How does PCT explain where our basic characteristics come from?

[_Boris_Hartman]

Hi Bill,

BP : When I read Koshland’s book on chemotaxis in E. coli, I found there another way to do reorganization which still relied on random changes rather than intelligent direction, but which was far more efficient – by a measured factor of about 70 in
one demonstration, and by much larger factors when multiple parameters were being changed at the same time and multiple dimensions of error were involved. While the initial model followed Ashby in most details, the new one brought in a principle that Ashby didn’t know about.

In the E. coli algorithm, parameter changes happen continuously at a rate determined by the amount of error in the system being reorganized. This corresponds to E. coli swimming in a straight line at some speed and some angle to the gradient of attractant or repellent.

BH : As we know I have troubles understanding your terminology for a long time. Do I understood right that you described some suitable environment of chemical substances and gradients as input (atractor and repellient) to E.coli. Further it seems to me that you described intermediate proces inside bacteria E. coli which you called “reorganization” and have some effect on behavior of E.coli. Is this right ? How “error in the system being reorganized” is presented ? What’s causing “error” ? What is excatly meant by reorganization ? What does reorganization do in the cell of E.Coli ?

BP : But I propose that the E. coli algorithm is enough.

BH : Is it possible to be more explicit in definition of algorithm. Can you put it in graphical presentation ?

I’ll try to help myself with “diagram of immediate effects” :

ENVIRONMENT – affect → INPUT (atractor or repellient) – affect → REORGANIZATION in E.coli – affect → OUTPUT (away or to substance) – affect → ENVIRONMENT…etc.

If this is right, I’m interested how this explains Rick’s statements :

1. When you get your copy of B:CP you will see that an important assumption of the theory is that organisms are born with a set of “intrinsic references”. These references are inherited and basically unchangeable.

2. Intrinsic reference are reference specifications for the values of intrinsic variables.

Best,

Boris

···

----- Original Message -----

From:
Bill Powers

To: CSGNET@LISTSERV.ILLINOIS.EDU

Sent: Saturday, December 10, 2011 5:59 PM

Subject: Re: How does PCT explain where our basic characteristics come from?

[From Bill Powers (2011.12.10.0857 MST)]

At 11:09 PM 12/9/2011 +0100, Boris Hartman wrote:

BH:  So it doesn't matter to me whether we are observing "E-colli" or any other single cell organism or an organism with billions of cells as I think the model should "cover" all living beings.

BP: It not E. coli that matters, but the principle that we are shown by the way it is able to move up and down gradients of attractants and repellants by randomly tumbling. A very systematic result is controlled through random variations of output. That’s what is important about what we learned from E. coli.
In the normal concept of random trial and error, or in genetics, mutation, the random effects cause step-changes in system parameters, each change being unrelated to previous or following changes. That method has some probability of finding a beneficial change. But not a very high probability, and the probability rapidly decreases as more kinds of errors are being increased at the same time by selection pressures.
That was my original concept of reorganization described in the 1960 paper, and it followed Ashby’s idea. It seemed too inefficient, and very little was done with reorganization for 20 years or so because of that.
When I read Koshland’s book on chemotaxis in E. coli, I found there another way to do reorganization which still relied on random changes rather than intelligent direction, but which was far more efficient – by a measured factor of about 70 in one demonstration, and by much larger factors when multiple parameters were being changed at the same time and multiple dimensions of error were involved. While the initial model followed Ashby in most details, the new one brought in a principle that Ashby didn’t know about.
In the E. coli algorithm, parameter changes happen continuously at a rate determined by the amount of error in the system being reorganized. This corresponds to E. coli swimming in a straight line at some speed and some angle to the gradient of attractant or repellent.
As long as the error being monitored by the reorganizing system is getting smaller, nothing else happens. The parameters simply go on changing, more slowly as the error gets smaller. I think this is what geneticists call “genetic drift.” But inevitably, the optimum point will be passed and the error will start increasing. E. coli swims past the point where it is closest to the source of the attractant. That’s when E. coli “tumbles” and randomly changes the direction in which it swims. In the reorganization model, the rates of change of the parameters being reorganized are randomly reset to new values, so now they are changing in new proportions relative to each other.
If the error is still increasing, another tumble occurs but eventually, if the system doesn’t crash first, the error starts to decrease again. Obviously this will continue until the remaining error is too small to drive the steady changes in parameters.
I’m explaining this again partly to make my presentations better, but partly to show that there is nothing special about E. coli other than the new principle of reorganization suggested by its method of locomotion. That is a vast improvement over any method that simply changes parameters directly at random.
This is important because certain objections to the idea of evolution and natural selection are based on calculations showing that the probabilties of the observed changes are simply too low to make that theory believable. And it is unbelievable: I have commented before on the way models of evolution like the “genetic algorithm” cheat by allowing organisms to survive that merely change in the right direction rather than getting all the way to the new organization actually needed for survival – they are given credit for moving a bit more in the direction of food, rather than actually getting to the food. Such models are actually given a critical amount of help by the intelligent programmer, who knows in advance what kind of change would be beneficial. The genetic algorithm, as used, is an instance of intelligent design.
With the E. coli algorithm, we now have a way for the organism to evolve all by itself so as to get to the food and eat it. At the same time, we do away with the old definition of “fitness,” which is merely survival to the age of reproduction. By that definition, the most fit human beings are successful serial rapists. Now we can say simply that the most fit organisms are those that keep their internal error signals the smallest – and, of course, who also survive long enough to reproduce to a reasonable degree and get along with their contemporaries while not devastating the resources they need. Natural selection does occur; it’s simply not enough by itself to account for evolution.
But I propose that the E. coli algorithm is
enough.

Best,

Bill P.

[Oliver]

[Oliver Schauman(2011.12.12.2310)]

See my comments below on evolution.

I highly recommend reading “The Greatest Show on Earth” by Richard Dawkins. In addition to being a vocal atheist he is first and foremost a brilliant evolutionary biologist. The book has many bits on evolution and computer modelling that might be very relevant to this topic. I might summarise them in a post later on. In the mean time, do check out “Boids” (http://www.red3d.com/cwr/boids/)

[From Rick Marken (2011.12.11.1030)]

Bill Powers (2011.12.10.0857 MST)
BP: It not E. coli that matters, but the principle that we are shown by the
way it is able to move up and down gradients of attractants and repellants
by randomly tumbling. A very systematic result is controlled through random
variations of output. That’s what is important about what we learned from E.
coli…
BP: This is important because certain objections to the idea of evolution and
natural selection are based on calculations showing that the probabilities of
the observed changes are simply too low to make that theory believable.
RM: This led me to realize that there might be another problem with 'Darwin’s Hammer" type natural selection besides the improbability of random changes getting to a “solution”. It’s the problem of staying at the solution once you get there.
Isn’t a “solution” in evolutionary terms is something that makes it more likely that the given gene is passed on? The way this gene would remain in the population (and thus it would seem as if the population is “staying at the solution”) is if it increased the likelihood of reproduction over other variants of that gene in the population.

Random changes getting to a solution is indeed improbable. People who give lectures on intelligent design tend to ask the audience the question whether it is more probable that millions of years of wind and erosion carved out Mount Rushmore OR that someone carved them there. This seems like the kind of argument about “Darwin’s hammer” you are referring to.

The changes that occur in a population over time are however NOT random. Evolution functions through non-random selection of randomly varying characteristics. There is nothing improbable about that. If you look at the variation in modern day dogs, which all have the wolf as a common ancestor very recently in evolutionary time, the power of this process becomes clear.

RM: If random mutations are always occurring at a constant rate in a population then it seems to me that the population is just as unlikely to remain in the “evolved” state as it was to get there in the first place. That is, I don’t think the constant mutation rate model would produce the “punctuated” fossil record that is actually observed. Instead, there would be continuous drifting from one form to another. (I know as well as anyone that God put those fossils there just to test our faith but it’s still fun to play with Godless models of this observation while everyone else is in church).
A “punctuated” fossil record is something that taxonomists have created. A continuous drifting from one form to another is exactly what we find in a given population and it is a founding principle why evolution by natural selection works. It is these drifts that are either selected or not for survival. However, a single mutation does not produce a new species or necessary provide a new “solution”. It would just slightly tweak something like how a given protein pumps in a cell would function or how some enzyme might catalyse a given chemical reaction. If another mutation occurred that made reproduction even more probably, that gene would eventually take over in the population. If it didn’t it would disappear after certain amount of time.

Remember that the mutated offspring of a given species would always be classified as the same species as its mother. The “punctuations” between species is something that is done to distinguish individuals that are a million+ years apart. Indeed, if we had a continuous fossil record it would be impossible to classify any animals into discrete species.

RM: I think I can demonstrate that this is the case using a variant of my “selection of consequences” demo at:

http://www.mindreadings.com/ControlDemo/Select.html

The Reinforcement button runs a model of dot movement that is equivalent to the natural selection model. The Control button runs a model of dot movement that is equivalent to the E. coli reorganization model of natural selection. The demo stops when the dot reaches the target or when the simulation has run for a threshold amount of iterations. The Control model always gets the dot to the target; the Reinforcement model rarely does, but sometimes the dot does get to the target by chance.

But if I let the simulation keep running indefinitely, the Control model would keep the dot on the target once it got it there for as long as the simulation ran; the Reinforcement model might eventually get the dot to hit the target but the dot would soon drift away from the target if the simulation kept running.

The evolutionary analog is that the target is an environmental niche to which the population (dot) is adapting. The changes in direction of movement of the dot are the phenotypic changes resulting from genetic mutation. The success of a mutation is measured in terms of the resulting direction of movement of the dot relative to the target; a “good” mutation is one which results in the dot moving toward the target. In the Reinforcement model the probability of a particular movement direction (mutation) becomes more probable if it produces a “good” result (survives longer). In the Control (E. coli) model the probability of a mutation (change in direction) decreases in proportion to whether the current direction is moving the dot closer or farther from the target.

I’m not sure that the Reinforcement model is a good analog of natural selection (Skinner thought it was) but I do think it would be interesting to see how existing models of natural selection do (in terms of keeping a population “adapted” to its niche) once the evolutionary process has brought the population to the adapted state. When I get a chance I’ll try to extend my “selection of consequences” demo to show that the most effective way to stabilize a result of random variation is via selection retention based on a control type (E. coli) process. But maybe evolutionary simulations already exist that show this. There must be some on the net. Does anyone know of any good ones?

