Chapter 11 Chapter 8 Waiting for Kano

Carnot's second law of thermodynamics, that is, the tendency of everything in the universe to cool down and decay.Farmer didn't know what the new second law would look like, but he believed it would describe the tendency of matter to organize itself and predict the general nature of organization in the universe. At the end of November 1988, the secretary of the Los Alamos Nonlinear Research Center handed Langton a sealed official envelope, which contained a memorandum signed by the director of the laboratory, Siegfred Hecker: We recently noticed that you have started your third year of a postdoctoral research fellowship, but have not yet completed your PhD.According to Article 40︱1130 of the Ministry of Energy, this institution shall not employ postdoctoral fellows without a doctorate for more than three years.In your case, due to an administrative error, we did not warn you in advance about possible violations of the regulations.Therefore, we have applied to the Department of Energy for an extension, and you do not have to return the scholarship for fiscal year 1989; however, since December 1, 1988, we will not be able to continue to employ you unless you have received a Ph.D. researcher.

Simply put, you are fired.Langton was shocked and ran to find Du Lun (Gary Doolen). Du Lun solemnly confirmed the matter, yes, there is indeed such a rule.And, yes, it was possible for Heike to do so. Langton still has lingering fears when he recalls this incident.The goddamn guys made him look bad for two full hours before throwing that surprise party.Farmer, who made up the letter and arranged the whole prank, said: The number specified by the Department of Energy should have been leaked long ago. Langton is almost forty years old, and his birthday is November 3rd. Ten days. Fortunately, once Langton recovered from the panic, the birthday party was a joy for both guests and hosts. After all, it is not every day that a doctoral candidate celebrates his fortieth birthday.Farmer also called on Langton's colleagues in the research center and theory department to chip in and buy a new electric guitar as a birthday present.I really want to stimulate him to finish his Ph.D., because I am worried that he will be blamed for not getting his degree for a long time, and maybe there is such a rule that restricts the laboratory from hiring people without a Ph.D.

Manifesto for Artificial Life Langton was well aware of Farmer's intentions, and no one wanted to finish his doctoral thesis more than him.Since the Artificial Life Symposium was held, his research has made a lot of progress.He has moved old cellular automata codes from Michigan to work on Los Alamos workstations, he has conducted countless computer experiments to investigate phase transitions at the edge of chaos, and he has even read the physics literature in depth. , learn how to analyze phase transitions statistically. But a year flew by and he hadn't actually started writing because of the many follow-up developments that took up his time after the Artificial Life workshop ended.Cowen and Paines asked him to organize the lectures and compile them into a volume as one of the Santa Fe Institute's series on complexity science.At the same time, Ke Wen and Paines also insisted that these articles must be rigorously reviewed by outside scientists like other academic papers.They told him that the Santa Fe Institute must not get its reputation for sloppiness, that it had to be science, not video game consoles.

Langdon didn't mind, because he had always held the same point of view himself.However, he has spent months editing; that is, reading each of the forty-five papers four times, sending each paper to a different reviewer, and sending the reviewer's comments back to Give the original author, ask them to revise the paper, and from time to time coax everyone to finish the revision with sweet words, and then he spends a few months writing the preface and general introduction himself.He sighed: It really took a lot of time. On the other hand, the whole process has benefited him a lot.He said: It's like preparing for a Ph.D. qualification exam.What makes a good thesis?After this experience, I became an expert in this field.Now that the book is finally finished, and it meets the standards required by Ke Wen and Paines, Langton feels that what he has created is not just a series of essays.His doctoral dissertation may still be in the mud, but the symposium proceedings may well have laid the foundation for artificial life science to become a serious science.More importantly, he integrated the ideas and insights of the seminar speakers into the preface and the forty-seven-page conclusion.Langton has written one of the clearest manifestos for what artificial life is all about.

Perspective of life from the perspective of abstract organization He writes: Basically, artificial life is the exact opposite of conventional biology.Artificial life does not understand life by analyzing the biological community into species, organisms, organs, tissues, cells, organelles, membranes, and finally molecules; artificial life attempts to understand life synthetically: in artificial systems, Combine simple pieces to create life-like behaviors.The tenet of artificial life science is that life is not just the properties of the surface of matter, but the organization of matter.The operating principle of artificial life is that the law of life must exist in the form of change.The vision of artificial life is to use new media such as computers and robots to explore other possible developments in biology.Artificial life scholars can be like space scientists studying other planets: because understanding the dynamics of other planets from the perspective of the entire universe, we can gain a deeper understanding of our own world.Only when we are able to look at the present form of life in terms of its possible form can we truly understand the nature of the beast.

Looking at life from the point of view of abstract organization was probably the most compelling insight from the workshop, he said.No wonder these insights are often closely related to computers, since both have the same intellectual origins. Since the time of the pharaohs, human beings have been searching for the secret of automatons. At that time, Egyptian craftsmen invented clocks by using water drippers.In the first century AD, Hero of Alexandria developed his gasology, describing how gases maintained at normal pressures produced simple motions in small machines imitating the shapes of animals and humans.A thousand years later in Europe, craftsmen in the Middle Ages and the Renaissance invented more and more sophisticated bell hammers, which would protrude from the inside of the clock to strike the time, and some public clocks even designed bell hammers of various shapes to perform a play .During the Industrial Revolution, clock automaton technology led to more complex process control technology. Factory machines have been guided by rotating cams and interconnected mechanical arms.In addition, after a more sophisticated combination of rotating cams, drums and mechanical arms, nineteenth-century inventors developed a controller that could generate different motion programs on the same machine.With the development of computers in the early 20th century, this kind of programmable controller became the beginning of computer development, Langton said.

At the same time, the procedure of logical steps gradually gained a clearer concept under the efforts of logicians, thus laying the foundation for the general theory of computation.In the early twentieth century, Alonzo Church, Kurt Godel, Tunning and others pointed out that no matter what the material of the machine is, the essence of the mechanical process is not a thing, but an abstract control structure that can A set of rules to represent a program.Indeed, Langton says, that's why you can take software out of one computer and run it in another; because the mechanics of the machine are not in the hardware, but in the software.Once you accept this idea (and this is exactly what Langton learned eighteen years ago at Massachusetts General Hospital), then you can easily see that the vitality of organisms also lies in soft bodies, in the organization of molecules. in, rather than the molecule itself.

