Good Parts: Introduction to a breadth of topics.
The reader who hasn't been introduced to the following topics might find the book enlightening and, at times, enjoyable: evolutionarily stable strategies, the red queen effect, genetic algorithms, self-modifying code, simulated annealing, the founder effect, game theory, the traveling salesman problem, the P vs. NP problem, auto-catalytic sets, strange attractors, fitness landscapes, and peptide chains. But there are far superior books out there for introduction to these topics. For example, see The Red Queen by Matt Ridley, The Selfish Gene by Richard Dawkins, and Sync by Steven Strogatz.
Thesis: This book has no thesis.
The book doesn't have a coherent thesis. Instead, it's just organized into chapters and sub headings. Now, I want to give the science part of this book a fair chance. I take no pleasure in attacking Kauffman's peculiar writing style, with its vague language mixed into meandering prose. I think he has some original ideas which deserve careful examination. But the book's disorganization poses a special challenge for me, because its hard to find a clear statement of the scientific claims. So, I've done my best to pick apart his argument into bite-sized chunks, individual claims, that can analyzed independently. There's plenty I hate about the book that isn't captured below, like how Kauffman wears ignorance like a badge of pride, spouts philosophy all over the place, and even ends the book on an anecdote about him and a friend practicing the art of dousing, ironically, like a couple of fucking hipsters. Anyway, lets dive into the slog.
Claim #1: Complexity rescues Natural Selection.
Since Darwin, we turn to a single, singular force, Natural Selection, which we might as well capitalize as though it were the new diety. Random variation, selection-sifting. Without it, we reason, there would be nothing but incoherent disorder. I shall argue in this book that this idea is wrong. For, as we shall see, the emerging sciences of complexity begin to suggest that the order is not all accidental, that vast veins of spontaneous order lie at hand. Laws of complexity spontaneously generate much of the order of the natural world. It is only then that selection comes into play, further molding and refining. Such veins of spontaneous order have not been entirely unknown, yet they are just beginning to emerge as powerful new clues to the origins and evolution of life... Life and its evolution have always depended on the mutual embrace of spontaneous order and selection’s crafting of that order.In this rambling quote, Kauffman makes takes pot shots at natural selection, in an attempt to bolster is ideas. But throughout the book, he fails to ever support this claim. Instead, he relies on a straw-man argument to fool the reader into paying attention to his other claims, which are cool, but fall short of explaining the origin of the framework that Natural Selection acts on. Anyway, this straw-man is a rigid, foolish Darwinist who only knows pure, unadulterated Natural Selection. In reality, the Neo-Darwinist appreciates the genetic code, and its origin in a primordial, non-living environment. In fact, much work has been done to find out just how life got started. But it is this sort of just-so story that Kauffman explicitly avoids:
Evolution is filled with these just-so stories, plausible scenarios for which no evidence can be found, stories we love to tell but on which we should place no intellectual reliance.Contrary to what Kauffman claims, plenty of evidence exists which will, in time, lead to ever better understanding of the origin of life. Rather than work on this frontier of science, Kauffman prefers software models of complexity. That's well and good. I love software models, too. But then he goes around taunting the reader with claims like:
We are not supposed to be here. Life cannot have occurred.He waves the straw-man in our faces and even names it George Wald. There's a whole sub heading about how Hoyle and Wickramasinghe discredited Wald's long-dead take on the origin of life! This is a neat bit of history, but it's meant to trick the reader into accepting it as evidence that the modern story is wrong. That's just not true. The modern story is uncertain, but Kauffman isn't adding to it. He's just waving his hands.
In the above quotations, you start to get an idea of the hyperbole and jargon used in the book. Soft words like order and complexity are thrown around, without defining them at all. Also, it shows the degree of sloppiness the author is willing put forth to make the prose sound profound. Since when do laws generate things in the natural world? Laws explain things in the natural world! But that doesn't sound as impressive, does it? I guess lying is okay, to make something sound cool, right?
This book is mostly about the origin of life, but Natural Selection is no damsel in distress, and this attempt at generating tension for the plot fails to impress me.
Claim #2: Complexity explains Democracy.