I am finding it difficult to understand why one needs to construct an “analog of evolution”. To me it is a beautiful and accurate model of itself.

However, I think it is definitely worthwhile applying the principles of control to evolution , but I do not think it would be valid to do this at a population level. Here is an idea: It would be interesting to have a population of E.coli with variations in their reorganizing systems. Some would tumble a lot and some very rarely, depending on their threshold (the kind of thing a given gene might code for). The E.coli that get to the target would reproduce and their offspring would appear at random places on the grid. If the “reproduction zone” can only be occupied by a given number of E.coli at a time, the environment will favour the gene that codes for tumbling that gets you to the target fastest. This could then me be compared to a model where the E.coli function through reinforcement.

Regards,

Oliver Schauman

[Bill_Powers1]

[From Bill Powers (2011.12.12.1520 MST)]

Hi Bill,

BH : As we know I have troubles understanding your terminology for a long
time. Do I understood right that you described some suitable environment
of chemical substances and gradients as input (atractor and repellient)
to E.coli. Further it seems to me that you described intermediate proces
inside bacteria E. coli which you called “reorganization” and
have some effect on behavior of E.coli. Is this right?

BP: As far as I know, E. coli is incapable of reorganizing, but even if
it can do this, the behavior of E. coli that inspired what I call the E.
coli algorithm has nothing to do with E. coli’s own ability to
reorganize. It is a PRINCIPLE that I am talking about, a principle that
allows a control system to control a variable by using a kind of behavior
that is basically random. E. coli controls the rate of change of
concentration of several substances it senses around it. It does so by
detecting whether the rate of change of concentration is positive or
negative (the internal biochemistry of this has been verified). It can do
this both to avoid unwanted substances, and to seek substances it wants
– Koshland said 27 different substances. If the rate of change is in the
wrong direction (relative to some internal reference signal), E. coli
does not just turn and go in the right direction. Apparently it can’t do
that. All it can do is change its direction of swimming at random.
Koshland noted that the randomness has been measured and verified: the
new direction is not related to the previous direction or to the
direction of the gradient. Nevertheless, E. coli can control the rate of
change of concentration that it experiences as it swims.

From this principle of behavior, I extracted a new way for
reorganization to occur that does not require any knowledge about the
environment. Its principle method is random. This means that
reorganization can happen before any intelligence has appeared, and in
any organism whether simple or complex. I thought that was important,
because we have to try to account for how all these levels of
organization arise that permit intelligent purposeful behavior – but
before the final product has appeared. And it has to do so in any
unpredictable environment.

The swimming of E. coli changes the organism’s X, Y, and Z coordinates in
space at different constant rates. This corresponds to a process in a
control system that changes three parameters of a control system – for
example, its reference signal, its output gain, and its time constant –
at three different constant rates. Or three input weights, or three
output weights.

For E. coli, the swimming affects the rate of change of some chemical
concentration that it senses. Furthermore, E. coli has an intrinsic
reference signal for this rate: the reference signal specifies that it
should be positive (for “good” substances) or negative (for
“bad” ones).

In the general control system, the changes in parameters affect how close
to some reference level the control system is keeping some variable. An
implicit reference level is for the error in the control system to be
zero, but any intrinsic variable affected by the control system can be
the basis of reorganization, as long as it can be compared with some
reference condition and the error can be detected. If the error is
decreasing, the changes simply continue. If it increases, a random change
of “direction” of changes in parameter space occurs,
corresponding to E. coli’s “tumbles.”

You can experience this with Demo 7-1 in LCS3, although you will be
imitating E. coli instead of changing parameters in some control system.
Other demos work by changing parameters – more specifically, input and
output weighting factors.

BH: How “error in the
system being reorganized” is presented ? What’s causing
“error” ?

BP: The error is simply the difference between the current reference
signal and the perception of the controlled variable, in the control
system that is being acted upon by the reorganizing system. If the
control system is not already organized perfectly, it will not be keeping
this difference or error as small as possible. If the controlled variable
is important, say the average rate of food intake to the body, this error
will set the reorganizing system into motion (that’s the basic postulate
of reorganization theory). It will start to change the parameters of the
control system, and that will change the control system’s ability to keep
the food input close to the (intrinsic) reference level. If the change
reduces the error, the changes will simply continue, slowing as the error
gets smaller. If the initial change increases the error (which is just as
likely, since the changes are being directed at random) another
“tumble” will occur right away and a new combination of rates
of change of parameters will be generated.

BH: What is excatly meant by
reorganization ? What does reorganization do in the cell of E.Coli
?

BP: Again, this is not about reorganization in the cell of E. coli. It is
about a principle that permits order to be created by a random and
therefore unintelligent and non-rational process. You have to understand
what it means to say that the parameters of a control system are being
changed by the reorganizing system. The amount of perception resulting
from a given input is changing. The comparison of reference and
perception is changing sensitivity. The gain of the output function is
changing. The weighting of the various effects of the output function on
lower-level reference settings is changing. These changes are being
caused by error, by stress in the system, and they will stop when the
error is corrected, when the stress drops below some threshold.

But it is not reorganization in E. coli that I am talking about. In E.
coli, we are seeing only the behavior by which this organism finds its
way up gradients of attractants and down gradients of repelllants. The
organization by which it does this is not changing; this is a fixed
behavior pattern.

BP earlier : But I propose that
the E. coli algorithm is enough.

BH : Is it possible to be more explicit in definition of algorithm. Can
you put it in graphical presentation ?

BP:I don’t see how I can make it any plainer than it is in Living Control
Systems III. The source code is given in Demo 8-1, in which 14 control
systems are simultaneously being reorganized, a total of 196 weights
being altered at random relative rates.

BH: I’ll try to help myself with
“diagram of immediate effects” :

ENVIRONMENT – affect → INPUT (atractor or repellient) – affect
→ REORGANIZATION in E.coli – affect → OUTPUT (away or to
substance) – affect → ENVIRONMENT…etc.

BP: It is not the input of attractor or repellent concentration that
affects reorganization, but the SENSED RATE OF CHANGE of that input in
relation to a reference rate of change. The link between reorganization
and output is not simply to make the output go toward or away from the
source of the substance; at that stage the relationship is random, and a
reorganization – that is, a random change in direction, not the steady
drift – is just as likely to increase the error as to reduce it. Ashby’s
diagram of immediate effects is useless here because the immediate effect
can be either unfavorable or favorable on successive trials. As I said
earlier, Ashby didn’t happen to come across the E. coli principle of
reorganization, even through his homeostat had a semi-random sort of
change mechanism in it in the form of stepper switches. His stepper
switches simply cycled systematically through a fixed sequence of
possible connections, stopping when a connection was found that reduced
error. That clever idea was the start of my ideas about reorganization.
But Ashby never saw that what was really needed was a random
stepper switch.

BH: If this is right, I’m
interested how this explains Rick’s statements :

1. When you get your copy of B:CP you will see that an important
   assumption of the theory is that organisms are born with a set of
   “intrinsic references”. These references are inherited and
   basically unchangeable.

2. Intrinsic reference are reference specifications for the values of
   intrinsic variables.

BP:His statement 1 refers to the need for a reference condition relative
to which the intrinsic error can be judged. The most basic reference
conditions are relative unchangeable except through evolution. That error
is what causes reorganization to start or continue, by hypothesis. His
statement 2 points to the controlled variables in the reorganization
loop: the values of intrinsic variables. When the intrinsic variables
match their reference conditions closely enough, reorganization
stops.

At this point I don’t know where to go next. I obviously don’t see what
it is about reorganization theory that you don’t understand, unless I
accidentally found it this time.

Best,

Bill P.

···

At 07:07 PM 12/12/2011 +0100, Boris Hartman wrote:

[Bill_Powers1]

[From Bill Powers (2011.12.12.1650 MST)]

[Oliver
Schauman(2011.12.12.2310)]
See my comments below on evolution.
I highly recommend reading “The Greatest Show on Earth” by Richard
Dawkins. In addition to being a vocal atheist he is first and foremost a
brilliant evolutionary biologist. The book has many bits on
evolution and computer modelling that might be very relevant to
this topic. I might summarise them in a post later on. In the mean time,
do check out “Boids”

BP: I have, and mentioned it recently on CSGnet. Boids works most in
terms of rules about how to response to certain stimuli, like distance to
the center of the flock. The critical variables are, of course,
controlled perceptions but they aren’t handled that way.

OS:
(
http://www.red3d.com/cwr/boids/)
Isn’t a “solution” in evolutionary terms is something that makes it
more likely that the given gene is passed on? The way this gene would
remain in the population (and thus it would seem as if the population is
“staying at the solution”) is if it increased the likelihood of
reproduction over other variants of that gene in the
population.

BP: I would give that a different slant, by saying that a solution is a
reorganization that reduces the chances that a given gene will fail to
reproduce accurately. Mutations, I am proposing, are actually E.
coli-style tumbles and are most initiated by the organism at the level of
DNA. As Rick Marken has just pointed out, the standard concept of
mutation and natural selection can’t explain why continued mutation
doesn’t lead to a random walk among all the alternative outcomes with
equal effects on fitness. The reorganization principle does explain that.

OS: Random changes getting to
a solution is indeed improbable. People who give lectures on intelligent
design tend to ask the audience the question whether it is more probable
that millions of years of wind and erosion carved out Mount Rushmore OR
that someone carved them there. This seems like the kind of argument
about “Darwin’s hammer” you are referring to.
OS: The changes that occur in a
population over time are however NOT random. Evolution functions through
non-random selection of randomly varying characteristics. There is
nothing improbable about that. If you look at the variation in modern day
dogs, which all have the wolf as a common ancestor very recently in
evolutionary time, the power of this process becomes clear.

Yes, that’s it.

···

At 11:30 PM 12/12/2011 +0000, Oliver Schauman wrote:
**

Non-random changes can still occur in the E. coli manner by trial and
error. The supposed selection process does not actually select for
specific changes of organization; it selects for favorable outcomes of
changes in organization – primarily survival to reproduce, but I think
other definitions of fitness are better.

