From: Eugene Leitl (Eugene.Leitl@lrz.uni-muenchen.de)
Date: Fri Mar 28 1997 - 13:35:07 MST
On Thu, 27 Mar 1997, Algimantas Malickas wrote:
> > [ cryonics is _the_ cornerstone of uploading ]
> Let us assume, that live brain uploading is more customary problem, than
> frozen brain (I believe, this is true). In this case the cryo-patients
I don't think gradual uploading is possible, nor do I buy the
nanoressurrection scenario (I'll believe in drextech when I see solid
evidence, like an experiment-validated STM tip chemistry calculation, and
full dynamics study of a realistic diamondoid cantilever). Cryopreservation
artefacts are far easier to DSP-filter, than refurbishing the actual
physical structures atom-by-atom, molecule-by-molecule, especially in
nonvitrified state, where you have to race deterioration, and introduce
new artefacts by just being there. (GIANT worms come in diminuitive
packages. That's nanotechology, in a nutshell).
In contrast, consider a parallel molecular-resolution scan of a myriad
vitrified brain specimens. By applying contrasting erosion techniques,
as vacuum sublimation (removes water, lets macromolecules intact), UV
flash photolysis etch (tune energy to lyse certain brand of bonds, e.g.
lyse C-C covalents, yet let ice relatively intact), plasma etch, and AFM
imaging of thusly abraded surface an arbitrarily detailed map of the
actual circuitry can be constructed. The resulting voxelset can be ANN
DSPeed to remove vitrification and scan artefact, do contrasting,
segmentation, etc. By processing voxel slabs on the fly, retaining
merely salient, efficiently encoded features (connectivity, geometry,
also both class and individual instance properties), the amount of fast
storage necessary can be reduced to relatively modest amounts (don't get
me wrong, it's still many GBytes). The processed data set would fit in a
roomful of next-generation optical tape cartridges. Now if we only knew
what the minimal salient feature set is...
Consider a neural net cultured on an electrode-grid silicon chip, which
can both activate and multi-channel-record the electrical activity of the
NN. By AFM in vivo, detailed local knowledge about geometry and
activity can be gathered. Now vitrify the specimen, and do either a
straight "ab initio" upload, or GA reverse-engineer a system with exactly
the same state space behaviour. A translation stage (which might consist
of GA breeding the code circuitry, target code to be run on the virtual
machine, which in turn is run on molecular hardware (which is so
braindead, it can be run directly by this spacetime's physics, whatever
that might run or not run on).
We will soon have Mrs Drosophila m.'s genome completely sequenced, and adult
specimens are of a handy size to be imaged by above means in one piece.
By modeling DNA/protein activations, part of cell's machinery, we can
unravel the morphogenetic code, causing neurons and glias to organize. I
think we might witness realtime insect emulations (uploads) in some 50
years, perhaps even sooner. How long after that to human-scale uploads? I
do not know. Frankly, building >H ALife Golems from scratch appears much
easier.
> uploading will be possible later, than lives patients. Why the
> super intelligent uploaded civiliation should upload the primitives
> cryo-patients?
An excellent question, to which I do not know the anwer. However, by
being preserved and not composted beyond any hope of retrieval, does seem
to increase the probablity of being at all uploaded. We're talking of future
posthuman entities (so treat current toddlers with respect, you may owe
them quite a favour) motivation, not acts of physical magick, like
spacetime engineering (anybody which is capable of that, is as much
beyond us as we are beyond a rhabdovirus), or Omega tinkering. I don't
like to rely on hypthetical gods, then rather give me meek mehum engineering.
gods'r'us, sure, but give it a few hundred years more time.
ciao,
'gene
> Regards,
> Algimantas
>
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:44:19 MST