250 thoughts on “What was the most significant scientific development in Intelligent Design in 2018?

  1. phoodoo: So the definition of fitness is relative fitness, and the definition of relative fitness is the ratio of the growth rate of the derived type to its ancestral competitor during direct competition?

    So humans that grow more are more fit. Does it matter if they grow vertically or horizontally, or what? Like do we weigh them? If so, America must be the most fit country by far. I guess Russia is next.

    The growth rate is in number of individuals in the population (for a bacterial population that would be the number of individual bacterial cells), not the growth-rate (as in the becoming bigger) of a single individual.

  2. phoodoo: So the definition of fitness is relative fitness

    No, the definition of absolute fitness (in a simple model) is viability x fertility. The definition of relative fitness is the ratio of absolute fitness of one type to the absolute fitness of the other. Discussion will be found early in Chapter 2 of my freely distributed text Theoretical Evolutionary Genetics.

  3. Joe Felsenstein: A quibble. For models with growth in continuous time, we compute relative fitness by taking the difference between the growth rates. Thus if genotype A is growing at rate 0.03 per unit time, and genotype B is growing at rate 0.02 per unit time, the relative fitness of A is 0.01, not 1.50.

    For models with discrete generations, the statement is correct provided we interpret “growth rate” as the fitness of the genotype, fitness being in a simple model the product of viability and fertility. Thus in an asexual case if a newborn A gives rise to an average of 1.5 newborn offspring in the next generation, and B gives rise to 1.2, the relative fitness of A is not the difference 1.5-1.2, but the ratio 1.5/1.2 = 1.25.

    Alright, that makes sense.
    Edit: I see this is actually also what they use in the Long-term evolution experiment with E coli (ratio in growth rate pr unit time): http://myxo.css.msu.edu/ecoli/srvsrf.html

  4. Joe Felsenstein,

    So relative fitness has dimensions of “per unit time” in the continuous case, and “per generation” [is that dimensionless?] in the discrete case?

  5. Alan Fox:
    Who do you think wrote:

    Darwinian evolution is obviously false from a mathematical perspective. No idea why it has stuck around so long, except as an atheist religious dogma.

    Eric Holloway!

    Nothing is so obvious as the failure of the ID movement to recruit new talent.

  6. Joe Felsenstein: No, EricMH was (in effect) questioning whether Dembski/Marks/Ewert

    A white noise fitness surface is very nonbiological, but all-possible-searches is utterly incomprehensible, and not the right thing for D/M/E to take as some average outcome of non-Designed evolution.

    I took him to be asking for citations for why fitness surfaces relevant to biology are not white noise.

    ETA: And if I understood that request correctly, it does show a recognition that the shape of the fitness surface is important. Now if he also admits it requires empirical science, not abstract math, to determine the shape, then that will be important.

  7. Rumraket: There’s an incredible amount of unproven and frankly highly implausible bullsights stuffed into these ID arguments.

    You realize that people think the same is true of many evolutionary arguments?

  8. Rumraket: If and when a GA works (which could be some model of evolution), you will argue it works because the GA was engineered in such a way that it works.

    This is something that we know to be true. I don’t understand why it should be in the least bit controversial. Well, actually, I do understand.

  9. Mung: You realize that people think the same is true of many evolutionary arguments?

    Yes, and most of these people are ignorant. And/or religiously crazy. And/or stupid.

  10. Rumraket: “Relative fitness is a dimensionless quantity, which is calculated as the ratio of the growth rate of the derived type to its ancestral competitor during direct competition.”

    It leaves out the niche!

  11. DNA_Jock:
    Joe Felsenstein,

    So relative fitness has dimensions of “per unit time” in the continuous case, and “per generation” [is that dimensionless?] in the discrete case?

    Yes. I hadn’t thought of that issue, but you are right. I guess that in both cases the relative fitness is not dimensionless. In the discrete case we always compute it for one generation, so we don’t think about the “per generation”.

    Just as we might say that the cost of a new car is a certain number of dollars, whereas it is actually given in dollars-per-car.