RM (Rick marken earlier): If
random mutations are always occurring at a constant rate in a population
then it seems to me that the population is just as unlikely to remain in
the “evolved” state as it was to get there in the first place.
That is, I don’t think the constant mutation rate model would produce the
“punctuated” fossil record that is actually observed. Instead,
there would be continuous drifting from one form to another. (I know as
well as anyone that God put those fossils there just to test our faith
but it’s still fun to play with Godless models of this observation while
everyone else is in church).
OS: A “punctuated” fossil record is something that taxonomists have
created. A continuous drifting from one form to another is exactly what
we find in a given population and it is a founding principle why
evolution by natural selection works. It is these drifts that are either
selected or not for survival.

BP: Be careful. The drifts are not selected for; only
survival is selected for under the principle of natural selection.
That means that all drifts that have equal effects on survival have the
same chance of being selected for. There is not just one solution for a
given problem; there are multitudes, as the variety of species and
subspecies attests.

I prefer to cast fitness measures in terms of successful control, which
is easy to define: minimum overall error or perhaps intrinsic error. If
increasing the number of members of a species will lead to less error,
that is the condition that will result from reorganization. But if
decreasing the number of members will make errors even smaller, that will
be the outcome. Increasing the population is not necessarily a
pro-survival strategy.

OS: However, a single
mutation does not produce a new species or necessary provide a new
“solution”. It would just slightly tweak something like how a given
protein pumps in a cell would function or how some enzyme might catalyse
a given chemical reaction. If another mutation occurred that made
reproduction even more probably, that gene would eventually take over in
the population. If it didn’t it would disappear after certain amount of
time.
OS: Remember that the mutated
offspring of a given species would always be classified as the same
species as its mother.

BP: All that depends on the hypothesis that a larger population is
better, which I contest.
**

BP: Like the offspring of a horse and a donkey? I thought that mutual
fertility was the main mark of being cospecific.

OS: he “punctuations” between
species is something that is done to distinguish individuals that are a
million+ years apart. Indeed, if we had a continuous fossil record it
would be impossible to classify any animals into discrete
species.

BP: I think you’re right about that.
*
…*

OS: I am finding it
difficult to understand why one needs to construct an “analog of
evolution”. To me it is a beautiful and accurate model of itself.

BP: But it is insufficiently effective, and evolutionary models that I
have seen mostly cheat by giving credit for a move toward survival
instead of actually getting there. The theory behind the phenomenon of
evolution is not a very good one – it takes a lot of suspension of
disbelief. It shows a reluctance to exhibit the necessary amount of
skepticism, perhaps for fear of being mistaken for a proponent of the
other side.

OS: However, I think it is
definitely worthwhile applying the principles of control to evolution ,
but I do not think it would be valid to do this at a population level.
Here is an idea: It would be interesting to have a population of
E.coli with variations in their reorganizing systems. Some would tumble a
lot and some very rarely, depending on their threshold (the kind of thing
a given gene might code for). The E.coli that get to the target would
reproduce and their offspring would appear at random places on the grid.
If the “reproduction zone” can only be occupied by a given number of
E.coli at a time, the environment will favour the gene that codes for
tumbling that gets you to the target fastest. This could then b be
compared to a model where the E.coli function through reinforcement.

BP: Yes, that would definitely be a good project. However, by the
basic design of the theory of reorganization, offspring do not appear at
random positions on the grid – they appear at only slightly different
positions, but perhaps in random directions relative to the starting
point. Genetic drift, it’s called. A mutuation, I have read, is known to
change the direction of genetic drift, which sounds an awful lot like the
E. coli reorganization principle.

If you play the Demo 7-1 from LCS3, you’ll see that varying the rate of
tumbling doesn’t by itself speed up the approach to the goal. Once you
react as quickly as you can to error, there’s not much you can do to
speed things up. If you r4eact too fast you’ll tumble when you shouln’t
and slow down the approach. Tricky.

Also, why should the environment care if you get the to target
faster?

Best,

Bill P.

[_Boris_Hartman]

Hi Bill,

I’ll try to go slowly forward as it’s to many information for me. Maybe I’ll put some statements as “random” trial, so that maybe once I will understand your “construct of reorganization”.

BP earlier : The reorganization process, as I imagine it and as my models demonstrate, involves the “E. coli” principle in which one more more PARAMETERS OF CONTROL, like output gain or input sensitivity or synaptic strength, are slowly varied at different speeds. If that reduces the error, the changes simply continue. At some point, normally, there will be a closest approach to the optimum settings of the parameter, and then the error will start to increase.

BP earlier :

It is about a principle that permits order to be created by a random and therefore unintelligent and non-rational process. You have to understand what it means to say that the parameters of a control system are being changed by the reorganizing system.

BH : It’s obviously to me that reorganization is something in your imagination, or we could probably say it’s your intelectuall construct, maybe to avoid physiological theory of homeostasis. As I don’t see exactly how reorganizing process or system is stabilizing “intrinsic quantities” in your Figure ¸14.1. p. 191. I’ll try to find comparing meaning of terms “reorganization” I saw so far.

If I understood right, your “reorganizing system” is something that’s happening (changing) in the same time in all elements of control loop. But I’m not sure.

Maybe I was mislead by the case how you used term “reorganization” in Figure 14.1. p. 191. As Rick is also refering to it, and we still didin’t find the answers to his statements, I try to compare term “reorganization” you used so far, with the term in Figure 14.1.

BP : E. coli controls the rate of change of concentration of several substances it senses around it. It does so by detecting whether the rate of change of concentration is positive or negative (the internal biochemistry of this has been verified).

BH : So can we say that some rate of changes in substances is “entering” E.coli through “input function” ? I can’t imagine in other way what could it mean SENSE “rate of changes of concentrrations” ? What does it mean “DETECTING” ?

If E.coli controls the rate of change by DETECTING, I think it’s important to know how this DETECTING process of E.coli could look like ?

So I’ll try to analyze this “DETECTING” process in E.coli, by using your proposal of genetic control (Figure 14.1. p. 191) which could look like this :

1. Genetic source of bacteria E.coli is giving the intrinsic reference signal to comparator (what we don’t know for now what it is).
2. The comparator is matching the rate of changes from input, coming from “intrinsic quantites” in E. coli, that are affected by “the rate of changes in concentrations” in environment.
3. “Reorganization” proces as output from comparator is “continuoulsy” changing and affect probably the first level of behavioral control hierarchy (flagella) as we can’t expect any behavioral hierarchy in E.coli cell organization. Bacteria seems to be first “Living control system” on the Earth (before 3.6 bilion years).
4. “Flagella” acts on environment as the result of “reorganizing process”
   So if understand right we have here control system with two inputs and one output which is called “reorganization” inside E.coli, and which is capable of “DETECTING” whether sensed “rate of change of concentration is positive or negative” and acting on environment in which “rates of changes in concentrations” is going on.“Intrinisic quantities” are in physiological limits again as this is the main goal of “homeostatical” behavior of bacteria E.coli.

Can you explain from all this meaninigs of “reorganization” that were presented, with using your Figure 14.1., what is it the meaning you imagined ? I still don’t understand even if I read your LCS III.

Best,

Boris

···

----- Original Message -----

From:
Bill Powers

To: CSGNET@LISTSERV.ILLINOIS.EDU

Sent: Tuesday, December 13, 2011 12:50 AM

Subject: Re: How does PCT explain where our basic characteristics come from?

[From Bill Powers (2011.12.12.1520 MST)]

At 07:07 PM 12/12/2011 +0100, Boris Hartman wrote:

Hi Bill,

BH : As we know I have troubles understanding your terminology for a long time. Do I understood right that you described some suitable environment of chemical substances and gradients as input (atractor and repellient) to E.coli. Further it seems to me that you described intermediate proces inside bacteria E. coli which you called "reorganization" and have some effect on behavior of E.coli. Is this right?

BP: As far as I know, E. coli is incapable of reorganizing, but even if it can do this, the behavior of E. coli that inspired what I call the E. coli algorithm has nothing to do with E. coli’s own ability to reorganize. It is a PRINCIPLE that I am talking about, a principle that allows a control system to control a variable by using a kind of behavior that is basically random. E. coli controls the rate of change of concentration of several substances it senses around it. It does so by detecting whether the rate of change of concentration is positive or negative (the internal biochemistry of this has been verified). It can do this both to avoid unwanted substances, and to seek substances it wants – Koshland said 27 different substances. If the rate of change is in the wrong direction (relative to some internal reference signal), E. coli does not just turn and go in the right direction. Apparently it can’t do that. All it can do is change its direction of swimming at random. Koshland noted that the randomness has been measured and verified: the new direction is not related to the previous direction or to the direction of the gradient. Nevertheless, E. coli can control the rate of change of concentration that it experiences as it swims.

From this principle of behavior, I extracted a new way for reorganization to occur that does not require any knowledge about the environment. Its principle method is random. This means that reorganization can happen before any intelligence has appeared, and in any organism whether simple or complex. I thought that was important, because we have to try to account for how all these levels of organization arise that permit intelligent purposeful behavior – but before the final product has appeared. And it has to do so in any unpredictable environment.

The swimming of E. coli changes the organism’s X, Y, and Z coordinates in space at different constant rates. This corresponds to a process in a control system that changes three parameters of a control system – for example, its reference signal, its output gain, and its time constant – at three different constant rates. Or three input weights, or three output weights.

For E. coli, the swimming affects the rate of change of some chemical concentration that it senses. Furthermore, E. coli has an intrinsic reference signal for this rate: the reference signal specifies that it should be positive (for “good” substances) or negative (for “bad” ones).

In the general control system, the changes in parameters affect how close to some reference level the control system is keeping some variable. An implicit reference level is for the error in the control system to be zero, but any intrinsic variable affected by the control system can be the basis of reorganization, as long as it can be compared with some reference condition and the error can be detected. If the error is decreasing, the changes simply continue. If it increases, a random change of “direction” of changes in parameter space occurs, corresponding to E. coli’s “tumbles.”

You can experience this with Demo 7-1 in LCS3, although you will be imitating E. coli instead of changing parameters in some control system. Other demos work by changing parameters – more specifically, input and output weighting factors.

BH:  How "error in the system being reorganized" is presented ? What's causing "error" ?

