In the “Evolving complex features” thread, TristanM asks:
I’m curious what the group’s thoughts are on Jonathan Bartlett’s (AKA “JohnnyB”) argument about open-loops in the AVIDA program?
I haven’t studied Jon’s argument yet, but I think some of you have. What did you think?
Why I Love AVIDA – Detecting Design in Digital Organisms
Thoughts on Parameterized vs. Open-Ended Evolution and the Production of Variability
Irreducible Complexity and Relative Irreducible Complexity: Foundations and Applications
I disagreed with johnnyb in that thread. Will be interesting to see what others here think.
Thank you Keiths! As a long time reader but infrequent commenter of TSZ I was surprised to see my name pop up in an OP, let alone help motivate one.
I bumped into Jonathan on a Facebook thread the other day, of which the subject of discussion was Lenski’s 2003 paper on AVIDA organisms. Allegedly, AVIDA can not realistically simulate the evolution of novel complex, functional features due to an open-ended loop that contributes to the function, but as I lack the programming knowledge to even begin to confidently address this assertion, I decided to try to bring JohnnyB’s argument to the attention of TSZ, where I know more discerning researchers frequent.
I am familiar enough with the ID movement to recognize that JohnnyB is proposing something unrelated to Dembski’s CSI as a metric for determining if an intelligent agency was involved in the emergence of complex systems.
Maybe that’s why I find it more interesting.
By the way I meant to say “open-ended loops” in my comment on the other thread. Not “open-loops” (sighs)
Good to see someone actually raising a defense!
Lizzie,
Who?
Wagner addresses this in Arrival of the Fittest.
Chemistry is not analogous to engineering or to programming. Alleles are common in biology. Biology is tolerant of change.
I’ve emailed Jonathan Bartlett the link to this thread.
I read:
1) I don’t think open ended loops can evolve. Open ended loops have lots of parts, and when things go wrong they go really wrong.
2) In AVIDA organisms, open loops supplied as part of the base capacity for replication. Whenever, upon inspection, we have encountered an open ended loop in an AVIDA organism, it has been traceable to the supplied (designed) portion of the software.
3) Therefore, open ended loops are a reliable indicator of design in biology. If we encounter a biological process computationally equivalent to an open ended loop in a biological organisms, it must have been designed.
4) Conclusion: I don’t think open ended loops can evolve. Therefore design.
Also, in no particular order:
– Open-ended-loops are IC, and provide a computational metric of IC.
– This metric indicates that there are degrees of IC. IC is on a continuum.
To which I would ask: If IC is on a continuum, can a system advance from just a little IC to a lot of IC by means of selection, moving small steps along the continuum? Or is each step along that continuum necessarily another instance of IC that can’t be traversed by means of selection?
TristanM raised Jonathan Bartlett’s.
TristanM:
You’re welcome.
It’s a good topic, and I’m glad you brought it up. We chide IDers for being short of arguments, so when one of them actually comes up with a new one, it’s only fair that we give it due consideration.
I’m setting some time aside tonight to read johnnyb’s paper, and then I’ll offer my two cents.
johnnyb:
Personally, I view Avida to be one of the best demonstrations of intelligent design on the planet.
Not sure what JB’s point is though. We all agree Avida is intelligently designed. Right? That’s not in dispute?
What’s interesting is a comparison of Avida with Darwinian evolution:
In Avida, the digital organisms are designed. In Darwinian evolution, organisms are not designed.
In Avida, the environment is designed. In Darwinian evolution, the environment is not designed.
In Avida, the organisms have an inbuilt potential to evolve logic functions, including complex logic functions. In Darwinian evolution, there are no such inbuilt teleological propensities.
My understanding on this ( and I hope people will correct me on this etc etc.) is that he claims that AVIDA itself could not come into existence without a designer. But its pretty clear that the program itself comes up with radically different solutions to the same problem that the designer wouldn’t have thought of. (this is almost a cliché now) and the designer doesn’t need to intervene during the process.