complexity comes from simplicity Langton admits, however, that this notion is not easy to accept when one considers the fluidity, spontaneity, and organic nature of life, and the complete control over the operation of computers and other machines.At first glance, it seems ridiculous to use these terms alone to talk about life systems. But the answer lies in the second insight that came up again and again in the workshop: Yes, living systems are machines, but the organization of such machines is very different from what we are used to.Unlike human designers who design machines from the top down, living systems seem to develop from the bottom up, with overall structures emerging from simple systems.Proteins, DNA, and other biomolecules make up cells, neurons make up the brain, cells that react with each other make up the embryo, ants make up the ant colony, and similarly, businesses and individuals make up the economic system.

This is, of course, the point of view of Hornan and the complex adaptive systems of the Santa Fe school.The difference is that Horan regards this group structure as a collection of basic units, and through the reorganization of basic units, an extremely efficient evolution can occur.Langton, however, sees it as an opportunity to generate a rich, life-like dynamic.One of the most surprising lessons we've learned from simulating complex physical systems on computers is that complex behavior doesn't need to come from complex sources.Indeed, interesting and fascinating complex behaviors can suddenly emerge from the combination of extremely simple elements.

Langton speaks from the bottom of his heart, and the passage fully reflects his own experience with the discovery of self-replicating cellular automata, but it also applies to Reynolds's most vivid presentation of the Bates Group Symposium on Artificial Life .Reynolds did not regulate the overall behavior of Bez in detail from top to bottom, or ask Bez to follow a specific leader. He only used three simple rules to partially regulate the interaction between individual Bez.It is this locality, however, that enables Betsy Group to adapt to changing conditions in an organized way.These rules will always pull Paz together, just as the invisible hand of Adam Smith will always maintain a balance between supply and demand.But, like the economy, this tendency to concentrate at one point is just a tendency, the result of each bez reacting to the actions of its neighbors, so when a group of bez hits an obstacle like a pillar, As long as each Bozi goes his own way, the Bozi group will have no difficulty in splitting into two groups and taking a detour.

Try, says Langton, telling each Baz how to respond in every conceivable situation with a bunch of rules, and the system becomes unwieldy and complicated.In fact, Langton had seen computer simulations like this one come out clumsy and unnatural, more cartoonish than animated.In addition, since it is impossible to program every possible situation, top-down systems will always encounter situations they do not know how to deal with, and thus become sensitive and fragile, and often hesitate to move forward. scientific vitalism The computer-generated plants published by Lindenmeier and Prezemyslaw Prusinkiewcz of Canada were also generated through bottom-up group thinking.These plants are not drawn on the computer screen, but grow out.Start with just one stem, and then use a few simple rules to tell each stem how to grow leaves, flowers, and more branches.Similarly, these rules do not mention what the final plant will look like, but only simulate how the numerous plant cells differentiate and interact with each other during the development of the plant, thus creating a shrub or flower that is almost unreal.In fact, with careful selection of the appropriate rules, it is even possible to create cyberplants that closely resemble known species.Twisting the rules slightly could produce radically different plants, proving that in evolution, slight changes in the development of organisms can cause drastic changes in appearance. Langton said that many people in the workshop mentioned that to produce life-like behavior, you must simulate groups of small units rather than complex large units; only control local behavior instead of global control.Don't make all kinds of regulations from the top down, but let the behavior emerge naturally from the bottom up.When experimenting, focus on the behavior in development, not on the end result.Like Horan's favorite argument, living systems never really settle down. Indeed, you could think of this bottom-up thinking as purely scientific vitalism.The so-called vitalism means that ancient people believed that life contains some kind of energy, force, or spirit that transcends matter.Langton said that life is indeed beyond matter, but not because there is an organism beyond the laws of physics or chemistry that animates these systems, but because groups that follow simple rules of