BP: The error is simply the difference between the current reference signal and the perception of the controlled variable, in the control system that is being acted upon by the reorganizing system. If the control system is not already organized perfectly, it will not be keeping this difference or error as small as possible. If the controlled variable is important, say the average rate of food intake to the body, this error will set the reorganizing system into motion (that’s the basic postulate of reorganization theory). It will start to change the parameters of the control system, and that will change the control system’s ability to keep the food input close to the (intrinsic) reference level. If the change reduces the error, the changes will simply continue, slowing as the error gets smaller. If the initial change increases the error (which is just as likely, since the changes are being directed at random) another “tumble” will occur right away and a new combination of rates of change of parameters will be generated.

BH: What is excatly meant by reorganization ? What does reorganization do in the cell of E.Coli ?

BP: Again, this is not about reorganization in the cell of E. coli. It is about a principle that permits order to be created by a random and therefore unintelligent and non-rational process. You have to understand what it means to say that the parameters of a control system are being changed by the reorganizing system. The amount of perception resulting from a given input is changing. The comparison of reference and perception is changing sensitivity. The gain of the output function is changing. The weighting of the various effects of the output function on lower-level reference settings is changing. These changes are being caused by error, by stress in the system, and they will stop when the error is corrected, when the stress drops below some threshold.

But it is not reorganization in E. coli that I am talking about. In E. coli, we are seeing only the behavior by which this organism finds its way up gradients of attractants and down gradients of repelllants. The organization by which it does this is not changing; this is a fixed behavior pattern.

BP earlier : But I propose that the E. coli algorithm *is* enough.

BH : Is it possible to be more explicit in definition of algorithm. Can you put it in graphical presentation ?

BP:I don’t see how I can make it any plainer than it is in Living Control Systems III. The source code is given in Demo 8-1, in which 14 control systems are simultaneously being reorganized, a total of 196 weights being altered at random relative rates.

BH: I'll try to help myself with "diagram of immediate effects" :

ENVIRONMENT -- affect --> INPUT (atractor or repellient) -- affect --> REORGANIZATION in E.coli -- affect --> OUTPUT (away or to substance) -- affect --> ENVIRONMENT.....etc.

BP: It is not the input of attractor or repellent concentration that affects reorganization, but the SENSED RATE OF CHANGE of that input in relation to a reference rate of change. The link between reorganization and output is not simply to make the output go toward or away from the source of the substance; at that stage the relationship is random, and a reorganization – that is, a random change in direction, not the steady drift – is just as likely to increase the error as to reduce it. Ashby’s diagram of immediate effects is useless here because the immediate effect can be either unfavorable or favorable on successive trials. As I said earlier, Ashby didn’t happen to come across the E. coli principle of reorganization, even through his homeostat had a semi-random sort of change mechanism in it in the form of stepper switches. His stepper switches simply cycled systematically through a fixed sequence of possible connections, stopping when a connection was found that reduced error. That clever idea was the start of my ideas about reorganization. But Ashby never saw that what was really needed was a random stepper switch.

BH: If this is right, I'm interested how this explains Rick's statements :

1. When you get your copy of B:CP you will see that an important assumption of the theory is that organisms are born with a set of "intrinsic references". These references are inherited and basically unchangeable.

2. Intrinsic reference are reference specifications for the values of intrinsic variables.

BP:His statement 1 refers to the need for a reference condition relative to which the intrinsic error can be judged. The most basic reference conditions are relative unchangeable except through evolution. That error is what causes reorganization to start or continue, by hypothesis. His statement 2 points to the controlled variables in the reorganization loop: the values of intrinsic variables. When the intrinsic variables match their reference conditions closely enough, reorganization stops.

At this point I don’t know where to go next. I obviously don’t see what it is about reorganization theory that you don’t understand, unless I accidentally found it this time.

Best,

Bill P.

[Oliver]

[Oliver Schauman(2011.13.12.2235)]

[From Bill Powers (2011.12.12.1650 MST)]

···

At 11:30 PM 12/12/2011 +0000, Oliver Schauman wrote:

[Oliver Schauman(2011.12.12.2310)]

OS: Isn’t a “solution” in evolutionary terms is something that makes it more likely that the given gene is passed on? The way this gene would remain in the population (and thus it would seem as if the population is “staying at the solution”) is if it increased the likelihood of reproduction over other variants of that gene in the population.

BP: I would give that a different slant, by saying that a solution is a reorganization that reduces the chances that a given gene will fail to reproduce accurately. Mutations, I am proposing, are actually E. coli-style tumbles and are most initiated by the organism at the level of DNA. As Rick Marken has just pointed out, the standard concept of mutation and natural selection can’t explain why continued mutation doesn’t lead to a random walk among all the alternative outcomes with equal effects on fitness. The reorganization principle does explain that.

OS: I suppose the main problem here is that I do not think this latter statement is true. Firstly, there ARE random mutations occurring in a population all the time. That is why evolution works. Secondly, the idea that two or more mutations that altered some biochemical process could have exactly the same effect on fitness is extremely improbable. Any solution has benefits and costs and this balance needs to be just slightly different to result in a one being favoured over the other over time. Am I making sense?

RM (Rick marken earlier): If random mutations are always occurring at a constant rate in a population then it seems to me that the population is just as unlikely to remain in the “evolved” state as it was to get there in the first place. That is, I don’t think the constant mutation rate model would produce the “punctuated” fossil record that is actually observed. Instead, there would be continuous drifting from one form to another. (I know as well as anyone that God put those fossils there just to test our faith but it’s still fun to play with Godless models of this observation while everyone else is in church).

OS: A “punctuated” fossil record is something that taxonomists have created. A continuous drifting from one form to another is exactly what we find in a given population and it is a founding principle why evolution by natural selection works. It is these drifts that are either selected or not for survival.

BP: Be careful. The drifts are not selected for; only survival is selected for under the principle of natural selection. That means that all drifts that have equal effects on survival have the same chance of being selected for. There is not just one solution for a given problem; there are multitudes, as the variety of species and subspecies attests.

OS: Yes, I should have said “through survival” rather than for survival. I agree that there are many solutions to a given problem, e.g. pin-hole camera eyes and compound eyes. I think what you need to be clear with is what you mean by “equal effect on survival” equal in comparison to what? Genetic mutations would occur in individuals of a given species that compete with each other and not directly with other species. As they say, you do not have to run faster than the bear, you just have to run faster than your friend.

I prefer to cast fitness measures in terms of successful control, which is easy to define: minimum overall error or perhaps intrinsic error. If increasing the number of members of a species will lead to less error, that is the condition that will result from reorganization. But if decreasing the number of members will make errors even smaller, that will be the outcome. Increasing the population is not necessarily a pro-survival strategy.

OS: I struggle to understand what this means in terms of changes that we can observe in a given population.

OS: However, a single mutation does not produce a new species or necessary provide a new “solution”. It would just slightly tweak something like how a given protein pumps in a cell would function or how some enzyme might catalyse a given chemical reaction. If another mutation occurred that made reproduction even more probably, that gene would eventually take over in the population. If it didn’t it would disappear after certain amount of time.

BP: All that depends on the hypothesis that a larger population is better, which I contest.
**

OS: Remember that the mutated offspring of a given species would always be classified as the same species as its mother.

BP: Like the offspring of a horse and a donkey? I thought that mutual fertility was the main mark of being cospecific.

OS: This would indeed not apply to your example. I can however guarantee that no species on Earth today has a sterile mule as its ancestor.

OS: he “punctuations” between species is something that is done to distinguish individuals that are a million+ years apart. Indeed, if we had a continuous fossil record it would be impossible to classify any animals into discrete species.

BP: I think you’re right about that.
*
…*

OS: I am finding it difficult to understand why one needs to construct an “analog of evolution”. To me it is a beautiful and accurate model of itself.

BP: But it is insufficiently effective, and evolutionary models that I have seen mostly cheat by giving credit for a move toward survival instead of actually getting there. The theory behind the phenomenon of evolution is not a very good one – it takes a lot of suspension of disbelief. It shows a reluctance to exhibit the necessary amount of skepticism, perhaps for fear of being mistaken for a proponent of the other side.

OS: I don’t tend to mix belief into it. Thinking of something like quantum theory, one would have to be mad to accept such a theory without evidence. I think any decent scientist would welcome evidence that challenged the current theory of evolution. I do not know of any arguments against the current model that hold water.

OS: However, I think it is definitely worthwhile applying the principles of control to evolution , but I do not think it would be valid to do this at a population level. Here is an idea: It would be interesting to have a population of E.coli with variations in their reorganizing systems. Some would tumble a lot and some very rarely, depending on their threshold (the kind of thing a given gene might code for). The E.coli that get to the target would reproduce and their offspring would appear at random places on the grid. If the “reproduction zone” can only be occupied by a given number of E.coli at a time, the environment will favour the gene that codes for tumbling that gets you to the target fastest. This could then b be compared to a model where the E.coli function through reinforcement.

BP: Yes, that would definitely be a good project. However, by the basic design of the theory of reorganization, offspring do not appear at random positions on the grid – they appear at only slightly different positions, but perhaps in random directions relative to the starting point. Genetic drift, it’s called. A mutuation, I have read, is known to change the direction of genetic drift, which sounds an awful lot like the E. coli reorganization principle.

If you play the Demo 7-1 from LCS3, you’ll see that varying the rate of tumbling doesn’t by itself speed up the approach to the goal. Once you react as quickly as you can to error, there’s not much you can do to speed things up. If you r4eact too fast you’ll tumble when you shouln’t and slow down the approach. Tricky.

Also, why should the environment care if you get the to target faster?

OS: The environment probably wouldn’t care. What I was thinking was setting up the premise that the target is a nutrient rich zone which allows the E.coli to reproduce. If we give each E.coli a life span of say 10 sec, then getting to the target and reproducing is extremely important for each E.coli. Wouldn’t this reaction time be something that could vary between individuals in the population? Let’s say they all start with the same reaction time and then in each generation, in one or two individuals, random change (“a tumble”) occurs in the parameter that affects reaction time. After say 100 generations, wouldn’t you expect the mean reaction time in the population to be closer to the optimum value than it was in the beginning?

Best wishes,

Oliver

[amatic]

[From Adam Matic 2011.12.14.23.30 CET]

Rick Marken (2011.12.11.1030)

I’m not sure that the Reinforcement model is a good analog of natural selection (Skinner thought it was) but I do think it would be interesting to see how existing models of natural selection do (in terms of keeping a population “adapted” to its niche) once the evolutionary process has brought the population to the adapted state. When I get a chance I’ll try to extend my “selection of consequences” demo to show that the most effective way to stabilize a result of random variation is via selection retention based on a control type (E. coli) process.

