Monday, May 25, 2020

A review of John Sanford's "Genetic Entropy"

1.  Introduction

(Feel free to skip this part.  It's just some context for what comes next.)

As regular readers will already know, I put a fair amount of effort into understanding points of view that I don't agree with.  I think if you're going to argue against a position it is incumbent upon you to understand what you're arguing against so that your arguments are actually on point and you're not just knocking down straw men.  So over the past few years I've taken a fairly deep dive into young-earth creationism.  I've gotten to the point where I'm pretty sure I could pass a YEC Turing-test.

One of the things I've noticed is that YEC arguments evolve (and yes, that is every bit as ironic as it sounds).  Old tropes like crocoducks and "If man evolved from monkeys why are there still monkeys?" have fallen out of favor.  In their place there are now a new crop of stock arguments that are not quite so transparently naive.

There are two arguments making the rounds nowadays that seem to be particularly fashionable: the "historical vs observational science" argument, and the "genetic entropy" argument.  The historical-vs-observational argument holds that there is some kind of fundamental difference when you do science about the way things were in the past vs about how they are in the present.  The argument goes something like this: we cannot time-travel into the past and so we cannot do repeatable experiments with regards to past events.  So the past is necessarily shrouded in a kind of mystery that the present is not.

This argument seems plausible on its face, but it is easily dispensed with: all of our data necessarily comes from the past (since none of it can come from the future -- duh!) so all science is in some sense "historical".  It is true that there are some singular events in the past that are inaccessible to scientific inquiry, and the further back you go the more such events there are.  There is probably no way to ever know, for example, what Julius Caesar had for breakfast the day after he crossed the Rubicon.  But there is a way to know (with very high confidence) that he did not cross the Rubicon by, say, flying across it.  How can we know?  Because we know a fair bit about the technology that was available in ancient Rome, and the constraints those would put on modes of travel.  Likewise, unless the laws of physics were different in the past than they are today, then it is extremely unlikely that biology was fundamentally different then than now.

The genetic entropy argument is not so easily dispensed with.  This is partly because the argument was advanced by John Sanford, a Cornell geneticist, in an eponymous book.  Sanford has credentials and that lends his argument some, well, credence.  Scott Buchanan wrote an extensive critique, to which Sanford responded, and then Buchanan responded to Sanford's response.  The exchange is very long and gets deep into the technical weeds.

This post is my attempt at a more accessible critique of Sanford's book.  It's not necessary to get very far into the details to see that Sanford is wrong.  By way of motivation I want to start with the story of another deep dive I did into a scientific controversy over twenty years ago.

2.  A parable

In 1996 a Berkeley biologist named Peter Duesberg published a book called "Inventing the AIDS Virus" whose thesis was that AIDS was not caused by the HIV virus but was instead caused by the drugs used to treat people who tested positive for HIV.  The book is 700 pages long, and it is dense with data, graphs, references... it is a genuine work of scholarship.  It has a forward written by Nobel laureate Kary Mullis, inventor of the polymerase chain reaction (PCR) which is a foundation of modern-day biological research.  The book is, by all appearances, a serious critique of the conventional wisdom, one that ought to command some respect.

It is also, of course, absolutely, 100%, catastrophically wrong.  But how can we tell?

Duesberg's work was never taken seriously by the scientific establishment, but it did launch a movement of HIV-denialism which persists to this day.  One of the leaders of this movement was a woman named Christine Maggiore, who founded an organization called Alive and Well.  She also wrote a short popular account of Duesberg's theory entitled, "What if everything you thought you knew about AIDS was wrong?"  Maggiore was diagnosed as HIV positive in 1992.  Her daughter, Eliza Jane, tested positive as well, having most likely contracted HIV through Christine's breast milk.  In accord with her belief in Duesberg's thesis, she refused treatment, both for herself and Eliza Jane.

Eliza Jane died of AIDS in 2005 at age 3.  Christine of course denied this, insisting that the coroner's office had gotten it wrong.  She even went so far as to sue them.  Three years later Christine also died of AIDS.  And that's how we know that Duesberg was wrong, because Christine and Eliza Jane did the crucial experiment, and the result was exactly what the scientific establishment predicted.  They even had a control in the form of Christine's son, Charlie, who tested negative for HIV, and is as of this writing still, as far as I have been able to determine, alive and well.

Keep all that in the back of your mind as you read the following.

