|Home | Intro | About | Feedback | Prev | Next||
Reactive or creative evolution?
Calculating the phenotype from the genotype
Calculating the power of evolution
Prediction in biology versus physics
Definition of life
Reply by the author Mark Ludwig
Similarities and Dissimilarities of Computer and Biological Viruses
Mark A. Ludwig as an early Intelligent Design Advocate
A review by Gert Korthof, 2 Feb 2006 ( updated 24 Aug 2017 )
Mark Ludwig's 'Computer Viruses, Artificial Life and Evolution' starts with a scientific study of computer viruses and Artificial Life, transforms into a defence of Intelligent Design and ends with a confession of belief in the supernatural.
Summary of the book:
Computer viruses are autonomous self-reproducing agents. They infect other computers just like viruses and parasites in the biological world do. Computer viruses are a form of artificial life. Mark Ludwig claims computer viruses are the only artificial life forms in the wild of which one can claim that their environments (operating systems) are not specially constructed to let them work. Therefore, naturally occurring as well as specially constructed computer viruses are the ideal objects to study the feasibility of computer virus evolution without philosophical or religious bias. Above that, they are easier to study and with more exact methods than real life. He finds that computer viruses can exhibit Darwinian evolution.
He uses the results to evaluate real-life biological evolution and especially neo-Darwinism. Experimental and theoretical evolutionary biology are in a abysmal state because they are not predictive. The problem of the neo-Darwinian theory of evolution and the claims of the Artificial Life community is that both ignore the role of intelligence and information injection. Darwinian evolution is not a creative process: it does not create information; it only reacts to environmental selection, which is the origin of the information. So evolution is only a reactive force. Ludwig suspects that the environment (laws of nature) is engineered to make life and evolution possible. Ludwig is also interested in the origin of life question.
He organised a contest to write the smallest possible functional self-replicating computer virus and concluded there is an absolute minimum size. He concludes from the minimum size that random processes inside a computer cannot create the smallest virus; even then such a virus could be an evolutionary dead end. All viruses we know about cannot arise by random processes, so must be intelligently designed by people. The same improbabilities apply to the spontaneous origin of a self-replicating molecule capable of progressive evolution.
Reactive or creative evolution?
In the world of computer viruses the hypothesis of intelligent design is unproblematic because we know
that humans create computer viruses (Ludwig himself included), but in the biological world things are different.
What argument does Ludwig advance for ascribing intelligent design to evolving organisms?
Ludwig's argument heavily relies on the distinction between 'reactive' and 'creative' evolution
'Reactive' means reacting on the selection pressures of the environment.
Darwinian mutation and selection do not create information,
because "the selection process is injecting information into the system",
just like Richard Dawkins injects information in his artificial selection experiments.
Creative evolution does not exist according to Ludwig (2).
Secondly, the environment of an organism often is another organism. More than half the world's species live in or on the bodies of other organisms. Carnivores and herbivores depend on other organisms. Prey-predator are in an evolutionary arms race. In insects (butterflies) mimicry is the result. Co-evolution of organisms means a two-way interaction. In the computer world: viruses and virus scanners co-evolve. Even in the virtual computer world viruses contain features of their environment because viruses are designed with full knowledge of the target Operating System (OS). So although Ludwig is right that the OS is not designed for viruses, the reverse is true. No wonder that viruses work. So, do organisms 'inject information' into each other? Again, yes, in an abstract sense and again by perfect natural Darwinian processes.
Ludwig's label 'reactive' is not an objection to evolution by natural selection at all. Darwin never claimed that evolution created species 'out of thin air'. The creative aspect of evolution is that species change by natural selection and thereby change the information content of their genomes. This change of information content is natural and not supernatural. There is no philosophical materialism necessary here. Since Ludwig accepts reactive evolution and his distinction between reactive and creative evolution melts away, his argument against creative evolution evaporates too.
Calculating the phenotype from the genotype
The first half of the book is based on the idea that viruses and Artificial Life are nothing but instructions to be executed (by a computer). Ludwig translates this view to biological systems. By analogy the genotypes (DNA) of living organisms are also viewed as instructions to be executed (by a cell). The name of the game is now: calculate the phenotype from the genotype. This seems reasonable if the genotype is available and can be computed (by a computer) and those instructions fully determine the phenotype. Let's first discuss the computation of the phenotype, then the computation of the power of evolution (Power).