There is a process called “stress induced mutations”, like here: http://www.fasebj.org/content/20/14/2476.full ?

I’m no geneticist, but it seems that some bacteria are generating more mutation in offspring DNA by “loosening” repair mechanisms in “bad environments”.

Is this in any way connected to what you’re talking about?

Best, Adam

[Bill_Powers1]

[From Bill Powers (2011.12.14.1443 MST)]

[Oliver
Schauman(2011.13.12.2235)]

OS: I suppose the main problem here is that I do not think this latter
statement is true. Firstly, there ARE random mutations occurring in
a population all the time.

BP: But what is causing them? The standard treatment is to assume that
they’re just a sort of background noise arising from thermal agitation or
other causes unconnected to the inner workings of the organism. I am
proposing that the main cause of significant amounts of mutation is some
mechanism in the organism itself, and that the mutations are part of a
method of adaptation that I call reorganization.

Try looking up stress-induced mutation. You’ll find a really surprising
amount of information on this quite real phenomenon. Mutation rates
increase with stress and subside again when the stress is removed. As we
in PCT know, the appearance that the stress causes the mutation rate to
change can easily be an illusion: the stress is quite likely a
disturbance, and the mutations could well be the action of some control
system trying to reduce the error caused by the stressor. At any rate,
that’s what I am proposing.

That is why evolution
works. Secondly, the idea that two or more mutations that altered some
biochemical process could have exactly the same effect on fitness is
extremely improbable.

Fitness is measured by the ability to survive to the age of reproduction
and successfully reproduce. How many paths to such an outcome do you
think could exist? I would guess that they are countless. Simply looking
right more often than left could make the difference to a visiting Brit
between stepping in front of a bus and not doing so, which means the
difference between not reproducing and reproducing. In birds, it can
depend on whether a bird happened to hear another bird, or a recording,
singing a particular kind of song, and learned to do the same
thing.

You may be able to reason from changes in a biochemical process to
changes in fitness, but you can’t go the other way. You can’t deduce the
particular biochemical changes that have occurred just from knowing about
reproductive success. Reproductive success doesn’t depend on any one
variable. It depends on a whole chain of parallel events, each one being
the starting point from which the next set of parallel events develops,
until at the end survival to reproduce does or doesn’t happen. That’s why
I say that a given degree of fitness can arise in an enormous number of
different ways. Does the fact that you arrive at work every day at 8:00
AM reveal what route you took on any one day?

*Any solution has benefits and
costs and this balance needs to be just slightly different to result in a
one being favoured over the other over time. Am I making
sense? *

BP: You may be making perfect sense, but fitness makes no sense as a
gauge of evolutionary success. It’s much too crude a criterion, as the
same degree of fitness (found mate at age 23, gave birth at age 24) can
result from, probably, an infinity of different histories of development
of an organism.

BP: Be careful. The
drifts are not selected for; only survival is selected for
under the principle of natural selection. That means that all drifts that
have equal effects on survival have the same chance of being selected
for. There is not just one solution for a given problem; there are
multitudes, as the variety of species and subspecies attests.

OS: Yes, I should have said “through survival” rather than for
survival. I agree that there are many solutions to a given problem, e.g.
pin-hole camera eyes and compound eyes. I think what you need to be clear
with is what you mean by “equal effect on survival” equal in comparison
to what?

BP: Not an equal effect on survival; equal amount of survival. Fitness
can’t be measured by looking at things that affect survival, because
there are too many of them working in both directions at the same time
and you can’t know which were critical. To measure fitness you simply
observe the survival and reproduction rates that have in fact happened.
That’s why it is such a weak criterion. There are simply too many things,
including luck, that affect it.

OS: Genetic mutations
would occur in individuals of a given species that compete with each
other and not directly with other species. As they say, you do not have
to run faster than the bear, you just have to run faster than your
friend.

BP: True, unless you trip because of a clumsiness gene, or fail to see
the bear because of a bad-eyesight gene, or mistake the bear for a friend
because of a defect in a form-recognition gene, and so on. And why would
such mutations occur only in individuals that compete with each other?
What keeps them from happening eventually, as Rick pointed out, in ALL
individuals, messing up whatever fitness advantage they had attained to
pass on to descendants? Are you saying that mutation rate depends on
selection pressure? If so, good, but that doesn’t lead to the conclusions
you’re drawing.

BP earlier: I prefer to cast
fitness measures in terms of successful control, which is easy to define:
minimum overall error or perhaps intrinsic error. If increasing the
number of members of a species will lead to less error, that is the
condition that will result from reorganization. But if decreasing the
number of members will make errors even smaller, that will be the
outcome. Increasing the population is not necessarily a pro-survival
strategy.
OS: I struggle to understand what this means in terms of changes that
we can observe in a given population.

What you observe in a population is the average of a large number of
characteristics that are present to widely varying degrees in
individuals. The group characteristics can easily be different from the
characteristics of every individual in the group, and even opposite to
them (see my paper in the Sept.Oct 1990 issue of the American Behavioral
Scientist, organized and edited by Rick Marken). You may be able, as I
said above, to discern certain kinds of changes that will affect the
general ability to reproduce, but that does not allow you to reason
backward, from observe ability to reproduce to the particular changes
that affect the observation. There are just too many that can have the
same effect. After all, fitness can’t change in very many directions: you
reproduce earlier or later, and more or less often.
*OS: I don’t tend to mix belief into it. Thinking of something like
quantum theory, one would have to be mad to accept such a theory without
evidence. I think any decent scientist would welcome evidence that
challenged the current theory of evolution. I do not know of any
arguments against the current model that hold water.*BP: Well, what about mine, above? And keep in mind that it’s not
evolution I’m arguing against, but the theory of natural selection that
supposedly explains it. I think that theory has rather large holes in it,
and the models I’ve seen require too much help from the modeler, as I
explained previously. I’m offering the theory of
mutation-as-reorganization not to debunk the idea of evolution, but to
provide a model that is actually effective enough to produce the kind of
evolution that we do in fact observe to have happened, and to keep
happening.

I think the standard theory of natural selection was accepted because its
proponents were eyeing not competing theories that had to be taken
seriously, but religious objections to the very idea of evolution. The
religious objections didn’t take a lot of deep thought to shoot down.
Anyone who criticized the idea of natural selection was naturally assumed
to be on the side of religion, so scientists didn’t work themselves up
into a lather over such criticisms. They had that problem taken care of.

I hope I’m offering criticisms with a little more substance to
them.

Best,

Bill P.

···

OS: However, I think it is
definitely worthwhile applying the principles of control to evolution ,
but I do not think it would be valid to do this at a population level.
Here is an idea: It would be interesting to have a population of
E.coli with variations in their reorganizing systems. Some would tumble a
lot and some very rarely, depending on their threshold (the kind of thing
a given gene might code for). The E.coli that get to the target would
reproduce and their offspring would appear at random places on the grid.
If the “reproduction zone” can only be occupied by a given number of
E.coli at a time, the environment will favour the gene that codes for
tumbling that gets you to the target fastest. This could then b be
compared to a model where the E.coli function through reinforcement.
BP: Yes, that would definitely be a good project. However, by the
basic design of the theory of reorganization, offspring do not appear at
random positions on the grid – they appear at only slightly different
positions, but perhaps in random directions relative to the starting
point. Genetic drift, it’s called. A mutuation, I have read, is known to
change the direction of genetic drift, which sounds an awful lot like the
E. coli reorganization principle.
If you play the Demo 7-1 from LCS3, you’ll see that varying the rate of
tumbling doesn’t by itself speed up the approach to the goal. Once you
react as quickly as you can to error, there’s not much you can do to
speed things up. If you r4eact too fast you’ll tumble when you shouln’t
and slow down the approach. Tricky.
Also, why should the environment care if you get the to target
faster?
*OS: The environment probably wouldn’t care. What I was thinking was
setting up the premise that the target is a nutrient rich zone which
allows the E.coli to reproduce. If we give each E.coli a life span
of say 10 sec, then getting to the target and reproducing is extremely
important for each E.coli. Wouldn’t this reaction time be something
that could vary between individuals in the population? Let’s say they all
start with the same reaction time and then in each generation, in one or
two individuals, random change (“a tumble”) occurs in the
parameter that affects reaction time. After say 100 generations, wouldn’t
you expect the mean reaction time in the population to be closer to the
optimum value than it was in the beginning?
Best wishes,
Oliver *

[rsmarken]

[From Rick Marken (2011.12.14.1615)]

Adam Matic (2011.12.14.23.30 CET)

Rick Marken (2011.12.11.1030)

RM: I'm not sure that the Reinforcement model is a good analog of natural
selection (Skinner thought it was) but I do think it would be interesting to
see how existing models of natural selection do (in terms of keeping a
population "adapted" to its niche) once the evolutionary process has brought
the population to the adapted state.� When I get a chance I'll try to extend
my "selection of consequences" demo to show that the most effective way to
stabilize a result of random variation is via selection retention based on a
control type (E. coli) process.

AM: There is a process called "stress induced mutations", like
here:�http://www.fasebj.org/content/20/14/2476.full�?
I'm no geneticist, but it seems that some bacteria are generating more
mutation in offspring DNA by "loosening" repair mechanisms in "bad
environments".
Is this in any way connected to what you're talking about?

You betcha! Coincidentally, Bill just mentioned it in the post I got
right after yours.

Best

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Bill_Powers1]

[From Bill Powers (2011.12.15.0919 MST)]

BP earlier:I haven’t spoken
about intrinsic variables inside E. coli, or its homeostatic systems if
it has any.

BH: I went through some titles in researches of E.coli. It says that this
bacteria is one of the most well-explained as there were so many
experiments speccially long term. There are articles about how some
substances are kept in steady-state by genetic codes.

BP: I agree that E. coli may have – does have – homeostatic systems,
but those aspect of it were not what I was talking about.

BH: From some articles I got an
impression that E.coli is normal cell maintaining it’s homeostasis and
showing Darwins characteristics of “natural selection”.

BP: What is observed is that E. coli does evolve new characteristics,
even inheritable characteristics. How that happens is not observed.
Natural selection is a theory about unobserved causes of changes in
characteristics.

