1000 humans in a grain of salt

Henri Kluytmans (hkl@stack.nl)
Thu, 22 Jan 1998 01:27:15 +0100


I had a discussion with somebody about the amount of
information required to store a persons state of mind.
I have written down my reasoning here...

By the way, Hans Moravec used this same kind of reasoning:
Estimate the amount of information required to describe
the brain by estimating the total amount of information
entering the brain.

---------------

First lets assume that every arbitrary state of the
human brain can be approached discretely.

The beginning state of the brain (at least before birth,
or better, before any actual inputs arrive) we should
in principle be able to deduct from the information
stored in the DNA. (The DNA contains about 10^10 bits)

Lets assume that the exact behavior of the brain can be
predicted. However, we do not have to understand how
the total system can behave as it will behave.
It is sufficient to be able to exactly predict how
the seperate elements from which the brain is constituted
(i.e. the neurons) will behave (when you know how they
are connected).

Now, every new successive state should be derivable from
the original state and the inputs it receives during the
interval in between. (The inputs to the brain also include
those originating from the rest of the internal body.)

The question now is : how much is the total information
input a person receives during his life. Or, on average,
how much input does a person receive during a day.

Of course this input enters via the sensory organs. You
could try to estimate the amount of digital information
per unit of time, required to discretely describe the
(ordinary average) input for ALL seperate sensory nerves.

However not all information that enters the brain (or the
sensory nerves) is actually being used by the brain. It
filters the information and uses only parts of it. We
should be able to do this filtering beforehand. For
example, the brain does not remember every detail of
a picture it has seen (or even one it is seeing).

Lets formulate it differently:

Assumed we would connect all sensory nerves to a computer.
Assumed the steps of the DACs are sufficiently small to
not be noticed by the brain. The question then is:
how much information per unit of time do I need to
simulate the real life environment of a human individual
in such a way that his brain will not notice the difference.

I guess that, for example, 50 gigabytes per day should
be sufficient to describe (simulate) all the sensory
impressions our brains receive during a day, without
our brains being able to notice the diffenrence between
the real impressions and the simulated ones.

An information input of 50 gigabytes per day results
in a total of approximately 10^16 bits after an average
human life time. The 10^10 of the DNA are much less and
can thus be neglected.

Of course these methods only indicate what could be
achieved when optimal compression methods are being
used. In what way (the algoritm by which) this compression
can be obtained is another (much more complex) problem.

However there is still some time left. After the arrival
of the singularity a posthuman jupiterbrain could try to
spend some time to come up with a good algoritm. :->

When molecular tape storage systems should arive that
can store 10^22 bits in a cubic centimeter, this would
mean that the identities of a 1000 human equivalent
individuals could be stored in a cubic millimeter (the
size of a grain of salt?).

Can somebody detect a fault in this reasoning?

=====================================================================>
>Hkl -------> Technology & Future at www.stack.nl/~hkl
Transcedo ---> Dutch Transhumanist Society at www.dse.nl/~transced
Because the future is where we will spend the rest of our lives ...
You see things and ask "Why?" ; I dream things and ask "Why not?"