3.  Genetic Entropy

In stark contrast to Deusberg's book, "Genetic Entropy" is not a scholarly work.  It is a popular book targeted at a lay audience, and as such it must be judged by looser standards than one would otherwise apply.  Unfortunately, even by the standards of popular accounts, "Genetic Entropy" runs off the rails almost immediately, in fact, in the third paragraph of the Prologue:
Modern Darwinism is built upon what I will be calling “The Primary Axiom”. The Primary Axiom is that man is merely the product of random mutations plus natural selection. Within our society’s academia, the Primary Axiom is universally taught, and almost universally accepted.
This is true, except for one salient detail: what Sanford calls the "Primary Axiom" is not an axiom.  An axiom is something that is taken to be true by assumption.  Axioms are, by definition, beyond question.  The idea that "man is merely the product of random mutations plus natural selection" is absolutely not an axiom of Darwinism, it is a conclusion, one that is supported by a mountain of evidence accumulated over 150 years of painstaking research.

But OK, maybe Sanford is applying Humpty-Dumpty's theory of language and is using the word "axiom" loosely?  Well, no.  In chapter 1 he writes:
An axiom is a concept that is not testable but is accepted by faith because it seems obviously true to all reasonable parties.
So Sanford really is attacking a straw man.  And that's really all you need to know about "Genetic Entropy."  Just as we can know that "Inventing the AIDS Virus" is wrong without wading into the details, so too can we know that "Genetic Entropy" is wrong because it starts with a false premise.  Garbage in, garbage out.

Still, Sanford is a Cornell professor, maybe there is something worthwhile in the book even if it's not his principal thesis?  Unfortunately, no.  The book is a hot mess of false premises layered on top of faulty reasoning resting on a foundtion of apparently willful ignorance.  There are far too many mistakes for me to go into them all, but a few stand out as particularly egregious, so I will talk about those in some detail.

First, let me summarize Sanford's argument to save you the trouble of actually having to read the book.

Sanford's thesis is actually pretty simple, and intuitively plausible: evolution happens when genes make copies of themselves.  Those copies are not perfect, but are subject to random mutation.  Because mutations are random, and because the machinery of life is very complicated and finely tuned, an arbitrary random mutation is vastly more likely to be harmful to an organism's reproductive fitness than beneficial.  Those harmful mutations will eventually overwhelm the beneficial ones through sheer force of numbers, and (he claims) he has the math to prove it.

Lest you think I'm being unfair in my paraphrase, here is Sanford's thesis in his own words, found at the end of Chapter 2:
Progressive evolution on the genomic level [is] virtually impossible. Adaptation to a special circumstance can still happen, due to extremely rare high-impact beneficials [sic] – which are isolated anomalies... These rare beneficial mutations almost always involve loss of function and are therefore unproductive in terms of “forward evolution”.
An example of a "high-impact beneficial" is a bacterium evolving a resistance to antibiotics.  Sanford has to carve that out as a special exception because such mutations obviously do occur, as evidenced by the existence of antibiotic-resistant bacteria.

Because harmful mutations are so much more likely than beneficial ones, Sanford goes on to argue, they must accumulate in the genome and eventually cause the species to go extinct.  This is the inevitable fate of all species.  So the fact that life still exists is evidence that it was all created fairly recently.

This is a not-entirely-implausible argument.  It is not even entirely incorrect.  It is true that, for certain kinds of mutations, harmful ones are much more likely than beneficial ones (with one very important caveat which I'll get to in a minute).  It is even true that harmful mutations can accumulate and eventually cause a species to go extinct.  This actually does happen -- in fact, it's not uncommon.  This is what has made is so popular in the YEC community: it's a position with a grain of truth and a veneer of respectability that YEC generally lacks.

Sadly (but not unexpectedly) it's just a veneer.  Underneath is one catastrophic mistake after another.

4. Mistake #1: Sanford does not appear to know what "information" is

Each chapter in "Genetic Entropy" starts with a "news flash", a pithy slogan that is supposed to set the stage for that chapter.  The first one is, "News flash: the genome is an instruction manual."

It is genuinely hard to tell how literally Sanford intends the reader to take this aphorism.  A plain reading of the following text seems to indicate that he means it to be taken quite literally: "A genome is an instruction manual that specifies a particular form of life. The human genome is a manual that instructs human cells how to be human cells and instructs the human body how to be the human body. There is no information system designed by man that can even begin to compare to the sophistication and complexity of the genome."