Reorgnaization is a different theory about the same thing.

BP earlier : Of course
reorganization theory is strange in another way because it claims that
mutations are driven by error signals inside the organism, not by cosmic
rays or mutagenic chemicals or other noxious causes from outside (though
it doesn’t rule out such effects). Selection pressures are surely caused
by environmental events, but they do not directly result in mutations.
The mutations are what the organism does in response to the selection
pressures, in an attempt to nullify their effects.

BH
: Are you saying that some “theoretical” process inside
organism is changing DNA ?

It has been pretty well established that E. coli itself initiates changes
in its own DNA when certain stresses occur. Here is an example from
Science magazine:

http://www.bichat.inserm.fr/equipes/emi0339/publications_pdf/exhaustiveliste/2003science.pdfhttp://www.bichat.inserm.fr/equipes/emi0339/publications_pdf/exhaustiveliste/2003science.pdf

For more, just Google “stress-induced mutation”. I know that
this sort of thing is still rejected by many people because it sounds
like Lamarkian theory, but apparently Lamark was not as wrong as he
seemed. You still can’t create mice with short tails by cutting off the
parent’s tails, but at least some organisms are capable of reorganizing
themselves, at the level of DNA, to counteract certain kinds of
disturbances. And the changes are inheritable.

However, I have to repeat that reorganization inside E. coli is not the
subject I was talking about, and I still don’t know if E. coli can do the
kind of reorganizing under discussion here: changing organization within
a single lifetime.

As
you are persistingly avoiding asnwers on some crucial questions which
usually involve intrinsic reference signal coming from “genetic
source” it’s maybe really useless to continue
conversation.

You have difficulty in seeing how a reference condition could be
genetically specified. How about body temperature, or all the other
homeostatically controlled variables you are interested in? Many of them
have variable reference conditions, but then we can ask where the
reference conditions at the next higher level come from. At some level we
will find a constant reference signal, zero or nonzero, and that will be
genetically specified. How else could it be specified?

All that is necessary to establish a reference condition is some
physiological characteristic that determines the level of perceptual
input to a control system at which its output will be exactly zero. A
simple threshold will suffice: no output until the input reaches some
magnitude. Or it could be a real neural signal. There are many neurons
that have a rate of spontaneous firing without any incoming signals.
Those neurons are clearly inheritable parts of the nervous system. If the
output of such a neuron enters a comparator neuron (as either the
inhibitory or excitatory input), it will establish a reference condition
that a control system will seek.

I didn’t realize that what bothered you was the idea of a genetically
specified reference level. I have taken that for granted, because there
are so many ways it can come into existence.

When you will find time and explain how does control work from
genetic source and how reference levels of intrinsic variables are set
(as that was the problem in my conversation with Rick) than I suppose our
converstaion will have any sense. It’s obviously that some members are
tyred of this kind of theorizing.

I guess we all are.
I wonder if the problem here isn’t in your concept of a reference signal.
Are you thinking that a reference signal has to have some specific
meaning in addition to having a specific magnitude? That the reference
signal for one controlled perception will somehow have a different nature
from a reference signal for a different controlled perception?
A reference signal is just a neural signal, if it’s anything more than a
threshold effect or some other built-in structural bias. All it does is
indicate a magnitude; it doesn’t even identify the variable that is to
have that magnitude. The nature of the controlled variable is set by the
organization of the perceptual input function, but the perceptual signal
contains no information about that organization. It’s just a neural
signal,. too. The comparator doesn’t know what the two signals entering
it mean: they’re just two signal magnitudes, one excitatory and the other
inhibitory. If the inhibition is less than the excitation, the net
excitation produces an output signal indicating the magnitude of the
difference, which we call the error signal.
So basically a reference signal, whether actual or only a structural
bias, can easily be inherited if it’s not the output of some higher-order
system. It is simply whatever properties have to exist to determine how
much perceptual signal is to produce zero error signal.
Is there some other problem with the ideas I’m offering that you
haven’t mentioned yet?

Best,

Bill P.

···

At 11:00 PM 12/14/2011 +0100, Boris Hartmann wrote:

[Bill_Powers1]

[From Bill Powers (2011.12.15.0930 MST)]

[From Rick Marken
(2011.12.14.1615)]

Adam Matic (2011.12.14.23.30 CET)

AM: There is a process called “stress induced mutations”,
like

here: http://www.fasebj.org/content/20/14/2476.full ?

I’m no geneticist, but it seems that some bacteria are generating
more

mutation in offspring DNA by “loosening” repair mechanisms
in "bad

environments".

Is this in any way connected to what you’re talking about?

You betcha! Coincidentally, Bill just mentioned it in the post I got

right after yours.

Thanks for the corroboration. I just finished writing to Boris Hartman
about the same thing. I haven’t been able to find a reference to the
original article which I read many years ago, probably in Science, so I
can’t cite the discoverers of this phenomenon, but it has been known for
many years that E. coli can mutate itself when the food supply is
switched between galactose and either fructose or glucose – can’t
remember which. The mutations are inherited by the next generation, and
can be reversed by switching back to the other sugar. Natural selection
kind of piggybacks on the spontaneous mutation, because the individuals
that don’t spontaneously mutate are weeded out because they can’t live on
the new nutrient. Clearly, one of the things selected for by the
“natural” method is the ability to mutate on purpose.
So we might think of reorganization as the next level of adaptive
processes, which works better than the simple random mutation of natural
selection. Maybe systematic problem solving using logical programs is
another level, which can be still more effective, though not as free to
try anything at all.
Here’s another:http://www.pnas.org/content/95/17/9997.full.pdf
ABSTRACT Increased spontaneous
mutation is associated
with increased cancer risk. Here, by using a model system, we
show that spontaneous mutation can be increased several hundred-
fold by a simple imbalance between the first two enzymes
involved in DNA base excision repair. TheSaccharomyces
cerevisiae
MAG1 3-methyladenine
(3MeA) DNA glycosylase, when
expressed at high levels relative to the
apurinicy
apyrimidinic
endonuclease, increases spontaneous mutation by up to’
600-
fold in S. cerevisiaeand
’
200-fold inEscherichia
coli. Genetic
evidence suggests that, in yeast, the increased spontaneous
mutation requires the generation of abasic sites and the processing
of these sites by the
REV1
y
REV3
y
REV7lesion bypass
pathway. Comparison of the mutator activity produced by Mag1,
which has a broad substrate range, with that produced by the
E.
coli Tag 3MeA DNA
glycosylase, which has a narrow substrate
range, indicates that the removal of endogenously produced
3MeA is unlikely to be responsible for the mutator effect of Mag1.
Finally, the human AAG3-MeA DNA glycosylase
also can
produce a small
(’
2-fold) but statistically
significant increase in
spontaneous mutation, a result which could have important
implications for carcinogenesisSo now we have a mechanism for causing spontaneous mutations.
Here it’s a bad thing, but it could also be good. Is cancer a result of
faulty or unlucky reorganization?

This is variously known as “spontaneous” or “induced”
mutation, depending on whether the speaker can bear to hint that the
organism itself might be the active agent causing the changes.

Best,

Bill P.

···

At 04:12 PM 12/14/2011 -0800, you wrote:

[tedcloak]

[From Ted Cloak (2011.12.15.1010 MST)]

The important thing to keep in mind is that while internal processes affect the mutation rates of some organisms, they have never been shown to affect the direction of mutations; i.e., mutations are always random with respect to their adaptive potential. Natural selection bats last.

[From Bill Powers (2011.12.15.0930 MST)]

[From Rick Marken (2011.12.14.1615)]

Adam Matic (2011.12.14.23.30 CET)

AM: There is a process called “stress induced mutations”, like
here: http://www.fasebj.org/content/20/14/2476.full ?
I’m no geneticist, but it seems that some bacteria are generating more
mutation in offspring DNA by “loosening” repair mechanisms in “bad
environments”.
Is this in any way connected to what you’re talking about?

You betcha! Coincidentally, Bill just mentioned it in the post I got
right after yours.

Thanks for the corroboration. I just finished writing to Boris Hartman about the same thing. I haven’t been able to find a reference to the original article which I read many years ago, probably in Science, so I can’t cite the discoverers of this phenomenon, but it has been known for many years that E. coli can mutate itself when the food supply is switched between galactose and either fructose or glucose – can’t remember which. The mutations are inherited by the next generation, and can be reversed by switching back to the other sugar. Natural selection kind of piggybacks on the spontaneous mutation, because the individuals that don’t spontaneously mutate are weeded out because they can’t live on the new nutrient. Clearly, one of the things selected for by the “natural” method is the ability to mutate on purpose.
So we might think of reorganization as the next level of adaptive processes, which works better than the simple random mutation of natural selection. Maybe systematic problem solving using logical programs is another level, which can be still more effective, though not as free to try anything at all.
Here’s another:http://www.pnas.org/content/95/17/9997.full.pdf
ABSTRACT Increased spontaneous mutation is associated
with increased cancer risk. Here, by using a model system, we
show that spontaneous mutation can be increased several hundred-
fold by a simple imbalance between the first two enzymes
involved in DNA base excision repair. The Saccharomyces cerevisiae
MAG1 3-methyladenine (3MeA) DNA glycosylase, when
expressed at high levels relative to the apurinicy apyrimidinic
endonuclease, increases spontaneous mutation by up to ’ 600-
fold in S. cerevisiae and ’ 200-fold in Escherichia coli. Genetic
evidence suggests that, in yeast, the increased spontaneous
mutation requires the generation of abasic sites and the processing
of these sites by the REV1 y REV3 y REV7 lesion bypass
pathway. Comparison of the mutator activity produced by Mag1,
which has a broad substrate range, with that produced by the E.
coli Tag 3MeA DNA glycosylase, which has a narrow substrate
range, indicates that the removal of endogenously produced
3MeA is unlikely to be responsible for the mutator effect of Mag1.
Finally, the human AAG 3-MeA DNA glycosylase also can
produce a small (’ 2-fold) but statistically significant increase in
spontaneous mutation, a result which could have important
implications for carcinogenesisSo now we have a mechanism for causing spontaneous mutations. Here it’s a bad thing, but it could also be good. Is cancer a result of faulty or unlucky reorganization?

This is variously known as “spontaneous” or “induced” mutation, depending on whether the speaker can bear to hint that the organism itself might be the active agent causing the changes.

