id or super-ego?
Mark & Roslyn Elkington
study of genetic algorithms reveal theoretical and empirical limitations to
the creative power of the trial and error mechanism?
A new site to watch re these questions is http://www.iscid.org IMO a source
of some well written and cogently argued essays, even if you don't agree
with the ID viewpoint (some from the usual suspects, of course).
For a creationist like myself this area seems crucial. If study of GAs
reveal no in principle limits to CSI creation and increase in novelty,
complexity and functionality, then it would appear that related arguments
against Darwinian evolution would be significantly weakened, and force a
retreat to secondary biological objections in practice, despite theoretical
possibility.
On the other hand, if evolutionary computing, GA problem-solving and
information studies etc do reveal such constraints, Darwinism is dead, and
there does not appear to be convincing naturalistic alternative.
What is really interesting is that this approach takes the argument out of
the hands of biologists and into the hands of mathematicians and computing
scientists (shock, horror...hallelujah?). Even aside from theoretical
analysis, I think at some stage we are going to hear the EC guys saying, "We
have found that in practice you can use GAs to discover solutions to
problems of class A, but not class B." (Life as found on planet earth being
decidedly class B.)
Either way, we seem to have a testable hypothesis, which is an essential
ingredient of a happy new year.
Regards,
Mark Elkington
Sydney
John Wilkins
> Is ID a serious challenge to Darwinian evolution? Specifically, will the
> study of genetic algorithms reveal theoretical and empirical limitations to
> the creative power of the trial and error mechanism?
>
> A new site to watch re these questions is http://www.iscid.org IMO a source
> of some well written and cogently argued essays, even if you don't agree
> with the ID viewpoint (some from the usual suspects, of course).
>
> For a creationist like myself this area seems crucial. If study of GAs
> reveal no in principle limits to CSI creation and increase in novelty,
> complexity and functionality, then it would appear that related arguments
> against Darwinian evolution would be significantly weakened, and force a
> retreat to secondary biological objections in practice, despite theoretical
> possibility.
Good on you, Mark, for accepting this. You are entirely correct in this
respect.
>
> On the other hand, if evolutionary computing, GA problem-solving and
> information studies etc do reveal such constraints, Darwinism is dead, and
> there does not appear to be convincing naturalistic alternative.
Not exactly, although it would weaken some of the arguments used in
favour of evolution as we know it (EAWKI, to avoid the loaded and
misleading term "Darwinism").
This is the main reason why Dembksi is now running his attack on GAs
with the No Free Lunch Theorem. As I understand that (this not being a
speciality of mine), that theorem only applies in restricted cases. If
you want to find algorithmic limitations, you can - Chaitin does not
think he can accommodate EAWKI in his AIT notion of complexity and
information. But Chaitin's response is not to say that EAWKI is false,
but merely that he hasn't yet managed to come up with an entirely
complete characterisation of the world. Dembski is, so far as I can
tell, grasping at straws.
And yet, GAs work, and they work in ways we never expected them to. Even
worse, *physical* approximations to natural selection work, in, say,
robotics and electronics. What do we make of that? Why, we must assert,
as Dembski dies, that this happens because we actually do know what we
were doing even if we could never have actually predicted the outcome.
Get real...
So what about if we find that the NFL theorem works out that in a class
of cases natural selection is ineffectual? What then? Is it ineffectual
in all cases? That is a logical error, and one Dembski had better not
make in print, or I'll have another paper on my hands :-) We already
know cases where NS is ineffectual - they are called constrained
situations and genetic drift. But selection still works to produce
complex cases.
>
> What is really interesting is that this approach takes the argument out of
> the hands of biologists and into the hands of mathematicians and computing
> scientists (shock, horror...hallelujah?). Even aside from theoretical
> analysis, I think at some stage we are going to hear the EC guys saying, "We
> have found that in practice you can use GAs to discover solutions to
> problems of class A, but not class B." (Life as found on planet earth being
> decidedly class B.)
EC? Anyway, the way the research is going, life on earth is decidedly
class A, and there's no reason to think otherwise, yet. If you or the ID
crowd can come up with some *solid* reasons why it must be in B, then we
might start to see some proper papers and research from them. I for one
will not hold my breath.
>
> Either way, we seem to have a testable hypothesis, which is an essential
> ingredient of a happy new year.
>
> Regards,
> Mark Elkington
> Sydney
--
John Wilkins
Occasionally making sense for over 46 years
Dana Tweedy
Mark & Roslyn Elkington <mar...@zeta.org.au> wrote in message
news:AvvY7.1509$ko4.1...@nasal.pacific.net.au...
> Is ID a serious challenge to Darwinian evolution?
Not until they produce a testable theory that explains the data better.
> Specifically, will the
> study of genetic algorithms reveal theoretical and empirical limitations
to
> the creative power of the trial and error mechanism?
My prediction would be no such limits will be found.
>
> A new site to watch re these questions is http://www.iscid.org IMO a
source
> of some well written and cogently argued essays, even if you don't agree
> with the ID viewpoint (some from the usual suspects, of course).
>
> For a creationist like myself this area seems crucial. If study of GAs
> reveal no in principle limits to CSI creation and increase in novelty,
> complexity and functionality, then it would appear that related arguments
> against Darwinian evolution would be significantly weakened, and force a
> retreat to secondary biological objections in practice, despite
theoretical
> possibility.
No such limits have been seen so far, and it's not likely such limits will
be found.
>
> On the other hand, if evolutionary computing, GA problem-solving and
> information studies etc do reveal such constraints, Darwinism is dead, and
> there does not appear to be convincing naturalistic alternative.
Any scientific alternative must be naturalistic. Science cannot accept
supernatural explanations.
>
> What is really interesting is that this approach takes the argument out of
> the hands of biologists and into the hands of mathematicians and computing
> scientists (shock, horror...hallelujah?).
It seems to me that the question should be decided by examining the
evidence, not by abstract computational models.
>Even aside from theoretical
> analysis, I think at some stage we are going to hear the EC guys saying,
"We
> have found that in practice you can use GAs to discover solutions to
> problems of class A, but not class B." (Life as found on planet earth
being
> decidedly class B.)
What are your class A and B?
>
> Either way, we seem to have a testable hypothesis, which is an essential
> ingredient of a happy new year.
Since there are no observations that would indicate that there are limits
that would prohibit evolutionary changes, I doubt your hypothesis would be
confirmed. Even if it was, you still haven't shown any evidence of a
designer, nor would ID be the winner by default.
DJT.
R. Baldwin
>
> Either way, we seem to have a testable hypothesis, which is an
essential
> ingredient of a happy new year.
>
In what way is ID falsifiable? Dembski kind of dances around this.
Steven J.
> study of genetic algorithms reveal theoretical and empirical limitations to
> the creative power of the trial and error mechanism?
>
questions.
It is quite conceivable -- it is probably quite certain -- that there
are important limits to the ability, of imperfect replication and
subsequent differential reproductive success, to produce certain kinds
of complex structures. It is equally conceivable that life on Earth
possesses such features -- although establishing the first would not
establish the second.
Indeed, common in discussions of the evidence for common descent is
the *absence* of certain features from life and the fossil record.
The entire "panda's thumb" and "vestigial organs" argument is based on
the idea that Darwinian mechanisms can't quickly and easily erase, and
are very unlikely to duplicate, the results of complex histories. The
whole "nested hierarchy" argument depends on limitations to genetic
algorithms.
But perhaps there are, indeed, features in life that cannot be
generated by Darwinian mechanisms. The advocates of ID try to to
demonstrate the inadequacy of such mechanisms. To the extent that the
succeed, they show that new mechanisms need to be proposed and tested.
Then they pull a shell game.
I'll cite Behe, but I see no evidence that Dembski or any other ID
theorist is different on this count. He argues at the start of
_Darwin's Black Box_ that most biologists ignore the "elephant in the
living room" of design, because they fear that, while the elephant
merely says "design" on one side, on the other it might say "God."
Later in the book, he deals with the questions of, shall we say,
questionable design choices in nature, by arguing that we are entitled
to no assumptions about how a Designer might work. That is to say,
the other side of the elephant surely doesn't say "God," at least not
with reference to any God to whom humans might be accountable. Any
God Who's interested in humans and how we live our lives presumably
has some specific character, purposes, and methods.
Thus, ID advocacy uses the inscrutability of the Designer, on the one
hand, to shield ID from testability. The ID hypothesis can't be
falsified by any set of observations, if we can't say what the design
philosophy of the Designer is. On the other hand, they show no
hesitation in implying that what is, at best, no more than evidence
for unknown but potentially quite naturalistic mechanisms for the
origin of species, is evidence for their specific theology.
In brief, disproving Darwinism will be zero evidence for ID, unless
"ID" is used so broadly that it has no particular theological *or*
scientific value. Why get all excited over a label for "we don't have
the complete answer?"
>
> A new site to watch re these questions is http://www.iscid.org IMO a source
> of some well written and cogently argued essays, even if you don't agree
> with the ID viewpoint (some from the usual suspects, of course).
>
> For a creationist like myself this area seems crucial. If study of GAs
> reveal no in principle limits to CSI creation and increase in novelty,
> complexity and functionality, then it would appear that related arguments
> against Darwinian evolution would be significantly weakened, and force a
> retreat to secondary biological objections in practice, despite theoretical
> possibility.
>
> On the other hand, if evolutionary computing, GA problem-solving and
> information studies etc do reveal such constraints, Darwinism is dead, and
> there does not appear to be convincing naturalistic alternative.
>
Again, it would have to be shown (as you implicitly note below) not
merely that such constraints exist, but that they would have to be
surpassed to produce some feature of life on Earth.
>
> What is really interesting is that this approach takes the argument out of
> the hands of biologists and into the hands of mathematicians and computing
> scientists (shock, horror...hallelujah?). Even aside from theoretical
> analysis, I think at some stage we are going to hear the EC guys saying, "We
> have found that in practice you can use GAs to discover solutions to
> problems of class A, but not class B." (Life as found on planet earth being
> decidedly class B.)
>
Biologists, though, have had some success, using bacteria in novel or
hostile media, in producing some quite exotic systems through
Darwinian mechanisms. I think they might conclude, if the
mathematicians announced that "our model says that your mechanism
won't work," that the mathematical model rather the Darwinian
mechanism is flawed. But that, as C.S. Lewis said in rather a
different context, is crossing the bridge before you've ascertained
that the river even exists.
>
> Either way, we seem to have a testable hypothesis, which is an essential
> ingredient of a happy new year.
>
> Regards,
> Mark Elkington
> Sydney
-- Steven J.
Bobby D. Bryant
> Mark & Roslyn Elkington <mar...@zeta.org.au> wrote:
>
>> On the other hand, if evolutionary computing, GA problem-solving and
>> information studies etc do reveal such constraints, Darwinism is
>> dead, and there does not appear to be convincing naturalistic
>> alternative.
> Not exactly, although it would weaken some of the arguments used in
> favour of evolution as we know it (EAWKI, to avoid the loaded and
> misleading term "Darwinism").
Let me put in a general word of warning re comparing GAs to biological
evolution. The comparison is useful up to a point, but there's a danger
of pushing the analogy too closely.
GAs were "inspired by" biological evolution (or perhaps merely dog
breeding), but they are not generally used to model biological
evolution. Most often they are used to find a "decent" solution for
some specific problem that would require too much time or space to solve
exactly. They do not usually follow the mechanisms of biological
evolution in anything but the most superficial form, and of course they
only use a trivial fraction of the resources that biological evolution
has used over the ages.
Biological evolution, IMO, is not trying to solve any problem at all,
any more than gravity or chemistry is. (I suppose you could say that
gravity and chemistry are "trying" to find the lowest energy state of
certain kinds of systems, but I would be very surprised to discover that
any scientist actually does think of them that way, at least as the
top-level understanding of the system under study.)
GAs can still be useful for demonstrating certain points, e.g. that
breeding, mutation, and weeding-out-by-fitness can indeed generate "new
stuff" that was not in the original population, and that random
mutations can sometimes be beneficial (as many ill-informed creationists
like to deny). However, at each and every invocation of GAs to
demonstrate a point regarding biological evolution, the reflexive
response should be to immediately ask yourself whether the analogy is
being pushed too far. This should save much grief resulting from
ill-gotten conclusions.
> This is the main reason why Dembksi is now running his attack on GAs
> with the No Free Lunch Theorem. As I understand that (this not being a
> speciality of mine), that theorem only applies in restricted cases. If
> you want to find algorithmic limitations, you can - Chaitin does not
> think he can accommodate EAWKI in his AIT notion of complexity and
> information. But Chaitin's response is not to say that EAWKI is false,
> but merely that he hasn't yet managed to come up with an entirely
> complete characterisation of the world. Dembski is, so far as I can
> tell, grasping at straws.
Dembski is going to get creamed if he stakes his credibility on using
the No Free Lunch theorem to refute the reality of biological evolution.
The essense of the NFL theorem is something like this -
"For two search algorithms A and B, if A outperforms B on n search
problems, then B likewise outperforms A on approximately n other search
problems."
An immediate corollary is that there is no universal "best" search
algorithm _without_reference_to_the_application_.
See Wolpert & Macready, "No Free Lunch Theorems for Search",
http://citeseer.nj.nec.com/wolpert95no.html. (If it turns out that I
only have access to that site by academic arrangement, let me know and
I'll see if I can find a direct link to his paper on line.)
There is a companion paper w.r.t. optimization (rather than search), but
I have not had time to read it.
Attempting to apply the NFL to biological evolution immediately
encounters one of the following problems:
1) Is biological evolution even a "search problem", in any meaningful
sense? If not, the NFL simply doesn't apply.
2) Even if you do think biological evolution is a search problem, notice
that no one is claiming that evolution is the "best" solution to *all*
search problems. Without such a claim, the NFL simply doesn't apply.
I personally don't care to think of evolution as a search problem
because of the teleological nature of that claim. (Granted, natural
selection does "improve the fit" over time, but I'm still a bit
reluctant to view it as a search or optimization problem; I prefer to
view the ToE as descriptive rather than prescriptive.)
For that matter, no one seems to be making even the very restricted
claim that biological evolution is the best way of solving the single
"search problem" of finding an organism that is best suited to the
environment. Clearly, evolution is the "algorithm" that nature actually
uses, but that is not the same as a claim that it is the best possible
algorithm. For that matter, evolution does not seem to actually be
finding the best fits between organisms and the environment, as
witnessed by bad backs, infections of the appendix, nerves looping down
a giraffe's neck and then back up, etc.
And there is *nothing* in the NFL that claims that a non-best search
algorithm can never find a decent solution to a problem.
I suspect Dembski is going to appeal to the fallacious notion that
humans are the "correct" answer to some search problem, add the
fallacious claim that the "correct" answer could not have been found by
anything but the "best" search algorithm, and then correctly invoke the
NFL to prove that no such "best" algorithm exists. Brings to mind a
parable about building towers on shifting sands...
> Even worse, *physical* approximations to natural selection work, in,
> say, robotics and electronics. What do we make of that? Why, we must
> assert, as Dembski dies
^ Typo, or Freudian slip?
Bobby Bryant
Austin, Texas
Bobby D. Bryant
> On the other hand, if evolutionary computing, GA problem-solving and
> information studies etc do reveal such constraints, Darwinism is dead,
> and there does not appear to be convincing naturalistic alternative.
Interesting. Notice that "Darwinism" is merely a model for
understanding the cause of the *observed* changes to the earth's
biosystem over time. Even if you somehow rule out "Darwinism" the
theory, you're still left with evolution, the fact of nature.
Bobby Bryant
Austin, Texas
Mark & Roslyn Elkington
Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
news:a0uo3m$mar$1...@geraldo.cc.utexas.edu...
> On Tue, 01 Jan 2002 21:54:58 -0600, Mark & Roslyn Elkington wrote:
>
> > On the other hand, if evolutionary computing, GA problem-solving and
> > information studies etc do reveal such constraints, Darwinism is dead,
> > and there does not appear to be convincing naturalistic alternative.
>
> Interesting. Notice that "Darwinism" is merely a model for
> understanding the cause of the *observed* changes to the earth's
> biosystem over time.
I'm using the term "Darwinism" as the neo-Darwinian component of the modern
synthesis, i.e. mutation and natural selection. If you lose that mechanism,
what are the serious naturalistic alternatives?
> Even if you somehow rule out "Darwinism" the
> theory, you're still left with evolution, the fact of nature.
Spoken like a true ontological naturalist.
Mark
Dunk
<mar...@zeta.org.au> wrote:
>
>Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
>news:a0uo3m$mar$1...@geraldo.cc.utexas.edu...
>> On Tue, 01 Jan 2002 21:54:58 -0600, Mark & Roslyn Elkington wrote:
>>
>> > On the other hand, if evolutionary computing, GA problem-solving and
>> > information studies etc do reveal such constraints, Darwinism is dead,
>> > and there does not appear to be convincing naturalistic alternative.
>>
>> Interesting. Notice that "Darwinism" is merely a model for
>> understanding the cause of the *observed* changes to the earth's
>> biosystem over time.
>
>I'm using the term "Darwinism" as the neo-Darwinian component of the modern
>synthesis, i.e. mutation and natural selection. If you lose that mechanism,
>what are the serious naturalistic alternatives?
Just how is this going to get lost? How can you stop it from
happening? Destroy all life on earth? Freeze time?
Dunk
Mark & Roslyn Elkington
John Wilkins <john.w...@bigpond.com> wrote in message
news:1f5e20o.19qaxcj1dhj04lN%john.w...@bigpond.com...
> Mark & Roslyn Elkington <mar...@zeta.org.au> wrote:
>
> > Is ID a serious challenge to Darwinian evolution? Specifically, will
the
> > study of genetic algorithms reveal theoretical and empirical limitations
to
> > the creative power of the trial and error mechanism?
> >
> > A new site to watch re these questions is http://www.iscid.org IMO a
source
> > of some well written and cogently argued essays, even if you don't agree
> > with the ID viewpoint (some from the usual suspects, of course).
> >
> > For a creationist like myself this area seems crucial. If study of GAs
> > reveal no in principle limits to CSI creation and increase in novelty,
> > complexity and functionality, then it would appear that related
arguments
> > against Darwinian evolution would be significantly weakened, and force a
> > retreat to secondary biological objections in practice, despite
theoretical
> > possibility.
>
> Good on you, Mark, for accepting this. You are entirely correct in this
> respect.
The sharpest swords are double-edged.
> >
> > On the other hand, if evolutionary computing, GA problem-solving and
> > information studies etc do reveal such constraints, Darwinism is dead,
and
> > there does not appear to be convincing naturalistic alternative.
>
> Not exactly, although it would weaken some of the arguments used in
> favour of evolution as we know it (EAWKI, to avoid the loaded and
> misleading term "Darwinism").
More precisely, Neo-Darwinism, meaning mutation plus natural selection. I
think that would be dead, and there are no serious alternatives I know of
(Kaufman's self-organisation, or teleology?).
> This is the main reason why Dembksi is now running his attack on GAs
> with the No Free Lunch Theorem. As I understand that (this not being a
> speciality of mine), that theorem only applies in restricted cases. If
> you want to find algorithmic limitations, you can - Chaitin does not
> think he can accommodate EAWKI in his AIT notion of complexity and
> information. But Chaitin's response is not to say that EAWKI is false,
> but merely that he hasn't yet managed to come up with an entirely
> complete characterisation of the world. Dembski is, so far as I can
> tell, grasping at straws.
>
> And yet, GAs work, and they work in ways we never expected them to.
They do work. My favourite example is
http://208.245.156.153/archive/output.cfm?ID=1455
But how this relates to the evolution of living things *is* the question at
hand. No-one is saying GAs don't work, but some people are saying they can't
work to produce the diversity and complexity of life as we know it, and are
attempting to find mathematical proof of this.
> Even
> worse, *physical* approximations to natural selection work, in, say,
> robotics and electronics. What do we make of that? Why, we must assert,
> as Dembski dies,
Freudian typo :-)
> that this happens because we actually do know what we
> were doing even if we could never have actually predicted the outcome.
> Get real...
>
> So what about if we find that the NFL theorem works out that in a class
> of cases natural selection is ineffectual? What then? Is it ineffectual
> in all cases? That is a logical error, and one Dembski had better not
> make in print, or I'll have another paper on my hands :-) We already
> know cases where NS is ineffectual - they are called constrained
> situations and genetic drift. But selection still works to produce
> complex cases.
Well, that's the whole the debate. Can NS in fact work to produce the set of
complex cases of living organisms, within the phase space and time/trials
available? Or can it only produce some mathematically bounded degree of
variation, i.e. microevolutionary adaptations? (with usual disclaimers for
the use of that problematic terminology).
> > What is really interesting is that this approach takes the argument out
of
> > the hands of biologists and into the hands of mathematicians and
computing
> > scientists (shock, horror...hallelujah?). Even aside from theoretical
> > analysis, I think at some stage we are going to hear the EC guys saying,
"We
> > have found that in practice you can use GAs to discover solutions to
> > problems of class A, but not class B." (Life as found on planet earth
being
> > decidedly class B.)
>
> EC?
Evolutionary computing.
> Anyway, the way the research is going, life on earth is decidedly
> class A, and there's no reason to think otherwise, yet. If you or the ID
> crowd can come up with some *solid* reasons why it must be in B, then we
> might start to see some proper papers and research from them. I for one
> will not hold my breath.
I'm holding my breath. As I've said, I think NFL or similar is potentially
decisive, in a double-edged way. It also side-steps much of the
rhetoric-as-science offered by the likes of Dawkins, Gould etc.
Of course, an inconclusive result is still possible, but it seems to me that
through empirical and/or theoretical study, there is room for optimism for a
definitive answer as to whether or not the postulated Neo-Darwinian
mechanism is sufficient.
rgds,
Mark
Mark & Roslyn Elkington
Dunk <pdu...@magicnet.net> wrote in message
news:3c3307d4...@fl.news.verio.net...
> >> > On the other hand, if evolutionary computing, GA problem-solving and
> >> > information studies etc do reveal such constraints, Darwinism is
dead,
> >> > and there does not appear to be convincing naturalistic alternative.
> >>
> >> Interesting. Notice that "Darwinism" is merely a model for
> >> understanding the cause of the *observed* changes to the earth's
> >> biosystem over time.
> >
> >I'm using the term "Darwinism" as the neo-Darwinian component of the
modern
> >synthesis, i.e. mutation and natural selection. If you lose that
mechanism,
> >what are the serious naturalistic alternatives?
>
> Just how is this going to get lost? How can you stop it from
> happening? Destroy all life on earth? Freeze time?
I'll put it more plainly: If the postulated Neo-Darwinian mechanism of
mutation and natural selection is demonstrated to be insufficient to explain
the diversity and complexity of life on earth, what other serious
naturalistic mechanisms do you know of?
Mark
Wade Hines
Mark & Roslyn Elkington wrote:
> John Wilkins <john.w...@bigpond.com> wrote in message
...