  12. DNA_Jock: But I was making the same mistake as Joe, in that I was addressing the idiocy that is the white noise landscape of No Free Lunch, whereas the latest foray into cluelessness involves averaging over all search algorithms, including hill-descending algorithms.

    So unless an algorithm is a hill climbing algorithm it is not an evolutionary algorithm?

    I hope you can see why natural selection is going to do better than the average of ALL algorithms…

    And they said Darwinism is dead. Hah.

  13. BruceS: I took him to be asking for citations for why fitness surfaces relevant to biology are not white noise.

    If the fitness surface were a white-noise surface, then a change of a single base in the genome (say replacing G at position 1073284 by C) would have the same average effect as changing every base in the genome simultaneously. Which
    would totally destroy the organism.

    Since we all have new mutations in us, that experiment is done every time an organism reproduces. In effect, if the reproduction is successful it rejects the white-noise theory,

    (In any case the issue between Dembski/Ewert/Marks and their critics is not the white-noise issue but the relevance of comparing evolution to randomly chosen “searches”).

  14. Allan Miller: Why would we expect to find one, if evolution is true? Competitive extinction of primitive organisms by more ‘advanced’ ones is a reasonable expectation.

    It makes me so pleased to see you put advanced in scare quotes. It tells me you don’t really mean what you just said. Can you say what you really mean? Thanks.

  15. Mung:

    Rumraket: If and when a GA works (which could be some model of evolution), you will argue it works because the GA was engineered in such a way that it works.

    This is something that we know to be true. I don’t understand why it should be in the least bit controversial. Well, actually, I do understand.

    And if a model of erosion of riverbanks and change of the course of a stream works, then that too is because of intelligent design of the simulation. But is that evidence that erosion of riverbanks is is controlled by an Intelligent Designer?

  16. DNA_Jock: CharlieM: First he makes an assumption, concludes assumes that his assumption is correct, and then asks questions based on his initial assumption. This is the type of science that Goethe was so against.

    I encourage you to compare the track record of Goetheans versus those, such as Sulston (or Smith or Venter for that matter…), who applied a less “gentle” empiricism. You might want to read up on what we know about the development and the nervous system of the worm C. elegans, all thanks to Sulston, Brenner and their collaborators.

    It is not a question of comparing Goethe with any of these scientists. It is more a question of the progress individuals make in their research and how prior assumptions affect this progress.

    Bruce Alberts on DNA replication:

    …all that time I was trying to replicate DNA with a single enzyme was crazy because the genetics of T4 bacteriophage showed that there were at least seven proteins, that is seven gene proteins needed to replicate DNA. And so all the prior frustration that I had had, working on this process, problem, set me up perfectly for recognizing immediately that this was a system that could be used to actually see how DNA replicated, rather than guessing how DNA replicated which is what I had done for five years as a graduate student. Unsuccessfully.

    He had assumed DNA replication to be more simple than it was. One gene, one protein, one function is more in keeping with the current paradigm. Groups of proteins working together is harder to explain, but it was found to be more realistic.

    Taking heed of the Goethean method allows researchers to carry on experimenting but without prior assumptions which may well hinder their progress. It is not a competition to find the best researcher, it is a question of individual attitudes.

    Both ID and believers in blind evolution of life out of dead matter begin from these respective assumptions. I’ve lost count of the times I’ve seen written, “It’s more complex than first thought”. But this only applies to those assuming simplicity in the first place.

  17. Mung: This is something that we know to be true.

    While I’m sure there are many GA’s which have been made, and then subsequently fine-tuned in order to reliably produce a particular class of results, this is not a necessary assumption in creation of a GA. Probably many GA’s used in the aerospace industry are fine-tuned so the landscape of solutions mirrors the physics of aerodynamics as closely as possible.

    It is possible to make a GA that simulates some sort of set of physical laws where the person making the GA doesn’t actually know anything about what kind of fitness landscape emerges from the combination of laws/rules used in the algorithm.

