Re: Evolution: A few questions

On Fri, 18 Jun 2004 19:37:20 +0200 Peter Ruest <pruest@pop.mysunrise.ch>
writes:
>
> D. F. Siemens, Jr. wrote:
>
> >On Wed, 16 Jun 2004 17:01:43 +0200 Peter Ruest
> ><pruest@mail-ms.sunrise.ch> writes:
> >
> <snip>
> I wrote "...the environment's 'answer' remains at most one bit -
> yes/no
> for each phenotypic result." Notice that "at most one bit" means
> anything between zero and one. In fact, usually it's much less than
> one
> bit. Each time a new (or just slightly changed) phenotypic feature
> of an
> organism occurs, it interacts with the environment, in that the
> changed
> feature is more or less "fit" than the original one - with respect
> to
> the current environment. By definition, this fitness difference
> results
> in a different number of progeny (or is defined by this result). To
> be
> somewhat more precise, this effect happens to the first individual
> organism changed, but the outcome is of course modified by a very
> large
> stochastic uncertainty. But in the long run, it happens to the
> species
> as a whole - if the change doesn't disappear (by negative selection
> or
> random drift) before it is fixed in the gene pool of the species.
> But in
> any case, the "response" of the environment is at most "yes, this is
> a
> very good change, we keep that" - which corresponds to one bit of
> information.
>
<snip>

In this I see that part of the problem lies in "bit." A bit encodes
either 0 or 1, not a value between 0 and 1. The latter is required for
probability. Indeed, to code exactly for a probability would involve an
infinity of bits--but we cut the sequence off and note something like
P<10^-3. In digital computers, we have moved from 4 bits through 8 bits
(a byte) to the current standard 32 bits, with 64 bits handled as a word
by the latest processors. 64 bits allow encoding a number up to about
18x10^18, larger than most of us ever require, but it can't be
manipulated to cover 0-1, which implies the mathematical continuum, which
is implicit in probability theory.

The response of nature to a change (mutation or "triggering" effect)
within a stable or changing environment is thus not a simple yes/no.
Advantage is never P=1. Even under artificial conditions, advantage seems
to lie in single percentages.

I also see a problem with using Shannon theory to measure information. A
duplication in the genome does not necessarily have the same information
content as the original entity if what I read about place effects holds.
It also appears that single genes produce more than one protein, making
life more messy than theory.
Dave
Received on Fri Jun 18 15:49:21 2004