> > And yet, GAs work, and they work in ways we never expected them to.
>
> They do work. My favourite example is
> http://208.245.156.153/archive/output.cfm?ID=1455
>
> But how this relates to the evolution of living things *is* the question at
> hand. No-one is saying GAs don't work, but some people are saying they can't
> work to produce the diversity and complexity of life as we know it, and are
> attempting to find mathematical proof of this.
Can you cite precedence for a "mathematical proof" by failed simulation?
Mark & Roslyn Elkington
Steven J. <stev...@altavista.com> wrote in message
news:127ccf2e.02010...@posting.google.com...
> "Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in message
news:<AvvY7.1509$ko4.1...@nasal.pacific.net.au>...
> > Is ID a serious challenge to Darwinian evolution? Specifically, will
the
> > study of genetic algorithms reveal theoretical and empirical limitations
to
> > the creative power of the trial and error mechanism?
> >
> Those are two separate questions. Possibly even three separate
> questions.
>
> It is quite conceivable -- it is probably quite certain -- that there
> are important limits to the ability, of imperfect replication and
> subsequent differential reproductive success, to produce certain kinds
> of complex structures. It is equally conceivable that life on Earth
> possesses such features -- although establishing the first would not
> establish the second.
Good opening definition.
I guess you might say there is "strong" ID and "weak" ID. Strong says we
have postively detected the action of an intelligent agency; weak merely
disconfirms known naturalistic mechanisms and concludes a designer by
default. Your criticism applies to latter.
My response is this: I think you are technically correct. In principle it is
humanly impossible to say we have exhausted all naturalistic explanations
and therefore deduce a designer. In practice though we may well reach the
provisional conclusion based on the weight of evidence that this is the most
*scientifically* rational option. True, ID can tell us nothing about this
deduced designer. And I acknowledge all the usual caveats about the
god-of-the-gaps, but to say that we may never even provisionally conclude
supernatural intervention is not science, it is ontological naturalism.
> >
> > A new site to watch re these questions is http://www.iscid.org IMO a
source
> > of some well written and cogently argued essays, even if you don't agree
> > with the ID viewpoint (some from the usual suspects, of course).
> >
> > For a creationist like myself this area seems crucial. If study of GAs
> > reveal no in principle limits to CSI creation and increase in novelty,
> > complexity and functionality, then it would appear that related
arguments
> > against Darwinian evolution would be significantly weakened, and force a
> > retreat to secondary biological objections in practice, despite
theoretical
> > possibility.
> >
> > On the other hand, if evolutionary computing, GA problem-solving and
> > information studies etc do reveal such constraints, Darwinism is dead,
and
> > there does not appear to be convincing naturalistic alternative.
> >
> Again, it would have to be shown (as you implicitly note below) not
> merely that such constraints exist, but that they would have to be
> surpassed to produce some feature of life on Earth.
Agreed.
> >
> > What is really interesting is that this approach takes the argument out
of
> > the hands of biologists and into the hands of mathematicians and
computing
> > scientists (shock, horror...hallelujah?). Even aside from theoretical
> > analysis, I think at some stage we are going to hear the EC guys saying,
"We
> > have found that in practice you can use GAs to discover solutions to
> > problems of class A, but not class B." (Life as found on planet earth
being
> > decidedly class B.)
> >
> Biologists, though, have had some success, using bacteria in novel or
> hostile media, in producing some quite exotic systems through
> Darwinian mechanisms.
But have they? What de novo, novel, complex, functional (and documented ;-)
systems have appeared in these experiments?
> I think they might conclude, if the
> mathematicians announced that "our model says that your mechanism
> won't work," that the mathematical model rather the Darwinian
> mechanism is flawed. But that, as C.S. Lewis said in rather a
> different context, is crossing the bridge before you've ascertained
> that the river even exists.
The postulated Darwinian mechanism is very well-characterised. If what you
suggest were to happen, it would rather prove that we should look for
another mechanism.
rgds,
Mark
Mark & Roslyn Elkington
Wade Hines
Yes. Can you answer the question?
Rodjk
> Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> news:a0uo3m$mar$1...@geraldo.cc.utexas.edu...
> > On Tue, 01 Jan 2002 21:54:58 -0600, Mark & Roslyn Elkington wrote:
> >
> > > On the other hand, if evolutionary computing, GA problem-solving and
> > > information studies etc do reveal such constraints, Darwinism is dead,
> > > and there does not appear to be convincing naturalistic alternative.
> >
> > Interesting. Notice that "Darwinism" is merely a model for
> > understanding the cause of the *observed* changes to the earth's
> > biosystem over time.
>
> I'm using the term "Darwinism" as the neo-Darwinian component of the modern
> synthesis, i.e. mutation and natural selection. If you lose that mechanism,
> what are the serious naturalistic alternatives?
Evolution is the observation. Darwinism, or Neo-darwinism, is the
theory that explains it.
Tell me, why turn to math and GA's?
Can't your friends do biology?
>
> > Even if you somehow rule out "Darwinism" the
> > theory, you're still left with evolution, the fact of nature.
>
> Spoken like a true ontological naturalist.
Seems you still do not understand the difference between the fact and
the theory. Even if you were to show that darwinism is wrong, you have
not shown that evoltution is wrong.
Try doing, or at least learning, some biology.
Rodjk
>
> Mark
mel turner
[Mark & Roslyn Elkington] wrote...
>
>Is ID a serious challenge to Darwinian evolution? Specifically, will the
>study of genetic algorithms reveal theoretical and empirical limitations to
>the creative power of the trial and error mechanism?
Observations from real biology would seem to trump any conclusions
from theoretical models of that reality.
[snip]
>For a creationist like myself this area seems crucial. If study of GAs
>reveal no in principle limits to CSI creation and increase in novelty,
>complexity and functionality, then it would appear that related arguments
>against Darwinian evolution would be significantly weakened, and force a
>retreat to secondary biological objections in practice, despite theoretical
>possibility.
I've often commented that YECs needn't propose any limits _at all_
to Darwinian evolution, since according to them there wouldn't have
been enough time for much evolution to take place [yet illogically,
they very often argue for such limits].
For that matter, an OEC needn't propose any _naturalistic_ limits
to evolution [after all, God might step in and limit it]. Or, an OEC
might simply choose to believe in special creation without having to
argue that there are any valid reasons why natural evolution _couldn't_
have done it.
>On the other hand, if evolutionary computing, GA problem-solving and
>information studies etc do reveal such constraints, Darwinism is dead, and
>there does not appear to be convincing naturalistic alternative.
No, since they'd still have to reconcile those findings with the
many observations of "Darwinism" at work from the real biological
world. No doubt they'd eventually find some ways in which the
computer models don't fit the biological reality. [That is, if
your postulated constraints _are_ found]
>What is really interesting is that this approach takes the argument out of
>the hands of biologists and into the hands of mathematicians and computing
>scientists (shock, horror...hallelujah?). Even aside from theoretical
>analysis, I think at some stage we are going to hear the EC guys saying, "We
>have found that in practice you can use GAs to discover solutions to
>problems of class A, but not class B." (Life as found on planet earth being
>decidedly class B.)
Or maybe they'd have to fix their simulations to make them be
closer to natural population genetics cases.
>Either way, we seem to have a testable hypothesis, which is an essential
>ingredient of a happy new year.
I note that looking for theoretical limits to evolution still
isn't looking for any testable hypotheses about ID and its
processes and mechanisms. [Is there ever going to be a scientific
creationism that's about positive evidence for creation, not
about attacking evolution?]
regards
David Jensen
>
>Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
>news:a0uo3m$mar$1...@geraldo.cc.utexas.edu...
>> On Tue, 01 Jan 2002 21:54:58 -0600, Mark & Roslyn Elkington wrote:
>>
>> > On the other hand, if evolutionary computing, GA problem-solving and
>> > information studies etc do reveal such constraints, Darwinism is dead,
>> > and there does not appear to be convincing naturalistic alternative.
>>
>> Interesting. Notice that "Darwinism" is merely a model for
>> understanding the cause of the *observed* changes to the earth's
>> biosystem over time.
>
>I'm using the term "Darwinism" as the neo-Darwinian component of the modern
>synthesis, i.e. mutation and natural selection. If you lose that mechanism,
>what are the serious naturalistic alternatives?
How do you get rid of these mechanisms? They exist and are occurring.
David Jensen
"Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in
<43EY7.1538$ko4.1...@nasal.pacific.net.au>:
Make the demonstration and lets go from there.
Chris Merli
"Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in message
news:AvvY7.1509$ko4.1...@nasal.pacific.net.au...
> Is ID a serious challenge to Darwinian evolution? Specifically, will the
> study of genetic algorithms reveal theoretical and empirical limitations
to
> the creative power of the trial and error mechanism?
>
> A new site to watch re these questions is http://www.iscid.org IMO a
source
> of some well written and cogently argued essays, even if you don't agree
> with the ID viewpoint (some from the usual suspects, of course).
>
> For a creationist like myself this area seems crucial. If study of GAs
> reveal no in principle limits to CSI creation and increase in novelty,
> complexity and functionality, then it would appear that related arguments
> against Darwinian evolution would be significantly weakened, and force a
> retreat to secondary biological objections in practice, despite
theoretical
> possibility.
>
> On the other hand, if evolutionary computing, GA problem-solving and
> information studies etc do reveal such constraints, Darwinism is dead, and
> there does not appear to be convincing naturalistic alternative.
>
> What is really interesting is that this approach takes the argument out of
> the hands of biologists and into the hands of mathematicians and computing
> scientists (shock, horror...hallelujah?).
Once again a post where the amazing mathematicians will in fact be taking
over all fields of science. While it is still impossible to mathematically
model a single one of even the simplest cells, mathematicians are now ready
to describe trillions of organisms evolving over billions of years. I
realize that what they are in fact planning is to derive more general
conclusions from far more simple experiments, but looking at the bigger
picture brings up the very important point that these experiments will
require a huge number of assumptions to create workable experiments. If you
have already read the previous attempts by people to describe such complex
biological systems mathematically (i.e. proof that it is impossible to form
a single DNA coding sequence for a protein by random chance) you must
realize that the issue is not whether they can create a mathematical proof
given certain assumption but if they can get the assumptions right. For
example in the case I mentioned the problem was not the math but the
assumption that the process of creating a coding sequence happened entirely
by chance. (There were other false assumptions but even a single poor
assumption invalidates a "proof".) For the situation described here, the
big question would be if the constraints appear is it because of poor
assumptions. I don't think mathematicians can reasonably hope to answer
that question.
Even aside from theoretical
> analysis, I think at some stage we are going to hear the EC guys saying,
"We
> have found that in practice you can use GAs to discover solutions to
> problems of class A, but not class B." (Life as found on planet earth
being
> decidedly class B.)
>
> Either way, we seem to have a testable hypothesis, which is an essential
> ingredient of a happy new year.
Just for those of us who were asleep could you state the testable
hypothesis. The only one I could come up with here is: "if evolutionary
computing, GA problem-solving and information studies etc do reveal such
constraints, Darwinism is dead, and there does not appear to be convincing
naturalistic alternative." Which boils down to the false dichotomy that if
evolution is wrong then creation is right. This is not a testable theory in
favor of creation.
By comparison look at the predictions made by Darwins theory. The existance
of genes (a contemparary but not yet accepted idea). The ability of these
genes to be passed on to offspring. The existence of a material that could
carry genetic information to the offspring in a highly conserved fashion.
The mutability of the genes. The idea that on the genetic level organisms
which are amzingly diverse in appearance would be related.
>
> Regards,
> Mark Elkington
> Sydney
>
mel turner
>I'll put it more plainly: If the postulated Neo-Darwinian mechanism of
>mutation and natural selection is demonstrated to be insufficient to explain
>the diversity and complexity of life on earth,
That's a big "if", of course... In your hypothetical
demonstration, is the 'Neo-Darwinian mechanism' [hardly
postulated, it really happens] something that occurs but is
found to be somehow inadequate as a complete explanation for
life's diversity, or is it one that somehow completely fails
to work at all [in which case a lot of earlier observations
of both nature and experiments will need some 'splainin']?
>what other serious
>naturalistic mechanisms do you know of?
Besides mutation and selection and drift [oh, my], and
speciation and extinction, if they were somehow shown to be
an incomplete explanation for life's diversity?
We might as well start out again with a very scientific
"we don't know [yet]", and try to do science from there.
Presumably your "demonstration" would point us to the
particular aspects of evolution that especially needed
additional explanations.
cheers
Wesley R. Elsberry
Mark & Roslyn Elkington <mar...@zeta.org.au> wrote:
[...]
ME>But how this relates to the evolution of living things *is*
ME>the question at hand. No-one is saying GAs don't work, but
ME>some people are saying they can't work to produce the
ME>diversity and complexity of life as we know it, and are
ME>attempting to find mathematical proof of this.
[...]
I think that an important aspect of Dembski's critique of EC
has been overlooked. Dembski makes claims about EC itself,
where these claims can be tested. One such claim is that the
information of the solution state found via EC techniques was
actually "infused" by an intelligent agent. I challenged
Dembski on this point in 1997, and he had no adequate response
(IMO). If Dembski's claim of "infusion" of information were
true, it should be possible to elucidate the precise route and
mechanism by which this "infusion" occurs within, say, a GA
solving the Traveling Salesman Problem (TSP). That was the
example I used in 1997. If Dembski's claims concerning EC
can be shown to be false on such basic points, I think that
effectively diminishes the purported applicability of his
EC critiques to larger issues, like evolutionary biology.
Dembski's "Intelligent design as a theory of information"
essay presented in 1997 also had a section critiquing natural
selection directly. Bill Jefferys made some quite cogent
criticisms of Dembski's claims regarding NS, with the effect
of apparently causing Dembski to dump an extended workup from
his book, "The Design Inference", and also to delete that
section from various derived essays which appear on the WWW.
Dembski's track record on making criticisms which withstand
even modest scrutiny on the topics of EC and NS seems to me to
be quite poor. We'll have to see whether Dembski has the
wherewhithal to cash in the various "promissory notes" he has
been handing out recently in talks with titles like "Darwin's
Unpaid Debt". I've found it somewhat ironic that Dembski will
castigate some people for their "promissory naturalism" when
he has long practiced a form of "promissory designism"
himself.
ME>I'm holding my breath. As I've said, I think NFL or similar
ME>is potentially decisive, in a double-edged way.
I don't see much cause for concern on the part of evolutionary
biologists due to consideration of NFL theorems. However,
I've given Dembski fair warning before that his use of "No
Free Lunch" as his next book title has considerable
opportunity for embarrassment on his part, since it seems
apparent to me that he has consistently misconstrued NFL
findings in prior work.
ME>It also side-steps much of the rhetoric-as-science offered by
ME>the likes of Dawkins, Gould etc.
I think that we will soon see whether Dembski's take on NFL
delivers anything but rhetoric.
ME>Of course, an inconclusive result is still possible, but it
ME>seems to me that through empirical and/or theoretical study,
ME>there is room for optimism for a definitive answer as to
ME>whether or not the postulated Neo-Darwinian mechanism is
ME>sufficient.
Marcel-Paul Schutzenberger thought he had a disproof of
Darwinian mechanisms when he attempted a form of EC and
failed. That was in the mid-1960's. He was quite definite,
but he was also definitely wrong.
--
Wesley R. Elsberry, Student in Wildlife & Fisheries Sciences, Tx A&M U.
Visit the Online Zoologists page (http://www.onlinezoologists.com)
CNS BBS FTP Archive: ftp://centralneuralsystem.com/pub/CNS/bbs/
"the main question is whether the stuff is literature or not." - archy
Steven J.
> news:a0uo3m$mar$1...@geraldo.cc.utexas.edu...
> > On Tue, 01 Jan 2002 21:54:58 -0600, Mark & Roslyn Elkington wrote:
> >
> > > On the other hand, if evolutionary computing, GA problem-solving and
> > > information studies etc do reveal such constraints, Darwinism is dead,
> > > and there does not appear to be convincing naturalistic alternative.
> >
> > Interesting. Notice that "Darwinism" is merely a model for
> > understanding the cause of the *observed* changes to the earth's
> > biosystem over time.
>
> I'm using the term "Darwinism" as the neo-Darwinian component of the modern
> synthesis, i.e. mutation and natural selection. If you lose that mechanism,
> what are the serious naturalistic alternatives?
>
be some sort of self-organizing principle in living matter, or even a
revival of Lamarckian mechanisms. Both of these, to be sure, would
depend on forces that have either not be shown to exist, or to be able
to produce the sort of effects needed. On the other hand, science
presumably hasn't discovered everthing yet, and there may be answers
no one has even vaguely thought of.
Now, I have argued before that "naturalism" is simply the assumption
that the forces which act in the universe (including those that
produced and shape the evolution of life) act according to
discoverable rules. An actual theory of intelligent design would have
to postulate a Designer with some sort of discoverable and (at least
partly) understandable nature. This "we can say nothing definite
about the Designer or His motives" nonsense can't be part of any
"theory" worthy of the name. The actions of such a Designer would be,
by definition, naturalistic, since they proceed from His
understandable nature.
The alternative is that some forces acting in nature are
undiscoverable and beyond our understanding. This is actually quite
conceivable, and may even be true. But that is simply a limit to our
ability to know. It is no sort of argument for any sort of
"intelligent designer" that would actually interest the Discovery
Institute or William Dembski. Some force, or Force, acting in ways
utterly beyond our ability to understand or predict, could give us no
guidelines in how to conduct our lives, or our worship, or our
scientific theorizing.
>
> > Even if you somehow rule out "Darwinism" the
> > theory, you're still left with evolution, the fact of nature.
>
> Spoken like a true ontological naturalist.
>
I do not believe that "ontological naturalism" compels a belief in
common descent or descent with modification, unless the evidence
points that way. Bobby Bryant's point is that we still have the
evidence of shared ancestry of humans and other primates, of mammals
and other vertebrates, of vertebrates and other animals, of animals
and other eurkaryotes, etc, whether we have any explanation for the
phenomena or not.
Darwin deduced common descent well before he figured out a mechanism
that he thought could account for it. If his mechanism (and the
elaborations and corrections of it which his successors devised) is
inadequate, this does not seem to mean that common descent did not
happen. So, do you acknowledge that the argument is simply over what
produced common descent with modification, or do you argue over
whether even that happened?
>
> Mark
-- Steven J.
Bobby D. Bryant
> John Wilkins <john.w...@bigpond.com> wrote in message
> news:1f5e20o.19qaxcj1dhj04lN%john.w...@bigpond.com...
>> And yet, GAs work, and they work in ways we never expected them to.
>
> They do work. My favourite example is
> http://208.245.156.153/archive/output.cfm?ID=1455
>
> But how this relates to the evolution of living things *is* the
> question at hand. No-one is saying GAs don't work, but some people are
> saying they can't work to produce the diversity and complexity of life
> as we know it,
FYI, in most cases the people who use GAs aren't in the least bit
interested in producing "diversity and complexity" for their own sakes.
> and are attempting to find mathematical proof of this.
Do they have any actual motivation for this? Just picking an idea at
random and trying to prove it mathematically does not usually work out
well.
>> Anyway, the way the research is going, life on earth is decidedly
>> class A, and there's no reason to think otherwise, yet. If you or the
>> ID crowd can come up with some *solid* reasons why it must be in B,
>> then we might start to see some proper papers and research from them.
>> I for one will not hold my breath.
>
> I'm holding my breath. As I've said, I think NFL or similar is
> potentially decisive, in a double-edged way. It also side-steps much
> of the rhetoric-as-science offered by the likes of Dawkins, Gould etc.
What motivates you to say that a NFL theorem is potentially decisive?
For that matter, what does a NFL theorem have to say about biological
evolution? Do you even know what the theorem says, or are you just
hanging you hat on it in hopes that it will come to the rescue for your
beliefs?
I suspect that Dembski is pushing it because he knows the public at
large doesn't have any idea what it says, and can be fooled into
thinking that it's relevant.
Bobby Bryant
Austin, Texas
Mark & Roslyn Elkington
I can, but I chose not to since it appeared to be blatantly misconstruing my
claims about ID. Perhaps you could elaborate the question and convince
otherwise?
>
Robt Gotschall
mar...@zeta.org.au says...
>
> More precisely, Neo-Darwinism, meaning mutation plus natural selection. I
> think that would be dead, and there are no serious alternatives I know of
> (Kaufman's self-organisation, or teleology?).
Kauffman's self-organisation had to do with abiogenesis.
Teleology: 2. the study of the evidences of design or purpose in nature.
3. such design or purpose. 4. the belief that purpose and design are a
part of or are apparent in nature.
Copyright © 1966-1994 by Random House Inc., All Rights Reserved.
Are you certain that teleology is not one of your serious alternatives.
> > Even
> > worse, *physical* approximations to natural selection work, in, say,
> > robotics and electronics. What do we make of that? Why, we must assert,
> > as Dembski dies,
>
> Freudian typo :-)
I won't mention that one.
> Well, that's the whole the debate. Can NS in fact work to produce the set of
> complex cases of living organisms, within the phase space and time/trials
> available?
According to theory:
Abiogenesis would work to produce living organisms.
Random mutations would work to produce variety in living organisms.
NS would work to eliminate all but the set of now living organisms.
By Kauffman's reckoning, if it's alive it's already in the "useful"
complex range and differentiation among living organisms by complexity
may not be as meaningful in an evolutionary sense as you may think.
Hint: Things _can_ get too complex(see edge-of-chaos).
> Or can it only produce some mathematically bounded degree of
> variation, i.e. microevolutionary adaptations? (with usual disclaimers for
> the use of that problematic terminology).
>
> > > What is really interesting is that this approach takes the argument out
> > > of the hands of biologists and into the hands of mathematicians and
> > > computing scientists (shock, horror...hallelujah?).
If they succeed at that I suppose they can go on to prove that white is
black and wind up being killed at a zebra crossing. I never trusted
zebras BTW.
> I'm holding my breath. As I've said, I think NFL or similar is potentially
> decisive, in a double-edged way. It also side-steps much of the
> rhetoric-as-science offered by the likes of Dawkins, Gould etc.
>
> Of course, an inconclusive result is still possible, but it seems to me that
> through empirical and/or theoretical study, there is room for optimism for a
> definitive answer as to whether or not the postulated Neo-Darwinian
> mechanism is sufficient.
>
> rgds,
> Mark
If EWAKI is a genuine theory, it is by definition falsifiable.
I assume Darwinism will someday suffer the same fate as Newtonism, and
continue to be taught a century afterward.
--
rg
remove spam to mail
http://home.att.net/~hobgots/wsb/html/view.cgi-home.html-.html
Mark & Roslyn Elkington
> On Wed, 02 Jan 2002 07:34:59 -0600, Mark & Roslyn Elkington wrote:
>
>
> > John Wilkins <john.w...@bigpond.com> wrote in message
> > news:1f5e20o.19qaxcj1dhj04lN%john.w...@bigpond.com...