Sanford makes a big deal about "information".  The "news flash" in chapter 2 is "Random mutations consistently destroy information."  And yet, he never defines information in the body of the book.  He seems to assume that the reader already knows, and I suspect most of Sanford's readers assume the same thing.

The book has a glossary, and this is how it defines "information":
The most useful definition of this word is its plain and ordinary sense – information is “that which is communicated through language”. Biological information takes on many forms, due to the labyrinth of communication networks which enable life.
This definition is hopelessly naive.  It's akin to defining "transportation" as "that which is provided by the Mercedes-Benz E class sedan."  It is true that information can be communicated through language, just as transportation can be provided by a Mercedes-Benz E-class sedan.  But to define these words in such a way is to miss the point rather badly.

This is not hard to see even without getting into technical details.  The focus of Sanford's book is genetic information encoded in DNA.  But that information has nothing to do with language.  DNA was storing and replicating information long before human brains came along and invented language.

There is a whole field of study called "information theory" of which Sanford seems to be completely unaware.  Information theory was invented by Claude Shannon.  No reference to Shannon's work appears in the references, an absolutely stunning omission in a book in which the loss of information is a core theme.

Just in case you're interested, the actual technical definition of "information" is that it is a measure of the degree of correlation in the states of two or more systems.  It's easy to see why Sanford might want to avoid that verbiage in a book directed at a non-technical audience, but it's really not that difficult a concept.  If your car has a warning light that turns on whenever a door is open and turns off when the door is closed then the light contains information about the state of the door.  Genes contain information about the organisms they inhabit because there is are similar correlations between the sequence of DNA nucleotides and physical characteristics of the organism.  It's a very simple concept, and it has absolutely nothing to do with language.

There is, however, one crucially important feature of the correct definition: information is a relative concept.  It is a measure of the correlation between two systems.  There is no sense in which a system can "contain information" in an absolute sense.  Information resident in one system can only be measured relative to another.  Information is always information about something.  A light that turns on and off does not in and of itself contain information.  It only contains information if its flashes are correlated with something else (like the state of your car door).  Saying that a system contains information in an absolute sense is not only wrong, it is non-sensical.

This is important because Sanford speaks as if information is absolute.  He talks about information being "created" and "destroyed", but again, he never defines what this means.  He just assumes that it's obvious.  Well, it's not obvious.  In order to talk about quantifying information at all you have to specify what correlations you are talking about, and Sanford doesn't.

Because he doesn't, we are forced to guess.  There is an obvious candidate: the correlations between an organisms genotype and its phenotype, i.e. the correlations between various genes and various physical characteristics of the biological organisms those genes produce.  OK, fair enough, but even then it is still not clear what it means to "create" and "destroy" this information.  Sanford claims that biology cannot "create" information.  But biology manifestly can create copies of information.  Every time an organism reproduces the result is correlations between systems that weren't there before.  Clearly this cannot be what Sanford means by "creating information".

So what can he mean?  I suspect that what he really means is that biology cannot produce novel information.  It can (obviously) produce copies of information that was already there, but it can't "invent" new things.  Sanford speaks of "new information" frequently, but he never defines what he means by "new".  Presumably, a fresh copy of a genome produced when a cell reproduces is not "new information".  Again we are left to guess that what he means by "new information" is "information that has never existed before" or something like that.

But this is also clearly wrong.  Every human -- indeed every organism that has ever lived (with the exception of identical twins and organisms that reproduce asexually) -- has had a unique genome.  (This, BTW, is a big clue as to why sexual reproduction evolved!)  You are unlike any human who has ever lived and (almost certainly) unlike any who will ever live.  Your genome may not have been "invented", but at some point in time you were conceived and the information that specifies how to make you (i.e. the sequence of DNA that would eventually correlated with the physical characteristics of the person you are now) came into being for the first time.  I think it's fair to say that this information was "created" at that point.

I suspect Sanford would readily concede this.  The kind of information-creation he is talking about is not at the level of an entire genome, it is at the level of an individual gene.  Yes, he would concede, individual genes can be mixed-and-matched through sexual reproduction, and he might concede that this "creates information" (I don't know).  But that is not what he's talking about.  What he's talking about is creating individual genes.  And here he makes an unambiguous claim.  The "news flash" that introduces chapter 9 is: "Mutation/selection cannot even create a single gene."

