Intelligent Evolution?

[Southwind17]

Any chance of a response to my Post #1650 jimbob?:

Jimbob, let's backtrack a little please. In my post above I asked what you see as the trigger for replication after the cheetah is born. I asked it like this because I'm tying to get you to realize that replication is dependent upon survival (obviously - a cheetah cannot breed if it's dead!), and survival is dependent upon how well equipped the cheetah is in the context of its environment, as cyborg clearly acknowledges above. I am puzzled, therefore, when you respond: 'conception', because at that stage it is too early to measure the effects of any mutations that the embryonic cheetah might undergo. It's critical to the analogy that we reconcile this fundamental difference in understanding.

Your duck-shooting analogy, to my mind, contradicts your 'conception' response, as, by your own admission in using the duck-shoot analogy, the replication trigger isn't pulled unless and until the duck makes it to the other end of the booth, i.e. the trigger is pulled at the end of the booth and not at the beginning, the fundamental difference being that only by making it to the end of the booth can we say that the duck was evidently suitably equipped to survive.

Could you please reconsider this and clarify your position jimbob. Thanks.

 

[Walter Wayne]

See above. They're not stepping stones (apart from any good aspects that are retained (evolution!)). They simply consitute the mutations that failed to proliferate because of weakness, and maybe gave the engineers clues as to what might actually work (again, the alternative being continuous trial and error).

No, the mutation to the original product may proliferate anyway. If one has design A, one can perform a mutation to design B, which fails, then continue performing changes to design B. In intelligent design, one isn't required to go back to design A, one can mutate design B to the next and so on.

Tech
A -> B (fails) -> C (fails) -> D (fails) ...... -> P (success)!
Bio
A -> B (fails)
A -> C'(fails)
A -> D'(fails)
....



Building on failures. Especially important in competitive environments.

Exactly. We keep what works and ditch what doesn't. Isn't that what nature does?

See above.

Irrelevant to the comparison. The case for increased complexity over time has already been proven - just look around you!

Relevant, being able to go back to long defunct designs and modify them to work with new info certainly changes the nature of evolution. It is why we can see such a nice chronological lineage in biological evolution.

What, exactly, do you mean by 'non-functional'?!

As an example, completely inferior to its contemparies.

 

[jimbob]

Jimbob, let's backtrack a little please. In my post above I asked what you see as the trigger for replication after the cheetah is born. I asked it like this because I'm tying to get you to realize that replication is dependent upon survival (obviously - a cheetah cannot breed if it's dead!), and survival is dependent upon how well equipped the cheetah is in the context of its environment, as cyborg clearly acknowledges above. I am puzzled, therefore, when you respond: 'conception', because at that stage it is too early to measure the effects of any mutations that the embryonic cheetah might undergo. It's critical to the analogy that we reconcile this fundamental difference in understanding.

Your duck-shooting analogy, to my mind, contradicts your 'conception' response, as, by your own admission in using the duck-shoot analogy, the replication trigger isn't pulled unless and until the duck makes it to the other end of the booth, i.e. the trigger is pulled at the end of the booth and not at the beginning, the fundamental difference being that only by making it to the end of the booth can we say that the duck was evidently suitably equipped to survive.

Could you please reconsider this and clarify your position jimbob. Thanks.

Yes.

The trigger for the self-replicator to start the process of self-replication is at its inception (i.e. conception in a cheetah). This process is usually incomplete.

To anthroppomorphise for a paragraph; the "raison d'être" of the self-replicator is to self-replicate. This is "why" it was conceived. Its "mission" begins when it begins, as does the process of self-replication. The fact that most organisms do not reproduce is because the process has been interrupted after it started, but before the process finished.

I agree that this is not the case in your analogy, where the replication trigger is the receipt of the sales information.

 

[jimbob]

What, exactly, do you mean by 'non-functional'?!

As an example, completely inferior to its contemparies.

The "function" of an organism is to reproduce, so a non-functional organism is one that doesn't reproduce.

Pretty difficult for such an organism to have descendents. However it is quite possible for failed prototypes to lead to successfull designs.

