Newsgroups: comp.ai.alife
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!uhog.mit.edu!bloom-beacon.mit.edu!eru.mt.luth.se!news.luth.se!sunic!news.chalmers.se!news.gu.se!gd-news!d6244
From: sa209@utb.shv.hb.se (Claes Andersson)
Subject: Re: Lamarckian Evolution
Message-ID: <1995Feb8.100831.10995@gdunix.gd.chalmers.se>
Sender: usenet@gdunix.gd.chalmers.se (USENET News System)
Nntp-Posting-Host: d6244.shv.hb.se
Organization: Dept. of economy and computer science.
X-Newsreader: News Xpress Version 1.0 Beta #2.1
References: <1995Feb1.200012.19562@gdunix.gd.chalmers.se> <3gs6v7$2og@laplace.ee.latrobe.edu.au> <1995Feb4.043915.6827@gdunix.gd.chalmers.se> <3h47mj$7db@laplace.ee.latrobe.edu.au>
Date: Wed, 8 Feb 1995 16:27:39 GMT
Lines: 113

khorsell@ee.latrobe.edu.au (Kym Horsell) wrote:
>In article <1995Feb4.043915.6827@gdunix.gd.chalmers.se> sa209@utb.shv.hb.se (Claes Andersson) writes:
>>I don't think you got my point there. Of course germline DNA can be affected
>>by genetically defines mechanisms but do you realise the tremendous complexity
>>of a genetic re-programmer? It should take a information processor that would
>>have to be much more advanced than the brain. We talk about super NP-complete
>>problems here! Take a look at the embryology of any species and get a hunch of
>>the way genes work, they don't build the body with blocks to say the least!
>
>You are still arguing along the lines of "I cant see how it would work"
>and are also supplying a few (what I shall call) straw men along the way.

 You are right, I cant see how it would work. The thing is: We know that
evolution can come up with tremendous adaptations. BUT we don't know
where to stop. A Lamarckistic reprogrammer is something completely else...
It's about a numbercruching of an infinit number of factors, It can't be.
>
>Why should a "genetic reprogrammer" be any more complex that the
>Drawinian model you already have claimed is understandable? You seem
>to be claiming that such a mechanism for Lamarckian evolution must
>be infalible and always correctly encode inforamation into the DNA.
>Obviously (to me) this is not required. It may make mistakes; it may
>also neglect to encode much information (e.g. injuries);  it doesn't
>need to encode it in a "straightforward" manner. It is also the
>case that Lamarckian effects might show up very strongly in
>simple creatures (e.g. when germ cells == all cells) but be
>ameliorated in more complex ones. Lamarck, after all, was mainly
>concerned with invertebrates. ;-)

  I can take an example:

  I  have made a program which optimise the strategy for the guidance of a vehicle
between two points in a course with random obstacles. To in an evolutionary way
evolve a good strategy takes up about five minutes and 100kb.

  Very well, it cannot be affected at all by it environment but we can refer to its
strategy as its phenotype and the genom as it genome of course.

Assume that:

Genome:   1,4,11,2,7

Gives it a part of its strategy like this:

PRIORITY 7: IF Target LEFT THEN TURN Left.

How would it in anyway be possible for anything that the vehicle does to affect
the genome in a meaningful way? Bumbing into a wall.. what would happen..

It would take that you had a look at all possible permutations of the environment
and even in the case with this simple disrete environment where a place can be
either obstacle or not obstacle the calculations would be ENORMOUS.

 You use evolution to deal with a large searchspace just because you don't have the
answers required for Lamarckism.

 If I, on the other hand, decides to add a check that reprogrammed vehicles with
stupid strategies like, IF target to the left then turn right. It wouldn't be Lamarckism
it would be more like divine intervention and if you don't believe this to be possible
it is impossible. A neural network that allows it to draw conclusions about the
course would be learning. Bumping into the wall and having this to change the
genes would be Lamarckism.

 Please read, for expample, The Selfish Gene by Richard Dawkins for a good
explaination about why Lamarckism is impossible.

>
>Contrarywise I could use evolutionary arguments in this line. Does
>Lamarckian evolution have any benefits over non-Lamarckian
>evolution? I know you have stated elsewhere that, effectively,
>it can not. But I rather think this goes against information theory.
>If a process uses more information -- in this case Lamarck uses
>data gathered from the environment as well as live/die information
>as does Darwinian model -- then it should allow faster adaptation
>to change. Does faster adaptation give a process an advantage over
>a more slowly-adapting one? Obviously this is arguable, but I would
>imaging there is SOME speed faster than straight Darwin that
>will win and have no nasty side-effects like "going too far"
>and have useful genotypes disappear too quickly for a pool. If such
>is the case then we might argue that WILL be the case in Nature.

  A faster rate of adatation is of course an advantage. (The humans don't
evolve sharp teeth, they create a spear instead.. that's the same thing)
In the same way we can do our maths faster in we instead of solving the
tasks look them up in the solutions. But as you know, there are no such
solutions in nature.

 I think you have a little bit too much faith in evolution, what can't be done
can't be done. Further I don't thing you know how information is coded
in the genes... you can't find ANYTHING but recipes on enzymes. You ought
to read soething about embryology to find out that any sort of mixing with
the genes is doomed. But as I've said, perhaps if a species lives in an
environment that rapidly change between two or three distinc states it could
be possible that it can choose between differenct types of gametes that
result in differently adapted offspring but it is still no sort of reprogramming.


>If you argue that Lamarckian evolution will be no "faster"
>than Darwinian then I'm quite prepared to do battle with
>some simple models. ;-)

 I am convinced you are. That's not the problem. As I said, I don't think
you are really aware of how the genes work. Evolution, since it is irreversible
"Dollo's law" creates a genome full of patches and, more obvious, alterations
by inserting alterations after the growth of what shall be altered. The lungs
, for example, grows from our guts! How would something be able to alter
the genome to accomplish a meaningful change to the lungs? Would inhaled
sulphurous oxid trigger a mechanism that first altered our digestive organs
early development, and then another patch  to change it back again so that
the right thing would occure. I'm sorry.. Lamarckism is impossible. People
found out that during the 19'th century for sure.

Claes Andersson. University of Bors. Sweden.