Best,

Bill P.

···

At 04:12 PM 12/14/2011 -0800, you wrote:

[rsmarken]

[From Rick Marken (2011.12.15.1050)]

Bill Powers (2011.12.15.0930 MST)--

RM: You betcha! Coincidentally, Bill just mentioned it in the post I got
right after yours.

BP: Thanks for the corroboration. I just finished writing to Boris Hartman about
the same thing. I haven't been able to find a reference to the original
article which I read many years ago, probably in Science, so I can't cite
the discoverers of this phenomenon...

I think you might be thinking of:

Cairns, J., Overbaugh, J and Miller, S. (1988) The origin of mutants.
Nature, 335, 142-145

Best

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Bill_Powers1]

[From Bill Powers (2011.12.15.1125 MST)]

[From Ted Cloak (2011.12.15.1010
MST)]

TC: The important thing to keep in mind is that while internal processes
affect the mutation rates of some organisms, they have never been
shown to affect the direction of mutations; i.e., mutations are
always random with respect to their adaptive potential. Natural selection
bats last.

BP (from here on): Reorganization theory is a replacement for natural
selection theory, just as it is a replacement for reinforcement theory.
In reorganization theory, the mutations are also unaffected by any prior
events and do not favor any one outcome over another. The changes are
truly random, just as DNA changes due to radiation damage are random.
They are just as likely to makes things worse as better.
The main difference is in what is mutated. In reorganization theory, it
is the direction of changes that is mutated, while in the
background all the things being mutated are changing very slowly even
without sudden mutations.
When an E. coli mutation takes place, all the variables being affected
simply change their slow rates of drift. They may start changing faster,
or more slowly, and change from increasing to decreasing, all at random.
This has actually been noticed in studies of natural selection and there
have been arguments about it: here’s a comment from a Wiki page

···

At 10:13 AM 12/15/2011 -0700, Ted Cloak wrote:

The effect of genetic drift is larger for alleles present in a smaller
number of copies, and smaller when an allele is present in many copies.
Vigorous debates occurred over the relative importance of
natural
selection versus neutral processes, including genetic drift.
Ronald Fisher
held the view that genetic drift plays at the most a minor role in
evolution, and this remained the dominant view for several decades. In
1968 Motoo Kimura
rekindled the debate with his

neutral theory of molecular evolution, which claims that most
instances where a genetic change

spreads across a population (although not necessarily changes in
phenotypes) are
caused by genetic
drift.
^[3] [

Genetic drift

](Genetic drift - Wikipedia)

There are also disagreements about the nature of the drift. Some authors
say it is just a sampling phenomenon, but others point to slow changes
from generation to generation in the same direction, and say that when a
mutation occurs, the direction of genetic drift changes. The latter, of
course, is the essence of my idea of reorganization. A mutation is a
“tumble”, and between tumbles, there are continuous slower
changes – “swimming in a straight line.”

The traditional view is that mutations simply alter parameters of an
organism at random – bang, and it’s randomly different. This is the
concept of mutation that makes any systematic effects highly unlikely. If
you change 10 characteristics at the same time, and each characteristic
has 10 levels at which it can be expressed, then the chance of a mutation
producing the best combination of characters and amounts of change is 1%
(crudely speaking).

I keep saying this and not getting any responses from anyone to defend
the traditional idea of natural selection: I claim that the genetic
algorithm and other models that I’ve seen give the model help that it
doesn’t deserve. An organism is allowed to live if it merely changes
something in the right direction, something that only a supervisor,
helper, teacher, or God can know. I claim that a realistic model has to
work by actually getting to the state that permits survival – actually
getting close enough to the food to eat it, not just moving in the
general direction of the food. In fact, the genetic algorithm seems to
work only because someone knows what the goal of the changes is and
whispers to the organism whether it guessed right or wrong before any
actual consequences happen. The algorithm is actually working like the
reorganization model, and that is why it succeeds.

In reorganization theory, the changes are based on detection of some kind
of error. In the case of food, it might be a hunger signal. To say
“based on,” however, is not to imply that the changes are
directed by the error. They are not; they are strictly random. Yes, there
is a way to know whether the goal-state is getting closer or farther way.
But that is all that needs to be known. The action that results from an
increase in the error is not a move toward the goal, or opposite to the
previous move. It is simply a random change in direction, unrelated to
the error or any previous action. And that turns out to be sufficient to
correct the error in the end.

Demo 7-1 in LCS3 allows you to play the part of E. coli and move a dot,
yourself, to a goal, a circle in the center of the screen. You do it by
clicking the mouse on the “tumble” button, or after you’ve done
that once, just tapping the space bar. A tumble sets the dot moving in a
new direction, with no relation to anything. The new direction is
obtained from the “random()” function which returns a random
number between 0.0 and 1.0. When you tap, the new direction, in radians,
is 2pirandom().

The program can also do this by itself, by running the same
reorganization model used in other demos. After it has run for a while,
resetting after each success in reaching the goal, you see the cumulative
number of random tumbles needed to reach the goal on each trial. It’s
usually about 12.

Also, the program can simply move the dot to any randomly-selected spot
on the screen each time there is a mutation. This is like the standard
concept of mutation. The goal circle has an area of about 133 square
millimeters (on my screen) and the active area is 400 x 280 or 11,200
square millimeters. On the average it would take 11200/133 or 842 trials
to hit the circle. In the automatic mode, the program does this very
fast, and sure enough, about 840 trials are needed to hit the circle, on
average. That’s why there is no manual mode for this variant.

Therefore the E. coli method is 840/12 or 70 times as efficient as the
model that just randomly chooses new positions.

If we now go to three dimensions, the circle becomes a sphere with a
volume of about 275 cubic millimeters, and there are about 400 x 280 x
(say) 350 or 39,200,000 cubic millimeters in the playing volume. Now the
dot will land in the target volume 275/39200000 of the time, or 1 time in
142,545 trials. Can someone work out what the E. coli probability will
be? The E. coli efficiency advantage, I’m pretty sure, would now be in
the thousands. That’s for 3 dimensions.

In Demo 8-1 we have 14 control systems each reorganizing 14 output
weights, so there are 196 dimensions in this case. It takes the E. coli
method about 30 minutes to reduce the total control error in all 14
systems to abut 3% of its initial value. How long would it take the
random-jump method? How many centuries, that is?

So what I’m proposing, to sum up, is that evolution actually works by the

E. coli method and not by the random-jump method that has always been
assumed. The only reason the random-jump method has seemed to work with
models is that the models, inadvertently, actually use something close to
the E. coli method. They give credit for moving toward a goal-state known
to the programmer but (supposedly) not to the test organism, and only
that fact can account for their success.

Best,

Bill P.

[Bill_Powers1]

[From Bill Powers (2011.12.15.1300 MST)]

Rick Marken (2011.12.15.1050)

I think you might be thinking
of:

Cairns, J., Overbaugh, J and Miller, S. (1988) The origin of
mutants.

Nature, 335, 142-145

Yay, that’s it! 1988, it was a long time ago. I guess I could have
looked through quite a few issues of Science without finding it,
since it was published in Nature.

Best,

Bill P.

[MartinT]

[Martin Taylor 2011.12.15.16.36]

[From Bill Powers (2011.12.15.1300 MST)]

    Rick Marken

(2011.12.15.1050)

    I think you might be

thinking
of:

    Cairns, J., Overbaugh, J and Miller, S. (1988) The origin of

mutants.

    Nature, 335, 142-145

  Yay, that's it! 1988, it *was* a long time ago. I guess I

could have
looked through quite a few issues of Science without
finding it,
since it was published in Nature.

  Best,



  Bill P.

I have two questions related to this thread, one long-standing, the

other new because I think Bill has introduced something new into the
conceptual space.

1. How does the e-coli process escape from a local optimum?

2. I understand what defines a direction in the reorganization of a

control network, but what defines a “direction” in genetic space?

Martin

[Bill_Powers1]

[From Bill Powers (2011.12.15.1645 MST)]

[Martin Taylor 2011.12.15.16.36]

MT: I have two questions related
to this thread, one long-standing, the other new because I think Bill has
introduced something new into the conceptual space.

1. How does the e-coli process escape from a local
   optimum?

BP: The basically random variations will occasionally introduce large
excursions that can help get out of local traps. I don’t know what the
limits are, or the kinds of local minima found in this application. I
would expect that if being stuck in a local minimum causes increasing
problems for the organism, reorganization will increase, the larger error
producing faster and coarser changes. That might work.

MT: 2. I understand what defines
a direction in the reorganization of a control network, but what defines
a “direction” in genetic space?

BP: I don’t know very much about genetics, but I’d guess the parameters
being varied would be relative rates of protein synthesis and any similar
continuous variables specified in DNA. Look up Savageau – he was
interested in control systems in DNA:

Savageau,
M. A. (1976) Biochemical systems analysis:

a study of function and
design in molecular
biology.

Reading, MA: Addison-Weseley.

The “direction,” of course is a metaphor for varying
relative rates of change in some set of parameters, each considered as a
different dimension ibn some hyperspace.

Also try this:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1460276/

Somehow I think you would be better than I at finding answers to these
questions!

Best,

Bill P.

[MartinT]

[Martin Taylor 2011.12.16.00.36]

[From Bill Powers (2011.12.15.1645 MST)]

  [Martin Taylor 2011.12.15.16.36]

    MT: I have two

questions related
to this thread, one long-standing, the other new because I think
Bill has
introduced something new into the conceptual space.

    1. How does the e-coli process escape from a local

optimum?

  BP: The basically random variations will occasionally introduce

large
excursions that can help get out of local traps. I don’t know what
the
limits are, or the kinds of local minima found in this
application. I
would expect that if being stuck in a local minimum causes
increasing
problems for the organism, reorganization will increase, the
larger error
producing faster and coarser changes. That might work.

So a bit a simulated annealing in conjunction with hill-climbing.

Yes, it might work.

    MT: 2. I understand

what defines
a direction in the reorganization of a control network, but what
defines
a “direction” in genetic space?

  BP: I don't know very much about genetics, but I'd guess the

parameters
being varied would be relative rates of protein synthesis and any
similar
continuous variables specified in DNA. Look up Savageau – he was
interested in control systems in DNA:

           Savageau,

M.
A. (1976) Biochemical systems analysis:

    a study of function an        d

design in molecular
biology.