>
> >> And yet, GAs work, and they work in ways we never expected them to.
> >
> > They do work. My favourite example is
> > http://208.245.156.153/archive/output.cfm?ID=1455
> >
> > But how this relates to the evolution of living things *is* the
> > question at hand. No-one is saying GAs don't work, but some people are
> > saying they can't work to produce the diversity and complexity of life
> > as we know it,
>
> FYI, in most cases the people who use GAs aren't in the least bit
> interested in producing "diversity and complexity" for their own sakes.
Why did you feel the need to make that statement?
>
>
> > and are attempting to find mathematical proof of this.
>
> Do they have any actual motivation for this? Just picking an idea at
> random and trying to prove it mathematically does not usually work out
> well.
Apparently some vague interest in the obscure question of the origin of
life...
> >> Anyway, the way the research is going, life on earth is decidedly
> >> class A, and there's no reason to think otherwise, yet. If you or the
> >> ID crowd can come up with some *solid* reasons why it must be in B,
> >> then we might start to see some proper papers and research from them.
> >> I for one will not hold my breath.
> >
> > I'm holding my breath. As I've said, I think NFL or similar is
> > potentially decisive, in a double-edged way. It also side-steps much
> > of the rhetoric-as-science offered by the likes of Dawkins, Gould etc.
>
> What motivates you to say that a NFL theorem is potentially decisive?
Because it scrutinises the bare Neo-Darwinian mechanism in a potentially
rigorous theoretical, mathematical way, and is testable over time via the
application and experience of evolutionary computing techniques.
>
> For that matter, what does a NFL theorem have to say about biological
> evolution? Do you even know what the theorem says, or are you just
> hanging you hat on it in hopes that it will come to the rescue for your
> beliefs?
http://iscid.org/papers/Dembski_WhyNatural_112901.pdf
Which incidently I think is only introductory, and does not provide nor
claim a proof. Science takes time.
>
> I suspect that Dembski is pushing it because he knows the public at
> large doesn't have any idea what it says, and can be fooled into
> thinking that it's relevant.
I'd have similar suspicions in your position, but don't let that lead to
close-mindedness.
rgds,
Mark
Mark & Roslyn Elkington
David Jensen <da...@dajensen-family.com> wrote in message
news:4l963uk3qtl6g84ai...@4ax.com...
> >> >> Interesting. Notice that "Darwinism" is merely a model for
> >> >> understanding the cause of the *observed* changes to the earth's
> >> >> biosystem over time.
> >> >
> >> >I'm using the term "Darwinism" as the neo-Darwinian component of the
> >modern
> >> >synthesis, i.e. mutation and natural selection. If you lose that
> >mechanism,
> >> >what are the serious naturalistic alternatives?
> >>
> >> Just how is this going to get lost? How can you stop it from
> >> happening? Destroy all life on earth? Freeze time?
> >
> >I'll put it more plainly: If the postulated Neo-Darwinian mechanism of
> >mutation and natural selection is demonstrated to be insufficient to
explain
> >the diversity and complexity of life on earth, what other serious
> >naturalistic mechanisms do you know of?
>
> Make the demonstration and lets go from there.
The demonstration is the subject of this thread. This post was concerned
with a *definition* within that subject.
Mark & Roslyn Elkington
Rodjk <rjk...@yahoo.com> wrote in message
news:dbe402.020102...@posting.google.com...
> "Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in message
news:<qxCY7.1530$ko4.1...@nasal.pacific.net.au>...
> > Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> > news:a0uo3m$mar$1...@geraldo.cc.utexas.edu...
> > > On Tue, 01 Jan 2002 21:54:58 -0600, Mark & Roslyn Elkington wrote:
> > >
> > > > On the other hand, if evolutionary computing, GA problem-solving and
> > > > information studies etc do reveal such constraints, Darwinism is
dead,
> > > > and there does not appear to be convincing naturalistic alternative.
> > >
> > > Interesting. Notice that "Darwinism" is merely a model for
> > > understanding the cause of the *observed* changes to the earth's
> > > biosystem over time.
> >
> > I'm using the term "Darwinism" as the neo-Darwinian component of the
modern
> > synthesis, i.e. mutation and natural selection. If you lose that
mechanism,
> > what are the serious naturalistic alternatives?
>
> Evolution is the observation. Darwinism, or Neo-darwinism, is the
> theory that explains it.
Wrong.
Evolution is one theory to explain the observation. Darwinism, or
Neo-darwinism are some proposed mechanisms within this theory, among others.
Your failure to correctly state these basic definitions is telling.
> Tell me, why turn to math and GA's?
> Can't your friends do biology?
>
> >
> > > Even if you somehow rule out "Darwinism" the
> > > theory, you're still left with evolution, the fact of nature.
> >
> > Spoken like a true ontological naturalist.
>
> Seems you still do not understand the difference between the fact and
> the theory. Even if you were to show that darwinism is wrong, you have
> not shown that evoltution is wrong.
>
> Try doing, or at least learning, some biology.
> Rodjk
See above.
chmc
"Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in message
news:43EY7.1538$ko4.1...@nasal.pacific.net.au...
Generally theories are created after seeing the evidence. It's kind of hard
to say what naturalistic mechanisms might be proposed instead, without
seeing how exactly the "postulated" mechanisms we now see as working are
insufficient.
Wade Hines
Mark & Roslyn Elkington wrote:
> Wade Hines <wade....@rcn.com> wrote in message
> > Mark & Roslyn Elkington wrote:
> > > Wade Hines <wade....@rcn.com> wrote in message
> > > > Mark & Roslyn Elkington wrote:
> > > > > John Wilkins <john.w...@bigpond.com> wrote in message
> > > > ...
>>>>>> And yet, GAs work, and they work in ways we never expected them to.
>>>>> They do work. My favourite example is
>>>>> http://208.245.156.153/archive/output.cfm?ID=1455
>>>>> But how this relates to the evolution of living things *is* the question at
>>>>> hand. No-one is saying GAs don't work, but some people are saying they can't
>>>>> work to produce the diversity and complexity of life as we know it, and are
>>>>> attempting to find mathematical proof of this.
>>>> Can you cite precedence for a "mathematical proof" by failed simulation?
> > > You mean "precedents"?
> > Yes. Can you answer the question?
> I can, but I chose not to since it appeared to be blatantly misconstruing my
> claims about ID. Perhaps you could elaborate the question and convince
> otherwise?
I don't see anything about ID, or evolution or creation in my question.
The question relates to using the failure of a simulation as a "mathematical
proof" of anything.
I think there is a basic flaw of logic in trying to show that something
is impossible because an attempt to simulate it fails. Thus I ask you
for precedent.
Maybe you can explain how I have blatantly misconstrued anything you
wrote about ID as I didn't reference ID at all.
Bobby D. Bryant
> Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> news:a0veqf$qkd$1...@geraldo.cc.utexas.edu...
>> On Wed, 02 Jan 2002 07:34:59 -0600, Mark & Roslyn Elkington wrote:
>> > I'm holding my breath. As I've said, I think NFL or similar is
>> > potentially decisive, in a double-edged way. It also side-steps
>> > much of the rhetoric-as-science offered by the likes of Dawkins,
>> > Gould etc.
>>
>> What motivates you to say that a NFL theorem is potentially decisive?
>
> Because it scrutinises the bare Neo-Darwinian mechanism in a
> potentially rigorous theoretical, mathematical way, and is testable
> over time via the application and experience of evolutionary computing
> techniques.
Per my post elsewhere in this thread, once you understand what NFLs
actually say and what the Theory of Evolution says, it is immediately
obvious that NFLs have nothing to say about ToE -- let alone do they
serve as a disproof.
>> For that matter, what does a NFL theorem have to say about biological
>> evolution? Do you even know what the theorem says, or are you just
>> hanging you hat on it in hopes that it will come to the rescue for
>> your beliefs?
>
> http://iscid.org/papers/Dembski_WhyNatural_112901.pdf
That "paper" is so bad that my gag reflex is making it difficult to
read. However, here are a couple of quick notes:
1) On page 1 he claims that "Life is both complex and specified". His
quick explanation of "specified" is "conforms to an independently given
pattern". In what sense does life "conform to an independently given
pattern"? It sounds to me like he's incorporating his conclusion into
his assumptions by claiming at the outset that there is "an
independently given pattern" for life. Who "gives" this pattern? (This
is the same trick the ID crowd want to sneak past their audience by
talking about "design" and pretending they're not making any claims
about a "designer".)
2) On page 11 he shows his ignorance of the basic concept of what NFLs
are all about (or else he twists that concept in a deceitful fashion).
He says "this means that the average performance of any evolutionary
algorithm is no better than blind search". That's actually true, but
what he doesn't tell the reader is that that _average_ is when averaging
across _all_possible_optimization_problems_. All that tells us is that
lots of problems have fitness functions with a "nice" geography (from
the POV of an EA), but lots of them have FFs with "not nice"
geographies. It is perfectly obvious that an EA does better than BS
[sic!] on *some* problems. It is also perfectly obvious that NFL
theorems don't claim otherwise. Finally, it is a question that is
amenable to experiment: for example, try optimizing the path for a
Travelling Salesman using a GA and a blind search, and see which gives
you the best result after running for a couple of hours on a 1000-city
problem.
I cannot emphasize this enough: NFLs are claims about algorithms when
applied to *all*possible* problems. They say *nothing* about whether a
specific algorithm works well on a specific problem. They also say
*nothing* about whether all those *all*possible* problems are actually
interesting as real-world problems. Dembski's claim about "the average
performance" is true _if_interpreted_correctly_, but it is *not* true if
you twist "average performance" to mean a blind search is just as good
as an evolutionary algorithm on any particular problem. (Lurkers who
doubt this should do the TSP experiment that I described above.)
> Which incidently I think is only introductory, and does not provide
> nor claim a proof. Science takes time.
Yes, especially when the "scientist" works from the conclusion back
toward the evidence, and finds a mismatch.
>> I suspect that Dembski is pushing it because he knows the public at
>> large doesn't have any idea what it says, and can be fooled into
>> thinking that it's relevant.
>
> I'd have similar suspicions in your position, but don't let that lead
> to close-mindedness.
You could always read Wolpert's paper and find out for yourself what it
says.
If you do decide to read it, make sure you don't miss the BOLD TEXT on
page 3, where the authors say "Finally, we cannot emphasize enough that
NO CLAIMS WHATSOEVER are being made in this paper concerning how well
various search algorithms work in practice." (I guess Dembski missed
that part. Maybe the bold print threw him off?)
I'll try again to read Dembski's "paper" later, though that sort of
a-prioristic logic twisting really annoys me. (FWIW, it's exactly what
made me get out of linguistics.)
Bobby Bryant
Austin, Texas
David Jensen
>
>Rodjk <rjk...@yahoo.com> wrote in message
>news:dbe402.020102...@posting.google.com...
>> "Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in message
>news:<qxCY7.1530$ko4.1...@nasal.pacific.net.au>...
>> > Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
>> > news:a0uo3m$mar$1...@geraldo.cc.utexas.edu...
>> > > On Tue, 01 Jan 2002 21:54:58 -0600, Mark & Roslyn Elkington wrote:
>> > >
>> > > > On the other hand, if evolutionary computing, GA problem-solving and
>> > > > information studies etc do reveal such constraints, Darwinism is
>dead,
>> > > > and there does not appear to be convincing naturalistic alternative.
>> > >
>> > > Interesting. Notice that "Darwinism" is merely a model for
>> > > understanding the cause of the *observed* changes to the earth's
>> > > biosystem over time.
>> >
>> > I'm using the term "Darwinism" as the neo-Darwinian component of the
>modern
>> > synthesis, i.e. mutation and natural selection. If you lose that
>mechanism,
>> > what are the serious naturalistic alternatives?
>>
>> Evolution is the observation. Darwinism, or Neo-darwinism, is the
>> theory that explains it.
>
>Wrong.
>
>Evolution is one theory to explain the observation. Darwinism, or
>Neo-darwinism are some proposed mechanisms within this theory, among others.
What are you talking about? Evolution is change in allele frequency in
populations over time. That is an observation, not a theory.
John Wilkins
> John Wilkins <john.w...@bigpond.com> wrote in message
> news:1f5e20o.19qaxcj1dhj04lN%john.w...@bigpond.com...
> > Mark & Roslyn Elkington <mar...@zeta.org.au> wrote:
> >
....
> > > On the other hand, if evolutionary computing, GA problem-solving and
> > > information studies etc do reveal such constraints, Darwinism is dead,
> and
> > > there does not appear to be convincing naturalistic alternative.
> >
> > Not exactly, although it would weaken some of the arguments used in
> > favour of evolution as we know it (EAWKI, to avoid the loaded and
> > misleading term "Darwinism").
>
> More precisely, Neo-Darwinism, meaning mutation plus natural selection. I
> think that would be dead, and there are no serious alternatives I know of
> (Kaufman's self-organisation, or teleology?).
I notice that the thread has a slew of responses. I shall confine myself
to a few comments.
Darwinism is way more than mutation + selection. It includes
recombination, drift, punctuation, sexual selection, and a host of other
things. "Neo-Darwinism" is an anachronistic term used by some as a badge
of pride and others as a slur. In neither case does it denote anything
much.
Kaufmann's ideas are not an *alternative* to NS, any more than the basic
laws of chemistry or physics are. Any self-organised system still has to
undergo the rigours of selection and drift. Merely being self-organised
is no guarantee that the system is viable.
>
> > This is the main reason why Dembksi is now running his attack on GAs
> > with the No Free Lunch Theorem. As I understand that (this not being a
> > speciality of mine), that theorem only applies in restricted cases. If
> > you want to find algorithmic limitations, you can - Chaitin does not
> > think he can accommodate EAWKI in his AIT notion of complexity and
> > information. But Chaitin's response is not to say that EAWKI is false,
> > but merely that he hasn't yet managed to come up with an entirely
> > complete characterisation of the world. Dembski is, so far as I can
> > tell, grasping at straws.
> >
> > And yet, GAs work, and they work in ways we never expected them to.
>
> They do work. My favourite example is
> http://208.245.156.153/archive/output.cfm?ID=1455
>
> But how this relates to the evolution of living things *is* the question at
> hand. No-one is saying GAs don't work, but some people are saying they can't
> work to produce the diversity and complexity of life as we know it, and are
> attempting to find mathematical proof of this.
Others have commented. In my view, GAs are a nice illustration of some
aspects of EAWKI. Certainly not all (for a start, they only apply to
toyworlds).
>
> > Even
> > worse, *physical* approximations to natural selection work, in, say,
> > robotics and electronics. What do we make of that? Why, we must assert,
> > as Dembski dies,
>
> Freudian typo :-)
Why are people so unkind? :-) I and O are right next to each other on
the keyboard.
>
> > that this happens because we actually do know what we
> > were doing even if we could never have actually predicted the outcome.
> > Get real...
> >
> > So what about if we find that the NFL theorem works out that in a class
> > of cases natural selection is ineffectual? What then? Is it ineffectual
> > in all cases? That is a logical error, and one Dembski had better not
> > make in print, or I'll have another paper on my hands :-) We already
> > know cases where NS is ineffectual - they are called constrained
> > situations and genetic drift. But selection still works to produce
> > complex cases.
>
> Well, that's the whole the debate. Can NS in fact work to produce the set of
> complex cases of living organisms, within the phase space and time/trials
> available? Or can it only produce some mathematically bounded degree of
> variation, i.e. microevolutionary adaptations? (with usual disclaimers for
> the use of that problematic terminology).
OK, let's carefully consider two of the concepts here:
Phase space - there is no set phase space of any dimensionality other
than the conditions required by the laws of physics. There are no
optimal solution coordinates in any space that organisms *have* to find.
A phase space is how *we* - primates in lab coats - choose to model the
system. It exists as a semantic construct only in our heads and our
tools (computers, journals, internet, etc).
Time/trials: although there is unfortunately a measure (invented by
Haldane, I think) called a "darwin" (relative change by a factor of _e_)
per million years, the *proper* measure of evolutionary change is
relative to the number of generations of cells in a lineage. Wemight
think that little can be accomplished in a million years, and it may be
this is true if we are considering large animals, for example (although
it is actually not true). But the number of biochemical novelties that
can be achieved depends very much on the recombinatorial complexities
possible in the number of mitotic divisions and recombinations. I think
that we tend to underestimate the amount of possible change in the time
available. It has been said that what requires explanation is that there
is so little change, not so much.
>
> > > What is really interesting is that this approach takes the argument out
> of
> > > the hands of biologists and into the hands of mathematicians and
> computing
> > > scientists (shock, horror...hallelujah?). Even aside from theoretical
> > > analysis, I think at some stage we are going to hear the EC guys saying,
> "We
> > > have found that in practice you can use GAs to discover solutions to
> > > problems of class A, but not class B." (Life as found on planet earth
> being
> > > decidedly class B.)
> >
> > EC?
>
> Evolutionary computing.
Of course. I got it when I read Wesley's response.
>
> > Anyway, the way the research is going, life on earth is decidedly
> > class A, and there's no reason to think otherwise, yet. If you or the ID
> > crowd can come up with some *solid* reasons why it must be in B, then we
> > might start to see some proper papers and research from them. I for one
> > will not hold my breath.
>
> I'm holding my breath. As I've said, I think NFL or similar is potentially
> decisive, in a double-edged way. It also side-steps much of the
> rhetoric-as-science offered by the likes of Dawkins, Gould etc.
You think so? See my and others' responses in this thread. Personally I
think that Dembski's use of NFL is all rhetorical flourish.
>
> Of course, an inconclusive result is still possible, but it seems to me that
> through empirical and/or theoretical study, there is room for optimism for a
> definitive answer as to whether or not the postulated Neo-Darwinian
> mechanism is sufficient.
>
> rgds,
> Mark
Just be careful to make sure that the mechanisms you test are not
strawmen.
--
John Wilkins
Occasionally making sense for over 46 years
John Wilkins
> On Tue, 01 Jan 2002 22:18:37 -0600, John Wilkins wrote:
>
> > Mark & Roslyn Elkington <mar...@zeta.org.au> wrote:
> >
> >> On the other hand, if evolutionary computing, GA problem-solving and
> >> information studies etc do reveal such constraints, Darwinism is
> >> dead, and there does not appear to be convincing naturalistic
> >> alternative.
>
> > Not exactly, although it would weaken some of the arguments used in
> > favour of evolution as we know it (EAWKI, to avoid the loaded and
> > misleading term "Darwinism").
>
> Let me put in a general word of warning re comparing GAs to biological
> evolution. The comparison is useful up to a point, but there's a danger
> of pushing the analogy too closely.
>
> GAs were "inspired by" biological evolution (or perhaps merely dog
> breeding), but they are not generally used to model biological
> evolution. Most often they are used to find a "decent" solution for
> some specific problem that would require too much time or space to solve
> exactly. They do not usually follow the mechanisms of biological
> evolution in anything but the most superficial form, and of course they
> only use a trivial fraction of the resources that biological evolution
> has used over the ages.
>
> Biological evolution, IMO, is not trying to solve any problem at all,
> any more than gravity or chemistry is. (I suppose you could say that
> gravity and chemistry are "trying" to find the lowest energy state of
> certain kinds of systems, but I would be very surprised to discover that
> any scientist actually does think of them that way, at least as the
> top-level understanding of the system under study.)
>
> GAs can still be useful for demonstrating certain points, e.g. that
> breeding, mutation, and weeding-out-by-fitness can indeed generate "new
> stuff" that was not in the original population, and that random
> mutations can sometimes be beneficial (as many ill-informed creationists
> like to deny). However, at each and every invocation of GAs to
> demonstrate a point regarding biological evolution, the reflexive
> response should be to immediately ask yourself whether the analogy is
> being pushed too far. This should save much grief resulting from
> ill-gotten conclusions.
All good points. EAWKI is much more than the dynamics captured by GAs.
Michod's book on Darwinian Dynamics covers some of the rest, I believe.
>
>
> > This is the main reason why Dembksi is now running his attack on GAs
> > with the No Free Lunch Theorem. As I understand that (this not being a
> > speciality of mine), that theorem only applies in restricted cases. If
> > you want to find algorithmic limitations, you can - Chaitin does not
> > think he can accommodate EAWKI in his AIT notion of complexity and
> > information. But Chaitin's response is not to say that EAWKI is false,
> > but merely that he hasn't yet managed to come up with an entirely
> > complete characterisation of the world. Dembski is, so far as I can
> > tell, grasping at straws.
>
> Dembski is going to get creamed if he stakes his credibility on using
> the No Free Lunch theorem to refute the reality of biological evolution.
>
> The essense of the NFL theorem is something like this -
>
> "For two search algorithms A and B, if A outperforms B on n search
> problems, then B likewise outperforms A on approximately n other search
> problems."
>
> An immediate corollary is that there is no universal "best" search
> algorithm _without_reference_to_the_application_.
>
> See Wolpert & Macready, "No Free Lunch Theorems for Search",
> http://citeseer.nj.nec.com/wolpert95no.html. (If it turns out that I
> only have access to that site by academic arrangement, let me know and
> I'll see if I can find a direct link to his paper on line.)
>
> There is a companion paper w.r.t. optimization (rather than search), but
> I have not had time to read it.
>
>
> Attempting to apply the NFL to biological evolution immediately
> encounters one of the following problems:
>
> 1) Is biological evolution even a "search problem", in any meaningful
> sense? If not, the NFL simply doesn't apply.
>
> 2) Even if you do think biological evolution is a search problem, notice
> that no one is claiming that evolution is the "best" solution to *all*
> search problems. Without such a claim, the NFL simply doesn't apply.
I agree also with this - see my comment elsewhere in this thread on
satisficing.
>
> I personally don't care to think of evolution as a search problem
> because of the teleological nature of that claim. (Granted, natural
> selection does "improve the fit" over time, but I'm still a bit
> reluctant to view it as a search or optimization problem; I prefer to
> view the ToE as descriptive rather than prescriptive.)
At best, a search problem is a *model* of some aspect of evolution.
Strictly speaking evolution has no problems or solutions, just organisms
that do, or do not, reproduce using the economic resources around them.
>
>
> For that matter, no one seems to be making even the very restricted
> claim that biological evolution is the best way of solving the single
> "search problem" of finding an organism that is best suited to the
> environment. Clearly, evolution is the "algorithm" that nature actually
> uses, but that is not the same as a claim that it is the best possible
> algorithm. For that matter, evolution does not seem to actually be
> finding the best fits between organisms and the environment, as
> witnessed by bad backs, infections of the appendix, nerves looping down
> a giraffe's neck and then back up, etc.
>
> And there is *nothing* in the NFL that claims that a non-best search
> algorithm can never find a decent solution to a problem.
Yup.
>
> I suspect Dembski is going to appeal to the fallacious notion that
> humans are the "correct" answer to some search problem, add the
> fallacious claim that the "correct" answer could not have been found by
> anything but the "best" search algorithm, and then correctly invoke the
> NFL to prove that no such "best" algorithm exists. Brings to mind a
> parable about building towers on shifting sands...
