[MURG] A repost

[Joseph J. Strout]

At 5:32 PM -0500 9/16/03, Joseph M. Graham Jr. wrote:

>There has been an idea around for some time that the
>complete contents of a human brain could somehow be
>"downloaded" into a computerlike device, then to either
>be "uploaded" into another brain or to be run as a
>simulation by the computer directly - an electronic
>"you" in a cyber-universe or "android" body.

Actually, the standard terminology here is that copying the brain 
into an artificial device is "uploading" (not "downloading").  We can 
go into the nuances if you like.

>The main frustration to this goal is that our brain
>tissues are not constructed in a way that makes it
>easy to deduce their functional parameters in high
>detail.

True, I believe that serial sectioning is the only approach that will work.

>  Indeed, the electrochemical nature of a
>nerve impulse gives very little signal to detect
>by any means other than the direct insertion of
>an electrode into discrete nerve cells. Getting
>nerves to respond to induced signals is also very
>difficult - again the only viable solution is to
>directly insert an electrode.

No, electrodes are irrelevant to uploading.  We need to capture the 
complete structure of the brain; electrodes disrupt structure rather 
that capture it.

>It is obviously not going to be possible to stick
>an electrode into each and every cell in the brain
>at the same time so we can record the minds
>functional parameters, much less to input new
>information.

Quite so.

>  Some have suggested freezing brains
>and then taking them down molecular layer by
>molecular layer so we can find the telltale signs
>of synapse weights and connections. Yucky, to
>say the least.

Open heart surgery is yucky too.  And yet this is how it'll be done.

>Instead of trying to "scan" synapses and neural
>activity, let's think about having the nerve itself
>play the tattletale. Can an organic chemical be
>fabricated which will lodge inside neurons, or
>be placed there indirectly via a retrovirus, which
>will momentarily change physical shape or even
>give off a high-frequency electromagnetic signal
>when when the cell membrane depolarizes ???

This is a variation on an old idea (it must be, since it's on the 
mind uploading home page, which I haven't updated in ages! -- see 
<http://sunsite.unc.edu/jstrout/uploading>).  A neat variation, I'll 
admit.  It doesn't sound very plausible to me, though...

>In the case of a spontaneous gigahertz radio
>transmission, we could then very easily listen
>to each nerve chatter away with conventional
>radio receiving apparatus.

But there's no reason to think that listening to the neural chatter 
would in any way help us to construct a functional duplicate of the 
brain.

>As for "uploading" ... a vital element in any
>future human/machine interfacing ... we can
>revisit the idea of inserting engineered
>"antennas" into the nerve membranes which,
>when hit with an appropriate external RF signal
>will cause the nerve to fire off.

I don't follow you here.

>Anyway ... if the brain exhibits low compatibility
>with our data and engineering technology, let's
>think about ways to change the brain at the cellular
>level to make it more compatible.

It's pretty compatible already.  The main advances needed are MEMS 
devices to automate the sectioning and scanning, while minimizing 
damage (i.e. information loss) due to the sectioning and scanning 
process.

>  What I have
>suggested is much more readily within our immediate
>capabilities than, say, nano-machines or scanners
>which can easily determine the disposition of every
>atom in a brain.

First, you don't need to know about every atom in the brain.  My 
estimate is that resolution of 10 nm or so will do just fine (though 
that's a question that needs more research in the future).  Second, 
you don't need nano-machines either.  Ordinary MEMS (microtechnology) 
will do just fine.

>  We don't really HAVE to know the
>disposition of every brain atom in order to "download"
>the person, but we do need to know the connection
>diagrams, synapse weightings and chemoreceptor
>responses of each discrete neuron.

Right.  Now you've got it.

>The computing power to run a gross simulation of a
>living brain is promised within 30-50 years - which
>likely means 15-20 years.

If you're an optimist, I suppose.  Or it might mean 60-100 years.  :)

>Wouldn't it be nice to trade a worn-out old body for a fresh new one ?
>You only have to get "you" from one skull into the other ...

Yes, that's why we're all here.

Cheers,
- Joe

-- 
,------------------------------------------------------------------.
|    Joseph J. Strout         Check out the Mac Web Directory:     |
|    joe at strout.net           http://www.macwebdir.com/            |
`------------------------------------------------------------------'