[MURG] RFC - necessary neuron info

[Eugen Leitl]

On Sat, Oct 04, 2003 at 09:51:30PM -0500, Joseph J. Strout wrote:

> Well, in general, anything that changes on the timescale of minutes 
> or hours is probably not essential; loss of that should at worst 
> produce a small retrograde amnesia (which is quite common, e.g. due 
> to head trauma or anaesthesia, and poses no problems for survival).

Considering the amount of structural information erasure occuring with a
typical cryonics patient, we're talking about one hell of a retrograde
amnesia. I'm not sure how many here realize how far the worst and the best
case are spread apart. It's not very pretty.
 
> >People seem to be able to survive seizures. As proof,
> >consider all the people that survive seizures.
> 
> Yes, this is a very good example -- these screw up the brain in all 
> sorts of transient ways, yet you can recover just fine with identity 
> intact.

Ditto appears for 30 sec of normothermic ischaemia or hypothermia, or even
plain drugs (such as barbiturates). Effectively flat EEG in all of the above,
and trivially reversible (you'll get some hyperactivity in normothermic
ischaemia case, but that can be blocked with proper medication. None of it
results in visible damage over short episodes).
 
> I think you're making it too complicated.  All neurons (and virtually 
> all cells) are intrinsically active.  And by the time we have the 
> technology to do such a scan, we'll know exactly what all the types 
> of neurons are in the brain, be able to identify them, and know what 
> their typical "default" state is.  We'll simply initialize the 
> emulation to default states, turn it on, and let itrun -- this 
> certainly will be a much smoother startup process than, say, 
> recovering from a seizure.

CNS activity attractors emerge very nicely from system noise. If we can't
simulate that very basic, trivial function we can as well go home now.
 
> None of those worry me too much except one: gene activation states. 

Gene networks of course have a state and of course have a very definite
external activity (see embryomorphogenesis for an illustration what the cells
can actually do). These are of course not magic, and can be parametrized. 
However, for time being we're looking at the lowest feasible level of theory
simulation imaginable, so we know where the cutoff is, and where
we can start parametrizing. Everything else is completely ad hoc.

> It's possible that there will be genes which are 
> activated/deactivated as an important part of the information-storage 
> process, but which do not manifest in any more obvious phenotype. 

I think that's a safe bet. Even if they result in phenotype change, the
simplest way to generate a phenotype change is to model the gene activity
network.

> That would not be a show-stopper, but it would be a great nuisance, 
> and I hope it's not true.  We do know that learning involves gene 
> activation, but they're mostly genes known to result in protein 
> synthesis, and this is probably just the machinery of forming new 
> synapses (or strengthening existing ones, or retracting them, etc.) 
> -- i.e., all things which ultimately result in changes in morphology. 
> Or at least, that's my hope.

You're rather optimistic here, but it might be true. Right now we're lacking
data to even calibrate the model. Both in vivo imaging and activity recording
are at their infancy.

-- Eugen* Leitl <a href="http://leitl.org">leitl</a>
______________________________________________________________
ICBM: 48.07078, 11.61144 http://www.leitl.org
8B29F6BE: 099D 78BA 2FD3 B014 B08A  7779 75B0 2443 8B29 F6BE
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
Url : http://askja.bu.edu/pipermail/murg/attachments/20031008/f1471cf1/attachment.pgp