interaction can consistently exhibit unexpected behavior.Life may indeed be a kind of biochemical machine, but to give life to this machine is not to inject life into the machine, but to organize the machine and make the dynamics of mutual reactions among machine groups come alive. life is a calculation Finally, Langton said, the third insight from the conference was that life might not just be like a way of computing (since life is not just molecules but properties of tissues), but life is a way of computing at all. To know why, we must start with the traditional biology based on carbon.For more than 100 years, biologists have continuously pointed out that the most outstanding characteristics of any living organism are its genotype (the genetic blueprint encoded in DNA) and phenotype (phenotype, also known as phenotype, which is the genetic blueprint). The difference between observable traits created by instructions).Of course, the actual activities of living cells are extremely complex. Every gene is a blueprint for a protein molecule, and there are countless proteins interacting in countless ways in the cell.But, in fact, you can think of a genotype as a collection of little computer programs running in parallel, and a gene is a computer program.In functioning, each program cooperates or competes with other active programs, and taken together, the overall computation performed by these interacting programs is the phenotype, the structure that an organism assumes as it develops. Next, let's look at the general biology of artificial life. The concept is exactly the same.Langton used pan-genotype or GTYPE to refer to any collection of low-level rules, and pan-phenotype or PTYPE to generally refer to the structure or behavior produced by the interaction of these rules in a specific environment.For example, in traditional computer programs, the pan-genotype itself is obviously the computer code, while the pan-phenotype is the program response caused by the user's input information.In Langton's self-replicating cellular automata, the pan-genotype is the rules that dictate how each cell interacts with its neighbors, while the pan-phenotype is the overall shape.In Reynolds' Betz program, the pan-genotype is the rule of three that guides Betz's flight direction, and the pan-phenotype is the grouping behavior of the Betz group. More broadly speaking, the concept of pan-genotype basically coincides with Hornan's concept of internal model, the only difference is that Langton emphasizes the role of pan-genotype as a computer program more than Horan.So, naturally, the idea of ​​a pan-genotype can be fully applied to Hornan's classifier system, as a set of classifier rules.The same idea applies to the Eco model, in which an organism's pan-genotype includes both offensive and defensive chromosomes.In Arthur's glass house economic model, the pan-genotype of artificial actors is the economic behavior rules learned through hard work.Basically, this concept applies to any complex adaptive system in which agents can interact according to a set of rules.And when their pan-genotype develops into a pan-phenotype, it presents a computational approach. disobedient computer software The beauty, says Langton, is that once you combine the idea of ​​life with computation, you can generate a wealth of related theories, such as why is life so amazing? In general, you cannot start with a set of pan-genotypes and predict how their pan-phenotype behavior will be.This is the undecidability theorem of computer science; that is, unless a computer program is trivial, the fastest way to know the outcome is to run the program and see what happens.No general-purpose program can scan a computer's passwords and enter information faster, and then tell you the answer.People used to like to say that computers only listen to programmers. This statement is true on the one hand, but on the other hand, it is false, because any complex and interesting computer password will produce unexpected results.That's why computer software is endlessly tested and bug-fixed before it's released to the market, yet users find bugs quickly enough.Most importantly for artificial life, this is why living systems can be fully programmed (pan-genotyped) biochemical machines on the one hand and still produce unexpected, spontaneous behaviors in the pan-phenotype. As you can tell from many other computer science theorems, you can't do it the other way around.You can't set the behavior you want (pan-phenotype) and hope to find a set of rules that will produce that outcome (pan-genotype).Of course, none of the theorems prevents programmers from using well-tested algorithms to solve well-defined problems under well-defined circumstances.But living systems often face ill-defined, ever-changing environments, Langton said, and the only way to go seems to be trial and error (aka Darwinian natural selection). He noted that the road seemed brutal and time-consuming.Nature's program is to