If we transpose this to the natural world we conclude that evolution ( ns,drift, mutation etc) are fully capable of producing complex IC living things but the process of evolution itself couldn’t have come into existence without a designer.
This seems to me to be an attempt at a rigorous justification for theistic evolution, though that’s probably not what JohnnyB intended.
What do you mean by “intelligent design”?
Be as specific as you are capable.
Funny how you go from “potential” to “teleological” without drawing breath.
Hello everyone – thanks for taking some time to discuss my work! I’ll try to answer questions as best as I can, but if the thread goes too long it will be beyond the time I can give it.
Reciprocating Bill:
“1) I don’t think open ended loops can evolve. Open ended loops have lots of parts, and when things go wrong they go really wrong.”
I think you’ve oversimplified. It is *fundamental* to computer science that results are unpredictable, the reason they are unpredictable is because there is not a smooth correlation between program and output, and this lack of smoothness prevents selective evolvability. It isn’t just that they “have a lot of parts”. Open-ended loops are especially interesting because they are the cause of *both* the non-smooth program/output correlations *and* the ability of the program to do genuinely new, interesting tasks.
“2) In AVIDA organisms, open loops supplied as part of the base capacity for replication. Whenever, upon inspection, we have encountered an open ended loop in an AVIDA organism, it has been traceable to the supplied (designed) portion of the software.”
Yes. More specifically, an open-ended loop that contributes to function.
“3) Therefore, open ended loops are a reliable indicator of design in biology.”
Actually, this is the result of #1. #2 is simply a separate confirmation of #1.
“– This metric indicates that there are degrees of IC. IC is on a continuum.”
I think this is misunderstood. It isn’t that it is on a continuum, but rather that there can be existing code that smooths out part of the mapping. If you find pieces that are evolvable that involve open-ended loops, that means that the design resided somewhere else, and we should look to see where that information came from.
“If IC is on a continuum, can a system advance from just a little IC to a lot of IC by means of selection, moving small steps along the continuum? Or is each step along that continuum necessarily another instance of IC that can’t be traversed by means of selection?”
Dembski has a great paper called “Searching Large Spaces”. He points out that you can have information that helps you evolve, but that makes the evolving system be *more* complex, not less. You can’t evolve complex systems by simply ramping up steps (see #1 – the steps do not form a smooth enough surface to do so). You *can* traverse specific steps or specific classes of steps by adding information to the equation at the beginning. This is what the “relative” is relative to – it is irreducible complex relative to the amount of starting information. So sure, you can always posit that there was a pre-existing source of information, and that’s logical. However, it doesn’t help you get out of needing design. This is much like what they’ve done with evo-devo – they have said that the information was already there, and evolution just played with a few pieces. That is a logical possibility, but it makes a bigger, not smaller problem for the earlier stages of life.
Another possibility that is often left out is that it could be that organisms themselves have non-material methods of design. That is, biology could be include conscious at the level of cells. I don’t hold to this, but it is a logical possibility to keep in mind.
Rodw –
“My understanding on this ( and I hope people will correct me on this etc etc.) is that he claims that AVIDA itself could not come into existence without a designer. ”
This is true, but it is not my claim. My claim is that (a) there are types of structures in computer programming which are basically unevolvable (technically, they are evolvable within really narrow parameters, but I think that it is a good enough approximation to say that they aren’t). (b) These structures are actually *required* to solve arbitrary (as opposed to specially-selected) problems. Therefore (c) whenever you find these structures, they have been designed, not evolved.
Then, for Avida, (a) we find these structures, and (b) they are in the originally-designed part of the organism (Avida organisms all start from designed organisms).