This claim should already be a little bit suspect because, as we've just seen, sexual reproduction can "produce new information" (whatever that might actually mean).  And note that the mixing-and-matching that goes on during sexual reproduction is random, so the mere presence of randomness is clearly not a show-stopper.

But even this last bastion of Sanford's position falls to a second critical mistake.

5.  Mistake #2: the benefit of a mutation is not absolute

Sanford's argument that natural selection cannot produce new genes goes something like this: biological systems are incredibly complicated.  Even the simplest bacterium is more complex than the most complex computer man has ever invented.  All of the components have to work together in perfect harmony in order to sustain life.

A random mutation is vastly more likely to disturb this delicate balance than to enhance it.  In fact, the odds of a random mutation being beneficial is so small that it never actually happens in reality.  Therefore, it must be the case that all of the genes that are required to sustain life must already be in existence.  Moreover, the process of random mutation necessarily depletes this store of beneficial genes.  Eventually, we will run out.  Life is on an inevitable path of deterioration and decay that will lead to its eventual extinction.  And, more to the point, the only way that our current stock of beneficial genes could have come about is through a process of intelligent design which bestowed on us an initial endowment of beneficial genes that we are now in the process of frittering away.

Sanford's mistake here is similar to the one he made when defining information.  Just as information is inherently a relative concept, so are "beneficial" and "harmful".  A gene cannot be "beneficial" in an absolute sense.  Genes are only beneficial (or harmful) with respect to an environment.  A gene that is beneficial in one environment will always be harmful in some other environment.

And that is exactly how evolution works.  There are random changes.  Some of those changes are beneficial with respect to the environment in which the organism finds itself, and some are harmful.  Some are so harmful that they stop the reproduction process altogether and those are immediately self-correcting: a mutation that cannot copy itself will obviously leave no copies of itself to further contaminate the gene pool.  Of the ones that remain, the ones that are beneficial will make more copies of themselves because that is how evolution measures benefit.  Evolutionary benefit is not just relative to an environment, it is relative to reproductive fitness in that environment.

And the relativity of benefit doesn't stop there.  Not only is evolutionary benefit measured relative to reproductive fitness in an environment, it is measured relative to its competitors in that environment.  It's like the old joke about the two hikers who encounter a bear.  They start to run.  One hiker says, "This is silly, we can't outrun a bear."  The other says, "I don't have to outrun the bear, I just have to outrun you."

Consider this example from chapter 2:
[In] mutations that lead to antibiotic resistance in bacteria[,] cell functions are routinely lost. The resistant bacterium has not evolved. In fact it has digressed genetically and is defective."
Defective by whose standard?  Certainly not by the standards of a bacterium living in the presence of antibiotics.  Whatever it is that leads a bacterium to be resistant to antibiotics -- whether it is a "loss of function" or otherwise -- it is manifestly beneficial to a bacterium living in the presence of antibiotics.

But, one might counter, it is a loss of function.  If you take a strain of bacteria that has evolved antibiotic resistance and put them back in an environment without antibiotics, those bacteria will be less fit than those that never evolved resistance.  That's true, but it in no way refutes the point because benefit can only be assessed relative to an environment.  Antibiotic resistance is beneficial (to a bacterium) in the presence of antibiotics, and harmful otherwise.

Another example: chihuahuas are descended from wolves.  The evolutionary changes that led from wolves to chihuahuas could arguably be considered "loss of function".  Wolves are stronger, faster, better able to defend themselves.  And yet chihuahuas vastly outnumber wolves in today's world.  Why?  Because wolves are a threat to humans and chihuahuas are not, so we kill wolves but feed and shelter chihuahuas.  In an environment that includes humans, the functional loss of features that are a threat to humans is a net reproductive advantage.  If humans were ever to disappear from the face of the earth, wolves would very quickly regain the upper hand.

And yet, to paraphrase an old creationist trope, if chihuahuas are so much better at survival then wolves, why are there still wolves?  And the answer, of course, is that chihuahuas are better at survival than wolves in the presence of humans.  So chihuahuas thrive where humans are plentiful and wolves thrive where they are scarce.

This "relativity of benefit" is ubiquitous.  At the risk of beating a dead horse, there is no such thing as a beneficial mutation in an absolute sense.  A gene that produces sharp claws and teeth is beneficial to a lion living in the Serengeti, not so much to a human living in on the upper East side.  A gene that produces resistance to malaria and a concomitant risk of sickle-cell disease can be a net win if you live somewhere where malaria is prevalent, otherwise not so much.