Whatever the computability of Artificial Life and Evolution, biological research is not done and cannot be done by computing complete phenotypes from genotypes (with or without interaction with the environment). Richard Dawkins (2009) wrote:
"Nobody, reading the sequence of letters in the DNA of a fertilized egg, could predict the shape of the animal it is going to grow into. The only way to discover that is to grow the egg, in the natural way, and see what it turns into. No electronic computer computer could work it out, unless it it was programmed to simulate the natural biological process itself, in which case you might as well dispense with the electronic version and use the developing embryo as its own computer." (28).
"Even if I knew the complete molecular specification of every gene in an organism, I could not predict what that organism would be."So, Ludwig is on the wrong track. His drawing (page 155) of the relation between genotype, phenotype and environment is extremely simplistic. Remarkably, in a sense Ludwig knows that one cannot compute a complete phenotype (15). He claims that the phenotype is 'emergent' because it cannot be calculated from the genotype. This is important for him: emergent behaviour is part of the definition of life he uses (see: Definition).
One cannot even begin to ask the question how a genotype produces a phenotype, if one does not know the genotype. Biologists rarely know the complete genotype (genome) of a species, although knowledge is rapidly growing. Before DNA sequencing (genomics) was discovered, a (partial) genotype could only be indirectly inferred from the phenotype. Gregor Mendel was the first who inferred genotypes. Today, more than 1,800 genes are known to cause hereditary disorders in humans (Online Mendelian Inheritance in Man). The first direct access to the genotype was possible through the study of aberrant chromosomes (cytogenetics). For example, prenatal diagnosis can predict the Down's syndrome phenotype from the trisomy-21 genotype with nearly 100% certainty. A hundred years after the rediscovery of Mendel in 1900, DNA sequencing techniques opened for the first time direct access to complete genotypes. Biologists and biochemists needed a hundred years to arrive at a point where AL-scientists simply started: knowlegde of complete genotypes. As if biologists simply needed to load the genotype into a computer and run it!
Only in 2012 computational biologists were able to do just that for the smallest organism known to us: the single celled Mycoplasma genitalium with 'only' 525 genes. They developed a computer model with 1,900 experimentally derived cellular parameters. The model is able to predict phenotypes (biological properties, cell behaviour) from genotypes (gene sequences) (31). This is an impressive achievement, but it is far from predicting the phenotype of a multicellular organism.
The difference between biology and AL is that:
To construct a complete 'fitness landscape' one needs all the data of the phenotype effects of all possible protein variants of the whole organism. So it is clear that a accurate fitness landscape cannot be calculated. In practice geneticists isolate and study well defined phenotypic effects of well defined mutations. This works fine. In the same way evolutionary biologists (population geneticists) calculate what will happen with well defined mutations in a population. This works fine too.
It is easy to construct a hypothetical Fitness Function unfriendly to evolution as Ludwig did. To construct a realistic fitness landscape one needs a lot of data. Ludwig's figure (page 215) is just an imaginary illustration. It is not based on data. No conclusions about Darwinian selection can be based on that.
Calculating the 'Power of Evolution'
Ludwig asks the BIG question:"Are the mechanisms proposed by biologists powerful enough to produce all life on Earth?". Ludwig observes that "most biologists believe evolution is powerful enough to create all the complexity and diversity of life we see on earth over the period of about a billion years." (4). Then he concludes:"we need a scientific theory of the power of evolution.", "Today, we don't even have a theory" (p.153), "nobody can really prove it is powerful enough to do the job." (4). A different question is: Is evolution not too powerful for the job? Is evolution infinitely powerful like God? (14).
Ludwig's expectations for a theory of evolution are strongly influenced by the field of Artificial Life. Evolutionary processes run on a computer, so are fully computable. In analogy with Artificial Life, Ludwig thinks an evolutionary theory in biology must compute the generation of complex life. "A theory, though, ought to give me the tools to start with a set of initial conditions and predict what is going to happen." (4). Then he concludes that evolutionary theory fails in making these kind of predictions and" Experimental and theoretical evolutionary biology are in a abysmal state"! (4).