The edge of chaos may even provide a deep new understanding of the logic of democracy.One of the points Kauffman drives home over and over is how well his pet complexity theory works at solving a variety of problems. He supports this by throwing the idea of patches at politics. In the introduction, he makes the above claim, as well as this one:
Democracy has evolved as perhaps the optimal mechanism to achieve the best attainable compromises among conflicting practical, political, and moral interests.Near the end of the book, he mumbles something about patches, receiver-based optimization, and mechanisms allowing complex systems to reach excellent compromises. I won't bother the reader of this review with the details, because they aren't important. Patches are cool, and I tried really hard to understand what he was talking about, and I think I understand why he sees something important in this political analogy. Kauffman is confused about how models relate to the scientific process. If two things have the same mathematical model, Kauffman says they have a deep, underlying connection. When in reality, they have in common only one superficial property: that we model them using the same mathematics, the same computer code. There is no law underlying all things that can be added, that just sounds silly! But you can model the act of joining any piles of one type of discrete object using addition. The same computer code models joining piles of pennies as does joining piles of grapes. This isn't mystical or magical, its just how computer modeling works. The networks of neurons in the human brain kind of look like galaxy filaments spread over megaparsec. That doesn't imply some deep connection between brains and the universe. Pot smoking hippies might see a connection, but it isn't real. It's a faaake! Kauffman can't learn anything from political patch/receiver-based models that Robert Axelrod didn't program into them, because there is no deep, underlying theory. Just some grandiose claims, and a few jargon words which, by themselves, fail to predict anything.
In the same chapter, Kauffman talks about power-laws in learning curves, extinction events, etc. He's got this unhealthy obsession with them. He even tries to somehow connect these ideas with technological progress, too:
Tissues and terra-cotta may indeed evolve in similar ways. General laws may govern the evolution of complex entities, whether they are works of nature or works of man.The general law he's talking about is his own invention, called the NK model, which, as its name suggests, isn't a law at all, but a model. This model is cool, and does find application with learning curves and technological progress, evolution, and piles of sand. But there's no general law there. Maybe his book should have been called Applications of the NK Model...? No, too boring, wouldn't sell. Gotta lie to people, that's how you sell books, right?
Claim #3: Life emerged whole.
Life emerged whole, not piecemeal, and has remained so. Thus unlike the dominant nude RNA view of the origin of life, with its evolutionary just-so stories, we have a hope of explaining why living creatures seem to have a minimal complexity, why nothing simpler than the pleuromona can be alive.This is probably Kauffman's most obvious blunder. By life, he means that an auto-catalytic set that contains the same metabolism life uses today, stuffed inside a lipid bi-lair. So, like, a cell, basically. But without the genetic code? He is less than clear on this point. By emerged whole he means that this was not a consequence of joining or competition between primitive replicators, but a singular event. In Kauffman's eyes, a certain threshold of molecular complexity guarantees that life will arise! No, I'm not making this up:
Life, in this view, is an emergent phenomenon arising as the molecular diversity of a prebiotic chemical system increases beyond a threshold of complexity. If true, then life is not located in the property of any single molecule, in the details, but is a collective property of systems of interacting molecules.Kauffman cites the minimum complexity of life as evidence for his point of view. His argument assumes that, since no simpler life exists on the Earth today, it cannot exist. But this is just not true. The reason for a minimum complexity on Earth, today, is that an organism needs to compete with other organisms. Imagine that I engineered the tiniest cell possible, optimized its DNA, got rid of organelles and DNA that isn't critical for survival in my Petri dish. Then, I set lose some e-coli in there. The tiny cell may divide more rapidly, but it will quickly be killed off, because it lacks basic defenses. The slimmed-down cell would have had its biological weaponry optimized away. It would also be poor at adapting to changing environment, because introns (often called junk DNA) are useful for fast evolution. Replication can occur below this level of complexity, as evidenced by viruses. So, which is more likely? That life emerged whole, but got simpler only once it became possible to parasitize other life, or that life started simply, and grew in complexity until a stable design, the single cell, became mostly ubiquitous, but leaving a niche for virus?