 

[cyborg]

In technology you do. Because if you can deduce why it failed, you can take a step based on that failed design rather than going back to square one.

That is NOT a design based on failure. That is failure analysis.

These are different things - you do not design to failure, you avoid designing failure.

You can produce something which is completely inferior to anything out there, or doesn't work at all, and still base the next generation on that.

Sure you can.

I'd have to wonder what bloody good intelligence is if it is applied to producing things that are completely inferior to everything out there and what good it is basing inferior products on this inferiority.

This to me basically sounds like: "Intelligence can reinvent the wheel!"

That is a design anti-pattern - i.e. not what intelligent people should do.

I can only guess what you mean by flagell. In the context would I guess correctly that it is the action of a flagellum?

What else could it possibly be?

Thats my point. In addition, there are still many produced that are completely inferior to current receivers, or altogether useless. They still serve as stepping stones in innovation.

So how is this different to evolution again?

We build on those technologies even though they are now extinct.

If I based my future technologies on extinct technologies when there are better technologies developing out there how exactly is this a good thing?

Every time Ford brings out a new car they DO NOT go back to the Model T designs - that would be stupid.

You really seem to be arguing that the great thing intelligence can do that evolution is not is make really, really stupid decisions.

Biological evolotion won't produce a variation on the t-rex now, because it sees a niche.

What niche Model T now other than to fulfil the arbitrary whims of car enthusiasts?

I think so far we can say some intelligences are nostalgic whilst evolution is not.

I should have said relied on unviable designs as well. It should have been noted from context that unviable stepping stones can exist in the technological tree.

How do you know you have a flagellum until it functions?

Are not the "unviable" flagella the stepping stones?

Are you really thinking about what you are arguing must be something found in intelligence design not found in evolution?

P.S. Can you provide an example of a living organism that counts a non-functional organism amongs in predecessors?

No - organisms that are "non-functional" are also typically called, "soft gooey organic matter that does nothing much of interest".

Now can you please give me an example of a technology with non-functional predecessors?

 

[cyborg]

No, the mutation to the original product may proliferate anyway. If one has design A, one can perform a mutation to design B, which fails, then continue performing changes to design B. In intelligent design, one isn't required to go back to design A, one can mutate design B to the next and so on.

Tech
A -> B (fails) -> C (fails) -> D (fails) ...... -> P (success)!
Bio
A -> B (fails)
A -> C'(fails)
A -> D'(fails)
....

Your tech way seems stupid to me.

Pool of designs: A, B, C, D, E, F, G

Right. Try A. No good? Okay, try B. Ugh. Try C... no go...

Well if you WANT to waste time doing it sequentially.

The evolutionary approach is:

Try A, B, C, D, E, F and G at the same time. D and G work? Fantastic - let us build on those...

Parallel processing is more efficient. Do you think we'd have the technology we do if we waited for ONE designer to continually crank out duds?

 

[cyborg]

The "function" of an organism is to reproduce, so a non-functional organism is one that doesn't reproduce.

At this level we cannot really say anything more about the structure of an organism then can we?

However it is quite possible for failed prototypes to lead to successfull designs.

If a failed prototype isn't reproduced how will it ever be successful?

If "success" is defined as "reproduction" then the prototype wasn't a failure.

 

[jimbob]

They are all playing their own game in their own head. It just gets weird trying to even figure out what the point is. And then you realize they have no point. Their requests for information are just means of keeping their game alive-- the one they are always winning but no one else is even aware of. If they're not chatbots, they may as well be for all the communication they are capable of.

My point is that it is a misleading analogy because:

1) It is superficially attractive.

2) In some statements e.g.

Photocopiers evolve.

the role of random mutation is completely ignored and equated with the results intelligently-driven source of variation.
If all "mutations" were beneficial, then that would be evidence for IDiots.

Random variation is important, as that is the mechanism that allows the creation of beneficial traits, and increasing complexity combined with section to remove deleterious traits.

Random variation allows one to dispense with top-down design; if all variations were beneficial, then something would be directing these variations, and complexity would be limited by the "analytical ability" of this agency.

