[ExI] The size of individual self

[Rafal Smigrodzki]

On Sun, Apr 28, 2013 at 12:48 PM, Eugen Leitl <eugen at leitl.org> wrote:

>
> I would be interesting to see how much compression starting with
> raw voxelset we can get. Arguably wavelet can give you at least 1:10
> if not 1:100 (though latter could introduce enough artifacts to
> throw a monkeywrence into the feature segmentation), and the
> trace will be even much more compact. However, I think you will need
> the fully annotated feature trace at least at 1-5 nm (with the annotation
> potentially derived from selective sampling to much higher resolution)
> as a point of departure for these hypothetical more compact
> representations. And I feel that encoding will be quite computation-intensive/
> time consuming (but highly paralellizable).

### Yes, the process of generating a compact representation from
scanning data is likely to be very computationally intensive,
involving detailed synapse-level modeling of function, to extract the
important details. By important details I mean the kind of declarative
memory engrams that let me construct my individual high level mental
imagery (e.g a girl named Marzena I interacted with some years ago)
from primitives (a compact mental representation of her face using
links to lists of relative distances between facial landmarks plus a
lot of other specialized encodings we use to store information about
persons and to connect them with a time-stamped narrative of our
existence). There may be billions of synapses involved in generating
the specific mental representation of Marzena in my currently existing
mind but an uploaded and compacted version might do with some very
short lists of links to standard face- and person-representations.

I am assuming here that a hierarchical temporal memory model of our
mind is largely correct, and it relies on the generation of multiple
hierarchical lists of various features of the experienced world. The
vast majority of these lists are very similar between most or many
people - the lists of spatially oriented lines in my V1 area are
equivalent to lists in the minds of all non-blind people, the lists I
use to recognize English words are shared with most English-speaking
persons, etc. etc. Only a small fraction of these lists, describing my
biographical memory, are truly unique and grafting these individual
lists on standardized mental modules should yield subjective
experience not appreciably different from mine (including e.g. all the
reactions and associations that might be evoked by an email from an
old flame).

This means going far beyond compacting a scanned document as an image,
it's more like reading a book and producing a precis. Sort of like
Cliff's notes for the soul.

Rafal