[MURG] Uploading via memory implants?

[Eric Zilli]

It's a minor point, but old information isn't so much deleted as
overwritten or interfered with.

You're confusing two uses of the word information, though. In the
first half of what you say, information seems to mean only memories.
But it isn't memories we try to directly extract from the brain, it's
synaptic weights and neural activities: a different sort of
information. Selectively recording from only particular parts of the
brain (i.e. only things immediately accessable from the surface)
doesn't selectively record memories, it selectively records neurons.
Memories are stored rather component-wise: all the visual aspects of
memories are in one place, auditory aspects in another, temporal
aspects in yet another, and some of these aren't reachable from the
surface (to name one example, the most important part of V1, the area
representing the fovea, is located along the medial wall of the
occipital lobe, out of reach of the electrode array). Thus recording
from a particular part of the brain can record the aspects of memories
associated with that region, but not any other sorts of aspects.
An example: say your understanding of formal social situations (i.e.
you must wear pants, you shouldn't shout, vulgarities tend to be
unwelcome, etc) is encoded along the medial wall of prefrontal cortex,
along with your understand of all other sorts of social situations (I
don't know that this is accurate, but it's a valid example of the sort
of things you'd miss with your method of uploading). Your method may
be able to record memories such as the words for different social
situations, sensory memories of past social events, and so forth, but
no matter how long you record, you won't get the understanding of such
events.

I hope that makes clear the problem of recording only from cortical
surface: you may get a lot of information, but you simply won't get
all the information you'd need.

You can't duplicate a hot dog factory just by studying its walls, no
matter how closely you study them: vital information will remain
unobserved, and thus unaccounted for in the duplicate factory. Even
duplicating the doors in all their detail won't make hot dogs come out
of them. (Worst example ever).

Of course, given the definition of a successful upload that you
offered in your previous email, this won't matter, since you don't
require personality to be the same before and after uploading. The way
I read what you said, it seems the upload need only inherit what would
amount to cortical activity patterns, but that it needn't "act" the
"same" as the original.


> Thomas:
> I agree that transfering all data from the mind will
> pose a serious problem. In my late contemplations on
> the subject I have come to believe that perhaps we
> don't need to transfer all information contained in
> the brain, as long as we transfer the sense of self.
> Notice that in the course of our lives we constantly
> learn new information and delete (in human language we
> call it "forget") old information. So we lose plenty
> of information on daily basis without even thinking
> about being uploaded, and this does not affect our
> sense of self at all. So I believe there must be some
> maximum information we may lose, and some minimum we
> must keep to maintain the sense of self. If that
> minimum is reached in the brain computer transfer -
> the mind uploading experiment may be considered
> successful, and if it is not reached .... we will have
> a problem with a comatose underupload.
> 
> Kind Regards,
> Thomas Weber
> 

-- 
Eric Zilli
Hasselmo Lab - Computational Neurophysiology
Center for Memory and Brain
Boston University
2 Cummington St.
Boston MA, 02215
digfarenough at gmail.com -- www.digfarenough.com