3) Nowhere in the anaogy has the importance of self-replication to biological evolution been highlighted, nor the fact that this leads to natural-selection without the need for intelligently defined "goals" or "direction".

4) IDiots are trying to equate evolution with technological development; they are also trying to muddy the distinction between evolutionary approaches to engineering and evolution, because they claim (in this case correctly) that engineering using evolutionary approaches still requires a guide to "direct" the evolution evolutionary approach towards their intelligently defined goal.

This can be countered by highlighting the fundamental differences between evolutionary approaches to engineering and actual biological evolution. Namely the lack of self-replication and consequent need for an intelligent agency in the evolutionary approach, and the converse for actual biological evolution.


​

An example of technical development which is not akin to evolution:

Did Remotely Operated Vehicles (ROVs) evolve?

Which ROV was the ancestor for this ROV's design?

The design used features of previous ROVs, and intelligent analysis of examples from nature to help create the initial design. The iterative process was nothing like evolution. I am not claiming that an evolutionary algorithm couldn't have created such a ROV, but that a it didn't.

Because technology is not only reliant on descent of the current generation, products can be cloned without worry that a change in environment will result in extinction.

Technologies have gone extinct by virtue of obsolescence.

Hadn't the hand-cranked gatling gun "gone extinct" by the 1920's (or even WWI)? Because technological development is not evolution, essential features of the design were later "resurected" in a way that has no analogy in biological evolution.

 

[jimbob]

I design transistors. A key parameter is their blocking voltage. As well as the transistor's active area (the switch) one needs other structures (edge-terminations) around the outside of the device.

The active area has a characteristic blocking voltage, and the edge-termination also has a characteristic blocking voltage. Both these can be altered.

We produce test structures consisting of only the edge-termination, and others with a very conservative edge termination. This way we can determine the blocking voltage entitlement of both parts of the structure. Neither of these test structures would be any good in any applications, so they are "failures" in that sense, but they provide invaluable information that allows us to alter the design of the actual device.

This sort of practice is common in engineering.

There is no analogous process in evolution.

 

[cyborg]

It is superficially attractive.

The argument for "intellegence" being a unique creative force seems rather superficially attractive to me.

I am not claiming that an evolutionary algorithm couldn't have created such a ROV, but that a it didn't.

Then we're still waiting to know what it is evolution cannot achieve that makes intelligence special.

Hadn't the hand-cranked gatling gun "gone extinct" by the 1920's (or even WWI)? Because technological development is not evolution, essential features of the design were later "resurected" in a way that has no analogy in biological evolution.

No analogy in biological evolution?

What about the fact that although "teeth" are obsolete to birds the gene for their creation can be switched on?

The information is still there even if it is not used.

If you have it you can use it. If you don't then creating the information over again is simply hard: intelligent or not.

 

[cyborg]

We produce test structures consisting of only the edge-termination, and others with a very conservative edge termination. This way we can determine the blocking voltage entitlement of both parts of the structure. Neither of these test structures would be any good in any applications, so they are "failures" in that sense, but they provide invaluable information that allows us to alter the design of the actual device.

This sort of practice is common in engineering.

There is no analogous process in evolution.

Jimbob - that is the very essence of evolution - you start off with a sub-optimal system and gradually remove the defects.

 

[jimbob]

Jimbob - that is the very essence of evolution - you start off with a sub-optimal system and gradually remove the defects.

No, that is the very essence of an iterative process.

With evolution, only the descendents of those organisms that reproduce, continue to evolve. There is no "targeted" variation, and no information from failures.

A sterile cat won't have any descendents for the sterility to be "corrected".

 

[cyborg]

With evolution, only the descendents of those organisms that reproduce, continue to evolve.

I'm not particularly interested in those that aren't good enough not to in any type of evolutionary system by default.

There is no "targeted" variation,

Yep. There is. Sexual selection is "targeted" variation and a whole slew of evolutionary mechanisms have evolved to maximise the potential genetic strength of offspring and minimise detrimental effects from plants evolving various mechanisms to prevent self-pollination to the complex pheromonic messages that can signal subtle information about the nature of the genome of two individuals such as immune system types.

