It will be a bottom-up approach, an attempt to unravel higher brain 
functions from neurochemistry (and possibly HUGO data) from the ground 
up. The focus is not only on clinical, construction of machines utilizing 
the brain's architecture is also targeted. Even now Japan is doing a lot 
this, check out e.g. Hugo de Garis' "Japan's brain builder" page.

> been a lot of advances based on the electronic variety of neural net:
> most work on recognition of any kind depends on them. And I've read a fair

> bit about them myself, in my persona as a computer person. Ideally, of course,
> they should not just be imitated on a PC, but made into fully parallel 
> computers. Some companies have neural net boards which do exactly that.

Not only that, we even need to abandon semiconductor lithography for 
critters much beyond that of an insect (about 1 Mneuron).

> (I've been interested in parallel computing particularly). 
> 
> My problem with Graham Clark's statement that electronic replacement will 
> become available relatively soon comes simply from the fact that these neural

Relatively to fully reversible cryopreservation, it can be considered to be 
"soon".

> nets fail to imitate the actual behavior of brains or nets of neurons --- not
> just in a minor way, but in major ways. Perhaps we WILL have some kind of 
> replacement for some brain structures --- but it will be a replacement in the

Since hardware-wetware interfacing sets up a major worm farm, a more 
straightforward concept would seem to emulate a wetware system in silico 
in toto.

> same sense as a wooden peg leg is a replacement for a lost leg. Sure, you
> can hobble around on it, but no one would be mad enough to claim it was a 
> perfect replacement. In terms of what I would want to be uploaded into, the
> last thing I would want would be to be uploaded into such a clumsy device ---
> it would probably imitate me in the same sense as the peg leg imitates
> a real leg.

An emulation which is even slightly off-key will produce garbage. Thus I 
think it is an all-or-nothing game.
 
> The really major difference --- one which dwarfs the others --- is that 
> nets of neurons involve GROWTH. Not so much growth of new neurons as growth

...which is not harder to represent than any 
electrophysiology/neurotransmitter diffusion magicks. Easier in fact, 
since happening on a much larger timescale.

> of new or modified connections. Again, to some extent that can be imitated
> on a computer --- but with billions of neurons and lots of memories and 
> experiences, that looks like a losing proposition. Do I believe that someday

Currently, yes. Assuming bleeding-edge technology, we are talking about some 
10 tons of chips, dissipating 0.1-1 GW juice. And that's a conservative 
estimate.

> we will be able to make neural nets which fully emulate nets of neurons?
> Certainly. But it is a far harder problem than many on Cryonet seem to 
> believe. And of course I haven't discussed all the lesser differences:

It is hard, no mistake. Yet nanoresurrection is even harder.

> response to a wide variety of chemical transmitters, gap junctions between
> neurons, and on and on.

A lot of data yes, which must be gathered, and a representing 
coding/manchinery which must be developed, sure.

> 			Long long life,
> 
> 				Thomas Donaldson
> 
> ----------------------------------------------------------------------
> 
> Message #7841
> From: "Jan Coetzee" <>
> Subject: Re: timescale of cryonics & uploading
> Date: Wed, 12 Mar 1997 16:22:11 -0500
> 
> Joe writes:
> 
> 
> "I find the prospect of repairing the brains
> and bodies of those now frozen to be much more daunting than copying their
> brains into artificial devices. " 
> 
> Is this not a contradiction? After all in order to copy a neural structure
> to a machine it must be repaired first? 

It is much easier by far using ANN DSP artefact filters operating upon 
voxelsets, and tracing/segmentation, than sending out hypothetical 
nanomachines into a vitrified chunk of tissue, pushing actual atoms 
around. I am having great respect of the promethean Prometheus project, but 
fully reversible cryopreservation being its goal, the grapes might be 
hanging a bit too high. But drastically enhancing brain cryopreservation 
is surely doable, and well worth the effort.

> J.C.
> 
> ----------------------------------------------------------------------
> 
> Message #7844
> Date: Wed, 12 Mar 1997 23:09:46 -0500
> From: Garret Smyth <>
> Subject: Timescale of cryonics & uploading
> 
> Thomas Donaldson:
> "As for uploading yourself, I personally doubt that will become possible
> until some time after those frozen now have been revived."
> 
> 
> Joe Strout:
> "I find the prospect of repairing the brains
> and bodies of those now frozen to be much more daunting than copying their
> brains into artificial devices...  ...So if you would kindly provide a
> bit more explanation and references, I would be much obliged."
> 
> 

> The question here is *how* are you going to copy brains into artificial 
devices?

By writing bit patterns into a piece of molecular circuitry, of course. 
;) Seriously, though, a destructive scan at molecular resolution seems to 
be inevitable.


> Learning is not simply a matter of whether there is a connection between one 
cell

> and another, but how easily the synapses are triggered. The threshhold changes
with

Whatever they do, they are still physical structures. Complex structures, 
with complex tasks, but fundamentally modelable.

> learning so affecting both "memory" and "processing" (in the Von Neumann 

> architecture sense) are affected. To copy a brain in fine detail - and to 
upload

> a "person" as I'm sure even die hard uploaders will admit, will require a 
pretty high

> definition copy - would need a scan of the brain almost molecule by molecule. 
This

Possibly, _literally_ molecule by molecule.

> would be so that ion gates and chemical receptors would not only be recorded 

> numerically but also in terms of position. On top of this, current suspensions
are

> pretty disruptive, so the scanning will have to be done on an even finer scale
so 
> that calculations can be made as to where things should have been.

Resolution which current AFM does achieve, especially if the speciment is 
not flexible (vitrified).


> Okay, MRI might be able to do this scanning, but the knowledge needed to 
calulate 

NMR microscopy is limited to um voxels solely, alas, and then only on 
mice/insects.


> what the information about fine brain structure means would need a knowledge 
of
> so much greater than we have know that it we'll probably need to have done a 
> plethora of nano-scale experiments to get there.

Proximal probe microscopy is an established discipline by now. Many are 
now doing AFM in vivo, though AFM of vitrified tissue afaik has not been 
done yet (hint, hint). CryoAFMs, and even UHV cryoAFMs are a quite recent 
invention, and not exactly cheap yet.
 

> On top of this, gung ho neuro-patient that I am, I remain to be convinced that
an 

> uploaded copy of me would actually be me. Wimpy though you may think this 
point

The identity dilemma has been discussed to death on diverse relevant 
list, it is a basic FAQ item. Personally, I believe in it intelectually, 
but not emotionally.


> of view, you have to agree that my uploaded copy (version) would agree with me
and

> so consider its ancestor cruelly murdered by the people that did the upload. 
So 
> beware of angry robots that answer to my name seeking vengeance!

;)

> TTFN
> 
> 
> Garret

> PS Sorry for the lack of references, but I don't think the information I drew 
on
> was too controversial (and its ages since I left University!)
> 
> ----------------------------------------------------------------------

ciao,
'gene


Rate This Message: http://www.cryonet.org/cgi-bin/rate.cgi?msg=7853