    I think you’re mistaking the general and trivial statement that GA’s are created, and so the fitness landscapes that emerges from the laws or rules of that GA is an indirect byproduct of those rules (and thus could be said to be “engineered in such a way that it works”), with the more direct case where the fitness landscape itself is deliberately set up with intent such that the GA operating in this landscape can reliably find and move up hills (which could happen by first creating the rules of the algorithm, and then subsequently fine-tuning them a lot so the landscape is smoothed and has lots of hills).

    I think there’s a crucial distinction between those two [engineered “in ignorance” with he results being emergent] vs [engineered with intent to fine-tune for particular classes of results] situations.

  18. Joe Felsenstein: And if a model of erosion of riverbanks and change of the course of a stream works, then that too is because of intelligent design of the simulation. But is that evidence that erosion of riverbanks is is controlled by an Intelligent Designer?

    GA’s are not models of evolution.

  19. What I find interesting is the relationship between behavior learned through evolution, and behavior learned through individual experience. We have called evolved behavior instinct.

    There’s a useful analogy to be made between the tweaking of GA algorithms, and the learned configuration. The tweaking, although done by designers, is also a product of trial and error.

    GAs have become an industry. The emergency safety sensors in cars are trained. The details of their behavior are learned rather than programmed. Autonomous cars are trained. No one programs the details of their behavior.

  20. CharlieM: He had assumed DNA replication to be more simple than it was. One gene, one protein, one function is more in keeping with the current paradigm. Groups of proteins working together is harder to explain, but it was found to be more realistic.

    Your first and last sentences are correct, but I don’t know what you mean by the “current” paradigm. You appear to be saddling today’s biologists with a view that went out of fashion over fifty years ago. No matter.

    Taking heed of the Goethean method allows researchers to carry on experimenting but without prior assumptions which may well hinder their progress. It is not a competition to find the best researcher, it is a question of individual attitudes.

    Yet Alberts’s breakthrough was the result of explicitly reductionist thinking. The fact that there were seven different genes that, when knocked out, caused T4 to fail to replicate. From this fact he realized that there must be seven different polypeptides involved in the process. It is positively anti-Goethean.
    The point that Bruce is making here is the same point that Ptashne used to make when lecturing grad students. I paraphrase:

    Imagine a world in which there are no Frenchmen. What would we be doing? We would be grinding cells up to produce goop, and then we would be fractionating the goop to make more goop, and fractionating that goop, to make purer and purer goop.

    The point, in both cases, being that it was the combination of good old-fashioned steam biochemistry (Hershey, Kornberg) with genetics (Jacob, Monod) that produced the insanely successful field of molecular biology.
    I did take Bruce Alberts to task for some of his statements re “factories”, which I thought could be misconstrued by creation scientists.
    The only Goethean biologist is Rupert Sheldrake.
    So IT IS about comparing the success rates of the empiricists (including Alberts) versus the holistic woo-meisters. The sense of being stared at, anyone?
    Mea culpa, mea maxima culpa

  21. Joe Felsenstein

    Since we all have new mutations in us, that experiment is done every time an organism reproduces.In effect, if the reproduction is successful it rejects the white-noise theory,

    (In any case the issue between Dembski/Ewert/Marks and their critics is not the white-noise issue but the relevance of comparing evolution to randomly chosen “searches”).

    Thanks Joe.

    I guess I just assumed that EricMH was at least familiar with all your work and its citations, eg from your one of your recent posts here on why he seemed to need help.

    Sadly, you may be right: either he has forgotten or we are just going to revisit the same issues.

  22. Joe Felsenstein:

    CharlieM: You say genotypes have phenotypes. Isn’t that a bit like saying, for example, cardiovascular systems have organisms?

    We are considering the outcome in simple models that have different genotypes, each of which comes with a phenotype, which comes with a fitness value.

    You are welcome to go off to into a great big discussion of what “causes” what, and it might be enlightening to someone, but I will not join you there, as the issue is irrelevant to the discussion of the behavior of these simple model schemes. Dembski/Marks/Ewert have used such models, and I agree that they capture enough behavior of evolutionary processes to enable us to make their argument for, and Tom’s and my argument against, their conclusions.

    So models are made and argued over while nature and life are pushed into the background. For me one question is: how close to reality are the models? We need to guard against the likes of Richard Dawkins getting all excited at a few computer generated two-dimensional lines which have a superficial resemblance to insects.