    Reading, MA: Addison-Weseley.

I haven't read the book, but later papers by him available from

Google Scholar that cite it seem to be dealing with the feedback
systems involved with gene regulation, not with altering the genes
themselves. Maybe I haven’t looked closely enough at the Savageau
papers.

        The "direction," of course is a metaphor for varying

relative rates of change in some set of parameters, each
considered as a
different dimension ibn some hyperspace.

It shouldn't be a metaphor. It isn't in e-coli reorganization of the

hierarchy, because if one move is in a direction {delta x1, delta
x2, … delta xn}, the next move (if things don’t get worse) is
easily made to be k*{delta x1, delta x2, … delta xn}. There’s no
problem with that, in my mind. It just is a continuation in the same
direction. But what I find difficult is to see how you continue the
direction if you have changed a C to a T in a DNA strand and “things
get better”. How do you do more of the same? Change more Cs to Ts?
That would usually create more of the same effect, in the way that
changing the weights in the hierarchy would usually do (until you
got to a dynamic bifurcation point).

The way I look at evolution is that even if you are right, and I

think you probably are, about stress changing mutation rates (I saw,
but can’t put my hands on, a recent Science or Nature squib
suggesting a mechanism, which I meant to post here), it is the
population that matters. An effective mutation affects only one of
the trillions of cells in the organism’s body, so it can’t do much
about changing the protein production rates in that individual, but
it could influence the rates in the offspring of that individual. If
they are more likely to pass on the mutated gene than others of the
species are to pass on the unmutated gene, then it is likely that
the mutated gene will be seen many generations down the road. To me,
that’s all that survival of the fittest means (as Dawkins says).

Anyway, my question wasn't about all of that. It was just about how

one would specify a direction in gene space so that subsequent
mutations could expand on the benefits of a good mutation. I don’t
mean specify for our analytic purposes, but specify in a way that
would make practical sense within the organism, in the way that it
does for reorganizing the control hierarchy.

Martin.

···

On 2011/12/15 6:47 PM, Bill Powers wrote:

[Bill_Powers1]

[From Bill Powers (2011.12.16.-220 MST)]

Martin Taylor 2011.12.16.00.36

[From Bill Powers
(2011.12.15.1645 MST)]

[Martin Taylor 2011.12.15.16.36]

MT: I have two questions related
to this thread, one long-standing, the other new because I think Bill has
introduced something new into the conceptual space.

1. How does the e-coli process escape from a local
   optimum?

BP: The basically random variations will occasionally introduce large
excursions that can help get out of local traps. I don’t know what the
limits are, or the kinds of local minima found in this application. I
would expect that if being stuck in a local minimum causes increasing
problems for the organism, reorganization will increase, the larger error
producing faster and coarser changes. That might work.

So a bit a simulated annealing in conjunction with
hill-climbing.

I don’t know what “annealing” means in this context. Another
far-fetched physicist’s metaphor? There’s another point, however, that I
missed. We’re talking about evolution here, and there’s no guarantee of
escaping from any local minimum, is there? The ultimate cure for a local
minimum is that you die, or your species ends.

There is probably some relationship between the probability of
encountering an inescapable local minimum and the average number of
offspring per generation. If the former is small enough and the latter
large enough, the species survives.

MT: 2. I understand what defines
a direction in the reorganization of a control network, but what defines
a “direction” in genetic space?

BP: I don’t know very much about genetics, but I’d guess the parameters
being varied would be relative rates of protein synthesis and any similar
continuous variables specified in DNA. Look up Savageau – he was
interested in control systems in DNA:

Savageau, M. A. (1976) Biochemical systems
analysis:

a study of function and design in molecular biology.

Reading, MA: Addison-Weseley.

I haven’t read the book, but later papers by him available from Google
Scholar that cite it seem to be dealing with the feedback systems
involved with gene regulation, not with altering the genes themselves.
Maybe I haven’t looked closely enough at the Savageau
papers.

This point seems to be addressed by the paper Rick cited, which gave rise
to a lot of other studies.

Cairns, J., Overbaugh, J and Miller, S. (1988) The origin of
mutants.

Nature, 335, 142-145

Here the genes themselves mutated so E. coli became able to live on
galactose or glucose alone when that was all that was available. Of
course I don’t know if those specific mutations were done in the E. coli
style, but they were definitely generated inside the bacterium as a
stress response. There’s little doubt about mutation rate being varied as
a stress response in many organisms. Those with the shortest generation
time are the best known, of course.

My only justification for saying that evolution must use the E. coli
strategy of reorganizing is the enormous advantage of doing so, relative
to the other alternative of just making step-changes at random – at
least as far as my demos have shown.

I went to the Denver Museum of Nature and Science recently to take some
pictures of an evolutionary sequence I had noticed a few years ago, only
to find that all the exhibits were different. I couldn’t find anything
like that one, but I’m sure there are ways to find examples.

The sequence showed some small creature’s fossils over a long stretch of
time, tens of millions of years. What was striking was how non-random the
changes were. The bones got thicker and longer and moved around a bit in
relation to each other, but they were clearly the same bones changing
shape and size through a continuum. It was only some years later that I
came across the concept of genetic drift and was struck with the idea
that I had been looking at the “swimming” phase of E. coli
reorganization. The sequence ended either when that species became
extinct – or when there was a tumble and it turned to a different
direction of development and became some new species.

There are some well-known sequences of this sort starting with Eohippus,
the evolution of the Equus horse line. Over 55 million years, starting
with a dog-sized organism (I shouldn’t say "creature,’ should I?),
the size and shape changed, but each feature morphed smoothly into the
next with most bones existing through the whole sequence. You never saw a
leg sprouting from the back, or the jaw starting to open sideways instead
of up and down. The hypothesis would be that these changes resulted in
continuously decreasing error signals so there was no need for a tumble.
Or, if we assume local reorganization rather than global, that only a few
features required a tumble now and then to get the local errors
decreasing again.

It’s a little odd that there is so little mention of the smooth
continuous changes. Maybe I haven’t read the right literature, but it
seems to me that this sort of change undermines the idea of purely random
mutations. It’s so obvious, whichever museum you visit. Just look at the
progression of dinosaurs! At some point some of them suddenly started
heading off in the direction of birds, but most of them just seemed to
get bigger, until they were big enough to succumb to an asteroid, all but
the ones who turned into something else and disappeared for that
reason.

BP earlier: The
“direction,” of course is a metaphor for varying relative rates
of change in some set of parameters, each considered as a different
dimension in some hyperspace.
MT: It shouldn’t be a metaphor. It isn’t in e-coli
reorganization of the hierarchy, because if one move is in a direction
{delta x1, delta x2, … delta xn}, the next move (if things don’t get
worse) is easily made to be k*{delta x1, delta x2, … delta xn}.
There’s no problem with that, in my mind. It just is a continuation in
the same direction.

BP: The concept of direction starts, for most people, with X, Y, and Z in
three-dimensional space. It’s not easy to extrapolate to any three
independent variables, and from there to any number of independent
variables. For most people, the extrapolations are metaphors.

MT: But what I find difficult is
to see how you continue the direction if you have changed a C to a T in a
DNA strand and “things get better”. How do you do more of the
same?

BP: That’s a discrete change, of course. But what about insertions and
deletions? These have the effect of changing the relationships between
sequences that occur before and after the changed part. And there’s also
the effect on folding of the DNA and the proteins synthesized according
to the sequences. We’re talking about 20,000 genes, and each gene
consists of a lot of base pairs, so the discrete changes quickly begin to
look smooth as you back away from the close-up view.

MT: Change more Cs to Ts?
That would usually create more of the same effect, in the way that
changing the weights in the hierarchy would usually do (until you got to
a dynamic bifurcation point).

BP: Yes, that’s what I thought, too.

MT: Anyway, my question wasn’t
about all of that. It was just about how one would specify a direction in
gene space so that subsequent mutations could expand on the benefits of a
good mutation. I don’t mean specify for our analytic purposes, but
specify in a way that would make practical sense within the organism, in
the way that it does for reorganizing the control
hierarchy.

BP: This is probably a good place to switch from the reinforcement
concept to the reorganization concept. Reorganization doesn’t try to
“expand on the benefits of a good mutation.” It simply causes
mutations when errors increase, and stops changing them when they
decrease. Introducing direction as the mutated variable, we can see that
as long as error is decreasing, the new direction of change will be
followed. Sooner or later, the continuing change will stop reducing the
error and will start making the error larger. Then there will be a change
of direction, another mutation. Some dinosaurs will start getting smaller
again, and growing more feathers.

Best,

Bill P.

···

On 2011/12/15 6:47 PM, Bill Powers wrote:

[tedcloak]

[Ted Cloak (2011.12.16.1030 MST)]

[Martin Taylor 2011.12.16.00.36]

MT: The way I look at evolution is that even if you are right, and I think
you probably are, about stress changing mutation rates (I saw, but can't put
my hands on, a recent Science or Nature squib suggesting a mechanism, which
I meant to post here), it is the population that matters. An effective
mutation affects only one of the trillions of cells in the organism's body,
so it can't do much about changing the protein production rates in that
individual, but it could influence the rates in the offspring of that
individual. If they are more likely to pass on the mutated gene than others
of the species are to pass on the unmutated gene, then it is likely that the
mutated gene will be seen many generations down the road. To me, that's all
that survival of the fittest means (as Dawkins says).

TC: Bingo.

TC: You might add that successful mutations are extremely rare. In a
well-adapted species, in a very slowly changing environment, mutant genes
are practically certain to be eliminated. This phenomenon is sometimes
called "stabilizing selection" but it's really NatSel* under special
conditions.

When, eventually, environmental change "speeds up", and more mutant genes
are successful, the species in question does perceptibly evolve, that's
sometimes called "punctuated equilibrium", but again, it's really just
NatSel under special conditions.

TC: Natural Selection, unlike any other process proposed as a cause of
evolution, does not require an underlying _mechanism_. All NatSel requires
is abiotic background conditions that support life, and unimaginably deep
time. (And, of course, the fundamental laws of physics.)

···

_________________
*Darwin's abbreviation in his notebooks.

================================
"Please tell me, if we used Republican Eisenhower's top tax rate of 90% for
billionaires, how many billionaires do you think would decide to stop being
billionaires?" -- David Pease