Systems that "target" improved variation inevitably arise in an evolutionary system that can produce them. They self-represent their own chemical rules for what constitutes a more successful genome in the environment in which they exist.

This is a pretty subtle thing to appreciate.

and no information from failures.

Apart from the negative space of the successes of course.

A sterile cat won't have any descendents for the sterility to be "corrected".

Unless that sterility is inexorably linked to some "improvement" I really don't see why that's an issue.

Either you want to see the trend or you are going to continue to grasp at the ever decreasing edge cases. If you want to demonstrate some significant difference then a more significant aggregate would be better.

 

[Walter Wayne]

Your tech way seems stupid to me.

Pool of designs: A, B, C, D, E, F, G

Right. Try A. No good? Okay, try B. Ugh. Try C... no go...

Well if you WANT to waste time doing it sequentially.

It isn't just sequentially, it is cumulatively. If a technological designer finds that his mutation, while conceptually good fails due to some unforseen parameter, he can keep the mutation he made while adding in another to fix the shortcoming.

One reason mimicing biological evolution will never produce anything resembling what we have in technology is it learns nothing from design B. It can't mutate a defunct design, and instead goes back A. When it produces the next mutation, it applies to our successful design A, and gets C', while technological innovation can, if there is potential, apply that mutation directly to B to get C (note the C and C' to denote the designs are not the same).

Try A, B, C, D, E, F and G at the same time. D and G work? Fantastic - let us build on those...

The primes I noted were to show difference in resulting design. That is design B in both cases is the same. But evolution applies the second mutation back to A, and get C' as opposed to innovation which applies the new mutation to B and get a different design C. (I definitely should have chosen a different notation).

The designer, seeing that the concept of the mutation is still good but is not compatable with other elements in the design applies a mutation to an inferior product.

This is the point I was trying to make earlier with my point about wholesale changes that can happen in a design. Note that the above example was 6 mutations. To try out 6 different mutations there are 63 possibilities to test (i.e just mutation 1, or mutation 1 and mutation 3, or any combination of the 6).

Parallel processing is more efficient. Do you think we'd have the technology we do if we waited for ONE designer to continually crank out duds?

Going from one product generation to the next almost every part on 1000+ part products can be "mutated" by the designer. The number of major alterations done to a product when making the "Next-Gen" is large. For biological type alogorithm the number of combinations of those alterations, not including those that a designer would never even try, grows exponentially.

So is doing several mutations sequentially more efficient that trying out hundreds or thousands of mutations in parallel? No.

Walt

 

[Walter Wayne]

Sure you can.

I'd have to wonder what bloody good intelligence is if it is applied to producing things that are completely inferior to everything out there and what good it is basing inferior products on this inferiority.

This to me basically sounds like: "Intelligence can reinvent the wheel!"

That is a design anti-pattern - i.e. not what intelligent people should do.

It can be very intelligent. Suppose a designer sees that an android could benefit from a complete reconstruction of its leg design, a very radicle change. He foresees that the new leg will require a new pneumatic system to keep it supplied with fluids, and a more powerful pump is needed to keep the fluids going.

He builds the unit, and it fails completely. It is useless because the power plant can not provide the new and improved pump with the power it needs. Should he stop here and go back to the original design. Or should he mutate this design to incorporate, the new power plant and new cooling system to keep it cool. The final android lineage includes an ancestor that was a failure.

Leave out any one of those mutations and the new robot is inferior to the original because it doesn't function, over-heats or takes excess energy to build because it is over designed for what it does. In two iterations five systems were mutated.

Assuming that no other mutations are possible, there are 31 combinations that could have been attempting, and through foresight and learning from a failure we got there in two steps.

Still think it is more effiicient to design "blindly"?

Now can you please give me an example of a technology with non-functional predecessors?

I can point to a number of proto-types sitting around lab bins in product manufacturing companies.