Are Ludwig's requirements for a theory of evolution reasonable? Not really. Ludwig knows that real biological organisms are too complex to be modelled on a computer, because the phenotype cannot be calculated from the genotype (Ludwig's 'emergence' is the defining property of life!), let alone that evolution of a population of those organisms can be predicted. It does not make sense to blame biologists for this when complete phenotypes are fundamentally unpredictable. There are several good reasons for the difficulty of calculating the power of natural selection. One reason is that species have histories. That makes them unique. "Evolutionary biology is a historical science" (10). Additionally, a complicating factor is the role of chance in evolution: "The outcome of an evolutionary process is usually the result of an interaction of numerous incidental factors." (10). Despite this, some evolutionary biologists claimed things as "The All-Sufficiency of Natural Selection" (11). But even if the majority of biologists would claim something like the above, the answer still depends on the precise meaning of the claim. What precisely has to be explained? And in what detail? Is the theory of evolution required to predict all the details of all species that ever lived on Earth (the existence of the duck-billed platypus, giraffe, flying fish, panda's thumb, chromosome number of humans, blind spot in human eye, introns in genes, etc) in contrast to general principles (adaptation, the power of natural selection, mimicry, parasitism, sex, sexual selection)? According to historian Roger G. Newton, the thing that has gradually separated physics from the other branches of science over the past 6,000 years is "the ability to predict future events with some confidence of success." (26).
Prediction in biology versus physics
However, do the combined laws of physics and cosmology explain all the details of the non-living universe?
Newton's laws of gravity predict with such precision which orbits planets can have, that eclipses of the sun and moon can be made with amazing accuracy and they are routinely successful. A possible explanation for the predictive power is that "Newtonian physics is essentially timeless" (33). (In biology the direction of time is essential). The perfect predictability of Newton's laws might be undermined simply by the presence of too many mutually interacting bodies. For even just three bodies (40) following Newton's laws, vanishingly small differences in the initial conditions can lead to widely different outcomes over long times – giving an appearance of randomness even though the process is in principle entirely predictable. This kind of 'deterministic chaos' is now known to be present in the orbits of planets in the Solar System (38).
The 'timeless character' of the physical laws causes also problems. "The planets' orbits are chaotic over longer timescales. This means that the position of a planet along its orbit ultimately becomes impossible to predict with any certainty (so, for example, the timing of winter and summer become uncertain), but in some cases the orbits themselves may change dramatically" (35, 36).
All eight planets in the Solar System orbit the Sun in the direction that the Sun is rotating. Six of the planets also rotate about their axis in this same direction. The exceptions–the planets with retrograde rotation–are Venus and Uranus. Why?
Apart from the long-term predictability of the solar system, Newton's (and Laplace's) laws cannot predict number, size, and distances of the planets of our own solar system. And in which direction around the sun must the planets go? (clockwise? anti-clockwise?) Neither do they predict which planets have moons or rings, there sizes and distances. For example, can the origin of our own moon be explained? Its size and composition? How many craters the moon has? Their sizes? Why is the orbit of the moon around the earth elliptical? Why does the Moon show 59% of her surface to Earth? (source). Why does Saturn have 62 moons and rings? Why does Mars have only 10 percent of the mass of the Earth? etc., etc., etc. Can it all be derived from Newton's and Laplace's laws? The actual orbits of planets, moons and rings are only one of the many possibilities. Which possiblity is realised depends on many historical and environmental factors which are not part of the laws of Newton or: historical contingencies (23). Neither do the laws of Newton predict Mercury's makeup (mostly iron core, with a thin veneer of rock) and Uranus's skewed magnetic field (30) and so on. The physicist Paul Davies stated "We know now that the arrangement of the planets is largely a historical accident" (27). Even the modern general theory of planet formation does not explain the detailed structure of the Solar System (29), (32), (39), (41), (42).
Similarily, physics cannot predict the famous constellations (Ursa Major, Orion, Cassiopeia, etc). Since historical accidents are a big factor in evolution, it is even the more unreasonable to require detailed predictions of the theory of evolution. At the moment, physicists cannot rigorously deduce the structure of the helium atom from basic physics, a non-historical problem, let alone that of a living organism. Even more fundamentally, the standard model of physics depends on 19 numerical parameters. Their values are known from experiment, but the origin of the values is unknown.
"In theory, the Navier–Stokes equations, developed almost 200 years ago, describe the physics of fluids well. But these equations are devilishly hard to solve. So engineers and scientists usually come up with simplified theoretical models or resort to numerical simulations when they want to predict fluid flow. This approach has its limits: modelling turbulence bogs down even supercomputers." (43).