There is one exception to this: a mutation that kills the organism before it is able to reproduce is unambiguously harmful.  Such mutations do happen.  But think about it: such mutations immediately eliminate themselves from the gene pool!  And that is the key insight into how evolution works.  Sanford is actually correct when he says that most mutations are harmful (with respect to the environment in which they occur).  But evolution is not just random mutation, it also crucially includes non-random selection which acts to amplify the prevalence of beneficial mutations and dampen the effects of harmful ones.  The more harmful a mutation is (relative to its environment of course), the quicker it gets eliminated from the gene pool (in that environment), never to be seen again (unless it should happen to arise again through random chance, which is, of course, extremely unlikely, and even if it does it will just get snuffed out again).

I've pointed this out to a few YECs and their response has been: but this is a tautology.  You are just defining benefit and harm in terms of reproductive fitness, so of course beneficial mutations are going to increase reproductive fitness.  And they're right, it is a tautology.  But here's the thing about tautologies: they are actually true.  The fact that random mutation + selection for reproductive fitness results in improved reproductive fitness is a tautology is in no way a refutation of evolutionary theory.  To the contrary, it is just the observation that evolutionary theory is obviously true because it is a logical consequence of two processes (random mutation and natural selection) that we know actually do occur.

The question is not whether evolution occurs.  Even Sanford concedes that it does:
Yet we all know that micro-evolution (adaptive selection) does happen. How can this be? Most adaptation is due to fine-tuning, not creation of new information.
The question is whether evolution can account for all of the variety of life on earth.

6.  Mistake #3: Evolution optimizes for the reproductive fitness of genes, not species or individuals

The reproduction of the human genome, like many (but not all) multicellular sexually-reproducing species, is an all-or-nothing affair: a fertilized egg either produces a baby which grows up to reproduce itself, or it doesn't.  There is no middle ground.

Because of this it seems intuitively plausible that evolution optimizes for the reproductive fitness of individuals, or maybe of species.  Indeed, this was the conventional wisdom for over 100 years, but it is wrong.  It was shown to be wrong by Richard Dawkins in his book "The Selfish Gene" and the associated scientific papers that back it up.  (This is actually the work that made him famous, not his activism as an atheist.)  Evolution optimizes for the reproductive fitness of genes.  The existence of things like species and individuals is a side-effect of the fact that cooperation among genes turns out to be an extremely effective reproductive strategy.

The easiest way to see that evolution does not optimize for the reproduction of individuals is to consider hive insects like ants, bees and termites.  The vast majority of ants are sterile.  They never reproduce as individuals.  And yet somehow ants have manifestly not gone extinct.

This is because evolution does not optimize for the reproduction of individual ants, it optimizes for the reproduction of ant genes.  A sterile ant is useless for its own reproduction as an individual, but it can be extremely helpful for the reproduction of its genes.  The same genes that reside in one individual ant also reside in countless other individual ants, and if any of those reproduce then it's a win for that ant's genes.  Indeed, an ant is more accurately seen as an organ rather than organism.  The organism is not the ant, it's the whole ant colony, which consists of parts that just happen not to be physically connected to one another.

This is even true for humans.  A single human individual cannot reproduce.  At the very least it requires a mating pair.  More realistically, it takes a village: a single mating pair of humans cut off from all civilization will almost certainly not be able to survive and reproduce.  The minimum viable unit of human reproduction is a tribe of at least a few dozen individuals.

This is no accident.  Diversity is a clear reproductive advantage.  The wider the range of available traits within a population, the wider the range of environmental challenges that population will be able to respond to without going extinct.

This is in fact the reason that evolution invented sexual reproduction, because it is a much more efficient way of producing diversity than asexual reproduction and waiting for cosmic rays to produce mutations.  Most of the variation that occurs in sexually reproducing organisms does not come from the creation of genes de novo (Sanford is actually right about that).  It comes from the mixing-and-matching of existing genes during sexual reproduction.  No sexually-reproducing organism ever replicates itself exactly.  But its genes are replicated (more or less) exactly (or at least half of them are) on every iteration.

Sanford is actually correct when he points out that the invention of new proteins is a very inefficient process.  But in the presence of sexual reproduction, it is also completely unnecessary in order for evolution to proceed as long as there is a sufficient pool of biodiversity to draw on.  It is analogous to the manner in which new information is created in natural language.  It is rare for new words to be created.  Most of the information communicated by natural language is created simply by putting existing words and phrases together in new combinations.