The origin of life at a threshold of chemical diversity follows the same logic as a theory of economic takeoff at a threshold of diversity of goods and services. Above that critical diversity, new species of molecules, or goods and services, afford niches for yet further new species, which are awakened into existence in an explosion of possibilities.Oh well, I can't argue with that logic! Wait, no, I can. A model that happens to work for goods and services most certainly need NOT apply to the set of molecules which happen to exist on the primordial Earth! For example, what if the most commonly formed dipeptide (chain of 2 amino acids) formed an enzyme that would cut up any other dipeptide? This isn't the case, but if you are building a model, this is the kind of thing you have to know! Either by empirical testing, or by assumption. In this case, Kauffman assumes life as we know it, and goes on to prove that he can model it, and with the same computer code, also model economics. Then, he hopes we didn't notice as he ties the cart before the horse, and touts his phony discovery.
What Kauffman calls unrepentant holism is just laziness meets an anti-reductionist philosophy. Of course the details matter! The outcome of complex phenomena something like life, star formation, and cellular automaton are a consequence of the underlying details. Statistical laws can be useful for studying chaotic and complex system, but statistical laws can't explain things that happened once, like the origin of life.
Claim #4: Self-organization allows for Natural Selection.
Selection, then, confronts twin limitations: it is trapped or frozen into local regions of very rugged landscape, and, on smooth landscapes, it suffers the error catastrophe and melts off peaks, so the genotype becomes less fit.Kauffman takes another swing at Natural Selection. The guy just won't give up! There is a lot of information packed into the above quote, which deserves some explanation. He's talking about fitness landscapes, which is an theoretical/imaginary space where peaks represent high fitness, and valleys represent poor fitness. Each sequence of genetic code is a point on the landscape. Evolution works by slowly climbing uphill, which eventually leads to a peak. From there, it's all downhill. If evolution worked in baby steps on a huge, smooth landscape, life would be trapped around one or a few peaks, and never spread out. But the obvious fact of the diversity of life is proof that this does not happen in the extreme. In fact, proposed limitation is actually necessary for the origin of species! A species separated by some geological barrier (in real space on Earth) is split into to populations that are free to explore the (theoretical) landscape in two directions. Then, as each encounters different conditions, represented a gradual change in the fitness landscape, they diverge, as each follows their own peak away from the other! You see, there is not just one, but multiple fitness landscapes. Alas, this seems to be something that never occurred to poor Kauffman.
The second proposed limitation he calls the error catastrophe, where errors accumulate with each cell division, and pull the organism away from its peak. I think he later addresses this in the book, but far from being a limitation, this phenomena is responsible for the best part of life! Sex! You see, if one of my genes has a mutation in it that lowers my overall fitness, but another of my genes has a mutation of vital importance, sex allows these genes to become disentangled, because my child can inherit their copy of the good gene from me, and the copy of the bad gene from mommy. Mommy and Daddy got some recombining to do! Oh, my. There are other reasons for sex, too, but this explanation at least removes the catastrophe from Kauffman's error catastrophe.
Evolution may be impossible without the privilege of working with systems that already exhibit internal order, with fitness landscapes already naturally tuned so that natural selection can get a foothold and do its job.Wait, what? Evolution by Natural Selection isn't the whole universe, self contained and unassailable? Better not tell Kauffman's army of straw-men he calls most biologists! But seriously, who is he trying to fool? What does he even mean by internal order? He isn't talking about entropy, which is the true measure of order for all living (thermal) systems. He's certainly not buying it that life could start with simple replicators. But he takes it one-step further:
Self-organization may be the precondition of evolvability itself. Only those systems that are able to organize themselves spontaneously may be able to evolve further.Ever since I saw the subtitle of the book, The Search for Laws of Self-Organization and Complexity, I was curious what he means by self-organization. My ears perked up at every mention of self-organization, and I paid very, very close attention. But not ONCE does Kauffman offer a definition, or even a hint at what self-organization means to him! He just assumes we know all about it! Now, in Conway's Game of Life, I know what is meant by self-organization; gliders and what not. But what the hell does Kauffman mean by it?! I don't think I'll ever know. I don't think he knows. I think he gets a mystical, warm feeling when he thinks about his NK models. He just doesn't know how to prove that they matter to the reader. So he makes shit up, and tries to make it sound important.