Here’s my challenge. Find *any* evolved algorithm (on any system, not just Avida) such that:
(a) It includes open-ended loops that contribute to function
(b) it is a Turing-complete language/system (or at least a decent approximation)
(c) the mutating system does not in itself guide the creation of open-ended loops
(d) it doesn’t loop on a favored variable (i.e. such that the variable is implicitly encoded)
I once thought I had found one, but it turned out that when you read the details that it wasn’t actually a Turing-complete language (it had recursion, but only 2 possible variables). (d) is actually a restatement of (c), but from a language perspective (i.e. the number of symbols guides the creation of open-ended loops). What (c) and (d) are, really, is just implicit information – it makes a certain class of open-ended loops close-ended. Anyway, the main purpose of (c) and (d) are to avoid trickery and slight-of-hand by the experimenter. For most purposes (a) and (b) are sufficient.
Anyway, if you know of *any* evolved algorithm published *anywhere* that meets the above criteria, I would love to see it. It isn’t just that I say that these things aren’t evolvable. It is both that (a) they are unevolvable based on the fundamentals of computer science, and (b) they have not been shown to evolve anywhere within computer science.
A quick note on my technical note, for those interested:
“technically, they are evolvable within really narrow parameters, but I think that it is a good enough approximation to say that they aren’t”
Because open-ended loops behave chaotically, there is no selectable path to their creation. Therefore, one has to arrive at the initially-selectable loop entirely by chance. Therefore, the chances of evolving an open-ended loop that is initially selectable is 2^x, where x is the smallest size of the loop that is selectable, in bits. So, if my smallest selectable loop took 5 bits to code, then the chances of this happening are 1 in 32. However, truly open-ended loops require quite a bit more bits. For instance, the loop in Avida, is, I believe, 19 bits. (Additionally, Avida technically violates rule (d) above, so it should normally be about 5-10 bits more). That means that it will only occur 1 in 500,000 organisms. But it is worse than that. It only has that probability *if* the organisms on the way aren’t being selected out. If natural selection is in play, then the probability goes down even faster, because, since it is not a smooth path, the predecessors will all be selected away. Therefore, the only way to generate these is to turn the mutation rate up to a sufficient level to microwave the digital organisms.
In my main paper (the third link in the OP), this is why I gave the exception for “trivially small” open-ended loops. Some programming systems may allow for the creation of trivially small open-ended loops, but that doesn’t really do anything to disprove the overall argument, because there are a fixed number of trivially small open-ended loops, and an infinite number of larger ones. Thus, for any non-trivial system, the chance of hitting these is effectively zero.
Admins, please check for posts by johnnyb stuck in moderation. thank you
Four posts approved.
A welcome to JohnnyB. And thanks to Mung for pointing out the issue.
Welcome johnnyb!
I just yesterday finished your chapter on calculating software complexity in Engineering and the Ultimate
Hey johnnyb,
glad you are here
Would you agree that a designed object is something that can not be arise by algorithmic processes and is therefore non-computable?
Peace
I’ve only skimmed your paper; I understand that your approach involves ideas that Jeff Shallitt would have the appropriate expertise to comment on in detail.
Have he or any of his peers provided any feedback?
I did not find a reference to you using a quick search at his blog.
He does sometimes show up at TSZ.
Fifthmonarchyman –
Not quite. Designed objects *may* be things that can be algorithmically approached – it depends on the task. For a simple enough task it may not be possible to distinguish the two approaches. However, for tasks that require open-ended loops to solve, then an intelligence would have had to perform non-algorithmic work somewhere in the causal chain (either by direct design or by adding information to the system at an earlier point that allows it to traverse certain classes of problems.
BruceS –
There have been very few who have read and responded to my papers. I believe Nick Matzke gave a thoughtful response at one point. However I doubt that Shallitt has read it.
Can anyone supply an example of a biological feature that could not have evolved?
Welcome to TSZ johnny B! On a train right now but I hope you will get some substantive response. Will try to respond myself later
I am willing to accept — arguendo — that AVIDA is a flawed model of biochemistry and evolution. I simply don’t see the point.
The model is not the object. It is simply stupid to argue that because a feature does not appear on the map it does not exist.
Weasel and AVIDA model the effects of selection. They do not model chemistry.