build many different machines out of many arbitrarily compartmentalized pan-genotypes, and weed out the ones that don't work.In fact, this messy and wasteful process may be the best that nature can find.Moreover, Hornan's genetic algorithm may be the only viable way to enable computers to deal with messy and ill-defined problems.This is probably the only efficient program that allows you to find pan-genotypes from specific pan-phenotype cues. Are computer viruses alive? In writing his general essay, Langton was careful not to claim that the entities studied by artificial life scholars were truly alive.It is obvious that these entities are not really alive. No matter whether it is a bezier, a plant, or a self-replicating cellular automaton, they are all just computer simulations, and they are simplified life models that do not exist without computers.Still, since the whole point of artificial life is to capture life's most fundamental principles, it's hard to escape the question: Can humans finally create truly artificial life? Langton found this question difficult to answer, in part because no one knew what real artificial life would look like.Maybe some kind of genetically engineered superorganism?Or a self-replicating robot?Or an overeducated computer virus?What is life?How can you be sure that you have really found life? Needless to say, the issue was widely discussed at the seminar, where it was debated loudly and enthusiastically, not only in the halls, but also in the corridors or at the dinner table. Computer viruses in particular are a hot topic.Many attendees felt that a computer virus had crossed the line, a nuisance that fit nearly every life condition anyone could think of.Computer viruses can reproduce and spread by copying themselves to another computer or disk; computer viruses can be stored in the form of computer codes, just like DNA; computer viruses can also commandeer host (computer) to perform their own functions, just as real viruses commandeer the molecular metabolism of infected cells; computer viruses respond to environmental (computer) stimuli; Thanks, computer viruses can even mutate and evolve.Although a computer virus cannot survive independently in the material world, it cannot be denied that it is a living thing.If, as Langton claims, life resides in organization, then a properly organized entity is alive, no matter what it is made of. Regardless of the identity of the computer virus, however, Langton has no doubts that true artificial life will one day emerge, and that day may be in the not-too-distant future.Furthermore, due to the development of biotechnology, robotics and advanced software technology, artificial life will be applied in commercial and military applications.However, the study of artificial life is therefore extraordinarily important, and if we are indeed heading towards a brave new world of artificial life, then at least we will keep our eyes open as we go. Playing the role of God? Langton writes: By the middle of this century, humanity had the capability to wipe out life on Earth.Humans will be capable of creating life before the middle of the next century.Of the two, it is difficult to say which imposes a greater responsibility on us.In the future, not only artificial life will appear, but the process of evolution will be more and more under the control of human beings. With this vision in mind, he feels that every scientist who devotes himself to the field should read Frankenstein right away.In the book (though not in the movie), Frankenstein claims that he is not responsible for his creation.Langton pointed out that we will never allow this to happen. We cannot predict how what we do now will affect the future, but we must take responsibility for the consequences no matter what.In other words, we must openly discuss the meaning of artificial life. Beyond that, assuming you could actually create life, you'd suddenly have problems bigger than a technical definition of life or non-life, and you'd quickly get bogged down in some kind of positive theology.For example, after creating a living creature, do you have the right to command it to worship and worship you?Do you have the right to play the role of God?Do you have the right to destroy it if it doesn't listen to you? Those are good questions, Langton said.Regardless of whether we can find the right answer, we should discuss these issues honestly and openly.Artificial life challenges not only science and technology, but also our most fundamental social, moral, philosophical and religious beliefs.Artificial life, like Copernicus' theory of the solar system, will force us to re-examine our position in the universe and the role we play in nature. new second law If Langton's tone is higher than the average scientific paper, he's not the only exception at Los Alamos; Farmer won't make him stand out. The best example is Farmer's non-technical paper, Artificial Life: The Coming Evolution, published in 1989 with his