>
>
>
> > Even worse, *physical* approximations to natural selection work, in,
> > say, robotics and electronics. What do we make of that? Why, we must
> > assert, as Dembski dies
> ^ Typo, or Freudian slip?
I'd love to say FS, but I sincerely do not want anyone to face their
dies irae...
>
>
> Bobby Bryant
> Austin, Texas
John Wilkins
> On Wed, 02 Jan 2002 14:27:50 -0600, Mark & Roslyn Elkington wrote:
>
>
> > Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> > news:a0veqf$qkd$1...@geraldo.cc.utexas.edu...
....
Ian and I were discussing this over a latte last night. It also seems to
me that the NFL is a theorem about finding an *optimum* solution in the
solution space. But selection does not "seek" optimal solutions. It
seeks "satisficing" solution, to use Herbert Simon's lovely word. All
that NS has to generate is a solution that is adequate to the task at
hand and is slightly better than the other alternatives already found.
In that respect, and also since the solution phase space is not
specified in advance, the NFL is inappropriate to apply to NS. Dembski
has overdrawn his case (again).
Incidentally, I must quote Simon's comments in introducing this term on
page 35 of his book _The Sciences of the Artificial_. The context is
economics, which in a deeper sense I think is entirely appropriate to
selection:
"Of course, the decision that is optimal in the simplified model [of
economics] will seldom be optimal in the real world. The decision maker
has a choice between optimal decisions for an imaginary simplified world
or decisions that are "good enough", that satisfice, for a world
approximating the complex real one more closely. The technique of
heuristic search makes much weaker demands on the problem structure than
do linear programming or linear decision rules, but it can generally
only find satisfactory solutions, not the optimal one."
This was written in 1969...
>
> > Which incidently I think is only introductory, and does not provide
> > nor claim a proof. Science takes time.
>
> Yes, especially when the "scientist" works from the conclusion back
> toward the evidence, and finds a mismatch.
>
>
> >> I suspect that Dembski is pushing it because he knows the public at
> >> large doesn't have any idea what it says, and can be fooled into
> >> thinking that it's relevant.
> >
> > I'd have similar suspicions in your position, but don't let that lead
> > to close-mindedness.
>
> You could always read Wolpert's paper and find out for yourself what it
> says.
>
> If you do decide to read it, make sure you don't miss the BOLD TEXT on
> page 3, where the authors say "Finally, we cannot emphasize enough that
> NO CLAIMS WHATSOEVER are being made in this paper concerning how well
> various search algorithms work in practice." (I guess Dembski missed
> that part. Maybe the bold print threw him off?)
Do a google for No Free Lunch and this paper in PDF will be abut the
third item..
>
> I'll try again to read Dembski's "paper" later, though that sort of
> a-prioristic logic twisting really annoys me. (FWIW, it's exactly what
> made me get out of linguistics.)
And me out of purely abstract philosophy.
Steven J.
> Steven J. <stev...@altavista.com> wrote in message
> news:127ccf2e.02010...@posting.google.com...
> > "Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in message
> news:<AvvY7.1509$ko4.1...@nasal.pacific.net.au>...
> > > Is ID a serious challenge to Darwinian evolution? Specifically, will
> > > the study of genetic algorithms reveal theoretical and empirical
> > > limitations to the creative power of the trial and error mechanism?
>
hypotheses about the proposed intelligent agent. To positively detect
the action of (one particular out of many imaginable) intelligent
agents, we would need such a hypothesis, to see if the evidence was
actually explained by the particular proposed agent. Thus, there is
to date no actual "strong ID" for any criticism to apply to. My
complaint, here, is that what are actually "weak ID" arguments are
presented (to the public and creationists) as though they were "strong
ID" arguments.
>
> My response is this: I think you are technically correct. In principle it is
> humanly impossible to say we have exhausted all naturalistic explanations
> and therefore deduce a designer. In practice though we may well reach the
> provisional conclusion based on the weight of evidence that this is the most
> *scientifically* rational option. True, ID can tell us nothing about this
> deduced designer. And I acknowledge all the usual caveats about the
> god-of-the-gaps, but to say that we may never even provisionally conclude
> supernatural intervention is not science, it is ontological naturalism.
>
naturalistic mechanisms, that doesn't justify positing a *designer*,
until you can tell us something about the *design*. Not just that
it's "irreducibly complex" or that its complexity is "specified," but
that it's designed *for* something, and serves some goal. The goal
will be a clue to the nature and design philosophy of the designer.
Above, I left in my point about the nested hierarchy as the result of
limitations of the ability of Darwinian mechanisms to optimize design.
I think this is important. Human designs -- known examples of
intelligent design -- freely borrow elements from one design to use in
unrelated designs, and in various other ways produce artifacts that
don't fit into consistent, unambiguous nested hierarchies. It seems
to me that a mechanism which can produce nested hierarchies either
must possess limitations that an intelligent designer would not
suffer, or else must want to LOOK as if it possessed such limitations.
>
-- [snip]
>
> > >
> > > What is really interesting is that this approach takes the argument out
> > > of the hands of biologists and into the hands of mathematicians and
> > > computing scientists (shock, horror...hallelujah?). Even aside from
> > > theoretical analysis, I think at some stage we are going to hear the EC
> > > guys saying, "We have found that in practice you can use GAs to discover > > > solutions to problems of class A, but not class B." (Life as found on > > > planet earth being decidedly class B.)
> > >
> > Biologists, though, have had some success, using bacteria in novel or
> > hostile media, in producing some quite exotic systems through
> > Darwinian mechanisms.
>
> But have they? What de novo, novel, complex, functional (and documented ;-)
> systems have appeared in these experiments?
>
The standard example in this newsgroup (this ability has been
generated more than once, and using more than one mechanism) is the
ability of bacteria to metabolize nylon. That's functional (aren't
_de novo_ and novel synonymous?); whether it's complex is a matter of
taste. It involved gene duplication followed by a mutation to the
duplicate gene.
>
> > I think they might conclude, if the
> > mathematicians announced that "our model says that your mechanism
> > won't work," that the mathematical model rather the Darwinian
> > mechanism is flawed. But that, as C.S. Lewis said in rather a
> > different context, is crossing the bridge before you've ascertained
> > that the river even exists.
>
> The postulated Darwinian mechanism is very well-characterised. If what you
> suggest were to happen, it would rather prove that we should look for
> another mechanism.
>
I think I meant to say that, genetic algorithms run on computers might
be inadequate models of the modern synthetic theory of evolution.
>
> rgds,
> Mark
-- Steven J.
Bobby D. Bryant
> Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote:
>
>> I cannot emphasize this enough: NFLs are claims about algorithms when
>> applied to *all*possible* problems. They say *nothing* about whether
>> a specific algorithm works well on a specific problem. They also say
>> *nothing* about whether all those *all*possible* problems are
>> actually interesting as real-world problems. Dembski's claim about
>> "the average performance" is true _if_interpreted_correctly_, but it
>> is *not* true if you twist "average performance" to mean a blind
>> search is just as good as an evolutionary algorithm on any particular
>> problem. (Lurkers who doubt this should do the TSP experiment that I
>> described above.)
>
> Ian and I were discussing this over a latte last night. It also seems
> to me that the NFL is a theorem about finding an *optimum* solution in
> the solution space. But selection does not "seek" optimal solutions.
> It seeks "satisficing" solution, to use Herbert Simon's lovely word.
> All that NS has to generate is a solution that is adequate to the task
> at hand and is slightly better than the other alternatives already
> found.
And of course, when you get down to cases the ToE doesn't actually say
that NS "has to" do even that much -- there's nothing in the ToE that
says a species can't go extinct due to failure to find an adequate
solution to the task at hand.
> In that respect, and also since the solution phase space is not
> specified in advance, the NFL is inappropriate to apply to NS. Dembski
> has overdrawn his case (again).
As I said earlier, it appears that Dembsi is following the IDer
tradition of going through the forms of "being scientific", but relying
on the ignorance of the public at large (and also of legislators and
justices) to keep their sleight-of-hand from being exposed.
> Incidentally, I must quote Simon's comments in introducing this term
> on page 35 of his book _The Sciences of the Artificial_. The context
> is economics, which in a deeper sense I think is entirely appropriate
> to selection:
>
> "Of course, the decision that is optimal in the simplified model [of
> economics] will seldom be optimal in the real world. The decision
> maker has a choice between optimal decisions for an imaginary
> simplified world or decisions that are "good enough", that satisfice,
> for a world approximating the complex real one more closely. The
> technique of heuristic search makes much weaker demands on the problem
> structure than do linear programming or linear decision rules, but it
> can generally only find satisfactory solutions, not the optimal one."
>
> This was written in 1969...
Yes, the whole raison d'être for AI is that some problems can, from the
mathematical perspective, be solved exactly -- but would in reality
require turning the universe to ash in order to actually do the
calculations. For example, an instance of the TSP problem with 1000
cities admits of 10^2565 unique solutions. (If that number's not big
enough for you, try the Galactic Surveyer Problem, where you have to
optimize a path for visiting each of the stars in a galaxy.)
So we have the field of AI, which for the most part works on methods of
finding a "good" solution to a hard problem in short order, rather than
a "perfect" solution when the universe grows cold.
Bobby Bryant
Austin, Texas
Bobby D. Bryant
> Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> news:a0veqf$qkd$1...@geraldo.cc.utexas.edu...
>> On Wed, 02 Jan 2002 07:34:59 -0600, Mark & Roslyn Elkington wrote:
>>
>> > John Wilkins <john.w...@bigpond.com> wrote in message
>> > news:1f5e20o.19qaxcj1dhj04lN%john.w...@bigpond.com...
>>
>> >> And yet, GAs work, and they work in ways we never expected them
>> >> to.
>> >
>> > They do work. My favourite example is
>> > http://208.245.156.153/archive/output.cfm?ID=1455
>> >
>> > But how this relates to the evolution of living things *is* the
>> > question at hand. No-one is saying GAs don't work, but some people
>> > are saying they can't work to produce the diversity and complexity
>> > of life as we know it,
>>
>> FYI, in most cases the people who use GAs aren't in the least bit
>> interested in producing "diversity and complexity" for their own
>> sakes.
>
> Why did you feel the need to make that statement?
To forstall anyone from concluding that "because GAs don't generally
produce a certain kind of result means that they *can't* produce a
that kind of result".
For example, a guy in my research group just produced an unusual GA that
tags solutions according to the mutational path of their ancestors, and
introduces a barrier to breeding between solutions with different tags.
The immediate result is that he gets speciation in his breeding
population; a nice side effect is that you could use the tags to draw a
"tree of life" at the end of the run, if you wanted to.[1]
This still falls far short of being a simulation of biological
evolution, but it shows how a minor tweak to a traditional GA can
produce results that are interesting in a direction that evolution
deniers like to claim as impossible (i.e., "diversity of life").
See
http://www.cs.utexas.edu/users/nn/pages/publications/abstracts.html#stanley.utcstr01.ps.gz
(Despite the poorly chosen name for the tag, the link will get you to an
HTML abstract with additional links for downloading his paper in .ps or
.pdf format.)
Bobby Bryant
Austin, Texas
[1] I suspect that in biological evolution the barriers to breeding
arise from "external" factors such as geographic isolation, rather than
from a single mutation. This could be addressed in a GA too, if anyone
felt a motivation to do so. (In fact there are things such as the
"landscape model" and the "island model" that already do something very
similar to that.)
Mark & Roslyn Elkington
Okay, I'll answer your original question.
I wasn't suggesting that a single failed simulation proves anything as such.
A formal proof will need to be based on rigorous mathematics, correct
assumptions etc, which I think is where Dembski is trying to head. From
other things he's published he seems to me to be reasonably equipped in this
area.
However, I was suggesting that as evolutionary computing matures a
discipline, an empirical understanding of the problem solving capabilities
and limits of evo GAs will emerge. My contention is that this result will be
strongly confirming of the theoretical result, either way.
rgds,
Mark
>
Mark & Roslyn Elkington
Robt Gotschall <hobgo...@worldnet.att.net> wrote in message
news:MPG.169d0ab7e...@netnews.worldnet.att.net...
> In article <V_DY7.1537$ko4.1...@nasal.pacific.net.au>,
> mar...@zeta.org.au says...
>
>
> >
> > More precisely, Neo-Darwinism, meaning mutation plus natural selection.
I
> > think that would be dead, and there are no serious alternatives I know
of
> > (Kaufman's self-organisation, or teleology?).
>
> Kauffman's self-organisation had to do with abiogenesis.
I thought though some of this had been extended to evolution as well.
>
> Teleology: 2. the study of the evidences of design or purpose in nature.
> 3. such design or purpose. 4. the belief that purpose and design are a
> part of or are apparent in nature.
>
> Copyright © 1966-1994 by Random House Inc., All Rights Reserved.
>
> Are you certain that teleology is not one of your serious alternatives.
Rather "teleology" in the Paul Davies sense, some unknown organising force
in nature.
>
>
> > > Even
> > > worse, *physical* approximations to natural selection work, in, say,
> > > robotics and electronics. What do we make of that? Why, we must
assert,
> > > as Dembski dies,
> >
> > Freudian typo :-)
>
> I won't mention that one.
>
>
> > Well, that's the whole the debate. Can NS in fact work to produce the
set of
> > complex cases of living organisms, within the phase space and
time/trials
> > available?
>
> According to theory:
>
> Abiogenesis would work to produce living organisms.
>
> Random mutations would work to produce variety in living organisms.
>
> NS would work to eliminate all but the set of now living organisms.
>
> By Kauffman's reckoning, if it's alive it's already in the "useful"
> complex range and differentiation among living organisms by complexity
> may not be as meaningful in an evolutionary sense as you may think.
>
> Hint: Things _can_ get too complex(see edge-of-chaos).
I would have thought Homo sapiens, for example, would have exceeded any and
all edge-of-chaos boundaries.
>
> > Or can it only produce some mathematically bounded degree of
> > variation, i.e. microevolutionary adaptations? (with usual disclaimers
for
> > the use of that problematic terminology).
> >
> > > > What is really interesting is that this approach takes the argument
out
> > > > of the hands of biologists and into the hands of mathematicians and
> > > > computing scientists (shock, horror...hallelujah?).
>
> If they succeed at that I suppose they can go on to prove that white is
> black and wind up being killed at a zebra crossing. I never trusted
> zebras BTW.
If they get their assumptions and model right, then they can potentially go
on prove the Darwinian mechanism either way.
> > I'm holding my breath. As I've said, I think NFL or similar is
potentially
> > decisive, in a double-edged way. It also side-steps much of the
> > rhetoric-as-science offered by the likes of Dawkins, Gould etc.
> >
> > Of course, an inconclusive result is still possible, but it seems to me
that
> > through empirical and/or theoretical study, there is room for optimism
for a
> > definitive answer as to whether or not the postulated Neo-Darwinian
> > mechanism is sufficient.
> >
> > rgds,
> > Mark
>
> If EWAKI is a genuine theory, it is by definition falsifiable.
>
> I assume Darwinism will someday suffer the same fate as Newtonism, and
> continue to be taught a century afterward.
It already has, it's just that some are slow to catch on.
rgds,
Mark
Mark & Roslyn Elkington
Yes, by one limited definition. I think though that in the context of this
discussion we all understand that the term "evolution" means much, much more
than that. Don't we?
Mark & Roslyn Elkington
....
> I notice that the thread has a slew of responses. I shall confine myself
> to a few comments.
>
> Darwinism is way more than mutation + selection. It includes
> recombination, drift, punctuation, sexual selection, and a host of other
> things. "Neo-Darwinism" is an anachronistic term used by some as a badge
> of pride and others as a slur. In neither case does it denote anything
> much.
Agreed, Neo-Darwinism (mutation + selection) is only a subset of the modern
synthesis, but in the origins debate it's the only part of the synthesis
worth a penny. What, apart from mutation + selection (sexual is a subset of
natural), can attempt to explain the creation of increasingly ordered,
complex, functional forms?
There is not necessarily any single optimal solution(s) that "organisms
*have* to find," agreed. But, the *viable* solutions within the phase space
remain a minute fraction of the total permutations.
Why do we marvel at a butterfly? Because for every butterfly, there are a
squadrillion other completely uninteresting permutations of the same bunch
of atoms. The butterfly "solution" is so utterly unlikely that a re-run of
evolution would not produce one (cf Gould). Moreover, I contend that the
fraction of permutations of those same atoms of which would form a viable
organism of any kind or shape would be infinitesimal. For example, Yockey
estimated that only 1 in10^40 possible ctyochrome c variants would be
functional.
Of course.
rgds,
Mark
Mark & Roslyn Elkington
Chris Merli <clm...@soltec.net> wrote in message
news:P6GY7.45761$a56....@atlpnn01.usenetserver.com...
You're right in sayin that getting the assumptions and model right is
fraught and vital. However, for this in-principle analysis we only need to
examine a simple and well-characterised mechanism.
Mark
Mark & Roslyn Elkington
> On Tue, 01 Jan 2002 22:18:37 -0600, John Wilkins wrote:
>
> >
> >> On the other hand, if evolutionary computing, GA problem-solving and
> >> information studies etc do reveal such constraints, Darwinism is
> >> dead, and there does not appear to be convincing naturalistic
> >> alternative.
>
> > favour of evolution as we know it (EAWKI, to avoid the loaded and
> > misleading term "Darwinism").
>
> Let me put in a general word of warning re comparing GAs to biological
> evolution. The comparison is useful up to a point, but there's a danger
> of pushing the analogy too closely.
>
> GAs were "inspired by" biological evolution (or perhaps merely dog
> breeding), but they are not generally used to model biological
> evolution. Most often they are used to find a "decent" solution for
> some specific problem that would require too much time or space to solve
> exactly. They do not usually follow the mechanisms of biological
> evolution in anything but the most superficial form,
Not at all. They can follow it in rather substantial form:
http://208.245.156.153/archive/output.cfm?ID=1455
> and of course they
> only use a trivial fraction of the resources that biological evolution
> has used over the ages.
Important distinction: we are checking in-principle limitations, not trying
to model all evolution on a computer.
> Biological evolution, IMO, is not trying to solve any problem at all,
> any more than gravity or chemistry is. (I suppose you could say that
> gravity and chemistry are "trying" to find the lowest energy state of
> certain kinds of systems, but I would be very surprised to discover that
> any scientist actually does think of them that way, at least as the
> top-level understanding of the system under study.)
>
> GAs can still be useful for demonstrating certain points, e.g. that
> breeding, mutation, and weeding-out-by-fitness can indeed generate "new
> stuff" that was not in the original population, and that random
> mutations can sometimes be beneficial (as many ill-informed creationists
> like to deny). However, at each and every invocation of GAs to
> demonstrate a point regarding biological evolution, the reflexive
> response should be to immediately ask yourself whether the analogy is
> being pushed too far. This should save much grief resulting from
> ill-gotten conclusions.
Valid caveat. I hope and expect Dembski et al will apply it.
> > This is the main reason why Dembksi is now running his attack on GAs
> > with the No Free Lunch Theorem. As I understand that (this not being a
> > speciality of mine), that theorem only applies in restricted cases. If
> > you want to find algorithmic limitations, you can - Chaitin does not
> > think he can accommodate EAWKI in his AIT notion of complexity and
> > information. But Chaitin's response is not to say that EAWKI is false,
> > but merely that he hasn't yet managed to come up with an entirely
> > complete characterisation of the world. Dembski is, so far as I can
> > tell, grasping at straws.
>
> Dembski is going to get creamed if he stakes his credibility on using
> the No Free Lunch theorem to refute the reality of biological evolution.
>
> The essense of the NFL theorem is something like this -
>
> "For two search algorithms A and B, if A outperforms B on n search
> problems, then B likewise outperforms A on approximately n other search
> problems."
>
> An immediate corollary is that there is no universal "best" search
> algorithm _without_reference_to_the_application_.
>
> See Wolpert & Macready, "No Free Lunch Theorems for Search",
> http://citeseer.nj.nec.com/wolpert95no.html. (If it turns out that I
> only have access to that site by academic arrangement, let me know and
> I'll see if I can find a direct link to his paper on line.)
Got it, thanks.
>
> There is a companion paper w.r.t. optimization (rather than search), but
> I have not had time to read it.
>
>
> Attempting to apply the NFL to biological evolution immediately
> encounters one of the following problems:
>
> 1) Is biological evolution even a "search problem", in any meaningful
> sense? If not, the NFL simply doesn't apply.
Most definitely: it's a search of an extremely sparse fitness landscape.
> 2) Even if you do think biological evolution is a search problem, notice
> that no one is claiming that evolution is the "best" solution to *all*
> search problems. Without such a claim, the NFL simply doesn't apply.
>
> I personally don't care to think of evolution as a search problem
> because of the teleological nature of that claim. (Granted, natural
> selection does "improve the fit" over time, but I'm still a bit
> reluctant to view it as a search or optimization problem; I prefer to
> view the ToE as descriptive rather than prescriptive.)
>
>
> For that matter, no one seems to be making even the very restricted
> claim that biological evolution is the best way of solving the single
> "search problem" of finding an organism that is best suited to the
> environment. Clearly, evolution is the "algorithm" that nature actually
> uses, but that is not the same as a claim that it is the best possible
> algorithm. For that matter, evolution does not seem to actually be
> finding the best fits between organisms and the environment, as
> witnessed by bad backs, infections of the appendix, nerves looping down
> a giraffe's neck and then back up, etc.
>
> And there is *nothing* in the NFL that claims that a non-best search
> algorithm can never find a decent solution to a problem.
>
> I suspect Dembski is going to appeal to the fallacious notion that
> humans are the "correct" answer to some search problem, add the
> fallacious claim that the "correct" answer could not have been found by
> anything but the "best" search algorithm, and then correctly invoke the
> NFL to prove that no such "best" algorithm exists. Brings to mind a
> parable about building towers on shifting sands...
I'll be interested to see what he does in his upcoming book -- advance blurb
claims it address these kinds of issues.
rgds,
Mark
Mark & Roslyn Elkington
<smirk>
> I'll try again to read Dembski's "paper" later, though that sort of
> a-prioristic logic twisting really annoys me. (FWIW, it's exactly what
> made me get out of linguistics.)
Fair enough. He has given a good advance summary of his new book which
gives some idea about what he's trying to do:
http://www.sunflower.com/~jkrebs/ID/DI/Dembski/No_Free_Lunch.html
Mark
Mark & Roslyn Elkington
I think they're attempting to support a "strong" version.
all read but must sleep...
Mark & Roslyn Elkington
I argue that the same data can equally point to design re-use as common
descent.
> >
> > Mark
>
> -- Steven J.
>
John Wilkins
> John Wilkins <john.w...@bigpond.com> wrote in message
> news:1f5fmr4.ky4cxw17xs7t8N%john.w...@bigpond.com...
>
> ....
> > I notice that the thread has a slew of responses. I shall confine myself
> > to a few comments.