    He exclaimed in “The Blind Watchmaker”:

    Nothing in my biologist’s intuition, nothing in my 20 years experience of programming computers, and nothing in my wildest dreams, prepared me for what actually emerged on the screen. I can’t remember exactly when in the sequence it first began to dawn on me that an evolved resemblance to something like an insect was possible. With a wild surmise, I began to breed, generation after generation, from whichever child looked most like an insect. My incredulity grew in parallel with the evolving resemblance. You see the eventual results at the bottom of figure 4. Admittedly they have eight legs like a spider, instead of six like an insect, but even so! I still cannot conceal from you my feeling of exultation as I first watched these exquisite creatures emerging before my eyes. I distinctly heard the triumphal opening chords of Also Sprach Zarathustra (the ‘2001theme) in my mind. I couldn’t eat, and that night ‘my’ insects swarmed behind my eyelids as I tried to sleep.

    This is Owen Barfield’s idolatry taken to the extreme. I’m not saying that you view your models in the same way, but there are many like Dawkins who have a hard job seeing the yawning chasm between models and reality.

    Myself, I think we can learn a great deal by studying living nature directly.

  23. CharlieM: So models are made and argued over while nature and life are pushed into the background.
    …there are many like Dawkins who have a hard job seeing the yawning chasm between models and reality.

    Good point: that would be wrong, were biologists to do that. It is essential that one tests how well one’s model comports with reality.
    Please let Dembski, Marks, Ewert, Sanford and Holloway know.

  24. Mung: You realize that people think the same is true of many evolutionary arguments?

    Everything is there waiting to be toppled. Nothing is sacrosanct. Have you not learnt that by now?

    I’m not even sure I know what evolutionary arguments really means. Can you given an example of such? And point who thinks what parts are bullshit?

  25. Mung: GA’s are not models of evolution.

    We make models, simpler or more complicated, of evolutionary processes. No, we’re not modeling the genome or the environment in all their complexities. But for the present arguments, the important thing is to model enough so that we can evaluate Marks, Dembski, and Ewert’s arguments. This we can do. And the fact is that MDE accept these models as relevant — they’re the ones they use themselves, the idea being that if their counterargument works for these simple schemes of hill-climbing on constant fitness surfaces, then evolutionary biology is in trouble more generally.

    The result is that we can see why, in these simple models, their arguments don’t work.

  26. Joe Felsenstein,

    While I agree with the substance of your arguments about models of evolution, I do think it’s important in these discussions to keep clear the distinction between GA’s and models of evolution.

    You seem to be saying that GA’s are models of evolution, just simple ones.

    But the difference between GA’s (any algorithm, really) and models of evolution (any model, really) is not a matter of simplicity; it’s more fundamental than that. An algorithm is a procedure for solving a problem. A mathematical model (any model) is a representation of something being studied.

    The only job a model has is to in some way be like something else. In the case of a process model, the model has to in some way act like the real-world process being modeled. Checking that this is so is called model validation (as you no doubt know already.)

    But one can’t criticize an algorithm for not doing a good job of representing a real-world process. Algorithms aren’t meant to do that; they are just computation. We don’t validate algorithms like we do models.

    This distinction becomes important when Marks, Dembski, and Ewert talk about what it takes for a model to work. What they do is change the subject from models to algorithms so their math is not accountable to the real world. So, for example, when they talk about a “search for a search” they don’t have to accept any criticism for introducing wacky alternative universes; they aren’t modeling anything. And of course they object to modelers “smuggling information” about the real world into their mathematical models.

  27. Mung: You realize that people think the same is true of many evolutionary arguments?

    Yes but the large majority of those people tend to be scientifically untrained and unknowledgeable laymen. ID’s target audience in other words.

  28. Adapa,

    Yea, but what about the fact that the large majority of Darwinian evolution proponents tend to be furry fandom loving water bed salesmen, with large goiters who collect breast pumps to trade at Confurence O ?