wife, environmental lawyer Alletta Belin, The paper was presented at a symposium at Caltech celebrating Gorman's 60th birthday.They write: With the advent of artificial life, we may be the first beings capable of creating our own heirs.If we fail to play creator, our heirs may be grim and malevolent.But if we succeed, they may be superior creatures far surpassing us in knowledge or wisdom.It is likely that when future conscious life looks back on this century, we will be the ones with the most attention; not because of ourselves, but because of the life we ​​have created.Artificial life is quite possibly the most beautiful creation of mankind. Regardless of his rhetoric, Farmer is serious about artificial life as a new science, so he is also serious about supporting Langton.After all, it was Farmer who first brought Langdon to Los Alamos, and although he was angry that Langdon hadn't finished his doctoral dissertation, he had no regrets.He said: Langton definitely has his value. Everyone likes him very much. There are too few people like him who really have dreams and goals in life.Langton isn't very efficient, but I think he's insightful and has a way of realizing his vision, and he's not afraid of the details. Indeed, Farmer was a mentor and friend to Langton, even though Langton was actually five years older than him.Farmer, one of the very few young scientists in Santa Fe's decision-making inner circle, persuaded Cowan to donate $5,000 to sponsor Langton's Artificial Life Symposium in 1987, and also arranged for Langton to be in Santa Fe. He spoke at meetings of the Philippine Institute, advocated at the Santa Fe Science Council for the hiring of scientists to study artificial life, and encouraged Langton to hold regular small seminars at Los Alamos.Most importantly, when Farmer agreed in 1987 to chair the newly formed Complex Systems Group in the Los Alamos Theory Division, he listed artificial life as one of the group's three major research projects. Farmer was not a natural administrative talent.Thirty-five-year-old Farmer, a tall, bony New Mexican with the look of a graduate student until now, with a ponytail and a T-shirt, shouted: Question Authority!The hectic bureaucracy hurts him, but writing proposals and asking for money from the dorks in Washington hurts him even more.Farmer, however, was gifted both for funding and intellectual enthusiasm.He is famous for his mathematical forecasting, and he was the first to find ways to predict the future behavior of seemingly random and chaotic systems, including the future movement of the stock market, which people are most interested in studying.Moreover, Farmer has no regrets in giving Langton and a small group of artificial life scholars more than half of the group's general funding, while his own nonlinear prediction and other research work is left to his own devices.Forecasting yields real results, so I can promise the sponsoring agency a payback within a year.But artificial life research will take a long time before it can produce practical results.In the current environment, it is almost impossible for artificial life to apply for research funding. In the long run, the current situation is not ideal.Farmer loved the work of forecasting, but outside of administrative responsibilities and forecasting work, he had little time to study artificial life, a topic that struck a chord with him more than any other subject.Artificial life, he says, cuts right into the deep questions of emergence and self-organization that have been on his mind. humanity's last question Farmer said: I have been thinking about self-organization since I was in middle school.Although when I first started, my idea was very vague and inspired by science fiction.He especially remembers a novel by Isaac Asimov, The Final Question.In the novel, future humans ask the cosmic supercomputer how to eliminate the second law of thermodynamics, which is the tendency of everything in the universe to cool down and decay.How, they ask, can we reverse increasing entropy (transliteration for entropy).As a result, after the death of mankind and the cooling of the planet for many years, the computer finally learned how to accomplish this great task, so it announced: the light of creation has returned!A new, low-energy universe was born. Farmer was fourteen when he read Asimov's novel, and even then he sensed that it pointed to a profound problem.Why, he asked himself, did the universe still produce stars, clouds, and trees, if it could grow stronger and stronger, if disorder and disorder at the atomic scale were immutable?Why does matter become more and more organized on a large scale and less and less organized on a small scale?Why didn't everything in the universe disintegrate into chaos and miasma before ancient times? Farmer said: Honestly, interest in these questions drove me to become a physicist.Wooters (William Wooters, a physicist) and I used to sit on the grass after our physics class at Stanford