> >
> > Darwinism is way more than mutation + selection. It includes
> > recombination, drift, punctuation, sexual selection, and a host of other
> > things. "Neo-Darwinism" is an anachronistic term used by some as a badge
> > of pride and others as a slur. In neither case does it denote anything
> > much.
>
> Agreed, Neo-Darwinism (mutation + selection) is only a subset of the modern
> synthesis, but in the origins debate it's the only part of the synthesis
> worth a penny. What, apart from mutation + selection (sexual is a subset of
> natural), can attempt to explain the creation of increasingly ordered,
> complex, functional forms?
You've already mentioned one - self-organisation (autopoiesis has a much
nicer ring to it). Things can be functional because the way physical
objects work results in such functionality (eg, stable orbits,
convection cells, hydrological cycle, possible Gaianisms). However I do
think that selection is the proximal cause of most functionality.
Two things - mutation is not the source of most variation. Donald
Campbell used to characterise what I think you are referring to as
neo-Darwinism as BVSR: blind variation and selective retention. But most
of that blind variation is recombination of already existing variants.
Of course mutation is the ultimate source of variation over which
selection acts, but the dimensionality of extant variation since, say,
the late Cambrian would be enought o drive a lot of evolution.
Generally mutations are merely excisions or duplications of nucleotides.
A simple duplication of a sequence can give a high dimensionality of
variation.
>
> >
> > Kaufmann's ideas are not an *alternative* to NS, any more than the basic
> > laws of chemistry or physics are. Any self-organised system still has to
> > undergo the rigours of selection and drift. Merely being self-organised
> > is no guarantee that the system is viable.
....
I leave this in to illustrate the point made 3 paras above.
> > > Well, that's the whole the debate. Can NS in fact work to produce the
> set of
> > > complex cases of living organisms, within the phase space and
> time/trials
> > > available? Or can it only produce some mathematically bounded degree of
> > > variation, i.e. microevolutionary adaptations? (with usual disclaimers
> for
> > > the use of that problematic terminology).
> >
> > OK, let's carefully consider two of the concepts here:
> >
> > Phase space - there is no set phase space of any dimensionality other
> > than the conditions required by the laws of physics. There are no
> > optimal solution coordinates in any space that organisms *have* to find.
> > A phase space is how *we* - primates in lab coats - choose to model the
> > system. It exists as a semantic construct only in our heads and our
> > tools (computers, journals, internet, etc).
>
> There is not necessarily any single optimal solution(s) that "organisms
> *have* to find," agreed. But, the *viable* solutions within the phase space
> remain a minute fraction of the total permutations.
In a rugged landscape they may be local to indefinitely many regions of
the phase space. So a search of a small fraction of the space can find a
satisficing solution easily in such cases. See Gavrilet's work:
Gavrilets, S. and J. Gravner (1997). "Percolation on the fitness
hypercube and the evolution of reproductive isolation." J. Theor. Biol.
184(1): 51-64.
In a smooth landscape, however, where a peak is distant and abrupt in an
otherwise ungradiented field, selection will clearly not find the peak
because neither a random search will find anything, nor will selective
coefficients overcome the noise of mutation and drift.
>
> Why do we marvel at a butterfly? Because for every butterfly, there are a
> squadrillion other completely uninteresting permutations of the same bunch
> of atoms. The butterfly "solution" is so utterly unlikely that a re-run of
> evolution would not produce one (cf Gould). Moreover, I contend that the
> fraction of permutations of those same atoms of which would form a viable
> organism of any kind or shape would be infinitesimal. For example, Yockey
> estimated that only 1 in10^40 possible ctyochrome c variants would be
> functional.
Yockey is full of shit. Sorry, but that "calculation" was so biased and
question begging as to lack intellectual rigor. There are *many*
cytochrome C variants that are functional in modern organisms. But that
is not to exhaust the variants that could be slightly less functional,
or variants in organisms with different metabolic pathways, or variants
that could fulfil another function until they were coopted for a novel
function, and so forth. Yockey loads the bases in his favour.
Moreover, that butterflies or humans would not necessarily result (ie,
specific forms and adaptive strategies might be contingent) is not to
say that *some* forms would not result. Once you stop restricting the
outcomes to whatever is a priori "interesting" to modern human
biologists with a zoological prejudice, the field exapnds enormously.
....
> > Just be careful to make sure that the mechanisms you test are not
> > strawmen.
>
> Of course.
>
> rgds,
> Mark
Wesley R. Elsberry
Mark & Roslyn Elkington <mar...@zeta.org.au> wrote:
>Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
>news:a0unt8$mai$1...@geraldo.cc.utexas.edu...
[...]
BDB>Let me put in a general word of warning re comparing GAs to
BDB>biological evolution. The comparison is useful up to a point,
BDB>but there's a danger of pushing the analogy too closely.
BDB>GAs were "inspired by" biological evolution (or perhaps merely
BDB>dog breeding), but they are not generally used to model
BDB>biological evolution. Most often they are used to find a
BDB>"decent" solution for some specific problem that would require
BDB>too much time or space to solve exactly. They do not usually
BDB>follow the mechanisms of biological evolution in anything but
BDB>the most superficial form,
ME>Not at all. They can follow it in rather substantial form:
ME>http://208.245.156.153/archive/output.cfm?ID=1455
[...]
There do exist some EC instances which are undertaken in order
to pursue theoretical questions in biology. These, though,
seem to be quite a small fraction of the total number of
instances of EC.
The URL brings up a page describing Adrian Thompson's
evolvable FPGA chip experiments. Thompson's work is very
interesting, but I know of no one who scores it highly on a
measure of biological verisimilitude. As such, citing it
seems to confirm Bryant's stance. (This practice of citing
references which support a stance opposite one's own is
commonly called "shooting yourself in the foot".) It should
also be noted that Bryant's claim is a statistical one, where
the evidence to overturn it would have to somehow substantiate
that over half of all instances of EC have high biological
verisimilitude. I find it highly doubtful that any such
demonstration will be forthcoming.
--
Wesley R. Elsberry, Student in Wildlife & Fisheries Sciences, Tx A&M U.
Visit the Online Zoologists page (http://www.onlinezoologists.com)
CNS BBS FTP Archive: ftp://centralneuralsystem.com/pub/CNS/bbs/
"three great virtues of a programmer: laziness, impatience, & hubris"-PP
Chris Merli
Apparently that is what those who are planning this analysis would have us
believe but if there was anything that 10 years of study has taught me, this
is not so simple an issue. My point is that until Biologists understand the
issues much better there is no way a mathematician is going to develop and
effective model.
Noelie S. Alito
> Steven J. <stev...@altavista.com> wrote in message
> news:127ccf2e.02010...@posting.google.com...
> > "Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in message
> news:<qxCY7.1530$ko4.1...@nasal.pacific.net.au>...
> > > Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> > > news:a0uo3m$mar$1...@geraldo.cc.utexas.edu...
> > > > On Tue, 01 Jan 2002 21:54:58 -0600, Mark & Roslyn Elkington wrote:
> > > > Even if you somehow rule out "Darwinism" the
> > > > theory, you're still left with evolution, the fact of nature.
> > >
> > > Spoken like a true ontological naturalist.
> > >
> > I do not believe that "ontological naturalism" compels a belief in
> > common descent or descent with modification, unless the evidence
> > points that way. Bobby Bryant's point is that we still have the
> > evidence of shared ancestry of humans and other primates, of mammals
> > and other vertebrates, of vertebrates and other animals, of animals
> > and other eurkaryotes, etc, whether we have any explanation for the
> > phenomena or not.
> >
> > Darwin deduced common descent well before he figured out a mechanism
> > that he thought could account for it. If his mechanism (and the
> > elaborations and corrections of it which his successors devised) is
> > inadequate, this does not seem to mean that common descent did not
> > happen. So, do you acknowledge that the argument is simply over what
> > produced common descent with modification, or do you argue over
> > whether even that happened?
>
> I argue that the same data can equally point to design re-use as common
> descent.
That would be "design re-use along vertical lines only, and not
transferred laterally", to distinguish from the known human
practice of design reuse, which has no such arbitrary constraint.
Then again, there's no way of knowing how our alien designers
think--lateral transfer of design ideas may have been against
department policy (which is in its own way like human design,
in that there are clueless managers and department directors
to screw up engineering decisions).
Noelie
--
And if thy oblation be a meat offering baken in a pan, it shall be
of fine flour unleavened, mingled with oil. Thou shalt part it in
pieces, and pour oil thereon: it is a meat offering. And if thy
oblation be a meat offering baken in the fryingpan, it shall be
made of fine flour with oil. And thou shalt bring the meat
offering that is made of these things unto the LORD: and when
it is presented unto the priest, he shall bring it unto the altar.
--Lev 2: 5-8
howard hershey
----------
In article <yRXY7.1588$ko4.1...@nasal.pacific.net.au>, "Mark & Roslyn
Elkington" <mar...@zeta.org.au> wrote:
>
> John Wilkins <john.w...@bigpond.com> wrote in message
> news:1f5fmr4.ky4cxw17xs7t8N%john.w...@bigpond.com...
>
[snip]
>>
>> OK, let's carefully consider two of the concepts here:
>>
>> Phase space - there is no set phase space of any dimensionality other
>> than the conditions required by the laws of physics. There are no
>> optimal solution coordinates in any space that organisms *have* to find.
>> A phase space is how *we* - primates in lab coats - choose to model the
>> system. It exists as a semantic construct only in our heads and our
>> tools (computers, journals, internet, etc).
>
> There is not necessarily any single optimal solution(s) that "organisms
> *have* to find," agreed. But, the *viable* solutions within the phase space
> remain a minute fraction of the total permutations.
That depends upon the relevant phase space. For a butterfly, the relevant
phase space that it will or can explore will be on the fringes of its
current environment. A butterfly will not explore the phase space of a
mussel.
>
> Why do we marvel at a butterfly? Because for every butterfly, there are a
> squadrillion other completely uninteresting permutations of the same bunch
> of atoms.
And the mechanism of self-reproduction (imperfect self-reproduction is a
requisite assumption of evolution) ensures that most of the next
generation's genetic sequences are repeats of this generations. And
permutations that do occur always undergo a selective filter that both weeds
out deleterious changes and retains "interesting" ones that allow exploring
unoccupied nearby phase space.
> The butterfly "solution" is so utterly unlikely that a re-run of
> evolution would not produce one (cf Gould).
Agreed. But evolution does not work by changing every one of the
butterfly's atoms to produce a different species of butterfly or to
magically convert the butterfly so that it can explore the phase space of a
mussel. It does change the three genes that are responsible for all the
different wing patterns. Evolution is a much more stepwise process than one
that magically converts one organism to another. At some point these
changes might lead to a creature that we would no longer "name" butterfly,
just as there is an ancestral species we also would no longer "name"
butterfly.
> Moreover, I contend that the
> fraction of permutations of those same atoms of which would form a viable
> organism of any kind or shape would be infinitesimal. For example, Yockey
> estimated that only 1 in10^40 possible ctyochrome c variants would be
> functional.
At what level of "function"? The biosphere has certainly explored many (if
not most) of the cytochrome c variants that do function at a high level of
function. But it did so in a stepwise fashion with filtering at each step
to retain or locally improve function. Most of the changes we see are
selectively neutral changes.
>
[snip]
George Acton
>
> Steven J. <stev...@altavista.com> wrote in message
> news:127ccf2e.02010...@posting.google.com...
>
> >
> > that he thought could account for it. If his mechanism (and the
> > elaborations and corrections of it which his successors devised) is
> > inadequate, this does not seem to mean that common descent did not
> > happen. So, do you acknowledge that the argument is simply over what
> > produced common descent with modification, or do you argue over
> > whether even that happened?
>
> I argue that the same data can equally point to design re-use as common
> descent.
This argument fails to deal with the pattern of neutral mutations
in the camparative DNA data. These have no functional significance,
since they do not affect the activity of the gene product. therefore,
they would be of no interest to a designer. A designer could
reasonably make them all the same or let them vary randomly. There
is no reason for a designer to produce the minutely detailed
nisted hierarchy we see, unless the designer were a trickster who
wanted us to infer common descent when the reality is separate
creation. The DNA sequence data now comprises billions of bases,
all of which conform to the pattern consistent with common descent.
--George Acton
Rodjk
> news:<qxCY7.1530$ko4.1...@nasal.pacific.net.au>...
> > > Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> > > news:a0uo3m$mar$1...@geraldo.cc.utexas.edu...
> > > >
> > > > > On the other hand, if evolutionary computing, GA problem-solving and
> > > > > information studies etc do reveal such constraints, Darwinism is
> dead,
> > > > > and there does not appear to be convincing naturalistic alternative.
> > > >
> > > > understanding the cause of the *observed* changes to the earth's
> > > > biosystem over time.
> > >
> > > I'm using the term "Darwinism" as the neo-Darwinian component of the
> modern
> > > synthesis, i.e. mutation and natural selection. If you lose that
> mechanism,
> > > what are the serious naturalistic alternatives?
> >
> > theory that explains it.
>
> Wrong.
>
> Evolution is one theory to explain the observation. Darwinism, or
> Neo-darwinism are some proposed mechanisms within this theory, among others.
No. Darwinism is the explanation for what we see, the observation of
evolution.
>
> Your failure to correctly state these basic definitions is telling.
LOL. Your failure to correctly state these basic definitions is
telling.
>
>
> > Tell me, why turn to math and GA's?
> > Can't your friends do biology?
I notice you did not answer this.
I wonder why. :-)
> >
> > >
> > > > Even if you somehow rule out "Darwinism" the
> > > > theory, you're still left with evolution, the fact of nature.
> > >
> > > Spoken like a true ontological naturalist.
> >
> > Seems you still do not understand the difference between the fact and
> > the theory. Even if you were to show that darwinism is wrong, you have
> > not shown that evoltution is wrong.
> >
> > Try doing, or at least learning, some biology.
> > Rodjk
>
> See above.
I still see no sign that you understand basic science, like the
difference and relationship between a fact and a theory.
Rodjk
Rodjk
Which just happens to be the definition used by scientist.
We are talking about science, right?
>I think though that in the context of this
> discussion we all understand that the term "evolution" means much, much more
> than that. Don't we?
The definition describes the basics of what happens. That it has
larger results does not change the definitions.
Rodjk
> > Is ID a serious challenge to Darwinian evolution? Specifically, will the
> > study of genetic algorithms reveal theoretical and empirical limitations
> to
> > the creative power of the trial and error mechanism?
> >
> > A new site to watch re these questions is http://www.iscid.org IMO a
> source
> > of some well written and cogently argued essays, even if you don't agree
> > with the ID viewpoint (some from the usual suspects, of course).
> >
> > For a creationist like myself this area seems crucial. If study of GAs
> > reveal no in principle limits to CSI creation and increase in novelty,
> > complexity and functionality, then it would appear that related arguments
> > against Darwinian evolution would be significantly weakened, and force a
> > retreat to secondary biological objections in practice, despite
> theoretical
> > possibility.
> >
> > information studies etc do reveal such constraints, Darwinism is dead, and
> > there does not appear to be convincing naturalistic alternative.
> >
> > the hands of biologists and into the hands of mathematicians and computing
> > scientists (shock, horror...hallelujah?).
>
> Once again a post where the amazing mathematicians will in fact be taking
> over all fields of science. While it is still impossible to mathematically
> model a single one of even the simplest cells, mathematicians are now ready
> to describe trillions of organisms evolving over billions of years. I
> realize that what they are in fact planning is to derive more general
> conclusions from far more simple experiments, but looking at the bigger
> picture brings up the very important point that these experiments will
> require a huge number of assumptions to create workable experiments. If you
> have already read the previous attempts by people to describe such complex
> biological systems mathematically (i.e. proof that it is impossible to form
> a single DNA coding sequence for a protein by random chance) you must
> realize that the issue is not whether they can create a mathematical proof
> given certain assumption but if they can get the assumptions right. For
> example in the case I mentioned the problem was not the math but the
> assumption that the process of creating a coding sequence happened entirely
> by chance. (There were other false assumptions but even a single poor
> assumption invalidates a "proof".) For the situation described here, the
> big question would be if the constraints appear is it because of poor
> assumptions. I don't think mathematicians can reasonably hope to answer
> that question.
>
> > analysis, I think at some stage we are going to hear the EC guys saying,
> "We
> > have found that in practice you can use GAs to discover solutions to
> > problems of class A, but not class B." (Life as found on planet earth
> being
> > decidedly class B.)
> >
> > Either way, we seem to have a testable hypothesis, which is an essential
> > ingredient of a happy new year.
>
> Just for those of us who were asleep could you state the testable
> computing, GA problem-solving and information studies etc do reveal such
> constraints, Darwinism is dead, and there does not appear to be convincing
> evolution is wrong then creation is right. This is not a testable theory in
> favor of creation.
>
> By comparison look at the predictions made by Darwins theory. The existance
> of genes (a contemparary but not yet accepted idea). The ability of these
> genes to be passed on to offspring. The existence of a material that could
> carry genetic information to the offspring in a highly conserved fashion.
> The mutability of the genes. The idea that on the genetic level organisms
> which are amzingly diverse in appearance would be related.
Chris, calm down. He was not talking about real mathematicians, he was
talking about Dembski.
Rodjk
> >
> > Regards,
> > Mark Elkington
> > Sydney
> >
Mark VandeWettering
>
> I wasn't suggesting that a single failed simulation proves anything as such.
> A formal proof will need to be based on rigorous mathematics, correct
> assumptions etc, which I think is where Dembski is trying to head. From
> other things he's published he seems to me to be reasonably equipped in this
> area.
> However, I was suggesting that as evolutionary computing matures a
> discipline, an empirical understanding of the problem solving capabilities
> and limits of evo GAs will emerge. My contention is that this result will be
> strongly confirming of the theoretical result, either way.
You are putting the cart before the horse.
There is no "theoretical result", merely the idle musings of those who are
not experts in either biological evolution or evolutionary computation.
Mark
> rgds,
> Mark
Gavin Tabor
>
> study of genetic algorithms reveal theoretical and empirical limitations to
> the creative power of the trial and error mechanism?
My god, finally a creationist who is prepared to accept GA's as a valid
way
of investigating Natural Selection. Wow!
>
> A new site to watch re these questions is http://www.iscid.org IMO a source
> of some well written and cogently argued essays, even if you don't agree
> with the ID viewpoint (some from the usual suspects, of course).
>
> For a creationist like myself this area seems crucial. If study of GAs
> reveal no in principle limits to CSI creation and increase in novelty,
> complexity and functionality, then it would appear that related arguments
> against Darwinian evolution would be significantly weakened, and force a
> retreat to secondary biological objections in practice, despite theoretical
> possibility.
Certainly GA's demonstrate the creative power of variation + selection.
So far there seem no limits demonstrated.
>
> On the other hand, if evolutionary computing, GA problem-solving and
> information studies etc do reveal such constraints, Darwinism is dead, and
> there does not appear to be convincing naturalistic alternative.
>
> What is really interesting is that this approach takes the argument out of
> the hands of biologists and into the hands of mathematicians and computing
> scientists (shock, horror...hallelujah?). Even aside from theoretical
> analysis, I think at some stage we are going to hear the EC guys saying, "We
> have found that in practice you can use GAs to discover solutions to
> problems of class A, but not class B." (Life as found on planet earth being
> decidedly class B.)
Why do you assume that?
>
> Either way, we seem to have a testable hypothesis, which is an essential
> ingredient of a happy new year.
And a very pleasant new year to you as well.
Gavin
>
> Regards,
> Mark Elkington
> Sydney
--
Dr. Gavin Tabor
School of Engineering and Computer Science
Department of Engineering
University of Exeter
Mark VandeWettering
> > Darwin deduced common descent well before he figured out a mechanism
> > that he thought could account for it. If his mechanism (and the
> > elaborations and corrections of it which his successors devised) is
> > inadequate, this does not seem to mean that common descent did not
> > happen. So, do you acknowledge that the argument is simply over what
> > produced common descent with modification, or do you argue over
> > whether even that happened?
>
> I argue that the same data can equally point to design re-use as common
> descent.
This should be a strong signal that design is unfalsifiable. Can you
present an instance of something which isn't designed?
Mark
Gavin Tabor
>
> On Tue, 01 Jan 2002 22:18:37 -0600, John Wilkins wrote:
>
> >
> >> On the other hand, if evolutionary computing, GA problem-solving and
> >> information studies etc do reveal such constraints, Darwinism is
> >> dead, and there does not appear to be convincing naturalistic
> >> alternative.
>
> > favour of evolution as we know it (EAWKI, to avoid the loaded and
> > misleading term "Darwinism").
>
> evolution. The comparison is useful up to a point, but there's a danger
> of pushing the analogy too closely.
>
> breeding), but they are not generally used to model biological
> evolution. Most often they are used to find a "decent" solution for
> some specific problem that would require too much time or space to solve
> exactly. They do not usually follow the mechanisms of biological
> evolution in anything but the most superficial form, and of course they
> only use a trivial fraction of the resources that biological evolution
> has used over the ages.
They provide a valid model of the processes believed to be underlying
evolution. Interestingly, they do provide the possibility of trying
other variants - 3 sexes for instance, or different crossover operators.
I don't know if anyone has ever applied this information to biology
though.
>
> Biological evolution, IMO, is not trying to solve any problem at all,
> any more than gravity or chemistry is. (I suppose you could say that
> gravity and chemistry are "trying" to find the lowest energy state of
> certain kinds of systems, but I would be very surprised to discover that
> any scientist actually does think of them that way, at least as the
> top-level understanding of the system under study.)
Prepare to be very surprised then. That is exactly how higher level
mathematical physics operates. Both Lagrangian and Hamiltonian dynamics
operate by constructing a function and then minimising it to find the
solution. For that matter, the classical interpretation of particle
propagation in QM is that the particle takes all possible routes between
A and B, but only those that give a stationary contribution to the
resulting phase (a form of extremum) interfere constructively at point
B.
I would suggest that the limitations in GA's as a mathematical
technique - difficult to find "the absolute" minimum, for instance
- are precisely the ones we find in biology.
>
> GAs can still be useful for demonstrating certain points, e.g. that
> breeding, mutation, and weeding-out-by-fitness can indeed generate "new
> stuff" that was not in the original population, and that random
> mutations can sometimes be beneficial (as many ill-informed creationists
> like to deny). However, at each and every invocation of GAs to
> demonstrate a point regarding biological evolution, the reflexive
> response should be to immediately ask yourself whether the analogy is
> being pushed too far. This should save much grief resulting from
> ill-gotten conclusions.
A fair point.
>
> > This is the main reason why Dembksi is now running his attack on GAs
> > with the No Free Lunch Theorem. As I understand that (this not being a
> > speciality of mine), that theorem only applies in restricted cases. If
> > you want to find algorithmic limitations, you can - Chaitin does not
> > think he can accommodate EAWKI in his AIT notion of complexity and
> > information. But Chaitin's response is not to say that EAWKI is false,
> > but merely that he hasn't yet managed to come up with an entirely
> > complete characterisation of the world. Dembski is, so far as I can
> > tell, grasping at straws.