Computer code is syntactical and fragile. A bit error in an instruction or a memory address will be fatal.
Biochemistry is not syntactical. There is no (current) way to predict the effects of a change in a protein coding or regulatory sequence.
But we know from observation that any given sequence has a number of possible synonyms and near synonyms. We also know from observation that when a sequence changes to a synonymous sequence, that this changes — in unpredictable ways — the list of synonyms. this is the subject of Wagners “Arrival of the Fittest”.
This fact allows sequences to vary indefinitely from the original type (borrowing a phrase from Wallace). Without crashing and burning. I daresay that Johnnyb cannot point to a computer program that can be varied one bit at a time until all, or the majority, of code has been replaced, and yet this can happen in biology.
So it is evident that computer programs are not a particularly good metaphor for biochemistry.
Edited for spelling and such.
“I am willing to accept — arguendo — that AVIDA is a flawed model of biochemistry and evolution. I simply don’t see the point”
I agree that it is not a model of biochemistry but my point in evolution is that it models it very well.
“Computer code is syntactical and fragile”
This is not a necessary feature of computer code (at least for the meanings I take you to be indicating). Many people have designed Turing-complete systems that have no syntactical fragility and no abortive error conditions. AVIDA is one such system.
However, all systems – computational and biological, have to deal with problems in recursion. This is where chaos theory comes from – the problem of recursively acting systems, and it is present everywhere that recursive systems exist, not just in modern computing.
“This fact allows sequences to vary indefinitely from the original type (borrowing a phrase from Wallace). Without crashing and burning. I daresay that Johnnyb cannot point to a computer program that can be varied one bit at a time until all, or the majority, of code has been replaced, and yet this can happen in biology.”
Do you have evidence that this is the case in biology? arbitrary point mutations are often highly damaging or even lethal. Most people point to the number of mutations we have each generation without realizing that most of our mutations are not arbitrary but guided by programming in the cell.
If you know this for a fact, why do IDist make such stupid bullshit arguments?
Read Wagner. Or anyone else who has been saying this since about 1970.
Johnny, please provide a before an after example from actual biology of a sequence change that was programmed, and while you are at it, please provide the program that caused the change.
And I’m still waiting for a specific example of a feature or structure that could not have evolved.
Hi Jon,
Welcome to TSZ!
You asked petrushka:
Petrushka isn’t talking about arbitrary point mutations. He’s saying that for any particular point in sequence space, there is practically always an adjacent point (accessible by a point mutation) that maintains the function.
You can navigate your way through sequence space one mutation at a time, maintaining the same function, until your sequence bears very little resemblance to the original.
Andreas Wagner’s book is a must-read for anyone involved in the evo debates. I highly recommend it to you.
Petrushka –
There is nothing that is blanket “unevolvable”. There are things that are unevolvable without prior information. That was the point of my main paper. If you find something that is apparently unevolvable, yet it evolves, that gives de facto evidence of the existence of information existing elsewhere. This is a way that ID can help guide evolutionary biology research.
As for examples of programs, the simplest is the usage of Simple Sequence Repeats as tuning knobs, and a more complex one is the adaptive immune system. You should read some books/papers by Caporale, especially “The Implicit Genome”.
Then if some future generation of AVIDA like programs could accurately model chemistry, we might have something to discuss. I suspect, however, that chemistry is faster than any likely simulation of chemistry. Being of limited intelligence, I cannot envision anything less than a quantum computer that could simulate protein folding in real time. And we already have protein folding.
Then there’s the small problem of simulating the behavior of a new protein or a new regulatory network. The design problem expands exponentially.
Give me an example from biology. A specific sequence that could not have evolved.
Perhaps we should all hike over to AtBC and read Gary Gaulin.
Petrushka and keiths –
I’ve read a lot of Wagner – I actually started a couple of discussions on his papers at UD. I have not read “Arrival” but I haven’t heard of anything in there that is not already in his earlier work. It’s on my to-read list, but towards the back.