and discuss these issues, with all kinds of ideas running through our minds.It was many years later that I discovered that other people had similar ideas and that they were documented, such as Norbert Wiener (1894︱1964) and regulation, Prigogian and the idea of ​​self-organization, and so on.In fact, the same issue can be found even in the writings of the British philosopher Herbert Spencer.In the 1860s, Spencer, who popularized Darwin's theory by inventing phrases like survival of the fittest, saw Darwin's theory of evolution as a special case of the vast forces driving the natural origin of the structure of the universe. So many people were thinking about these questions independently, Farmer says, but he was frustrated at the time that there wasn't a single discipline that was addressing it.Biologists are not working on this kind of question, they are stuck in the maze of which protein will react with which protein, and ignore the general rules.As far as I can see, physicists are not working on this kind of problem either.This is one of the reasons I plunged headfirst into Chaos Theory. In his best-selling book Chaos, Greyick dedicated a chapter to tell this story: In the 1970s, Farmer and his lifelong friend Packard were still studying physics at the University of California, Santa Cruz. How they became fascinated by the kinematics of roulette.Computing the trajectory of a ball rolling rapidly on a roulette wheel gave them a sense that small initial changes in the physical system could produce large changes in the final outcome.The book also describes how they and two other graduate students, Robert Shaw and James Crutchfield, began to understand the new science of so-called chaos or the theory of dynamical systems that are more familiar to ordinary people; Based on the research in this area, the so-called power system group was formed. After a while, however, I got tired of chaos theory, Farmer said.The basic theory of chaos has been formed, so I no longer have the fun of breaking new ground and exploring the unknown.In addition, chaos theory itself is not deep enough. Chaos theory tells you that simple rules of behavior can produce extremely complex changes; Not much, nor does it explain how a disjointed initial state self-organizes into a complex whole.What's more, chaos theory didn't answer his old question: Why is structure and order constantly forming in the universe? Farmer believes the answer has yet to be revealed.This is why he, together with Kaufman and Packard, studied the autocatalyst group and the origin of life, and enthusiastically supported Langton's research on artificial life.Like many in Los Alamos and Santa Fe, Farmer could feel that understanding, answer, principle or law, almost within reach. He said: I maintain that life and organization are as immutable as increasing energy tends to wear out; but, because life and organization are less regular, they appear to be more uncertain.Life reflects a universal phenomenon that I believe can be described by a law like the second law of thermodynamics, which describes the tendency of matter to organize itself and predicts the general nature of organization in the universe. Farmer had no idea what the new Second Law would look like.If we are clear, we know how to get there.It's purely speculation at the moment, just a hunch.In fact, he doesn't know if there will be one law in the result, or many laws?He does know, however, that people have recently discovered many clues, such as emergence, adaptation, the edge of chaos.They have at least begun to sketch the outlines of this hypothetical new second law. emergent First, Farmer says, this imaginary law must have a rigorous description of emergent: What exactly does it mean that the whole is greater than the sum of its parts?He said: "It's not magic, but to humans, in our little brains, it feels like magic.Flying beetles (and true flocks) adapt to the movements of their neighbors, thus forming flocks.Organisms cooperate and compete in a dance of coevolution, thus forming coordinated ecosystems.Atoms bond to each other in search of the lowest energy state, thus forming emergent structures called molecules.Human beings buy, sell or trade things with each other to satisfy material needs, thus creating emergent structures called markets.Humans also interact with each other for other unquantifiable factors, forming families, religions, and cultures.By constantly seeking mutual adaptation and self-unification, the agent transcends itself and composes something new.The trick is figuring out what makes sense without turning into dry philosophy or New Age mysticism. And that's the beauty of computer simulation and artificial life: you can experiment with simple models on your desktop to see how your ideas actually work, to see if you can accurately establish vague concepts, Also try to distill the essence of prominence at work in nature.Moreover, there are currently many models to choose from.In particular, Farmer's attention was drawn to the association theory's use of a network of interconnected nodes to represent a group of interacting agents.In the past ten years, associative models have appeared everywhere. The best example is the neural network movement. Scholars use artificial neurons to simulate perception and memory recovery. At the same time, they also launch a fierce attack on the symbol processing methods adopted by mainstream artificial intelligence. .