  29. Freelurker,

    I’m as guilty as anyone of using ‘GA’ as shorthand for a model of evolution. Though, I do find it striking that something entirely inspired by simple genetic mechanisms in populations – something deemed not to actually work ‘out there’ in any significant sense – finds widespread application in the bottom-line-driven world of commercial data processing.

  30. Allan Miller:
    Freelurker,

    I’m as guilty as anyone of using ‘GA’ as shorthand for a model of evolution. Though, I do find it striking that something entirely inspired by simple genetic mechanisms in populations – something deemed not to actually work ‘out there’ in any significant sense – finds widespread application in the bottom-line-driven world of commercial data processing.

    Only if you consider genetic mechanisms as having a target.

    Then yea sure they are very effective at finding that target.

    If no target, I guess they wouldn’t have much application.

  31. phoodoo,

    If they had a target, there would be no point in writing them, except as pedagological tools. I’m afraid you’ve been infected by the simplistic thinking that all GAs are implementations of Weasel.

    What they have is a space of possible genotypes, and some kind of fitness differential that influences likelihood of preservation. That differential could be by comparison to a target, but absolutely does not have to be.

  32. phoodoo: If no target, I guess they wouldn’t have much application.

    Most of them don’t have targets. Personally I only know of one that does, Dawkins’ Weasel.

    The most interesting GA (or simulation of evolution), because of it’s degree of realism, is Avida. There are no targets, there is not even a fitness function that “scores” different individuals according to some criterion and copies them to the next generation, as is typically what happens in a GA.

    Individual organisms are simulated, and they reproduce, and they suffer mutations that affect their ability to reproduce and carry out crucial functions (their code literally mutates), in turn leading to changes in their reproductive success. So changes in their reproductive success is emergent. So are the complex solutions that evolve to carry out those functions through competition for limited resources.

  33. On a related note, if GAs have targets, why even bother with the GA? Why not just straight up make the target?

    Why hire coders to code a GA to make a preconceived target, instead of just making the target?

  34. phoodoo: Solutions to what?

    The problems encountered by the simulated organisms when they attempt to replicate and consume resources before other organisms around them.

  35. Rumraket: The problems encountered by the simulated organisms when they attempt to replicate and consume resources before other organisms around them.

    And that’s not a target?

    They just create simulated organisms, and they can just consume or not consume, up to them really?

    And what are the problems-nightmares? Chainsaws?

    You have repeated that same false claim many times, that avida has no target.

  36. phoodoo: And that’s not a target?

    Correct.

    They just create simulated organisms, and they can just consume or not consume, up to them really?

    The organisms do what their genomic instructions do. Much like you.

    And what are the problems-nightmares? Chainsaws?

    Execution of instructions for replication costs resources, which have to be taken in, which is also a function encoded by instructions. So the problems are those encountered when replicating and taking in resources. Since there is only a limited amount of resources in the environment, organisms that can execute their instructions more efficiently, and/or tak up more energy for execution of those instructions outcompete those that don’t. So essentially the problems are all those that would make a lineage go extinct because it was outcompeted.

    You have repeated that same false claim many times, that avida has no target.

    I have repeated the true claim many times that avida has no target. If you think Avida has a target, then it should be trivial for you to go into the sourcecode of the program and point it out. Pick out the lines of code you think are the target.

    https://github.com/devosoft/avida

  37. BruceS: Sadly, you may be right: either he has forgotten or we are just going to revisit the same issues.

    Bill Cole Syndrome?

  38. Rumraket: I have repeated the true claim many times that avida has no target. If you think Avida has a target, then it should be trivial for you to go into the sourcecode of the program and point it out. Pick out the lines of code you think are the target.

    Isn’t Avida more like a framework that provides services?

  39. DNA_Jock:

    CharlieM: He had assumed DNA replication to be more simple than it was. One gene, one protein, one function is more in keeping with the current paradigm. Groups of proteins working together is harder to explain, but it was found to be more realistic.

    Your first and last sentences are correct, but I don’t know what you mean by the “current” paradigm. You appear to be saddling today’s biologists with a view that went out of fashion over fifty years ago. No matter.