Sanford actually comes tantalizing close to realizing this in chapter 5:
The fact that most mutations are recessive dramatically masks their negative fitness effects, and greatly hinders selection against them. Likewise, all interactions between genes (“epistasis”) will interfere with selective elimination of minor mutations. In smaller populations, the randomness of sexual recombination (chromosome-segregations and gamete-unions are both random and thus fluctuate randomly) can routinely override selection. These effects cause the fundamental phenomenon of genetic drift. Genetic drift has been extensively studied, and it is well known that it can override selection against all but the most severe mutations in small populations.
Notice how the assumption that all mutations can be determined to be either "good" or "bad" in an absolute sense is baked into his rhetoric here.  Remember, diversity is a reproductive advantage.  A gene's reproductive fitness can only be assessed relative to its environment, and a big part of the environment for a gene is the various collections of other genes that it has "teamed up" with in order to build organisms.  The wider the range of organisms a gene has been able to get itself into, the greater its odds of survival against external environmental variations.  The reason recessive genes exist is so that genes that are deleterious in one environment can be held "in reserve", simply as a store of diversity to be used in the future when environmental changes cause it to become beneficial again.

Sanford simply ignores all of this.  He barely refers to Dawkins at all, and not at all to the selfish gene theory that is his main contribution to the field.  Because of this, he's attacking a straw-man.  Like all straw-man arguments, it is actually correct as far as it goes: an evolutionary process such as the one that Sanford describes would indeed not work.  But real evolution simply doesn't work the way Sanford describes.

7.  Mistake #4: complexity does not require intelligence

Apart from ignoring elementary facts like how evolution actually works and what information actually is, Sanford's entire argument really boils down to a logical fallacy called the argument-from-incredulity.  Life is fantastically complex (true) and so it cannot possibly have come about without intelligence.  The reason for this is essentially that Sanford cannot imagine how it could have happened, and because Sanford cannot imagine it, it must be impossible.  After all, Sanford is not just any old shmoe, he is a well-read and learned man, a professor at a prestigious university.  If he can't figure it out, maybe it's because it actually cannot be figured out.

Again, this is not entirely-implausible.  At the beginning of the book Sanford says that "the genome is an instruction manual", which is not a horrible analogy, and it does serve to illustrate one important point: the instruction manual is not enough.  If I gave you an instruction manual for how to build a 747 you still probably would not be able to build a 747.  Building a 747 requires, in addition to the instruction manual, a whole slew of specialized equipment, jigs, material, skills and knowledge that is not in the manual.

Likewise, if I gave you a printout of the DNA sequence of the human genome and said, "Here, go make a human", you wouldn't be able to do it.  The information in DNA is rendered into organisms through an incredibly complex process which we are only beginning to understand.  We don't even fully understand the first step in this process, the construction of proteins.  We know how the sequences of nucleotides in DNA get translated into sequences of amino acids by ribosomes, but how those amino acids then fold themselves into the particular shapes that allow them to perform various biological functions is still entirely beyond our ability to predict.  And what happens after that has all the appearance of true magic.

But just because something looks like magic doesn't mean that it is.  There are a host of natural phenomena that we once thought were the work of deities: Earthquakes. Lightning.  Floods.  All of these are still beyond our ability to predict or reproduce but no one seriously argues that this entails that they are the product of anything other than natural processes.

Creationists argue that life is different because of its complexity.  Earthquakes and lightning and floods might be mysterious, but they are simple phenomena.  Life isn't simple.  It is, as I have just noted, unfathomably complex, and it is that unfathomable complexity that reveals it to be the work of some kind of intelligent designer.

But there are many examples of simple processes that produce vast complexity.  In fact, simple processes can be computationally complete.  It just doesn't take a lot of hardware to build a universal Turing machine -- it can be done with a few thousand transistors or a pile of wood.  These machines can compute any computable function.

Of course, this leaves open the question of how complicated of a program you need in order to accurately model biology.   Creationists, notably Michael Behe, claim that biological processes are "irreducibly complex".  There is no merit in this argument, and to show this it is not even necessary to examine the argument.  Just as it is not necessary to dig into the details of Peter Duesberg's book to know that it is wrong, nor is it necessary to dig into the details of the design of a perpetual motion machine to know that it is wrong, we can formally prove that irreducible complexity cannot possibly be demonstrated.