Walt

 

[jimbob]

Apart from the negative space of the successes of course.

Albino animals generally fail to reproduce, so albinism in wild animals shouldn't occur, because it has already been "tried"?

There is nothing stopping a harmful trait being "tried" many times. It is unlikely, but you were arguing that just because something is so unlikely that it wouldn't happen in the life of the universe doesn't mean that it couldn't.

Intelligent agencies can avoid repeating the same mistake. They can also analyse failure, and see why the failure occured. They can then corrrect the failure in the next design iteration, so a failed prototype provides information that a "failed" organism doesn't.

Intelligent agencies can deliberately reuse design features, whilst evolution can't.

In the case of the GM mouse that I talked about, it was not the functional part of the GFP gene that prevented the exact 700+ letter sequence from evolving independantly in a mouse, it was the non-functional part, for which there would not be any selective pressure.

 

[Southwind17]

It isn't just sequentially, it is cumulatively. If a technological designer finds that his mutation, while conceptually good fails due to some unforseen parameter, he can keep the mutation he made while adding in another to fix the shortcoming.

You have to realize that, for the purpose of this discussion, we can disregard detrimental mutations, as they serve no purpose in validating the analogy (and they do not invalidate it either; they can simply be considered an 'ancillary' process running in parallel with successful mutations).

So, we're left comparing indifferent or successful mutations with beneficial design changes. If a designer sees some 'potential' (or even indifference) in a design modification, even though a further modification is required to realize that potential, as you describe above, that is no different from natural evolution. Provided a mutation isn't detrimental there's no reason why it should disappear. It will, therefore, hang around until some further mutation occurs which complements it and the potential is then realized. No difference between natural evolution and technological development - just TIME, as I keep saying!

One reason mimicing biological evolution will never produce anything resembling what we have in technology is it learns nothing from design B. It can't mutate a defunct design, and instead goes back A. When it produces the next mutation, it applies to our successful design A, and gets C', while technological innovation can, if there is potential, apply that mutation directly to B to get C (note the C and C' to denote the designs are not the same).

I think you're getting yourself confused because of the almost inconceivably different timescales over which natural evolution and design technology occur. I commented earlier that if the Wright brothers had subsequently evolved their first successful attempts at flight through a process of trial & error, as they could have, instead of applying the 'convenience' of forethought, we might just be flying around in bi-planes by now. I suspect that might actually be over-optimistic by a factor of 1 followed by, oh, one or two zeros, so to speak!

The primes I noted were to show difference in resulting design. That is design B in both cases is the same. But evolution applies the second mutation back to A, and get C' as opposed to innovation which applies the new mutation to B and get a different design C. (I definitely should have chosen a different notation).

Covered off above.

The designer, seeing that the concept of the mutation is still good but is not compatable with other elements in the design applies a mutation to an inferior product.

Covered off above. By definition, the initial (poor) modification must have some potential benefit for the designer to retain it. If not, he would simply scrap it and return to go. If he keeps 'fiddling' with an otherwise detrimental modification until it becomes advantageous, then the first time that an advantage emerges can be considered to be THE mutation. Everything else was just trial and error, the same as detrimental natural mutations, which come to nought.

This is the point I was trying to make earlier with my point about wholesale changes that can happen in a design. Note that the above example was 6 mutations. To try out 6 different mutations there are 63 possibilities to test (i.e just mutation 1, or mutation 1 and mutation 3, or any combination of the 6).

Going from one product generation to the next almost every part on 1000+ part products can be "mutated" by the designer. The number of major alterations done to a product when making the "Next-Gen" is large. For biological type alogorithm the number of combinations of those alterations, not including those that a designer would never even try, grows exponentially.

So is doing several mutations sequentially more efficient that trying out hundreds or thousands of mutations in parallel? No.

Refer above re. inconceivably different timescales. Doesn't affect the principle, though!

It can be very intelligent. Suppose a designer sees that an android could benefit from a complete reconstruction of its leg design, a very radicle change.