>
> Dembski is going to get creamed if he stakes his credibility on using
> the No Free Lunch theorem to refute the reality of biological evolution.
>
> The essense of the NFL theorem is something like this -
>
> "For two search algorithms A and B, if A outperforms B on n search
> problems, then B likewise outperforms A on approximately n other search
> problems."
>
> An immediate corollary is that there is no universal "best" search
> algorithm _without_reference_to_the_application_.
>
> See Wolpert & Macready, "No Free Lunch Theorems for Search",
> http://citeseer.nj.nec.com/wolpert95no.html. (If it turns out that I
> only have access to that site by academic arrangement, let me know and
> I'll see if I can find a direct link to his paper on line.)
>
> There is a companion paper w.r.t. optimization (rather than search), but
> I have not had time to read it.
>
> Attempting to apply the NFL to biological evolution immediately
> encounters one of the following problems:
>
> 1) Is biological evolution even a "search problem", in any meaningful
> sense? If not, the NFL simply doesn't apply.
>
> 2) Even if you do think biological evolution is a search problem, notice
> that no one is claiming that evolution is the "best" solution to *all*
> search problems. Without such a claim, the NFL simply doesn't apply.
>
> I personally don't care to think of evolution as a search problem
> because of the teleological nature of that claim. (Granted, natural
> selection does "improve the fit" over time, but I'm still a bit
> reluctant to view it as a search or optimization problem; I prefer to
> view the ToE as descriptive rather than prescriptive.)
>
> For that matter, no one seems to be making even the very restricted
> claim that biological evolution is the best way of solving the single
> "search problem" of finding an organism that is best suited to the
> environment. Clearly, evolution is the "algorithm" that nature actually
> uses, but that is not the same as a claim that it is the best possible
> algorithm. For that matter, evolution does not seem to actually be
> finding the best fits between organisms and the environment, as
> witnessed by bad backs, infections of the appendix, nerves looping down
> a giraffe's neck and then back up, etc.
Precisely. Evolution is a crummy way of designing things. Thats why
biological systems don't show many signs (IMHO) of being designed.
Didn't the Discovery Institute start offering grants for ID-related
work? I'm wondering if one could apply for money to do valid scientific
work with it, or if the requirements would be too constrictive
to allow it. If not, then I can imagine the SETI community being very
interested in funding for projects to investigate ways of distinguishing
designed information from random noise.
>
> And there is *nothing* in the NFL that claims that a non-best search
> algorithm can never find a decent solution to a problem.
>
> I suspect Dembski is going to appeal to the fallacious notion that
> humans are the "correct" answer to some search problem, add the
> fallacious claim that the "correct" answer could not have been found by
> anything but the "best" search algorithm, and then correctly invoke the
> NFL to prove that no such "best" algorithm exists. Brings to mind a
> parable about building towers on shifting sands...
Sounds all too plausible.
Gavin
>
> > Even worse, *physical* approximations to natural selection work, in,
> > say, robotics and electronics. What do we make of that? Why, we must
> > assert, as Dembski dies
> ^ Typo, or Freudian slip?
>
> Bobby Bryant
> Austin, Texas
David Jensen
rayt...@yahoo.com (Mark VandeWettering) wrote in
<580c1a4a.02010...@posting.google.com>:
Or that it was designed by programmers who didn't understand the old
bugs/features, so they just left the crap in and did a workaround.
Mark & Roslyn Elkington
Wesley R. Elsberry <w...@cx33978-a.dt1.sdca.home.com> wrote in message
news:2002010313...@cx33978-a.dt1.sdca.home.com...
Rather, it is a willingness to bring in and consider all evidence, even that
which appears to oppose one's own conclusions.
George Acton
>
> Is ID a serious challenge to Darwinian evolution? Specifically, will the
> study of genetic algorithms reveal theoretical and empirical limitations to
> the creative power of the trial and error mechanism?
>
> of some well written and cogently argued essays, even if you don't agree
> with the ID viewpoint (some from the usual suspects, of course).
>
> For a creationist like myself this area seems crucial. If study of GAs
> reveal no in principle limits to CSI creation and increase in novelty,
> complexity and functionality, then it would appear that related arguments
> against Darwinian evolution would be significantly weakened, and force a
> retreat to secondary biological objections in practice, despite theoretical
> possibility.
>
> information studies etc do reveal such constraints, Darwinism is dead, and
> there does not appear to be convincing naturalistic alternative.
>
> the hands of biologists and into the hands of mathematicians and computing
> scientists (shock, horror...hallelujah?). Even aside from theoretical
> analysis, I think at some stage we are going to hear the EC guys saying, "We
> have found that in practice you can use GAs to discover solutions to
> problems of class A, but not class B." (Life as found on planet earth being
> decidedly class B.)
>
> ingredient of a happy new year.
It sounds like "let's-pretend" science, as if we were going to
decide some controversy in physics or chemistry by pure calculation,
without attending to any empirical data. The calculatios are
interesting but they aren't necessarily informative about the
correct description of nature. In this case a necessary constraint
is that we know common descent occurred, and we know the
aproximate time scale. This appears to remove much of the
theological motivation for the investigation.
I have a problem with calling an empirical claim a "Theorem".
Theorems are usually published in mathematics journals and are
are true without needing experimental confirmation. Hypotheses
about the physical world are usually called um hypotheses.
--George Acton
Wesley R. Elsberry
Mark & Roslyn Elkington <mar...@zeta.org.au> wrote:
>Wesley R. Elsberry <w...@cx33978-a.dt1.sdca.home.com> wrote in message
>news:2002010313...@cx33978-a.dt1.sdca.home.com...
>> In article <K7YY7.1590$ko4.1...@nasal.pacific.net.au>,
>> Mark & Roslyn Elkington <mar...@zeta.org.au> wrote:
>> >Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
>> >news:a0unt8$mai$1...@geraldo.cc.utexas.edu...
WRE> [...]
BDB>Let me put in a general word of warning re comparing GAs to
BDB>biological evolution. The comparison is useful up to a point,
BDB>but there's a danger of pushing the analogy too closely.
BDB>GAs were "inspired by" biological evolution (or perhaps merely
BDB>dog breeding), but they are not generally used to model
BDB>biological evolution. Most often they are used to find a
BDB>"decent" solution for some specific problem that would require
BDB>too much time or space to solve exactly. They do not usually
BDB>follow the mechanisms of biological evolution in anything but
BDB>the most superficial form,
ME>Not at all. They can follow it in rather substantial form:
ME>http://208.245.156.153/archive/output.cfm?ID=1455
WRE> [...]
WRE> There do exist some EC instances which are undertaken in order
WRE> to pursue theoretical questions in biology. These, though,
WRE> seem to be quite a small fraction of the total number of
WRE> instances of EC.
WRE> The URL brings up a page describing Adrian Thompson's
WRE> evolvable FPGA chip experiments. Thompson's work is very
WRE> interesting, but I know of no one who scores it highly on a
WRE> measure of biological verisimilitude. As such, citing it
WRE> seems to confirm Bryant's stance. (This practice of citing
WRE> references which support a stance opposite one's own is
WRE> commonly called "shooting yourself in the foot".)
ME>Rather, it is a willingness to bring in and consider all
ME>evidence, even that which appears to oppose one's own
ME>conclusions.
That's a commendable attitude to have regarding the evidence.
However, the evidence of the post indicates that Mark was,
in fact, citing the Thompson GA research as if it were a
datum disconfirming Bryant's claim. Note the "Not at all",
which in ordinary usage indicates opposition rather than
agreement, and the use of the colon, which indicates that
the following information is an instance of the general
claim made. I'm afraid that what Mark has done in his reply
is also known by a colloquial phrase: "damage control".
WRE> It should
WRE> also be noted that Bryant's claim is a statistical one, where
WRE> the evidence to overturn it would have to somehow substantiate
WRE> that over half of all instances of EC have high biological
WRE> verisimilitude. I find it highly doubtful that any such
WRE> demonstration will be forthcoming.
So, will we see Mark fulfill his "Not at all" claim? I bet not.
--
Wesley R. Elsberry, Student in Wildlife & Fisheries Sciences, Tx A&M U.
Visit the Online Zoologists page (http://www.onlinezoologists.com)
CNS BBS FTP Archive: ftp://centralneuralsystem.com/pub/CNS/bbs/
"clowns basting each other with clubs and cheap puns" - archy
Bobby D. Bryant
> Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> news:a0unt8$mai$1...@geraldo.cc.utexas.edu...
>> GAs were "inspired by" biological evolution (or perhaps merely dog
>> breeding), but they are not generally used to model biological
>> evolution. Most often they are used to find a "decent" solution for
>> some specific problem that would require too much time or space to
>> solve exactly. They do not usually follow the mechanisms of
>> biological evolution in anything but the most superficial form,
>
> Not at all. They can follow it in rather substantial form:
> http://208.245.156.153/archive/output.cfm?ID=1455
Notice your change of my "they do not usually" to "they can". You're
arguing past me already. As I said, they _are_not_generally_ used to
model biological evolution. Which brings us to...
>> and of course they
>> only use a trivial fraction of the resources that biological
>> evolution has used over the ages.
>
> Important distinction: we are checking in-principle limitations, not
> trying to model all evolution on a computer.
Sure; but my warning is to make sure people don't overgeneralize the
in-principle limitatons on one kind of system to a similar-but-different
kind of system where they may not apply.
It is, of course, for the purpose of in-principle demonstrations that
makes GAs relevant in this forum at all. For instance, I feel like it
is fair to invoke them as in-principle demonstrations of the fact that
blind chance filtered by selection-by-fitness can generate new
structures or behaviors, a fact that many creationists deny.
As for in-principle limitations, whether they lead to arguments relevant
for biology must await the identification of the limitations you wish to
invoke. So far you're merely arguing from the position of "there _may_
be relevant limitations". I don't dispute that claim, but I have to
point out that it's useless as a criticism of the ToE unless/until you
actually have some relevant limitations in hand.
>> Attempting to apply the NFL to biological evolution immediately
>> encounters one of the following problems:
>>
>> 1) Is biological evolution even a "search problem", in any meaningful
>> sense? If not, the NFL simply doesn't apply.
>
> Most definitely: it's a search of an extremely sparse fitness
> landscape.
As I mentioned earlier, I'm not sure I agree with that. Calling it "a
search" seems to import inappropriate teleology into the ToE as a
_description_ of natural processes.
>> I suspect Dembski is going to appeal to the fallacious notion that
>> humans are the "correct" answer to some search problem, add the
>> fallacious claim that the "correct" answer could not have been found
>> by anything but the "best" search algorithm, and then correctly
>> invoke the NFL to prove that no such "best" algorithm exists. Brings
>> to mind a parable about building towers on shifting sands...
>
> I'll be interested to see what he does in his upcoming book -- advance
> blurb claims it address these kinds of issues.
Yeah, well, "advance blurb" said the ID crowd were going to give us a
reliable design inference algorithm, but we're still eagerly waiting for
it. (A pity about those who held their breaths...)
Bobby Bryant
Austin, Texas
Bobby D. Bryant
> "Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in message
> news:AvvY7.1509$ko4.1...@nasal.pacific.net.au...
>> What is really interesting is that this approach takes the argument
>> out of the hands of biologists and into the hands of mathematicians
>> and computing scientists (shock, horror...hallelujah?).
>
> Once again a post where the amazing mathematicians will in fact be
> taking over all fields of science.
Since we get so many people here who don't know what science is (e.g.,
always demainding "proof" for a theory), it may be worth pointing out
that, while mathematics can be a useful tool for science, there is a
world of difference between mathematics and the other logical sciences
vs. the empirical sciences such as physics, chemistry, biology, etc.
Mathematics is wonderful for its ability to provide solid proofs.
However, those proofs exist in a world of arbitrarily chosen axioms.
Change your axioms and your carefully proven theorems may be invalidated
under the new system.
So there is a bit of a disconnect between the mathematical sciences and
the empirical sciences. If someone wants to use math to prove a theorem
about biology, that's well and good -- *iff* the mathematician's axioms
are actually relevant for biology. If the axioms don't make a good
model of the biological world, then the theorem and its proof are
entirely worthless for understanding biology, however rigorous the proof
might be in its own right.
Ultimately, if we have a theorem that says "X is impossible" but we
actually see X in nature, it's the theorem that has to go; the relevance
of the axioms that the theorem was derived from becomes a purely
empirical issue.
Bobby Bryant
Austin, Texas
w watts
> any more than gravity or chemistry is. (I suppose you could say that
> gravity and chemistry are "trying" to find the lowest energy state of
> certain kinds of systems, but I would be very surprised to discover that
> any scientist actually does think of them that way, at least as the
> top-level understanding of the system under study.)
I think as long as no one says that 'biological Evolution' finds
something (anything) better then random (without considering the full
distribution of the fitness landscape--past to present), then the nfl
does not apply.
> Dembski is going to get creamed if he stakes his credibility on using
> the No Free Lunch theorem to refute the reality of biological evolution.
He will definitly get creamed if he says that the No Free Lunch
theorem refutes biological evolution. For instance, its possible for
the fitness function could be matched to the fitness landscape(of the
entire universe). I am sure theistic evolutionists would take this
route.
The most he can say is biological evolution is **no better then
random**
>
> The essense of the NFL theorem is something like this -
>
> "For two search algorithms A and B, if A outperforms B on n search
> problems, then B likewise outperforms A on approximately n other search
> problems."
Remember thats one of the algorithms is random search. Thats what
puts the bite into it. People don't like it when someone says
your_favorite_GA = random when averaged over all distibutions of the
data. I didn't like it either because it seemed ludicrous given that
things like databases MUST perform searchs better then random,
otherwise any old configuration of the data would do. But even this
is handled in that your_favorite_oracle_search must reindex(read
consider the whole distribution of the data) in order to perform
better then random search. If we don't reindex, it actually performs
worse then random. When I considered that I was taken back by the NFL
theorem.
> 1) Is biological evolution even a "search problem", in any meaningful
> sense? If not, the NFL simply doesn't apply.
Not as long as it doesn't do anything better then random.
>
> 2) Even if you do think biological evolution is a search problem, notice
> that no one is claiming that evolution is the "best" solution to *all*
> search problems. Without such a claim, the NFL simply doesn't apply.
Nope, it applys merely if we say it finds **anything** better then
random. It may not be the best but even if it is better then random
the fitness landscape must be examined.
> For that matter, no one seems to be making even the very restricted
> claim that biological evolution is the best way of solving the single
> "search problem" of finding an organism that is best suited to the
> environment. Clearly, evolution is the "algorithm" that nature actually
> uses, but that is not the same as a claim that it is the best possible
> algorithm. For that matter, evolution does not seem to actually be
> finding the best fits between organisms and the environment, as
> witnessed by bad backs, infections of the appendix, nerves looping down
> a giraffe's neck and then back up, etc.
See above.
>
> And there is *nothing* in the NFL that claims that a non-best search
> algorithm can never find a decent solution to a problem.
Never find is absolutely correct. Never better then random without
considering the **whole** fitness landscape is what it says.
>
> I suspect Dembski is going to appeal to the fallacious notion that
> humans are the "correct" answer to some search problem, add the
> fallacious claim that the "correct" answer could not have been found by
> anything but the "best" search algorithm, and then correctly invoke the
> NFL to prove that no such "best" algorithm exists. Brings to mind a
> parable about building towers on shifting sands...
He would be absolutely wrong if he did that. But he may say something
like if you have a premise that evolution finds something (life for
example) better then random you can't hold that the No Free Lunch
theorem is correct.
Regards.
Robt Gotschall
mar...@zeta.org.au says...
>
> Robt Gotschall <hobgo...@worldnet.att.net> wrote in message
> news:MPG.169d0ab7e...@netnews.worldnet.att.net...
> > In article <V_DY7.1537$ko4.1...@nasal.pacific.net.au>,
> > mar...@zeta.org.au says...
> >
> >
> > >
> > > More precisely, Neo-Darwinism, meaning mutation plus natural selection.
> I
> > > think that would be dead, and there are no serious alternatives I know
> of
> > > (Kaufman's self-organisation, or teleology?).
> >
> > Kauffman's self-organisation had to do with abiogenesis.
>
> I thought though some of this had been extended to evolution as well.
Yes, but usually adjunctive to selection and other things.
> > According to theory:
> >
> > Abiogenesis would work to produce living organisms.
> >
> > Random mutations would work to produce variety in living organisms.
> >
> > NS would work to eliminate all but the set of now living organisms.
> >
> > By Kauffman's reckoning, if it's alive it's already in the "useful"
> > complex range and differentiation among living organisms by complexity
> > may not be as meaningful in an evolutionary sense as you may think.
> >
> > Hint: Things _can_ get too complex(see edge-of-chaos).
>
> I would have thought Homo sapiens, for example, would have exceeded any and
> all edge-of-chaos boundaries.
It may not even apply, or we could just be entering the chaos phase
transition.
> > > > > What is really interesting is that this approach takes the argument
> out
> > > > > of the hands of biologists and into the hands of mathematicians and
> > > > > computing scientists (shock, horror...hallelujah?).
> >
> > If they succeed at that I suppose they can go on to prove that white is
> > black and wind up being killed at a zebra crossing. I never trusted
> > zebras BTW.
>
> If they get their assumptions and model right, then they can potentially go
> on prove the Darwinian mechanism either way.
Even if you prove that bees can't fly, all you've done is demonstrate
that your assumptions about how bees fly are wrong.
You have not proven that bees are being held up by tiny little angels.
> > If EWAKI is a genuine theory, it is by definition falsifiable.
> >
> > I assume Darwinism will someday suffer the same fate as Newtonism, and
> > continue to be taught a century afterward.
>
> It already has, it's just that some are slow to catch on.
>
> rgds,
> Mark
But do you know _why_ Newtonian Mechanics is still being taught almost a
century after relativity was introduced?
--
rg
remove spam to mail
http://home.att.net/~hobgots/wsb/html/view.cgi-home.html-.html
John Wilkins
> And if thy oblation be a meat offering baken in a pan, it shall be
> of fine flour unleavened, mingled with oil. Thou shalt part it in
> pieces, and pour oil thereon: it is a meat offering. And if thy
> oblation be a meat offering baken in the fryingpan, it shall be
> made of fine flour with oil. And thou shalt bring the meat
> offering that is made of these things unto the LORD: and when
> it is presented unto the priest, he shall bring it unto the altar.
> --Lev 2: 5-8
Clearly this is the origin of the High Table tradition...
All it lacks is a prescription on the use of sage and thyme, and the
offering of a fine after dinner port.
Mark & Roslyn Elkington
....
> > If they get their assumptions and model right, then they can potentially
go
> > on prove the Darwinian mechanism either way.
>
> Even if you prove that bees can't fly, all you've done is demonstrate
> that your assumptions about how bees fly are wrong.
>
> You have not proven that bees are being held up by tiny little angels.
The difference, and it is fundamental, and I wish I didn't have to keep
mentioning it, is: we directly observe bees flying, but we do not directly
observe megavolution.
That's why the talk.bees-cannot-fly ng has never got off the ground, but t.o
is flying high.
> > > If EWAKI is a genuine theory, it is by definition falsifiable.
> > >
> > > I assume Darwinism will someday suffer the same fate as Newtonism, and
> > > continue to be taught a century afterward.
> >
> > It already has, it's just that some are slow to catch on.
> >
> > rgds,
> > Mark
>
> But do you know _why_ Newtonian Mechanics is still being taught almost a
> century after relativity was introduced?
Yes. It's rather handy and accurate approximation for all but the
fastest-moving problems.
Mark
Michael Altarriba
Mark & Roslyn Elkington wrote:
> Robt Gotschall <hobgo...@worldnet.att.net> wrote in message
> news:MPG.169e7c4de...@netnews.worldnet.att.net...
>
> ....
>
>>>If they get their assumptions and model right, then they can potentially
>>>
> go
>
>>>on prove the Darwinian mechanism either way.
>>>
>>Even if you prove that bees can't fly, all you've done is demonstrate
>>that your assumptions about how bees fly are wrong.
>>
>>You have not proven that bees are being held up by tiny little angels.
>>
>
> The difference, and it is fundamental, and I wish I didn't have to keep
> mentioning it, is: we directly observe bees flying, but we do not directly
> observe megavolution.
Is "megavolution" different from speciation, which has been observed
many times (and links have been mentioned here on talk.origins many,
many, many times)? Have you directly observed nuclear fusion going on
within our Sun? After all, the radiation being emitted is all well and
good, but it isn't a -direct- observation of nuclear fusion. How about
plate tectonics, or any of the other phenomena that occur over long
timescales?
Mark & Roslyn Elkington
Michael Altarriba <mik...@jps.net> wrote in message
news:3C35629C...@jps.net...
>
>
> Mark & Roslyn Elkington wrote:
>
> > Robt Gotschall <hobgo...@worldnet.att.net> wrote in message
> > news:MPG.169e7c4de...@netnews.worldnet.att.net...
> >
> > ....
> >
> >>>If they get their assumptions and model right, then they can
potentially
> >>>
> > go
> >
> >>>on prove the Darwinian mechanism either way.
> >>>
> >>Even if you prove that bees can't fly, all you've done is demonstrate
> >>that your assumptions about how bees fly are wrong.
> >>
> >>You have not proven that bees are being held up by tiny little angels.
> >>
> >
> > The difference, and it is fundamental, and I wish I didn't have to keep
> > mentioning it, is: we directly observe bees flying, but we do not
directly
> > observe megavolution.
>
>
> Is "megavolution" different from speciation, which has been observed
> many times (and links have been mentioned here on talk.origins many,
> many, many times)?
No.
> Have you directly observed nuclear fusion going on
> within our Sun? After all, the radiation being emitted is all well and
> good, but it isn't a -direct- observation of nuclear fusion. How about
> plate tectonics, or any of the other phenomena that occur over long
> timescales?
I didn't say that science by inference was invalid.
Mark & Roslyn Elkington
> Is "megavolution" different from speciation, which has been observed
> many times (and links have been mentioned here on talk.origins many,
> many, many times)?
I meant, yes, it is different. It's mega!
Mark
Mark & Roslyn Elkington
I agree, and it's worth drawing attention to.
Indeed, if you see a flying bumble bee, you must scrap your
bumble-bees-can't fly theorem.
But in the case of origins, we simple don't have this luxury.
Scientifically, we can only ever infer naturalistic evolution or design (as
with several other areas of science), and so we must live with the
disconnect and redouble our caution, but not throw out the tool entirely.
Mark
Mark & Roslyn Elkington
Gavin Tabor <G.R....@exeter.ac.uk> wrote in message
news:3C34A1CA...@ex.ac.uk...
> Mark & Roslyn Elkington wrote:
> >
> > Is ID a serious challenge to Darwinian evolution? Specifically, will
the
> > study of genetic algorithms reveal theoretical and empirical limitations
to
> > the creative power of the trial and error mechanism?