How?  There is a formal result from the theory of computational complexity called Chaitin's theorem.  To describe it in detail would require getting deep into some technical weeds, but the upshot is that once a system gets beyond a certain threshold of complexity (and that threshold is quite low, vastly lower than the complexity of biological systems) you can never prove that it is irreducibly complex.

Here is a sketch of some of the technical details just in case you're interested: Chaitin's theorem refers to a mathematical quantity called Kolmogorov complexity (KC) of a system S which is defined (roughly) as the size of the smallest formal description that can reproduce the behavior of S.  To claim that a system is irreducibly complex is essentially the same as claiming that its KC is large.  Chatin's theorem shows that it is impossible to prove that a system has a large KC.  It's a remarkable result, but the proof is actually pretty elementary.

In the face of Chaitin's theorem, any claim to a proof of irreducible complexity has the same status as a claim of perpetual motion: such a proof, if it were valid, would quite literally violate the known laws of physics.  It might be true that biological systems are irreducibly complex, but we can prove that we can never know this for certain.  We can know therefore without even looking that Behe's argument must be an argument-from-incredulity.  It cannot be a proof because such a proof cannot exist.

It might be the case that biological systems are irreducibly complex.  There is some evidence that this is the case in the form of our current inability to fully understand it.  But I'll give long odds against this lack of understanding persisting for very long (where "very long" is a few hundred years or so).  Betting against the continuation of scientific progress has never been a winning strategy and I see no reason to believe that is going to change.

8.  Mistake #N

I could go on and on.  I could talk about how fuzzing proves that random processes can produce information.  I could give examples of known beneficial mutations in biological systems.  I could talk about how some biological innovations are indeed hard to produce, and so are correspondingly rare (but others that you would think are rare actually happened multiple times!)  I could talk about how Sanford completely ignores the existence of error-correcting codes, and the fact that biological systems use them, when he dedicates an entire chapter to talking about noise.

But I won't because life is too short.  It's too short for me to spend the time writing about it, and it's too short for you to spend the time reading about it (unless you're a YEC, in which case you should definitely go educate yourself).

I will, however, close with one more anecdote: one of the examples of beneficial mutations in biological systems is the evolution of lactase persistence, or the ability to digest milk into adulthood.  This evolved in humans when we moved north out of Africa and into colder climates where our normal diet of plants and animals was not available year-round.  In an environment like that, the ability to digest milk from domesticated animals is a clear win.

Sanford does not talk about this at all, but Answers in Genesis does:
Mutations responsible for lactase persistence actually represent a loss of genetic information, a shut-down of normal regulation. If anything, the prevalence of lactase persistence is a testimony to the fact an all-knowing Creator designed the human genome with the ability to change.
This actually made me laugh because it fully exposes how vacuous this argument is.  The only reason that the shut-down of lactase production is considered "normal" is because of the order of events.  If humans had originated in cold climates and migrated to warmer ones, the argument on AIG would surely have been the exact opposite: "The mutations responsible for the loss of lactose tolerance in adulthood are a perfect example of a loss of function leading to the further degradation of the human genome."

And that is the core problem with Sanford's thesis: biological systems are complicated.  For any incremental change (which is the only kind there is) it is always possible to tell a just-so story that casts that change as some kind of loss or deficiency.  All you have to do is cherry-pick your environment so that it is the one relative to which the change is in fact harmful.

The opposite is, of course, also true: you can (almost) always tell the same kind of just-so story that casts any change as a benefit.  But here's the difference: cherry-picking those environments is exactly what evolution does!  It's not that beneficial mutations survive, it's that mutations that are beneficial relative to some environments survive in those environments and not in others.  In this way evolution gradually, in the fullness of time, fills all the available niches and, in so doing, creates new environments, new niches, in which new kinds of genes, like those that make chihuahuas, can begin to thrive.  That's how life works.  It's complicated.  It's messy.  And it has absolutely nothing to do with any kind of design.


ash said...

Thank you! Please do more of those. I've especially enjoyed Chaitin's theory proof.

Ron said...

@Ash: Thanks for the kind words. I'm happy to oblige. Anything in particular you'd like me to look at?

Luke said...

> This is true, except for one salient detail: what Sanford calls the "Primary Axiom" is not an axiom. An axiom is something that is taken to be true by assumption. Axioms are, by definition, beyond question. The idea that "man is merely the product of random mutations plus natural selection" is absolutely not an axiom of Darwinism, it is a conclusion, one that is supported by a mountain of evidence accumulated over 150 years of painstaking research.