緊迫其後的就是聖塔菲研究院所支持的許多模型的研究，包括賀南的分類者系統、考夫曼的遺傳網路、生命起源的自動催化組模型，以及派卡德在一九八○年代中期和羅沙拉摩斯的皮瑞森合作的免疫系統模型。 法默承認，其中有許多模型看起來不像結合論的模型，許多人第一次聽到這些模型被歸為結合論，都大吃一驚。但是，這只不過是因為不同的人在不同的時間，創造了這些模型以解決不同的問題，而且他們用不同的語言來描述這些模型。法默說：其實剖開來看，他們的本質都一樣。 當然，在神經網路中，節點及連結點的結構非常明顯。節點相當於神經元，而連結點則相當於連接神經元的突觸。如果程式設計師有一個視覺的神經網路模型，他可以藉著刺激能接受輸入信息的節點，使之反應，來模擬明暗不同的光線落在視網膜的形態，然後讓這種反應透過連結點，散播到神經網路的其他部分。這就好像把滿滿一船貨物送到沿海各個港口，再由不計其數的貨車把貨物沿著公路運送到內陸城市。如果妥善安排連結點，那麼網路的反應很快就會安定下來，對應於所看到的景觀（例如，那是一隻貓！），而形成自我統一的形態。而且，即使輸入的信息混雜而不完整，網路模型仍然會有同樣的表現。 在賀南的分類者系統中，節點︱連結點結構就沒有那麼明顯。節點組就是所有可能的內部信息的組合，例如l001001110111110。而連結點就是分類者規則，每一條規則都在系統的內在公布欄上找尋適當的信息，然後也在布告欄上張貼信息回應。程式設計師藉著刺激一些輸入節點，也就是把相關的輸入信息張貼在布告欄上，來刺激分類者發出更多的信息，然後又引起更多信息回應。結果就好像神經網路散播對刺激的反應一樣，分類者系統中會流瀉出大量信息。然後，也正像神經網路會安定下來形成自我統一的狀態；分類者系統也會安定下來，形成一組穩定的信息及分類者，以解決眼前的問題或是在賀南的眼中，代表一個突現的心智模型。 異曲同工 法默說，在他和派卡德、考夫曼合作完成的自動催化及生命起源模型中，也可以看到這種網路結構。在這個模型中，節點組也就是所有可能的聚合物物種的組合，例如abbcaad，連結點則是聚合物之間的模擬化學反應：聚合物A催化聚合物B，以此類推。藉著刺激某些節點的反應（也就是從模擬的環境中，讓小小的食物聚合物穩定的流入系統之中），將會引起一連串的反應，最後安定下來，形成活躍而且可以自給自足的聚合物和催化反應形態也就是自動催化組，一種模擬從太初渾湯中突現的原始有機體。 其他模型的分析也都殊途同歸，其中都暗藏著同樣的節點︱連結點結構。法默說，找到共同的架構令人安心不少，因為這表示四個瞎子至少是把手放在同一頭大象身上。此外，共同架構也幫助學者更容易溝通，不必再遭受不同術語的干擾。最重要的是，找出共同的架構可幫助學者提煉出模型的精髓，因此能更提綱挈領的討論突現的意義。而這些模型告訴我們的教訓就是：力量其實蘊藏於連結點之中。這是為什麼許多人對結合論如痴如醉，因為你可以從非常、非常簡單的節點著手線性的聚合物、只有二元的信息、或是只能開關的神經元，經過相互反應後，仍然產生令人驚訝而複雜的結果。 就以學習和演化為例吧。既然節點如此簡單，網路的整體行為就幾乎完全由連結點來決定。或是套句蘭頓的話，連結點把網路的泛基因型編碼了，因此要修正系統的泛表現型行為，你只要改變連結點即可。法默說，事實上，要改變行為的方式有兩種。第一種是連結點的位置固定不變，但是改變它們的強度。這就是賀南所謂的採掘式學習（exploitation learning），不斷改善你已有的知識。在賀南的分類者系統中，他藉著不斷獎勵能產生好結果的分類者規則，來達到這個目的。在神經網路中，則藉著各種學習演算法，在網路中呈現一系列已知的輸入信息，然後不斷加強或減弱連結點，直到出現正確的反應。 第二種比較激烈的調整連結點的方法，是改變網路的整個布線圖，把舊的連結點扯掉，放入新的連結點。這種方式就等於賀南所謂的探險式學習（exploration learning）冒大風險來取得高回收。例如自動催化組中發生的狀況就是如此，就好像在真實世界一樣，偶爾會自動形成新的聚合物，因此而產生的化學連結點會給自動催化組一個機會，來探索聚合物的全新領域。但是，神經網路就不會發生這種狀況，因為神經網路的連結點是模擬突觸的，不能被更動。但是，近來有一批神經網路迷所做的實驗中，神經網路在學習的過程也會重新布線，理由是任何固定的布線圖都是任意配置，應該容許改變。 法默說，所以簡單的說，結合論的想法顯示，即使節點（個別的作用體）沒有腦子，沒有生命，依然能突現出學習和演化的能力。更廣義而言，當力量是在於連結點，而非節點時，所代表的意義就和蘭頓及人工生命學者的理論一致。也就是說，生命的本質是在於組織而不在於分子。這同時也幫助我們對於生命和心靈在宇宙的起源，有了更深一層的了解。 混沌邊緣的魅力 法默說，但是儘管美景可期，結合論模型仍然無法解釋新的第二定律。首先，結合論模型無法告訴你突現如何在經濟、社會、或生態體系中運作，在這些體系中，節點都非常精明，會不斷彼此適應。要了解這類的系統，你必須先了解合作和競爭的共同演化之舞，也就是說，用過去幾年日漸流行的艾可之類的模型來研究共同演化。 更重要的是，無論是結合論模型或共同演化模型，都無法解釋生命和心靈最初的起源。宇宙中為什麼會出現生命和心靈？單單說突現不足以解釋一切。宇宙中充斥著各種突現的結構，例如銀河、雲、雪花等物體，但它們都沒有獨立的生命。一定還需要其他的條件，而這個假設的新第二定律必須告訴我們其他的條件是什麼。 顯然，必須由直接指向基本物理和化學原理的模型來完成這個工作，例如蘭頓最喜歡的細胞自動機，法默說。而蘭頓在細胞自動機中發現的混沌邊緣的奇怪相變，似乎正提供了大部分的解答。在人工生命研討會中，蘭頓對這個題目一直保持緘默；但是從一開始，聖塔菲和羅沙拉摩斯的許多人都發現混沌邊緣的觀念扣人心弦。 蘭頓的基本觀念是，產生生命和心靈的神祕東西是在秩序和失序之間的某種平衡。更明確的說，你應該從系統如何表現的角度來看系統，而不是只注意系統如何構成。如此一來，你就會發現秩序和混沌這兩個極端，就好像當原子被鎖定在定點所形成的固體，以及當原子任意互相顛覆時形成的液體一樣。但是在這兩種極端之間，在某種叫混沌邊緣的抽象相變中，你也可以找到複雜性也就是系統的組成元素從來不會鎖定在固定位置，但是也從來不會分崩離析，變成混沌一片。這類系統一方面穩定得足以儲存資訊，另一方面又鬆散得足以傳遞資訊。這類的系統能組織起來作複雜的計算，能對外界反應，能表現得自動自發、有適應性及生意盎然。 前蘇聯當然會解體！ 當然，嚴格的說，蘭頓只有在細胞自動機中證明了複雜和相變之間的關聯性。沒有人知道在其他模型中，或在真實的世界裏，是不是依然如此。但是法默說，另一方面，有很多線索顯示，或許這是真的。例如，你可以看到多年來，許多結合論模型會突然出現類似相變的行為。早在一九六○年，考夫曼在他的遺傳網路模型中最先發現的就是相變。如果連結點太鬆散的話，網路基本上就凍結不動，而如果連結點太緊密的話，網路就會在一片混亂中劇烈攪動。惟有在兩者之間，當每個節點恰好有兩個輸入信息時，網路才能產生考夫曼所要尋找的穩定的狀態循環。 法默說，在一九八○年中葉，自動催化組模型也發生了相同的狀況。這個模型有許多參數，例如反應的催化強度，以及食物分子供給的頻率都是。基本上，他和派卡德、考夫曼必須借助從嘗試與錯誤中獲得的經驗，以人工來設定所有的參數。他們首先發現的就是在參數到達某種程度之前，模型中沒什麼狀況發生，但是一旦跨越了某個門檻後，自動催化組就會迅速發展。法默說，這種行為又和相變大同小異。 他說：我們可以感覺到雷同之處，但又很難明確的解釋清楚。這是另外一個領域，需要有人作一些嚴謹的比較分析。 法默說，更混沌未明的是，混沌邊緣的觀念能不能應用在共同演化的系統上？當你探討生態系或經濟系統時，你不清楚是否能明確定義像秩序、混沌及複雜這些觀念，更遑論相變了。儘管如此，法默覺得混沌邊緣的原理仍然沒錯。就拿前蘇聯為例吧，很明顯，採取中央極權統治的社會組織是行不通的。長期以來，史達林所構築的體系太僵硬不變，控制嚴密，以致無法生存。或是看看七○年代底特律的三大汽車公司吧，他們規模擴張得太龐大，太嚴格的鎖定幾種做事方式，以致對日益增強的日本挑戰無動於衷，更不要說積極應變了。 另一方面，無政府主義也行不通，最好的例子就是前蘇聯瓦解後，部分小國的狀態。自由放任的經濟體系也行不通，狄更斯筆下英國工業革命時期的恐怖生活及近代美國儲貸銀行大災難，都是例證。最近的政治發展更顯示，健全的經濟和健全的社會都必須讓秩序和混沌保持平衡，但不是只求取和稀泥、折衷式的平衡而已。這些體系必須像活細胞一樣，一方面以嚴密的回饋及管制來自我規範，另一方面也要留下創造、改變及因應新狀況的空間。在由下而上、有彈性的組織中，演化勃然而興；但同時，演化必須導正由下而上的活動，以免摧毀整個組織，必須有某種控制的階層使資訊不但由下而上流動，同時也由上而下流動。法默說。混沌邊緣的複雜動力學似乎最適合這類的行為。 邁向複雜 法默說：我想我們隱約知道，這種有趣的組織現象活動的領域在哪裏。然而，這並不能解釋一切。即使為了辯解，你假設這種特別的混沌邊緣領域確實存在，想像中的新第二定律還是得解釋突現系統如何到達這個領域，如何繼續留在混沌邊緣，以及在那裏做什麼。 法默說，你很容易就安慰自己說，達爾文早就回答了這兩個問題。既然在競爭的世界裏，反應最複雜的系統總是最占上風，那麼固定不變的系統只要稍微放鬆一點，就能表現得更好，而紊亂的系統只要有組織一點，也同樣能表現得更好。所以如果這個系統還不在混沌邊緣的話，你期待學習和演化會把系統推向混沌邊緣。如果系統已經在混沌邊緣了，那麼你期待學習和演化會在它開始游離時，把它拉回來。換句話說，你期待學習和演化穩住混沌邊緣，因為那才是複雜、適應性系統生存的自然領域。 第三個問題：一旦到達了混沌邊緣，系統做什麼事情？這個問題比較微妙。 在所有可能的變動行為的廣大空間中，混沌邊緣就像一片超薄的薄膜，是分隔混沌和秩序的特殊複雜行為區域。就像你看，汪洋大海的表面只不過由一個分子那麼厚的界線，來分隔海水輿空氣。而且混沌邊緣也像汪洋大海一樣，廣闊得超乎想像之外，在其中，作用體可以經由無限種方式而變得複雜而有適應能力。的確，就像賀南提到的永恆的新奇，以及有適應能力的作用體逡巡於可能性的無垠穹蒼，他所討論的正是有適應能力的作用體游走於廣闊的混沌邊緣薄膜中。 所以，關於這點，新的第二定律可能會怎麼說呢？ 當然，新的第二定律可能會部分談到基本單位、內在模型、共同演化，以及賀南和其他人研究的所有適應機制。然而，法默猜想這條定律的核心或許不是關於機制，而是關於方向它在陳述一個簡單的事實：演化的結果總是令事物比演化之前更複雜、更精巧、更有架構。他說：雲比大霹靂後的混沌一片有架構，生命起始的太初渾湯又比雲有架構。