Oh, wow, evolution is just the tip of the iceberg, man. It wouldn't even be possible without self-organization there to, you know, organize the systems so they spontaneously, like, do stuff. Far out, man! It's just a bunch of made-up jargon to confuse the reader into submission. I'll have none of it. There is no evidence linking self-organization to Natural Selection. There isn't even any definition of the former, anywhere in the book! But there is a foundation which Natural Selection relies on to operate: the genetic code. It has some natural origin, and it would be really fucking cool to learn even a tiny bit about that origin! Kauffman comes to this party empty handed, so he starts waving them again, and passes you a joint. Perfect. Thanks a lot, asshole.
Claim #5: Complexity theory explains rates of cell division.
A law of biological organization? Using logarithmic scales, the number of elements in a random network is plotted against the length of the state cycle, and the estimated number of genes for various organisms, assumed proportional to DNA content per cell, is plotted against the cells' median replication times. In both cases, the result is a straight line with a slope of 0.5. This is the hallmark of a square-root relation.If you have no idea what he's talking about, don't worry. This graph is insane, and shows how far away Kauffman has traveled from scientific reality. Hey, at least it's quantified, for once! Now, to understand this claim, you first need to understand logarithmic scales. Everybody with me? Good. Now, this graph shows log of one thing vs. log of another thing. If the line is strait, that means we have a power law, and we know how much Kauffman loves power laws. Ignore the upper and lower lines. They are just there to remind you that he also found power laws in his models. What matters are the dots, and the line drawn through their median. Now, he says that the x-axis is the number of genes of various organisms, and the y-axis is the replication time of cells from those same organisms. If they obey a power law, what does that tell us about complexity theory? Not a whole lot.
However, It does show animals with more DNA take more time to replicate. Duh! They have more DNA to copy, so it takes longer. Also, if you double the length of the DNA, it takes maybe 4 times as long to copy, in general. That's not surprising, as a larger job takes more time to manage and error correct. A power law is expected here, without any distracting models of random networks. At that, I'd love to get my hands on his data, because it doesn't look like this is quite a power law at all. It looks DNA size has only a loose correlation to replication time. I also wonder what would happen if you controlled for cell size, because that must also play a role in cell division.
By the way, where are the plants? Maybe their tenancy to exhibit lots of polyploid made them not fit the chart, so he left them out. You know, to protect us from confusing evidence that doesn't fit his theory. That's nice of him. Oh, no wait, nice isn't the word. What's it called? Fraudulent. Yeah, that's the word. Fraudulent!
Claim #6: Number of cell types vs. log(DNA size) = 0.5
Another candidate law. The logarithm of the number of cell types in organisms across many phyla is plotted against the logarithm of the DNA content per cell. Again, the plot is linear, with a slope of 0.5, indicating that the number of cell types increases as the square root of the amount of DNA per cell. If the number of structural and regulatory genes is assumed proportional to the DNA content per cell, the number of cell types increases as a square-root function of the number of genes.This is more of the same. The model is selected using K = 2, but that is an heuristic value; he made it up in order to reproduce the power law, but gives no underlying explanation as to why each virtual gene in the model has two inputs. What would this mean for real genes? It would mean they have two different, other genes that can turn them on/off. I guess, maybe this is true, on average. There might be something about gene regulation that we can learn from this, but Kauffman fails to propose any such insight. Instead, he says:
It is hard not to be impressed. The theory that genomic systems should lie in the ordered regime is once again not only in the biological ballpark, but almost exactly on the mark.Boastful, but not very informative. And again, where are the planets? They have huge numbers of chromosomes, but relatively few different types of tissues.
Personally, I'm not surprised that the number of cell types can grow proportionally to the length of the DNA code. In fact, Kauffman made a mistake, this slope is 0.5 on a linear scale (because both axis are log base ten, you can cancel out the log, leaving it linear), meaning the number of cell types increases directly proportional to the DNA content! You don't need to have one of his K = 2 canalyzing random networks to understand why. More cell types means more code. Double the number of cell types, double the amount of code you need to make them work. Fraudulent!
My favorite quote from the book.
What if (11010100001010) were Armageddon?Kauffman was trying to make a point about formal undecidability Gödel incompleteness. But it's a funny quote. Just in case, don't go around reciting 11010100001010 at people. You never know...