So, if physicists are unable to predict the behaviour of dead systems, how reasonable is it to expect biologists to predict the behaviour of living systems?
Evolutionary biologists do not 'tacitly assume' that evolution is omnipotent and therefore a mathematical quantification is not necessary (p.154). They do not assume it at all, but because they know natural selection is not omnipotent (14).
fact - path - causes
There is an important distinction absent in Ludwig's evaluation of the theory of evolution, which blocks any meaningful result. That is the well-known distinction fact - path - causes (12). The evidence for common descent ('The Fact of Evolution') is so strong that biologists do not constantly try to prove or disprove it, just like physicists are not constantly trying to prove or disprove the second law of thermodynamics.
The historical path of evolution (phylogeny) may never be completely known in all its details.
The causes (mechanisms) of evolution can and are being experimentally investigated. Biologists do not test whether Darwinian mechanisms could produce all life forms on earth, they investigate the relative importance of different mechanisms and search for new mechanisms. Ludwig completely misses this three-part division. I recommend reading reference 12, but there are many others.
Furthermore, Ludwig completely misses the theory of population genetics (13). Despite all the theoretical restrictions, population genetics is the most mathematical thing in evolutionary biology: it comes closest to calculating the power of evolution (18). Population genetics is the most fundamental body of theory in evolutionary biology. It is the proving ground for almost all ideas in evolutionary biology (8). The difference of population genetics and Artificial Life is that population genetics makes useful abstractions. Typically, population genetics calculates what happens with genes, not with individual organisms.
Definition of Life
Although Ludwig is sceptical about the possibility to give a list of defining characteristics of life, ("We don't understand life well enough to give a set of hard rules to determine what is alive" and " The very concept cannot be put in terms accessible to science.") he works with the list used by Artificial Life researchers:
Furthermore, Ludwig does not distinguish between the properties 'self-reproduction' and 'information system' (DNA). This prevents him from making an information subsystem a primary life criterion, and self-reproduction a secondary criterion of life. This is important because an information system has also the function to inform metabolism (gene > enzyme > metabolic pathway). That's why an information system belongs to the primary characteristics of life. The metabolic function of DNA is most clearly present in the soma, while the genetic function of DNA is present in the germline.
Ludwig's goal was an unbiased, unprejudiced assessment of the theory of biological evolution (1). Did he succeed? His approach was to study the evolution of computer viruses and Artificial Life. This seems a sensible thing to do, especially if you are a physicist with good knowledge of computer viruses but without sufficient training in biology. Ludwig's knowledge of computer viruses is unique. He published several books on computer viruses. He demonstrated that computer viruses can and do evolve (Darwinian mutation engine). What he says about viruses and Artificial Life (AL) I trust to be largely correct. However, the success of his approach ultimately depends on whether the results say anything meaningful about biological evolution. How did he establish that his results are relevant for biological evolution? Surprisingly, he did not even attempt to answer that question. He did not realise that it was a crucial question for the success of his investigation. Yet, he found it appropriate to claim that 'AL holds the promise of a real theory of evolution' (my emphasis). This statement is wrong. The reason why the statement is wrong is that AL has abstracted everything that is crucial for the evolution of life on earth: having a body, getting and digesting food, urinate, respiration, metabolism, maintaining body temperature, adjusting blood sugar, blood pressure, diploidy, meiosis, getting a mate, getting pregnant, etc, and all complications that go with these things. Paradoxically, despite incorporating 'the essentials of life', whatever can be calculated in AL does have restricted value for biology. Above that, viruses are parasitic, so are not a good model for non-parasitic life. Remarkably, Ludwig's knows that biological objects are too complex and computer viruses and AL are too easy to study, but at the same time he beliefs that AL done properly could reveal insights about biological life forms. He did not resolve this dilemma: how to gain insight in complex life if your method eliminates complexity right from the start? I'm afraid there is simply no substitute for studying the messy, wet and dirty thing called 'life'. This does not mean that mathematics has no role to play. The secret is making the right abstractions. One of those magnificent and very useful abstractions is 'The Selfish Gene' (Richard Dawkins,1976), that is the idea of a replicator. Darwin's message can be translated in the language of today with one concept: replicators (25). If anyone could have understood the power of the idea of the replicator, it's Ludwig because viruses are selfish replicators.
The second question is the assessment of the theory of evolution itself. Ludwig claimed that 'the theory of evolution is in an abysmal state'. This statement is wrong, naive, arrogant and insulting. Two main reasons are: lack of relevant knowledge and presence of bias in the technical sense.