    The “current” paradigm is the belief that life has sprung from inanimate matter by the actions of chemistry and physical forces that just happened to arrange matter in a combination suitable for it to form.

    Taking heed of the Goethean method allows researchers to carry on experimenting but without prior assumptions which may well hinder their progress. It is not a competition to find the best researcher, it is a question of individual attitudes.

    Yet Alberts’s breakthrough was the result of explicitly reductionist thinking. The fact that there were seven different genes that, when knocked out, caused T4 to fail to replicate. From this fact he realized that there must be seven different polypeptides involved in the process. It is positively anti-Goethean.
    The point that Bruce is making here is the same point that Ptashne used to make when lecturing grad students. I paraphrase:

    Imagine a world in which there are no Frenchmen. What would we be doing? We would be grinding cells up to produce goop, and then we would be fractionating the goop to make more goop, and fractionating that goop, to make purer and purer goop.

    The point, in both cases, being that it was the combination of good old-fashioned steam biochemistry (Hershey, Kornberg) with genetics (Jacob, Monod) that produced the insanely successful field of molecular biology.
    I did take Bruce Alberts to task for some of his statements re “factories”, which I thought could be misconstrued by creation scientists.

    Any researcher who makes detailed, careful observations and does not make unjustified assumptions is doing science in the Goethean way.

    Bruce Alberts endued five years of failed experiments because of his prior assumption that DNA replication required just one enzyme. In fact in the early ’60s just about everyone made the same assumption. Another widespread assumption that was made was that the molecles involved in DNA replication were floating around randomly bumping into each other. But it has now been realised that this type of action would be far too inefficient to achieve the results needed to organise and sustain life.
    Why were these and other similar assumptions made? Because they seemed logical in light of the major assumption that the appearance of novel living forms was achieved by random molecular changes. The level of coordinated organisation that they found was not something that they would have predicted.

    The only Goethean biologist is Rupert Sheldrake.

    And here you are making an unjustified assumption.

    So IT IS about comparing the success rates of the empiricists (including Alberts) versus the holistic woo-meisters. The sense of being stared at, anyone?
    Mea culpa, mea maxima culpa

    Goethe was an empiricist. And if you think Sheldrake is a woo-meister, well that is between you and him. It’s not something I can be bothered discussing.

  40. Rumraket: Why hire coders to code a GA to make a preconceived target, instead of just making the target?

    The problem is that different commenters have different meanings for the word “target”. In a GA (or an evolutionary algorithm) there is some function that evaluates the fitness of a genotype. The objective of the GA, when used to solve engineering problems, is to find a genotype that has fitness that is high.

    Does that mean that the GA “has a target”, namely high fitness? If we don’t know which genotype that will be, then I would argue that no, there is no target. A case like Dawkins’ Weasel does know what that genotype would be, and thus does have a target. Creationists and ID advocates point to this, triumphally, and say that it shows that such algorithms only succeed because the information is already incorporated into the algorithm.

    But for cases like the evolved antenna and the moving vehicles in BoxCar2D there is only a function that evaluates fitness, but the information about which genotype has highest fitness is not in the algorithm. These examples refute the front-loading argument. (In a disingenuous response, people like Marks, Dembski, and Ewert, in Introduction to Evolutionary Bioinformatics, try to get out of this by arguing that parameters of the programs were tuned, but never manage to show that this is equivalent to the full information about the target genotype).

    So no, GAs and EAs do not always have the information about the target genotype front-loaded into them.

  41. Freelurker: But the difference between GA’s (any algorithm, really) and models of evolution (any model, really) is not a matter of simplicity; it’s more fundamental than that. An algorithm is a procedure for solving a problem.

    And it’s not uncommon to see evolution presented in just that way, as a procedure for problem solving (and as a search).

  42. BruceS: I hope Mung is paying attention.

    I am and I don’t see anything he wrote to disagree with or that undermines the things I’ve been saying here for years. 🙂

  43. Rumraket: Most of them don’t have targets. Personally I only know of one that does, Dawkins’ Weasel.

    What about that one that created an antenna out of nothing? The clock one. The cars one.

Leave a Reply