>
> My god, finally a creationist who is prepared to accept GA's as a valid
> way
> of investigating Natural Selection. Wow!
>
> >
> > A new site to watch re these questions is http://www.iscid.org IMO a
source
> > of some well written and cogently argued essays, even if you don't agree
> > with the ID viewpoint (some from the usual suspects, of course).
> >
> > For a creationist like myself this area seems crucial. If study of GAs
> > reveal no in principle limits to CSI creation and increase in novelty,
> > complexity and functionality, then it would appear that related
arguments
> > against Darwinian evolution would be significantly weakened, and force a
> > retreat to secondary biological objections in practice, despite
theoretical
> > possibility.
>
> Certainly GA's demonstrate the creative power of variation + selection.
> So far there seem no limits demonstrated.
One researcher suggests otherwise:
"Genetic algorithms are rather robust search methods for [simple problems]
and small design spaces. But for hard problems and very large design spaces,
designing a good genetic algorithm is very, very difficult. All the
expertise that human engineers would use in confronting a design
problem--their knowledge base, engineering principles, analysis tools,
invention heuristics and common senseSmust be built into the genetic
algorithm. Just as there is no general-purpose engineer, there is no
general-purpose genetic algorithm." Geoffrey Miller
(http://iscid.org/papers/Dembski_WhyNatural_112901.pdf)
He also notes:
"The fitness function must embody not only the engineer's conscious goals,
but also her common sense. This common sense is largely intuitive and
unconscious, so is hard to formalize into an explicit fitness function.
Since genetic algorithm solutions are only as good as the fitness functions
used to evolve them, careful development of appropriate fitness functions
embodying all relevant design constraints, trade-offs and criteria is a key
step in evolutionary engineering."
> > On the other hand, if evolutionary computing, GA problem-solving and
> > information studies etc do reveal such constraints, Darwinism is dead,
and
> > there does not appear to be convincing naturalistic alternative.
> >
> > What is really interesting is that this approach takes the argument out
of
> > the hands of biologists and into the hands of mathematicians and
computing
> > scientists (shock, horror...hallelujah?). Even aside from theoretical
> > analysis, I think at some stage we are going to hear the EC guys saying,
"We
> > have found that in practice you can use GAs to discover solutions to
> > problems of class A, but not class B." (Life as found on planet earth
being
> > decidedly class B.)
>
> Why do you assume that?
It's only my intuition on the subject. The ID approach may be able to
confirm or disconfirm it.
> > Either way, we seem to have a testable hypothesis, which is an essential
> > ingredient of a happy new year.
>
> And a very pleasant new year to you as well.
cheers,
Mark
Mark & Roslyn Elkington
George Acton <gac...@softdisk.com> wrote in message
news:3C34B9...@softdisk.com...
> > Either way, we seem to have a testable hypothesis, which is an essential
> > ingredient of a happy new year.
>
> It sounds like "let's-pretend" science, as if we were going to
> decide some controversy in physics or chemistry by pure calculation,
> without attending to any empirical data. The calculatios are
> interesting but they aren't necessarily informative about the
> correct description of nature. In this case a necessary constraint
> is that we know common descent occurred, and we know the
> aproximate time scale.
We do not "know" that common descent occured. It's just one inference from
the data. Another is design re-use.
You can argue about the relative strength of those hypotheses, but to state
common descent as observed fact is not science, it's a statement of your own
bias and presuppositions.
> This appears to remove much of the
> theological motivation for the investigation.
> I have a problem with calling an empirical claim a "Theorem".
> Theorems are usually published in mathematics journals and are
> are true without needing experimental confirmation. Hypotheses
> about the physical world are usually called um hypotheses.
Dembski is attempting a theoritical approach by your definition, indeed, his
is using the journaled NFL theorem in an attempt to arrive at a greater
theorem (id or super-ego? :-) . The results from GAs in evolutionary
computing may provide empirical support or otherwise for any such theorm he
or others come up with.
rgds,
Mark
> --George Acton
>
Mark & Roslyn Elkington
howard hershey <hers...@indiana.edu> wrote in message
news:a1223n$fqg$1...@flotsam.uits.indiana.edu...
I think we're in general agreement here - my main point was simply to
emphasise the sparsity of solutions within the phase space. Another question
is, What is the gradient and smoothness of the fitness functions within that
space?
Mark
Bobby D. Bryant
> Mark & Roslyn Elkington wrote:
>>>
>> The difference, and it is fundamental, and I wish I didn't have to
>> keep mentioning it, is: we directly observe bees flying, but we do
>> not directly observe megavolution.
>
>
> Is "megavolution" different from speciation, which has been observed
> many times (and links have been mentioned here on talk.origins many,
> many, many times)? Have you directly observed nuclear fusion going on
> within our Sun? After all, the radiation being emitted is all well and
> good, but it isn't a -direct- observation of nuclear fusion. How about
> plate tectonics, or any of the other phenomena that occur over long
> timescales?
The subtle thing you're missing is that Mark is invoking a "complex
specified argument", which immediately falls apart if you try to apply
it to natural phenomena that do not threaten the user's religious beliefs.
This class of argument is extremely popular among creationists, for
reasons that I will leave you to guess.
Bobby Bryant
Austin, Texas
Mark & Roslyn Elkington
> Mark & Roslyn Elkington <mar...@zeta.org.au> wrote:
>
> > John Wilkins <john.w...@bigpond.com> wrote in message
> > news:1f5fmr4.ky4cxw17xs7t8N%john.w...@bigpond.com...
> >
> > > I notice that the thread has a slew of responses. I shall confine
myself
> > > to a few comments.
> > >
> > > Darwinism is way more than mutation + selection. It includes
> > > recombination, drift, punctuation, sexual selection, and a host of
other
> > > things. "Neo-Darwinism" is an anachronistic term used by some as a
badge
> > > of pride and others as a slur. In neither case does it denote anything
> > > much.
> >
> > Agreed, Neo-Darwinism (mutation + selection) is only a subset of the
modern
> > synthesis, but in the origins debate it's the only part of the synthesis
> > worth a penny. What, apart from mutation + selection (sexual is a subset
of
> > natural), can attempt to explain the creation of increasingly ordered,
> > complex, functional forms?
>
> You've already mentioned one - self-organisation (autopoiesis has a much
> nicer ring to it). Things can be functional because the way physical
> objects work results in such functionality (eg, stable orbits,
> convection cells, hydrological cycle, possible Gaianisms). However I do
> think that selection is the proximal cause of most functionality.
>
> Two things - mutation is not the source of most variation.
Strictly speaking mutation is ultimately the source of all variation. Every
single base pair after abiogenesis was added by mutation. I guess I'd call
recombination only a secondary source, though it's largely a definitional
thing, and I take your point.
> Donald
> Campbell used to characterise what I think you are referring to as
> neo-Darwinism as BVSR: blind variation and selective retention. But most
> of that blind variation is recombination of already existing variants.
> Of course mutation is the ultimate source of variation over which
> selection acts, but the dimensionality of extant variation since, say,
> the late Cambrian would be enought o drive a lot of evolution.
>
> Generally mutations are merely excisions or duplications of nucleotides.
> A simple duplication of a sequence can give a high dimensionality of
> variation.
> >
> > >
> > > Kaufmann's ideas are not an *alternative* to NS, any more than the
basic
> > > laws of chemistry or physics are. Any self-organised system still has
to
> > > undergo the rigours of selection and drift. Merely being
self-organised
> > > is no guarantee that the system is viable.
> ....
>
> I leave this in to illustrate the point made 3 paras above.
>
> > > > Well, that's the whole the debate. Can NS in fact work to produce
the
> > set of
> > > > complex cases of living organisms, within the phase space and
> > time/trials
> > > > available? Or can it only produce some mathematically bounded degree
of
> > > > variation, i.e. microevolutionary adaptations? (with usual
disclaimers
> > for
> > > > the use of that problematic terminology).
> > >
> > > OK, let's carefully consider two of the concepts here:
> > >
> > > Phase space - there is no set phase space of any dimensionality other
> > > than the conditions required by the laws of physics. There are no
> > > optimal solution coordinates in any space that organisms *have* to
find.
> > > A phase space is how *we* - primates in lab coats - choose to model
the
> > > system. It exists as a semantic construct only in our heads and our
> > > tools (computers, journals, internet, etc).
> >
> > There is not necessarily any single optimal solution(s) that "organisms
> > *have* to find," agreed. But, the *viable* solutions within the phase
space
> > remain a minute fraction of the total permutations.
>
> In a rugged landscape they may be local to indefinitely many regions of
> the phase space. So a search of a small fraction of the space can find a
> satisficing solution easily in such cases. See Gavrilet's work:
>
> Gavrilets, S. and J. Gravner (1997). "Percolation on the fitness
> hypercube and the evolution of reproductive isolation." J. Theor. Biol.
> 184(1): 51-64.
>
> In a smooth landscape, however, where a peak is distant and abrupt in an
> otherwise ungradiented field, selection will clearly not find the peak
> because neither a random search will find anything, nor will selective
> coefficients overcome the noise of mutation and drift.
Ah, I just posed this exact question in another post. Dembski noted
somewhere the question landscape topology and search convergence, but I
can't recall if he made any claim about how much impact it would have on his
NFL approach.
> >
> > Why do we marvel at a butterfly? Because for every butterfly, there are
a
> > squadrillion other completely uninteresting permutations of the same
bunch
> > of atoms. The butterfly "solution" is so utterly unlikely that a re-run
of
> > evolution would not produce one (cf Gould). Moreover, I contend that the
> > fraction of permutations of those same atoms of which would form a
viable
> > organism of any kind or shape would be infinitesimal. For example,
Yockey
> > estimated that only 1 in10^40 possible ctyochrome c variants would be
> > functional.
>
> Yockey is full of shit.
You're a big fan, eh :-)
> Sorry, but that "calculation" was so biased and
> question begging as to lack intellectual rigor. There are *many*
> cytochrome C variants that are functional in modern organisms. But that
> is not to exhaust the variants that could be slightly less functional,
> or variants in organisms with different metabolic pathways, or variants
> that could fulfil another function until they were coopted for a novel
> function, and so forth. Yockey loads the bases in his favour.
>
> Moreover, that butterflies or humans would not necessarily result (ie,
> specific forms and adaptive strategies might be contingent) is not to
> say that *some* forms would not result. Once you stop restricting the
> outcomes to whatever is a priori "interesting" to modern human
> biologists with a zoological prejudice, the field exapnds enormously.
It is very difficult to estimate obviously, but I maintain that the
solutions according to even the broadest criteria are still very sparse.
rgds,
Mark
Mark & Roslyn Elkington
Not at all.
I have previously cited the Thompson GA result as the subject of another
thread. I did so because it is a strong result, apparently against my
position, and therefore demands attention.
My purpose for citing here was different: to counter with an example of a
close match between a GA and the mechanism of biological evolution.
I make no claims of virtue here, but you seem unwilling to admit a skerrick
of reasonableness.
> WRE> It should
> WRE> also be noted that Bryant's claim is a statistical one, where
> WRE> the evidence to overturn it would have to somehow substantiate
> WRE> that over half of all instances of EC have high biological
> WRE> verisimilitude. I find it highly doubtful that any such
> WRE> demonstration will be forthcoming.
>
> So, will we see Mark fulfill his "Not at all" claim? I bet not.
Fair enough, in the haste of making 101 replies here I overlooked the
statistical nature of the claim. Bobby has first-hand experience in this
field so I'm in not in much of a position to question his claim on that
basis.
Mark
Mark & Roslyn Elkington
> On Thu, 03 Jan 2002 06:29:21 -0600, Mark & Roslyn Elkington wrote:
>
>
> > Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> > news:a0unt8$mai$1...@geraldo.cc.utexas.edu...
>
> >> GAs were "inspired by" biological evolution (or perhaps merely dog
> >> breeding), but they are not generally used to model biological
> >> evolution. Most often they are used to find a "decent" solution for
> >> some specific problem that would require too much time or space to
> >> solve exactly. They do not usually follow the mechanisms of
> >> biological evolution in anything but the most superficial form,
> >
> > Not at all. They can follow it in rather substantial form:
> > http://208.245.156.153/archive/output.cfm?ID=1455
>
> Notice your change of my "they do not usually" to "they can". You're
> arguing past me already. As I said, they _are_not_generally_ used to
> model biological evolution. Which brings us to...
Yep, sorry, my error in haste...
>
>
> >> and of course they
> >> only use a trivial fraction of the resources that biological
> >> evolution has used over the ages.
> >
> > Important distinction: we are checking in-principle limitations, not
> > trying to model all evolution on a computer.
>
> Sure; but my warning is to make sure people don't overgeneralize the
> in-principle limitatons on one kind of system to a similar-but-different
> kind of system where they may not apply.
Okay, fair warning. Indeed, when a GA result comes up opposing my view,
that is my first defence :-)
> It is, of course, for the purpose of in-principle demonstrations that
> makes GAs relevant in this forum at all. For instance, I feel like it
> is fair to invoke them as in-principle demonstrations of the fact that
> blind chance filtered by selection-by-fitness can generate new
> structures or behaviors, a fact that many creationists deny.
>
> As for in-principle limitations, whether they lead to arguments relevant
> for biology must await the identification of the limitations you wish to
> invoke. So far you're merely arguing from the position of "there _may_
> be relevant limitations". I don't dispute that claim, but I have to
> point out that it's useless as a criticism of the ToE unless/until you
> actually have some relevant limitations in hand.
I'm waiting for Dembski and others to find these and fill out the details.
But for an appertiser:
"Genetic algorithms are rather robust search methods for [simple problems]
and small design spaces. But for hard problems and very large design spaces,
designing a good genetic algorithm is very, very difficult. All the
expertise that human engineers would use in confronting a design
problem--their knowledge base, engineering principles, analysis tools,
invention heuristics and common sense--must be built into the genetic
algorithm. Just as there is no general-purpose engineer, there is no
general-purpose genetic algorithm."
"The fitness function must embody not only the engineer's conscious goals,
but also her common sense. This common sense is largely intuitive and
unconscious, so is hard to formalize into an explicit fitness function.
Since genetic algorithm solutions are only as good as the fitness functions
used to evolve them, careful development of appropriate fitness functions
embodying all relevant design constraints, trade-offs and criteria is a key
step in evolutionary engineering."
--Geoffrey Miller, quoted in
http://iscid.org/papers/Dembski_WhyNatural_112901.pdf
>
>
> >> Attempting to apply the NFL to biological evolution immediately
> >> encounters one of the following problems:
> >>
> >> 1) Is biological evolution even a "search problem", in any meaningful
> >> sense? If not, the NFL simply doesn't apply.
> >
> > Most definitely: it's a search of an extremely sparse fitness
> > landscape.
>
> As I mentioned earlier, I'm not sure I agree with that. Calling it "a
> search" seems to import inappropriate teleology into the ToE as a
> _description_ of natural processes.
>
>
> >> I suspect Dembski is going to appeal to the fallacious notion that
> >> humans are the "correct" answer to some search problem, add the
> >> fallacious claim that the "correct" answer could not have been found
> >> by anything but the "best" search algorithm, and then correctly
> >> invoke the NFL to prove that no such "best" algorithm exists. Brings
> >> to mind a parable about building towers on shifting sands...
> >
> > I'll be interested to see what he does in his upcoming book -- advance
> > blurb claims it address these kinds of issues.
>
> Yeah, well, "advance blurb" said the ID crowd were going to give us a
> reliable design inference algorithm, but we're still eagerly waiting for
> it. (A pity about those who held their breaths...)
Overturning the Darwinian paradigm would be the twenty-first century
equivalent of Einstein knocking Newton out of his tree. These things take
time.
rgds,
Mark
Steven Carr
<big skip>
(About http://iscid.org/papers/Dembski_WhyNatural_112901.pdf )
> Ian and I were discussing this over a latte last night. It also seems to
> me that the NFL is a theorem about finding an *optimum* solution in the
> solution space. But selection does not "seek" optimal solutions. It
> seeks "satisficing" solution, to use Herbert Simon's lovely word. All
> that NS has to generate is a solution that is adequate to the task at
> hand and is slightly better than the other alternatives already found.
> In that respect, and also since the solution phase space is not
> specified in advance, the NFL is inappropriate to apply to NS. Dembski
> has overdrawn his case (again).
Let us assume that Dembski can use No Free Lunch to show that natural selection
cannot find optimal solutions. (It is unclear to me whether I also need to
assume that pigs can fly.)
Let us also assume that Dembski believes that the designs in nature are
not optimal
http://www.leaderu.com/offices/dembski/docs/bd-optimal.html gives us reason
to think that this assumption can be justified. For example, the sentence
'Optimal design is perfect design and hence cannot exist except in an idealized
realm
(sometimes called a "Platonic heaven").' is a clue here.
Why then would Dembski think that the failure of natural selection to
produce optimal design is important?
<big skip>
--
Posted via Mailgate.ORG Server - http://www.Mailgate.ORG
John Wilkins
> Mark & Roslyn Elkington wrote:
>
> > Robt Gotschall <hobgo...@worldnet.att.net> wrote in message
> > news:MPG.169e7c4de...@netnews.worldnet.att.net...
> >
> > ....
> >
> >>>If they get their assumptions and model right, then they can potentially
> >>>
> > go
> >
> >>>on prove the Darwinian mechanism either way.
> >>>
> >>Even if you prove that bees can't fly, all you've done is demonstrate
> >>that your assumptions about how bees fly are wrong.
> >>
> >>You have not proven that bees are being held up by tiny little angels.
> >>
> >
> > The difference, and it is fundamental, and I wish I didn't have to keep
> > mentioning it, is: we directly observe bees flying, but we do not directly
> > observe megavolution.
>
>
> Is "megavolution" different from speciation, which has been observed
> many times (and links have been mentioned here on talk.origins many,
> many, many times)? Have you directly observed nuclear fusion going on
> within our Sun? After all, the radiation being emitted is all well and
> good, but it isn't a -direct- observation of nuclear fusion. How about
> plate tectonics, or any of the other phenomena that occur over long
> timescales?
>
It is sometimes claimed that macroevolution is a creationist neologism.
It isn't - but as I pointed out when I wrote the macroevolution FAQ, it
merely applies to any evolution that involves the splitting of a single
species into two or more.
Nowadays we seem to be getting more creationists using terms like
"megaevolution" (a term coined, I believe, to mean large scale
evolutionary changes by GG Simpson in his 1944 Tempo and Mode. It is a
term that never flew, because, as Simpson later realised, it is entirely
subjective - one man's mega is another's macro. But here it is doing
exactly what its critics in the 40s knew it would - applying to any set
of goalpost locations the rhetorician wants it to. It's a Humpty Dumpty
term.
>
> >
> > That's why the talk.bees-cannot-fly ng has never got off the ground, but t.o
> > is flying high.
> >
> >
> >>>>If EWAKI is a genuine theory, it is by definition falsifiable.
> >>>>
> >>>>I assume Darwinism will someday suffer the same fate as Newtonism, and
> >>>>continue to be taught a century afterward.
> >>>>
> >>>It already has, it's just that some are slow to catch on.
> >>>
> >>>rgds,
> >>>Mark
> >>>
> >>But do you know _why_ Newtonian Mechanics is still being taught almost a
> >>century after relativity was introduced?
> >>
> >
> > Yes. It's rather handy and accurate approximation for all but the
> > fastest-moving problems.
> >
> > Mark
> >
> >
> >
John Wilkins
> John Wilkins <john.w...@bigpond.com> wrote in message
> news:1f5gkre.10hi9shwkhjk0N%john.w...@bigpond.com...
....
> >
> > In a smooth landscape, however, where a peak is distant and abrupt in an
> > otherwise ungradiented field, selection will clearly not find the peak
> > because neither a random search will find anything, nor will selective
> > coefficients overcome the noise of mutation and drift.
>
> Ah, I just posed this exact question in another post. Dembski noted
> somewhere the question landscape topology and search convergence, but I
> can't recall if he made any claim about how much impact it would have on his
> NFL approach.
It'll have a lot of impact if he claims that any search method is as
good or better in locating local satisficing "solutions". Landscape
structure is entirely the most important feature of the "solution phase
space of an NS search.
[Wesley - you emailed me on my work account, which I can't easily use at
home, that you thought NS was *not* satisficing; if you read this, can
you say why here?]
....
Derek Stevenson
news:CmXY7.1587$ko4.1...@nasal.pacific.net.au...
> David Jensen <da...@dajensen-family.com> wrote in message
> news:9q773us9q84e27qpp...@4ax.com...
> > On 2 Jan 2002 15:37:08 -0500, in talk.origins
> > "Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in
> > <2bKY7.1546$ko4.1...@nasal.pacific.net.au>:
> > >Evolution is one theory to explain the observation. Darwinism, or
> > >Neo-darwinism are some proposed mechanisms within this theory,
among
> others.
> >
> > What are you talking about? Evolution is change in allele
frequency in
> > populations over time. That is an observation, not a theory.
>
> Yes, by one limited definition. I think though that in the context
of this
> discussion we all understand that the term "evolution" means much,
much more
> than that. Don't we?
You mean, do we understand that when some people use the word
"evolution", what they really mean is "all of the bits of science
which I reject for religious reasons"? Yes, we understand that. We
also understand that it's an inappropriate use of the word.
Gavin Tabor
GA's are developed to find _specific_ solutions. We have at least some
idea as to what we are aiming to produce. I might be interested in
using GA's to optimise the airflow around a racing car, for instance.
If the GA decides to turn the racing car into an aircraft, I'm not
going to be too happy. However biological evolution has no such
constraints - if the best solution for species A is to evolve into
a bird, so be it.
GA's demonstrate that recombination, mutation + selection can generate
new information (however defined) and new functionality (again, however
defined). These are aspects of the theory of biological evolution that
c'ists repeatedly claim that it cannot, or that it is impossible to
prove that it can. Thus, I suggest that it does demonstrate no limits
on what can be achieved.
Gavin
>
> > > On the other hand, if evolutionary computing, GA problem-solving and
> > > information studies etc do reveal such constraints, Darwinism is dead,
> and
> > > there does not appear to be convincing naturalistic alternative.
> > >
> > > What is really interesting is that this approach takes the argument out
> of
> > > the hands of biologists and into the hands of mathematicians and
> computing
> > > scientists (shock, horror...hallelujah?). Even aside from theoretical
> > > analysis, I think at some stage we are going to hear the EC guys saying,
> "We
> > > have found that in practice you can use GAs to discover solutions to
> > > problems of class A, but not class B." (Life as found on planet earth
> being
> > > decidedly class B.)
> >
> > Why do you assume that?
>
> It's only my intuition on the subject. The ID approach may be able to
> confirm or disconfirm it.
>
> > > Either way, we seem to have a testable hypothesis, which is an essential
> > > ingredient of a happy new year.
> >
> > And a very pleasant new year to you as well.