I'm a little confused about how to operationalize "man is merely the product of random mutations plus natural selection", with focus on the bold. There are no comprehensive simulations of evolution based on this, which show this to be true. So it could easily be that a third element is required, rather like classical physics not being sufficient for the ultraviolet catastrophe, even though in his "Two Clouds" speech, Lord Kelvin was quite confident that nothing like GR + QFT would be required. How do we *know* that evolution won't require a similar scientific revolution? Without such *knowledge*, it seems that "it is a conclusion" is not justified. (Or: us accepting the conclusion is unjustified.)

To counter-balance my previous paragraph, I will also note that there is no known third element which, if we add to scientific theory, results in superior modeling power. Specifically, nobody has managed to posit a "God did it" or "God used purpose" which helps us better understand reality. But I don't see what warrant we have for believing that: (i) that the present mathematical patterns we know of will be sufficient for explaining evolution; and/or (ii) any additional mathematical patterns will not exceed the bounds set by "merely the product of random mutations plus natural selection".

My favorite analogy here is that of building a space elevator. Wouldn't it be neat if we could just lift cargo to space, rather than use such dangerous and polluting means such as the recent SpaceX launch? Well, take the fact that we've been able to build higher and higher buildings with steel-reinforced concrete. The Burj Khalifa has been the tallest building in the world since 2009, at 828 meters high. (Previously it was Taipei 101, at a mere 509 meters high.) Well, how about we just build higher and higher until we can elevate cargo to space? Unfortunately, steel-reinforced concrete would collapse under its own weight far before that. The X-Seed 4000, at 4000m, is about the highest you can go with steel-reinforced concrete. Do we know if there are any analogous 'scalability limits' of "merely the product of random mutations plus natural selection"? If we don't know either way, then what should we do with conclusions that say the answer is "none relevant"?

P.S. If you're going to go with Deutsch on scientific explanations being merely the best we have rather than something we can be confident will stand the test of time, I suppose this comment is irrelevant to you. But do we have empirical evidence that Deutsch's stance is optimal for all scientists to adopt, if we care about maximal scientific research?

Ron said...


It's Sanford who put the word "merely" in there, not me. I was just quoting him. I agree with you that it's not appropriate. But merely-ness is kind of in the eye of the beholder, so I didn't want to start quibbling over that.

Unknown said...

Watched your debate. SpeedOfSound here. Email me. We could have an interesting chat.

Luke said...


What would you add to:

     (1) random mutations
     (2) natural selection


Ron said...


Bacon. Everything is better with bacon.

Luke said...


Hmmm, that's less than satisfying for someone who likes EE&R. I prefer stuff like the following, from Massimo Pigliucci and Gerd B. Müller:

>> Rather, the majority of the new work concerns problems of evolution that had been sidelined in the [Modern [Evolutionary] Synthesis] and are now coming to the fore ever more strongly, such as the specific mechanisms responsible for major changes of organismal form, the role of plasticity and environmental factors, or the importance of epigenetic modes of inheritance. This shift of emphasis from statistical correlation to mechanistic causation arguably represents the most critical change in evolutionary theory today. (Evolution: The Extended Synthesis, 12)

This is far more satisfying to an engineer, who knows that you can't just build higher and higher with steel-reinforced concrete: the structure collapses on itself well before space elevator height. Likewise, merely having "random mutations × natural selection × many organisms × much time" doesn't necessarily get you "the variety of life we see on earth and in its fossil record". It fascinates me how science sometimes demands mechanisms and refuses to listen to you if you don't have them (I have an offline story about that if you'd like), and other times castigates anyone who says we don't have enough mechanisms specified to create an actual working simulation of what goes on.

Possibly, purpose actually does show up, maybe in the form Robert Rosen describes in Anticipatory Systems: Philosophical, Mathematical, and Methodological Foundations. Possibly, without something like this, life—or at least complex life—could not possibly exist. And yet, on my journey from YEC → ID → evolution, trying to bring up ontological purpose was always anathema. It's like I had committed heresy! People kept pushing me back to (1) & (2). Sanford's "merely" is something I regularly encountered with real-life "evolutionists" I found online. They did ultimately convince me to accept evolution as the most promising explanation, but boy howdy were they obnoxious in the differing rule sets that apply at different times. (e.g. see previous paragraph)