Now the claims of keiths and Petrushka on Wagner’s work are very different. I agree entirely with keiths that you can mutate a sequence and have a very different-looking sequence that continues to perform the same function. But this is largely a function of the degenerate coding of the genome, and makes pretty much no difference to the argument I’ve put forth here.
Petrushka’s claim is that you can get from any genomic configuration A to any genomic configuration B with mutation. Clearly that’s what Wagner believes, clearly that’s what every Darwinist believes. The question is, is there evidence that this is the case? Wagner believes it, and I agree he has been arguing for it since the 1970s. But what is the data that says these kinds of adaptations can come into existence?
“Give me an example from biology. A specific sequence that could not have evolved.”
Petrushka –
Go to my paper (the last link in the OP). Read section 3.1.
I should be so luck to have $150 to spend on a Bible Code tome, that has zero reviews among biologists and zero impact, and which ignores genetic load.
You seem to be assuming that features that exist were at one time goals or targets.
johnnyb,
Sequence similarity across functionally distinct proteins. It does not exclude design, but it is supportive of common descent between proteins. This is given further weight by ancestral protein reconstruction, using evolutionary models to (approximately) reverse the sequence of substitutions.
Petrushka –
“Then if some future generation of AVIDA like programs could accurately model chemistry, we might have something to discuss. I suspect, however, that chemistry is faster than any likely simulation of chemistry.”
This is hilarious. You are basically saying that you think that we should not do biological modeling. I bet you also say ID is anti-science. Science *relies* on imperfect models to teach us things. If there is something specific about the model you don’t like, say so, but simply saying that the model is not an exact replica of reality is just being silly. No model is equivalent to reality. That’s like being against using the ideal gas law because ideal gases don’t exist.
Alan Miller –
“Sequence similarity across functionally distinct proteins. It does not exclude design, but it is supportive of common descent between proteins.”
It is supportive of common descent, but that isn’t what is at issue here. The question is whether it took information to perform the evolution, whether that information was in the beginning (a la Behe), or introduced along the way (i.e., some form of special creation). Even as evidence of descent, it doesn’t on its own (i.e., without considering the specifics) evidence whether the descent was information-guided or guided only through selection.
johnnyb,
Degenerate coding is only a part of it. Function is highly robust even under amino acid substitution, which is quite significant for evolution.
I haven’t read your paper yet (I’m finishing up a weekend coding project), so I can’t comment on it, but I did want to point out why petrushka was referencing Wagner.
Natural Selection selects against evolution of certain complex features when the intermediates are malfunctioning. Avida doesn’t model that.
As far as Petrushka’s request for unevolvable features? How about tRNA-synthetases.
tRNA-synthetases are required for protein production and which included tRNA-synthetases themselves. Without tRNA-synthetase in one generation, there is no tRNA-synthetase in the next. Ergo a tRNA-synthetase dependent system without tRNA-synthetase to start with won’t evolve since it would be dead!
Some will say the evolution of tRNA-synthetases is an OOL problem. Fine! The fact it maybe an OOL problem is evidence it’s not evolvable under evolutionary theory.
Natural selection can’t select for features that don’t exist yet.
Of course one could invoke astronomically improbable events that are not distinguishable from miracles as Koonin did, but at that point the explanation doesn’t look that much different from creationism.
No, Johnny, Petrushka is not saying that, at least I don’t think so. I think he is making a muchore specific point than you are interpreting as, and one I wanted to male myself.
Biochemistry is difficult to model, specifically biochemistry, and AVIDA doesn’t attempt to do so. Instead it instantiates, in a computer model, the mechanism proposed by Darwin for adaptive evolution, given an initial population of self_replicating entities and an environment on which there is competition for resources and in which some phenotypes can access those resources better than others
As such, it falsifies Behe’s IC claims.
You, however make a different and potentially interesting claim. But if we want to investigate it we have to somehow map biochemistry on to computer code, something AVIDA doesn’t attempt to do, and which, as Petrushka says, is very difficult.