>
> cheers,
> Mark
--
Dr. Gavin Tabor
School of Engineering and Computer Science
Department of Engineering
University of Exeter
Derek Stevenson
> On 3 Jan 2002 13:27:22 -0500, in talk.origins
> rayt...@yahoo.com (Mark VandeWettering) wrote in
> <580c1a4a.02010...@posting.google.com>:
> >"Mark & Roslyn Elkington" <mar...@zeta.org.au> wrote in message
> >> I argue that the same data can equally point to design re-use as
common
> >> descent.
> >
> >This should be a strong signal that design is unfalsifiable.
>
> Or that it was designed by programmers who didn't understand the old
> bugs/features, so they just left the crap in and did a workaround.
So not just multiple designers, but successive generations of isolated
design teams who failed to document their work.
Bobby D. Bryant
> Let us assume that Dembski can use No Free Lunch to show that natural
> selection cannot find optimal solutions. (It is unclear to me whether
> I also need to assume that pigs can fly.)
>
> Let us also assume that Dembski believes that the designs in nature
> are not optimal
>
> http://www.leaderu.com/offices/dembski/docs/bd-optimal.html gives us
> reason to think that this assumption can be justified. For example,
> the sentence 'Optimal design is perfect design and hence cannot exist
> except in an idealized realm
> (sometimes called a "Platonic heaven").' is a clue here.
>
> Why then would Dembski think that the failure of natural selection to
> produce optimal design is important?
I think it's called "say whatever I have to to stay in the ring for one
more round".
Bobby Bryant
Austin, Texas
Derek Stevenson
> Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> news:a12ivg$j8j$1...@geraldo.cc.utexas.edu...
> > On Thu, 03 Jan 2002 06:29:21 -0600, Mark & Roslyn Elkington wrote:
[snip]
> > As for in-principle limitations, whether they lead to arguments
relevant
> > for biology must await the identification of the limitations you
wish to
> > invoke. So far you're merely arguing from the position of "there
_may_
> > be relevant limitations". I don't dispute that claim, but I have
to
> > point out that it's useless as a criticism of the ToE unless/until
you
> > actually have some relevant limitations in hand.
>
> I'm waiting for Dembski and others to find these and fill out the
details.
> But for an appertiser:
>
> "Genetic algorithms are rather robust search methods for [simple
problems]
> and small design spaces.
Which is all that any individual iteration of mutation and natural
selection requires. The problem is simple -- all you need is a slight
advantage over your competitor (be that a predator or another
individual of your own species) -- and the design space is typically
small -- most mutations do not involve a major morphological
reworking.
> But for hard problems and very large design spaces,
> designing a good genetic algorithm is very, very difficult.
Developing a vast, complex, interrelated biosphere would appear to
qualify as a hard problem involving a very large design space, but
that's not the "objective" that mutation and natural selection are
"pursuing". It's merely an effect that emerges out of the reiterated
operation of this algorithm.
> All the
> expertise that human engineers would use in confronting a design
> problem--their knowledge base, engineering principles, analysis
tools,
> invention heuristics and common sense--must be built into the
genetic
> algorithm. Just as there is no general-purpose engineer, there is no
> general-purpose genetic algorithm."
>
> "The fitness function must embody not only the engineer's conscious
goals,
> but also her common sense. This common sense is largely intuitive
and
> unconscious, so is hard to formalize into an explicit fitness
function.
> Since genetic algorithm solutions are only as good as the fitness
functions
> used to evolve them,
Which is one reason why jury-rigged design is seen as evidence of
evolution.
> careful development of appropriate fitness functions
> embodying all relevant design constraints, trade-offs and criteria
is a key
> step in evolutionary engineering."
>
> --Geoffrey Miller, quoted in
> http://iscid.org/papers/Dembski_WhyNatural_112901.pdf
[snip]
> > > I'll be interested to see what he does in his upcoming book --
advance
> > > blurb claims it address these kinds of issues.
> >
> > Yeah, well, "advance blurb" said the ID crowd were going to give
us a
> > reliable design inference algorithm, but we're still eagerly
waiting for
> > it. (A pity about those who held their breaths...)
>
> Overturning the Darwinian paradigm would be the twenty-first century
> equivalent of Einstein knocking Newton out of his tree.
You mean, a fine-tuning of principles that work well enough in most
cases ("near enough for government work", as the saying goes) but may
not always apply in extreme circumstances?
> These things take
> time.
Wesley R. Elsberry
WRE> [...]
WRE> [...]
WRE> That's a commendable attitude to have regarding the evidence.
WRE> However, the evidence of the post indicates that Mark was,
WRE> in fact, citing the Thompson GA research as if it were a
WRE> datum disconfirming Bryant's claim. Note the "Not at all",
WRE> which in ordinary usage indicates opposition rather than
WRE> agreement, and the use of the colon, which indicates that
WRE> the following information is an instance of the general
WRE> claim made. I'm afraid that what Mark has done in his reply
WRE> is also known by a colloquial phrase: "damage control".
ME>Not at all.
Here we go again.
ME>I have previously cited the Thompson GA result as the
ME>subject of another thread. I did so because it is a strong
ME>result, apparently against my position, and therefore
ME>demands attention.
That was then; this is now.
ME>My purpose for citing here was different: to counter with
ME>an example of a close match between a GA and the mechanism
ME>of biological evolution.
Yes, and Mark is *wrong* to do so. Thompson's work does not
provide an instance of what Mark claims here.
ME>I make no claims of virtue here, but you seem unwilling to
ME>admit a skerrick of reasonableness.
I will freely admit that Mark's previous mentioned citation
of Thompson, as a strong result demanding attention, does
demonstrate a degree of reasonableness on Mark's part.
However, that doesn't carry over to the current citation of
Thompson, for it is *false* that Thompson's work has high
biological verisimilitude, and thus provides no counter to
Bryant's claim.
I don't know why Mark should perseverate in trying to claim
that Thompson's work has high biological verisimilitude.
Either Mark doesn't understand Thompson's methods, or he
doesn't understand the biology, or both, or perhaps he's
just confused. But in any case, Mark is wrong on this
point.
[...]
--
Wesley R. Elsberry, Student in Wildlife & Fisheries Sciences, Tx A&M U.
Visit the Online Zoologists page (http://www.onlinezoologists.com)
CNS BBS FTP Archive: ftp://centralneuralsystem.com/pub/CNS/bbs/
"she told me it is merely a plutonic attachment" - archy
Bobby D. Bryant
> Mark & Roslyn Elkington <mar...@zeta.org.au> wrote:
>
>> John Wilkins <john.w...@bigpond.com> wrote in message
>> news:1f5gkre.10hi9shwkhjk0N%john.w...@bigpond.com...
> ....
>> >
>> > In a smooth landscape, however, where a peak is distant and abrupt
>> > in an otherwise ungradiented field, selection will clearly not find
>> > the peak because neither a random search will find anything, nor
>> > will selective coefficients overcome the noise of mutation and
>> > drift.
>>
>> Ah, I just posed this exact question in another post. Dembski noted
>> somewhere the question landscape topology and search convergence, but
>> I can't recall if he made any claim about how much impact it would
>> have on his NFL approach.
>
> It'll have a lot of impact if he claims that any search method is as
> good or better in locating local satisficing "solutions". Landscape
> structure is entirely the most important feature of the "solution
> phase space of an NS search.
That's the whole point of the NFL, at least if I understand it. For any
search algorithm you can design "solution space" landscapes that will
trap it in a region of bad solutions, though some other algorithms will
not fall for those same traps.
For instance, you can fool a simple hill-climbing solution by
surrounding the peak with a ditch that the hill-climber will not
descend. That doesn't mean hill-climbing sux; just that it isn't a
panacea for every possible problem. It happens to work quite well on
*some* problems.
Unfortunately for Dembski (and forgive me for repeating myself),
biological evolution isn't "a plan for finding man", or a plan at all,
let alone the best of possible plans or a plan that is guaranteed to
find the best of possible solutions. The NFL simply doesn't tell us
anything interesting about biological evolution. Dembski has to bend
the NFL *and* the ToE to make it look like they share some common ground
where the one can veto the other. (He's essentially imitating the
strategy of those pseudo-scientists who say the 2LoT vetoes the ToE on
the grounds of "information gain": a little twist here and a little
twist there, and presto! you've got a trainwreck.)
This is an excellent illustration of the importance of peer review. If
I wrote a paper based on such egregios fallacies I'd never get it past
my advisor, let alone past a journal's review panel.
Bobby Bryant
Austin, Texas
TomS
<1f5ic0b.1dy7ylv1pg7blvN%john.w...@bigpond.com>, john.w...@bigpond.com
stated..."
[...snip..]
>It is sometimes claimed that macroevolution is a creationist neologism.
>It isn't - but as I pointed out when I wrote the macroevolution FAQ, it
>merely applies to any evolution that involves the splitting of a single
>species into two or more.
>
>Nowadays we seem to be getting more creationists using terms like
>"megaevolution" (a term coined, I believe, to mean large scale
>evolutionary changes by GG Simpson in his 1944 Tempo and Mode. It is a
>term that never flew, because, as Simpson later realised, it is entirely
>subjective - one man's mega is another's macro. But here it is doing
>exactly what its critics in the 40s knew it would - applying to any set
>of goalpost locations the rhetorician wants it to. It's a Humpty Dumpty
>term.
I haven't noticed the term "megaevolution".
But I'd like to mention "subevolution" to refer to the process
of producing a new individual ... the mixing and mutations involved
that start off the developmental process. I like to mention it,
because nobody (to the best of my knowledge) denies its reality.
But when people give objections to evolution on such grounds as its
randomness, materialism, the lack of values, no mention of the soul
or divine intervention ... they don't object to "subevolution".
Why?
Tom S.
Bob Pease
Mark & Roslyn Elkington <mar...@zeta.org.au> wrote in message
news:e3dZ7.1644$ko4.1...@nasal.pacific.net.au...
I took Physics 511, Celestial Mechanics.
We occasionally used relativistic formulas in extreme cases.
the idea of substituting this as an alternative for Newtonian Mechanics in
Undergrad courses is substituting the needlessly complicated and tedious
for the simple and elegant.
Viva Newton!!
RJ Pease
Dana Tweedy
Ok, what is "Mega" evolution then? If it's "dogs giving birth to cats",
then that is a strawman. Evolution doesn't posit that kind of change. What
exactly do you mean by "mega"?
DJT
Noelie S. Alito
> Noelie S. Alito <noe...@deadspam.com> wrote:
>
> > And if thy oblation be a meat offering baken in a pan, it shall be
> > of fine flour unleavened, mingled with oil. Thou shalt part it in
> > pieces, and pour oil thereon: it is a meat offering. And if thy
> > oblation be a meat offering baken in the fryingpan, it shall be
> > made of fine flour with oil. And thou shalt bring the meat
> > offering that is made of these things unto the LORD: and when
> > it is presented unto the priest, he shall bring it unto the altar.
> > --Lev 2: 5-8
>
> Clearly this is the origin of the High Table tradition...
Oh, not ideas for what to bring to a potluck supper?
> All it lacks is a prescription on the use of sage and thyme, and the
> offering of a fine after dinner port.
Brandy.
Noelie
--
The Lord said to Moses and Aaron, "When a woman has her
regular flow of blood, the impurity of her monthly period will
last seven days, and anyone who touches her will be unclean
till evening." --Lev 15:19
Noelie S. Alito
Not necessarily. Individuals often come back to their [poorly-
documented] work and find they no longer understand it.
Noelie
--
A few weeks of debugging can save you *days* of up-front design.
howard hershey
----------
In article <XjfZ7.1666$ko4.1...@nasal.pacific.net.au>, "Mark & Roslyn
Elkington" <mar...@zeta.org.au> wrote:
>
> Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
> news:a12ivg$j8j$1...@geraldo.cc.utexas.edu...
>> On Thu, 03 Jan 2002 06:29:21 -0600, Mark & Roslyn Elkington wrote:
>>
>>
>> > Bobby D. Bryant <bdbr...@mail.utexas.edu> wrote in message
>> > news:a0unt8$mai$1...@geraldo.cc.utexas.edu...
>>
[snip]
>
> I'm waiting for Dembski and others to find these and fill out the details.
> But for an appertiser:
>
> "Genetic algorithms are rather robust search methods for [simple problems]
> and small design spaces. But for hard problems and very large design spaces,
> designing a good genetic algorithm is very, very difficult.
So what is the problem that the *real* genetic algorithm poses and tries to
solve (if one excludes teleological goals)? My answer would be the problem
of differential survival relative to the currently existing organisms in the
currently existing environment or a nearby open niche. That seems like a
small, rather than large problem and a narrow design space to me. That,
over multiple iterations, one can produce many variants and specializations
using this process working on that problem and its local design space is a
consequence and not a goal.
> All the
> expertise that human engineers would use in confronting a design
> problem--their knowledge base, engineering principles, analysis tools,
> invention heuristics and common sense--must be built into the genetic
> algorithm. Just as there is no general-purpose engineer, there is no
> general-purpose genetic algorithm."
Only for teleological (an _a priori_ defined single optimal) solution to a
large problem. And evolution did not *intend* to create humans or any other
organism from the common ancestor. That does not mean that one cannot
create such organisms using the process. *If* one starts off with the
teleological *goal* of creating a "human", say, with an evolutionary
algorithm, one must indeed specify many features in order to constrain the
process. With such constraints one could produce a "human" most of the
time. Without such constraints it is unlikely that one would produce a
"human" although one *might* produce something (perhaps an intelligent,
technologically oriented octopus-like organism) that fills the "human"
niche, once such a niche exists. Or one *might* never produce anything
other than clonal bacteria producing stromatolites. There is no evidence at
all to support the idea that "humans" were the teleological goal of
evolution, so designing a genetic algorithm to produce "humans" is not what
evolution did. The *real* genetic algorithm has only the local goals
described above: local advantage in local environments. The fact that there
are many local and unstable environments (including new ones generated by
existing organisms) is what ensures diversity of result.
>
> "The fitness function must embody not only the engineer's conscious goals,
> but also her common sense. This common sense is largely intuitive and
> unconscious, so is hard to formalize into an explicit fitness function.
> Since genetic algorithm solutions are only as good as the fitness functions
> used to evolve them, careful development of appropriate fitness functions
> embodying all relevant design constraints, trade-offs and criteria is a key
> step in evolutionary engineering."
>
> --Geoffrey Miller, quoted in
> http://iscid.org/papers/Dembski_WhyNatural_112901.pdf
>
>>
>>
>> >> Attempting to apply the NFL to biological evolution immediately
>> >> encounters one of the following problems:
>> >>
>> >> 1) Is biological evolution even a "search problem", in any meaningful
>> >> sense? If not, the NFL simply doesn't apply.
>> >
>> > Most definitely: it's a search of an extremely sparse fitness
>> > landscape.
It is a search of a *local* fitness landscape which is not, generally
speaking, uniform across its entire expanse or the same at all times. [This
means that variance *within* a population is favored. This pre-existing
variance can then be exploited to track environmental changes or to adapt to
conditions on the "fringe" of the niche that would not be useful in the core
of the niche.] The only goal is fitness relative to others in the local
environment, not some teleological end point.
>>
>> As I mentioned earlier, I'm not sure I agree with that. Calling it "a
>> search" seems to import inappropriate teleology into the ToE as a
>> _description_ of natural processes.
>>
It is certainly not a teleological search with its "one correct answer".
>>
>> >> I suspect Dembski is going to appeal to the fallacious notion that
>> >> humans are the "correct" answer to some search problem, add the
>> >> fallacious claim that the "correct" answer could not have been found
>> >> by anything but the "best" search algorithm, and then correctly
>> >> invoke the NFL to prove that no such "best" algorithm exists. Brings
>> >> to mind a parable about building towers on shifting sands...
>> >
>> > I'll be interested to see what he does in his upcoming book -- advance
>> > blurb claims it address these kinds of issues.
>>
>> Yeah, well, "advance blurb" said the ID crowd were going to give us a
>> reliable design inference algorithm, but we're still eagerly waiting for
>> it. (A pity about those who held their breaths...)
>
> Overturning the Darwinian paradigm would be the twenty-first century
> equivalent of Einstein knocking Newton out of his tree. These things take
> time.
And sometimes they don't happen at all. Einstein merely presented a broader
explanation that subsumed the special case that Newton explained quite well
(which is why Newtonian mechanics is still useful). Dembski has the much
harder task of completely destroying and replacing current explanation that
explains observation quite well (consistently with the observable mechanisms
and processes proposed) and Dembski must do so without actually presenting
any material evidence whatsoever to support the material existence of his
proposed mechanism (saying that a something did something that somehow
produced all current life exactly the way it currently looks is not exactly
a useful explanation capable of being tested or falsified).
>
> rgds,
> Mark
>
>
howard hershey
----------
In article <5JeZ7.1659$ko4.1...@nasal.pacific.net.au>, "Mark & Roslyn
Elkington" <mar...@zeta.org.au> wrote:
>
> howard hershey <hers...@indiana.edu> wrote in message
> news:a1223n$fqg$1...@flotsam.uits.indiana.edu...
>>
>>
>> ----------
>> In article <yRXY7.1588$ko4.1...@nasal.pacific.net.au>, "Mark & Roslyn
>> Elkington" <mar...@zeta.org.au> wrote:
>>
>>
>> >
>> > John Wilkins <john.w...@bigpond.com> wrote in message
>> > news:1f5fmr4.ky4cxw17xs7t8N%john.w...@bigpond.com...
>> >
>> [snip]
>> >>
>> >> OK, let's carefully consider two of the concepts here:
>> >>
>> >> Phase space - there is no set phase space of any dimensionality other
>> >> than the conditions required by the laws of physics. There are no
>> >> optimal solution coordinates in any space that organisms *have* to
> find.
>> >> A phase space is how *we* - primates in lab coats - choose to model the
>> >> system. It exists as a semantic construct only in our heads and our
>> >> tools (computers, journals, internet, etc).
>> >
>> > There is not necessarily any single optimal solution(s) that "organisms
>> > *have* to find," agreed. But, the *viable* solutions within the phase
> space
>> > remain a minute fraction of the total permutations.
>>
>> That depends upon the relevant phase space. For a butterfly, the relevant
>> phase space that it will or can explore will be on the fringes of its
>> current environment. A butterfly will not explore the phase space of a
>> mussel.
>> >
>> > Why do we marvel at a butterfly? Because for every butterfly, there are
> a
>> > squadrillion other completely uninteresting permutations of the same
> bunch
>> > of atoms.
>>
>> And the mechanism of self-reproduction (imperfect self-reproduction is a
>> requisite assumption of evolution) ensures that most of the next
>> generation's genetic sequences are repeats of this generations. And
>> permutations that do occur always undergo a selective filter that both
> weeds
>> out deleterious changes and retains "interesting" ones that allow
> exploring
>> unoccupied nearby phase space.
>>
>> > The butterfly "solution" is so utterly unlikely that a re-run of
>> > evolution would not produce one (cf Gould).
>>
>> Agreed. But evolution does not work by changing every one of the
>> butterfly's atoms to produce a different species of butterfly or to
>> magically convert the butterfly so that it can explore the phase space of
> a
>> mussel. It does change the three genes that are responsible for all the
>> different wing patterns. Evolution is a much more stepwise process than
> one
>> that magically converts one organism to another. At some point these
>> changes might lead to a creature that we would no longer "name" butterfly,
>> just as there is an ancestral species we also would no longer "name"
>> butterfly.
>>
>> > Moreover, I contend that the
>> > fraction of permutations of those same atoms of which would form a
> viable
>> > organism of any kind or shape would be infinitesimal. For example,
> Yockey
>> > estimated that only 1 in10^40 possible ctyochrome c variants would be
>> > functional.
>>
>> At what level of "function"? The biosphere has certainly explored many
> (if
>> not most) of the cytochrome c variants that do function at a high level of
>> function. But it did so in a stepwise fashion with filtering at each step
>> to retain or locally improve function. Most of the changes we see are
>> selectively neutral changes.
>
> I think we're in general agreement here - my main point was simply to
> emphasise the sparsity of solutions within the phase space.
The relevant phase space is whatever changes can be produced in whatever
sequence that organisms alive at that time had that eventually led to
current cytochrome c's. Remember that the *real* genetic algorithm only
requires that a variant be better than what currently existed at that time,
not be better (or even as good as) current cytochrome c's. And the original
cyt c need not necessarily have served the same function as the current
enzyme. But there does have to be a *detectable* (by nature) difference in
order for selection to choose the more favorable variant. Cyt c is a very
ancient protein, so it is hard to determine what the ancestor was and what
function it performed wrt binding heme moities and their metal ions. Other
proteins (such as hemoglobin) have a more clearly identifiable familial
ancestry (vertebrate hemoglobin is a modified myoglobin and beta globin is a
modified alpha globin). Of course, this involves duplication of genes. I
do not think that the sparsity of *all possible* amino acid sequences was
ever relevant to the evolution of cyt c. It was not poofed into existence
in one swell foop with the same function and activity of modern cyt c's. It
*evolved*. Stepwise. And each 'step' was useful in its own right.
> Another question
> is, What is the gradient and smoothness of the fitness functions within that
> space?
Likely not smooth.
>
> Mark
>
>
howard hershey
----------
In article <M0eZ7.1653$ko4.1...@nasal.pacific.net.au>, "Mark & Roslyn
Elkington" <mar...@zeta.org.au> wrote:
>
> Michael Altarriba <mik...@jps.net> wrote in message
> news:3C35629C...@jps.net...
>>
>>
>> Mark & Roslyn Elkington wrote:
>>
>> > Robt Gotschall <hobgo...@worldnet.att.net> wrote in message
>> > news:MPG.169e7c4de...@netnews.worldnet.att.net...
>> >
>> > ....
>> >
>> >>>If they get their assumptions and model right, then they can
> potentially
>> >>>
>> > go
>> >
>> >>>on prove the Darwinian mechanism either way.
>> >>>
>> >>Even if you prove that bees can't fly, all you've done is demonstrate
>> >>that your assumptions about how bees fly are wrong.
>> >>
>> >>You have not proven that bees are being held up by tiny little angels.
>> >>
>> >
>> > The difference, and it is fundamental, and I wish I didn't have to keep
>> > mentioning it, is: we directly observe bees flying, but we do not
> directly
>> > observe megavolution.
>>
>>
>> Is "megavolution" different from speciation, which has been observed
>> many times (and links have been mentioned here on talk.origins many,
>> many, many times)?
>
> No.
>
>> Have you directly observed nuclear fusion going on
>> within our Sun? After all, the radiation being emitted is all well and
>> good, but it isn't a -direct- observation of nuclear fusion. How about
>> plate tectonics, or any of the other phenomena that occur over long
>> timescales?
>
> I didn't say that science by inference was invalid.
>
So what is it about the evolutionary inference of common descent from
microevolution (and speciation) by mutation/selection (and neutral drift,
founder effects, etc.) that you have a problem with? Not enough time? That
there are limits to the possible variations?
[snip]
>