- Ludwig is unaware of relevant biological knowledge. Yes, life on earth is extremely complex, but it turned out that life is not too complex to make predictions of partial phenotypes. Gregor Mendel is a magnificent example of the success of biological science in isolating specific characteristics among thousands of them and to predict the frequency of them in the next generation with mathematical precision (17). Mendelian genetics gave rise to population genetics and out of the marriage of population genetics with Darwinism the neo-Darwinian Synthesis was born (more).
Ludwig failed to investigate the structure of Evolution Theory (more) and the evidence (I don't know any critics of evolution who sufficiently know the theory!). Apart from missing population genetics, he missed the argument for common descent. That led him to the idea that in evolutionary biology only the mechanism counts and additionally, when one cannot calculate the evolution of bacterium to humans, the whole theory fails (6). Ludwig is pessimistic about what wet biology has produced. Regrettably, Ludwig has no idea what biological research has produced. It is not the theory of evolution, but his knowledge of biology and evolution that are 'in an abysmal state'. It would be easy to make fun about the state of physics and mathematics (21).
- What about bias? Was Ludwig unbiased in the gathering of the necessary information and his judgement? If not, did he compensate for the bias? Bias is present when judgement is unfair. Applying standards of theory construction common in the physical sciences to biology is unfair, because biology is different from the physical sciences. Is there religious bias too? In the first half of the book (studying viruses and AL) he is admirably unprejudiced, but then he introduces Phillip Johnson. Phillip Johnson is neither a biologist, nor an Artificial Life expert, but a lawyer. Is it professional for a physicist to consult a lawyer to gain insight in the field of evolutionary biology? Ludwig entered unfamiliar territory with a non-expert guide. Above that, introducing and recommending Phillip Johnson is introducing religious bias into his investigation (9). This is not the same as making his whole book worthless, but it is the opposite of what he wanted to do. Further, he did read biologist Michael Denton. Denton is a critic of evolution (19). Ludwig's list of 'Selected References' is extremely one-sided. That is easy to establish. Nearly all his references are critical of evolution or Darwinism. Did Ludwig compensate for this bias in any way? Insight into the uniqueness of biological knowledge could have counterbalanced his judgement, but that is absent in his book.
A different question is whether Ludwig is an Intelligent Design Theorist (IDT); the question addressed in the title of this review. It is not meant to dismiss Ludwig, but it is a question of the history of ideas. His book pre-dates Michael Behe (1996) and William Dembski (1999) who made the ID concept popular. At the time Ludwig wrote his book, the word IDT was not common. As far as I know, Dembski and Behe do not refer to him. In his own summary of the book, Ludwig does not advertise himself as an IDT, nor does he mention that the book is about IDT. So, what evidence do I have for my claim that Ludwig is an early IDT? The signature of IDT is clear: belief that nature does not have the creative power to create species (conspicuous in Johnson: 7); information is the essence of life; information must be injected into the system from outside the system; anti-materialism; limitations of Evolution Theory; rejection old-style creationism; "science can never tell us whether life actually began as the result of a natural chemical process or a divine miracle" (p.145); evolution is (only) a theory and should be tested; belief in the supernatural ("I have to admit the supernatural into my worldview", page 330); endorsement of Phillip Johnson. For Ludwig ID is not a superficial idea: it deeply penetrates his thinking (20). I do think that this strongly influenced his ability to judge the current status of evolutionary theory. I do think that it substantially contributed to the outrageous claim that 'evolutionary biology is in a abysmal state'.
Despite all my criticism, I did enjoy reading and reviewing Ludwig's book, because he asks the big questions. Only an outsider is able to ask this charming and naive question: "I want to try to find out how likely it is that natural law could cause what is observed in the fossil record and in today's world." (p.146). This is nothing less than 'A Theory of Everything' in Biology! It is easy to forget the big questions because professional science usually is about small, manageable but more fruitful questions (5).
Mark A. Ludwig (1993) 'Computer Viruses, Artificial Life and Evolution'.
American Eagle Publications, paperback 373 pages
|guestbook (moderated)||home: wasdarwinwrong.com||wasdarwinwrong.com/korthof77.htm|
|Copyright ©G. Korthof 2006||First published 2 Feb 2006||Updated: 24 aug 2017 F.R./Notes 24 aug 2017|