When you write 'radical', are you taking into account the fact that, at the micro level, the technology is, at best, only small steps removed from the previous design? Does the designer, for example, suddenly and unexpectedly stumble upon new materials and components that appear from thin air, or are those material and components 'evolved' over time?

He foresees that the new leg will require a new pneumatic system to keep it supplied with fluids, and a more powerful pump is needed to keep the fluids going.

How does he 'foresee' these things. Do you, for example, foresee how man will colonize another planet in the future, such that you can go do it right now instead of waiting, or do you think that the reason we can't do it right now is because we don't yet have the technology? We need to wait for the technology to evolve. What you're saying is that we're only limited to what we can achieve by our imaginations. Clearly not true. Why do we not expect to put a man on the surface of Mars until possibly 2050? I can 'imagine' doing it. Hell, I can virtually visualize it right now. But what's stopping it happening tomorrow?!

He builds the unit, and it fails completely. It is useless because the power plant can not provide the new and improved pump with the power it needs. Should he stop here and go back to the original design. Or should he mutate this design to incorporate, the new power plant and new cooling system to keep it cool. The final android lineage includes an ancestor that was a failure.

As I wrote earlier in this thread, all he's doing is 'leap-frogging' the intermediates, simply because he can look farther ahead. That doesn't make what he's doing anything special. He can resort to trial and error, instead, and still get to where he's looking, but simply not during his generation!

Leave out any one of those mutations and the new robot is inferior to the original because it doesn't function, over-heats or takes excess energy to build because it is over designed for what it does. In two iterations five systems were mutated.

Covered off above.

Assuming that no other mutations are possible, there are 31 combinations that could have been attempting, and through foresight and learning from a failure we got there in two steps.

Covered off above. Nothing special about it.

Still think it is more effiicient to design "blindly"?

No - extremely inefficient, just like nature, but 'inefficiency' is a relative term, and you're using it in the context of a commercial environment, that's all. Again, nothing special about design development, in principle.

I can point to a number of proto-types sitting around lab bins in product manufacturing companies.

So what?

 

[Southwind17]

Yes.

The trigger for the self-replicator to start the process of self-replication is at its inception (i.e. conception in a cheetah). This process is usually incomplete.

To anthroppomorphise for a paragraph; the "raison d'être" of the self-replicator is to self-replicate. This is "why" it was conceived. Its "mission" begins when it begins, as does the process of self-replication. The fact that most organisms do not reproduce is because the process has been interrupted after it started, but before the process finished.

I agree that this is not the case in your analogy, where the replication trigger is the receipt of the sales information.

You are overlooking an ABSOLUTELY FUNDAMENTAL ASPECT of evolution jimbob. Where, in what you've described above, does 'selection' come into play? Selection is the process that weeds out inferior mutations and determines obsolescence when new, superior mutations arise. In other words, 'selection' establishes the parameters by which an organism (or a technological development) is 'tested' against the competition. That 'testing' can only occur AFTER the organism/device has come into being, and the mutations/developments are manifest. What constitutes 'passing' the test? In both cases it's REPRODUCTION.

Whilst conception is clearly a significant occurrence, it doesn't actually 'trigger' anything in the evolutionary process. The only way conception could be considered the alternative trigger to reproduction is if you're alluding to the fact that conception demonstrates that the previous generation survived long enough to effect conception. Is that what you mean? If so, then that would be equally valid, but in the AA (or Sam & Ollie analogy), the 'trigger' would simply shift from receipt of sales proceeds as an indicator of success to starting to assemble the next variant (conception). The two, as in nature, or inextricably linked, and as such, arise from the same root cause - survival of the fittest.

Are you able to clarify please?

 

[Belz...]

That is one reason why I am arguing that self-replication is important.

And this time, instead of forgetting about intelligence being irrelevant to the analogy, you forget that there is no such thing as "self-replication", and that it's ALSO irrelevant to the analogy, from an information standpoint.

 

[Belz...]

the role of random mutation is completely ignored and equated with the results intelligently-driven source of variation.
If all "mutations" were beneficial, then that would be evidence for IDiots.

Not all technological "mutations" are beneficial, even when they seem to be.