ah, but what you are doing here is answering the question(s)
a) could we make a processor out of DNA and
b) how big would such a processor have to be to process an AI's "program"
Basically, you are calculating the DNA necessary to squeeze a brain's
information content into a strand of DNA.
That would mean that each cell of the body thus created would each be a copy
of the AI, thus, (if we fully replicate the human), get billions of AI's
within the body.
And, as you clearly show, we still don't have the processor to handle the
chemical reactions, unless this is also somehow coded in the DNA
Perhaps a question to Trevor: what is the end product? An AI in a human
body? An AI coded in DNA that could somehow be accessed ( I remember a Star
Trek episode with a 5 minute, holographic, colour and sound message from the
past encoded and encrypted in a combination of Human and Klingon and other
DNA, that survived 2 million years with no alteration - ha!)
Emmanuel
----- Original Message -----
From: "Joshua Bell" <[EMAIL PROTECTED]>
To: "Brin-L" <[EMAIL PROTECTED]>
Sent: August 21, 2001 20:15 Fizzlin'
Subject: Re: A.I. Genetics
> [EMAIL PROTECTED] wrote:
> >
> >Could it be possible, then, to somehow merge the "Digital DNA" of an A.I.
> >with the double helix of a human genome, thereby creating a human being
> >with
> >the "soul" of an A.I.?
>
> In normal parlance:
>
> Given the software and/or data files necessary for a human-equivalent
A.I.,
> could you encode it in the base-pairs of DNA?
>
> Absolutely.
>
> It's the first part that's tricky. :) The limiting factors would be the
> length (since we have very little idea what what the size of an A.I.'s
> infoset is*) and the cohesion of the medium - DNA is pretty stable, but if
> the length ends up being millions of kilometers it might not hold together
> long enough for the encoding to take place.
>
> A more interesting question:
>
> Given a data file representing a state snapshot of a human-equivalent A.I.
> and a description of an automaton which processes the data - in other
words,
> a Turing machine description of an A.I. - can you encode the data on DNA
and
> construct a biological processor (probably a whole suite of enzymes) that
> acts as the automaton?
>
> Probably, but this would probably function unbearably slowly - like a
10^100
> slowdown in subjective time.
>
> Could you cram all of this into a human cell? My guess, no. But again, we
> have no idea what the scales involved are.
>
> >The idea is cool, but I have no idea if it is remotely plausible or
> >laughable
> >on a scientific level -- even in the realm of science fiction.
>
> The SFnal take would be to postulate a subset of this - some biological
> Turing machine that isn't an A.I. but still a useful tool - and injecting
it
> into existing life forms.
>
> ....
>
> * So how much data is necessary to encode an A.I.?
>
> A first crack at a guess for just the data, not counting the code: (# of
> neurons in the brain (1e11) ^ # of interconnections per neuron (1e4) ) * #
> bits to meaningfully represent connection strength/type (16) * # bits to
> address a neuron (26) / #bits per DNA pair (2) = 2e16 base pairs.
>
> For comparison, there are 3e9 base pairs in the human genome. So nominally
> representing the state of a human brain in DNA would take about six
million
> times as much DNA as we cart around in each cell.
>
> For reference, the DNA strand would be about 7e8cm long, or 4000 miles.
>
> Refs:
>
> http://faculty.washington.edu/chudler/what.html
> http://www.ornl.gov/TechResources/Human_Genome/home.html
> http://www.auhs.edu/netbiochem/nucacids.htm
> http://www.webcom.com/~legacysy/convert2/unitconvertIE.html
>
> Joshua
>
>
> _________________________________________________________________
> Get your FREE download of MSN Explorer at http://explorer.msn.com/intl.asp
>