我們又比太初渾湯有架構，現代經濟體系要比美索不達米亞的城邦有架構，就好像現代科技要比羅馬時代的科技複雜一樣。似乎學習和演化不止慢慢、不可遏止的把作用體拉向混沌邊緣，同時學習和演化也把作用體沿著混沌邊緣，帶向愈來愈複雜的發展方向。why? 什麼是進步？ 我們很難在生物學中說明進步的觀念。法默說。當我們說，這個生物比那個生物先進，究竟代表什麼意思？例如，蟑螂已經在地球上生存了幾億年，比人類的歷史還要久遠，牠們對於當蟑螂非常拿手。我們真的比牠們先進嗎？還是只是不同而已？六千五百萬年前，我們的靈長類祖先真的比殘暴霸王龍先進，還是只不過僥倖逃過彗星隕落的劫難而已？ 法默說，當我們無法清楚的定義適應度（fitness）時，適者生存就和生者生存（survival of the survivors）意義相同。 但是，我也不相信虛無主義也就是任何事物都不會比別的事物更好。如果你退後一步，綜覽整個演化過程，我相信你可以很有意義的討論進步的觀念，你會看到整個趨勢都是朝向日益增強的精巧、複雜和功能性發展；T型車（福特最早期製造的汽車）和法拉瑞車的差異比起最初的有機體和最近的有機體的差異，簡直是小巫見大巫。儘管這個觀念叫人難以捉摸，但是演化的設計確實日漸朝向質的提升，這種整體趨勢是有關生命意義最迷人而深奧之處。 他最喜歡的例子就是他和派卡德、考夫曼合作的自動催化組研究中的演化。自動催化的美妙之處在於你可以從頭看著突現產生，看著一些能互相催化的化學組，其濃度以極驚人的速度超越了均衡狀態的濃度。也就是說，整個自動催化組就好像突現的新個體，從均衡的背景中突顯出來，這正是你想用來解釋生命起源的現象。如果我們知道如何在真正的化學實驗中，重複這個過程，那麼我們就找到了生命和非生命的中介者。這些自動催化的個體沒有遺傳密碼，但是卻能自給自足，自我繁衍；它可能不像種子做得那麼好，但是卻比一堆石頭好一些。法默說。 在最初的電腦模型中，自動催化組沒有演化，因為當時自動催化組與外在環境之間沒有互動關係。模型假設所有的事情都發生在一鍋混合的化學溶液中，所以一旦自動催化組突現出來，就會變得很穩定。然而，在四十億年前真實的世界裏，環境會讓這種面貌模糊的自動催化個體陷於掙扎、波動之中。所以為了要了解在這種情境下會發生什麼狀況，法默和研究生巴格利讓自動催化組模型的食物供應產生波動。最妙的就是，有些自動催化組就像貓熊一樣，只吃竹子。如果你改變了它們的食物，它們就活不了。但是另外有一些自動催化組則像雜食動物一樣，有很多不同的新陳代謝方法，因此可以替換不同的食物分子。所以，當你改變供應的食物時，它們完全不受影響。法默說。像這種最強壯的自動催化組很可能就會生存於遠古的地球上。 近來，法默又和巴格利及羅沙拉摩斯的博士後研究員方塔那（Walter Fantana）再修正了一次自動催化組模型，使這個模型能像真正的化學系統一樣，容許間歇發生一些自然反應。這些自然反應引起很多的自動催化組分裂，但是分崩離析的自動催化組卻正好為演化的大躍進鋪路。它們引發了紛至沓來的新品種，有些變種變得更強大，然後又穩定下來，直到下一次大崩潰出現。我們看到了一系列的自動催化組變種相互取代。也許這就是線索。法默說：我很有興趣看看我們在談進步的概念時，是否能涵蓋突現的結構這種突現結構具備了某些前所未有的追求穩定的回饋環。關鍵在於必須發生一系列的演化來架構宇宙的物質，每一個層次的突現都為下一個層次的突現鋪路。 等待拼圖英雄 法默說：談這些令我沮喪，因為有語言上的問題。人們再三討論如何定義複雜和突現計算的傾向，但是如果不能以數學術語作清楚的定義，單單用這些字眼，只能在你的腦中激起模糊的影像。這就像熱力學誕生之前的一八二○年代，他們知道有一種東西叫作熱，但是他們討論的時候所用的名詞，後來聽起來簡直可笑極了。 事實上，當時他們甚至不確定熱是什麼，更不要談了解熱如何作用了。當時大多數德高望重的科學家都相信燒得紅熱的撥火鐵棒中密布著一種叫熱素的無重透明液體，只有少數人認為熱可能代表了撥火棒的原子中某種細微的運動。結果少數人的意見是正確的。 此外，當時也沒有人能想像如蒸汽機、化學反應及電池等複雜混亂的事物，竟然全都由一些簡單的通則所宰制。直到一八二四年，年輕的法國工程師卡諾（Sadi Carnot）發表了後來被稱之為熱力學第二定律的陳述：熱不會自然的從冷物體流到熱物體，當時卡諾正在寫一本關於蒸汽機的書，他正確的指出，這個簡單而普遍的事實使蒸汽機的效率大受限制，更不要提內燃機、發電廠的渦輪機，或是任何靠熱來發動的引擎了。不過要再過七十年，才出現第二定律在統計上的解釋：原子總是不斷嘗試回復隨機狀態。 法默說，同樣的，直到一八四○年代，英國釀酒商和業餘科學家焦耳（James Joule）才為熱力學第一定律奠定了實驗基礎。熱力學第一律又稱能量不滅定律，它說明了能量能夠從一種形式轉換到另一種形式，包括熱的、機械的、化學的或電的形式；但是你永遠無法創造或毀滅能量。到了一八五○年代，科學家才能以清晰的數學形式說明這兩條定律。 在自我組織的範疇中，我們正悄悄的邁向這個階段。但是，了解組織要比了解失序難多了，我們還無法以清晰量化的形式來說明自我組織的核心概念。我們需要像氫原子這樣的東西，讓我們能夠把它分解開來，清楚描述讓它發生作用的關鍵是什麼。但是，目前我們只了解這塊拼圖中的零星片段，每一片各自有其意義。例如，我們現在很清楚混沌和碎形的概念，也就是簡單的部分所構成的簡單系統能產生複雜的行為。我們也蠻了解果蠅的基因調節。在某些特定的狀況下，我們也依稀掌握了腦部的自我組織過程。而在人工生命的研究中，我們創造了小宇宙的新內涵。所有的這些行為只約略反映了自然系統中的真實狀況，但是我們已經能夠完全利用電腦來模擬上述的概念，並任意改變其中的情況，而且不但知其然，亦知所以然。我們希望有一天終於能組合所有的片段，形成完整的演化和自我組織理論。法默說。 他補充：喜歡問題定義明確的科學家對這門學問不會感興趣，但是沒有定見正是這個領域迷人之處。一切都還在發展之中，沒有人已經找到解答的途徑，但是點點滴滴的線索四處飄散，模擬的系統和模糊的概念紛紛湧現，因此可以預見二、三十年後，我們將會有一套真正的理論。 思考天擇問題 至於考夫曼，他衷心希望不需要等那麼久。 他說：我聽過法默說這好像熱力學誕生前的階段，我想他說得沒錯，我們在複雜科學中想要尋求的，是整個宇宙的非均衡系統中狀態形成的通則。有了像混沌邊緣之類的提示，我覺得我們已經瀕臨突破，就好像距離卡諾發明熱力學只有幾年的時間。 的確，考夫曼顯然希望新一代的卡諾會叫作考夫曼。他像法默一樣，預見新的第二定律會解釋突現的實體在混沌邊緣時，如何表現出最有趣的行為，以及這些實體如何藉著適應的過程，而愈趨複雜。但是，和法默不同的是，考夫曼沒有背負行政管理的重擔和挫折。自從抵達聖塔菲的第一天起，他就埋首於這個問題中。他十分迫切的想找到答案，彷彿要花三十年來探究秩序和自我組織的意義，令他有種咫尺天涯的痛楚。 他說：對我而言，在研究自我組織和天擇如何結合的辛苦過程中，下一步就是研究邁向混沌邊緣的演化過程。我覺得很煩，因為我依稀捉摸到一點輪廓了。我並不是太過小心翼翼，我的研究還未結束，我只是對許多事情都有了初步的認識。我覺得自己好像榴彈砲，射穿一面又一面的牆，留下滿目瘡痍。我覺得我匆匆跳過一個又一個題目，想要看到彈道弧線的終點，卻不知道回程時該如何清除那些殘骸。 這條弧線要回溯到一九六○年代，當時他剛開始琢磨自動催化組和基因組的網路模型。在那段日子裏，他確實希望能相信，生命完全經由自我組織而形成，天擇只不過是枝節而已。胚胎發育就是最好的證明，相互作用的基因自我組織成不同的形狀，對應於不同的細胞；而相互作用的細胞又在發育中的胚胎自我組織成不同的結構。他說：我從來不懷疑天擇的作用，我只是覺得在最深層，終究還是和自我組織有關。 然後，在一九八○年初，有一天我去拜訪史密斯。史密斯是著名的英國生物學家，也是考夫曼的老朋友。考夫曼在停頓了十年之後，這時候正重新開始認真思考自我組織的問題，在這十年間，他一直在研究果蠅的胚胎發育。他說：史密斯夫婦和我一起散步時，史密斯說我們離達爾文的家不遠，然後他就大發高論：一般說來，把天擇認真當一回事的人是英國鄉紳，就像達爾文一樣。接著他看著我，微微笑了一下，然後說：認為天擇和生物演化沒什麼關係的人是城市猶太人！史密斯說：你真的得好好思考天擇的問題，考夫曼。我卻不聽，我希望一切自然產生。 但是，考夫曼必須承認史密斯說得對，自我組織沒有辦法獨自完成這項工作，畢竟突變的基因和正常基因一樣可以自我組織，結果當產生的是腿長在頭上或沒有頭的畸形果蠅時，你仍然需要靠演化來去蕪存菁。 百味雜陳 他說：所以，一九八二年我坐下來寫書的大綱。（經過再三修訂，終於在一九九二年出版的這本著作秩序起源（The Origins of Order），是考夫曼三十年來思考的精華。）這本書打算探討自我組織和天擇：你如何整合這兩種理論？最初我的想法是兩者互相衝突，天擇要的是一種東西，但是系統中的自我組織行為又有其局限。所以，兩者相互較勁，直到達到某種均衡之後，演化再也無法推動改變。整本書的前三分之二，我都抱持著這種意念。考夫曼說。或說得更精確一點，他這種想法一直延續到一九八○年代中，直到他在聖塔菲聽到混沌邊緣的說法，才開始動搖。 最後，混沌邊緣的觀念大大改變了自我組織與天擇問題在他心目中的意義。但是，那時他卻百味雜陳，因為他自己不但從一九六○年代起，就在遺傳網路上看過近似相變的行為，而且在一九八五年，他幾乎就要領悟了混沌邊緣的想法，但終究擦身而過。 他帶著自責的語氣說：有很多篇論文，我一直懊悔沒能寫出來，這正是其中一篇。一九八五年夏天，他利用休教授年假的時間在巴黎做研究，這個想法正是在那個時候開始萌芽。當時，他和物理學家威斯巴可（Gerard Weisbuch）及研究生佛吉曼梭爾（Francoise Fogelman︱Soule）一起到耶路撒冷的哈達薩醫院（Hadassah Hospital）待了幾個月。一天早上，考夫曼正在思考網路中他稱之為凍結成分（frozen component）的問題。他在一九七一年首先注意到這種現象。在他的燈泡比喻中，這就好像網路中東一堆、西一堆相互連接的節點群，不是全都亮起燈光，就是全都熄滅，而且就一直保持那樣的狀態；然而網路中其他部分的燈泡卻一直持續閃爍。在緊密連結的網路中，燈泡集體混亂閃爍，根本不會出現凍結成分。然而凍結成分卻充斥於連結疏鬆的網路中，這也是為什麼這類的系統很容易就完全僵化。他很好奇，介於兩者之間會是什麼狀況？也就是最近似真實遺傳系統的不太緊密、也不太鬆散的網路，也就是既不完全僵化，也不是全然一片混沌的網路 還記得那天早上，我衝到威斯巴可和佛吉曼梭爾的面前說：你們瞧，就在凍結成分逐漸溶解、連結，未凍結的孤島也開始連結之處，能夠產生最複雜的計算！那天早上，我們熱烈的討論這個現象，我們都同意這個問題很有趣，但是我們手邊還有別的工作，此外，當時我的想法還是：沒有人會在乎這些東西。所以，我沒有繼續研究下去。 所以，當考夫曼聆聽有關混沌邊緣的演講時，他的心情混雜著興奮與懊悔。他禁不住覺得這個觀念有一部分為他所有；但是他也承認，蘭頓把相變、計算和生命的關聯性解析得扣人心弦，蘭頓下了苦功把這個想法琢磨得嚴謹而精確。更重要得是，蘭頓領悟到考夫曼所未見的道理：混沌邊緣不只是純然的秩序系統和混沌系統之間的簡單分界線。經過多次長談後，蘭頓終於讓考夫曼明白，混沌邊緣是自成體系的特殊區域，是你能找到近似生命的複雜行為的地方。 所以，蘭頓的研究出色而且重要。儘管如此，由於其他各種研究工作和寫書的羈絆，考夫曼幾年後才真正領悟了混沌邊緣的含義。事實上，那是在一九八八年夏天，派卡德從伊利諾到聖塔菲來，在研討會中說明他對混沌邊緣的研究。 配錯對遊戲 派卡德差不多和蘭頓同時有了相變的想法，因為他也一直在思考適應的問題，因此不免質疑：最有適應能力的系統是否也是能得出最佳演算結果，也就是位在這奇怪邊緣的系統？ 所以派卡德做了個簡單的模擬。開始的時候是很多的細胞自動機規則，他要求每個規則都作一些計算。然後，他採用賀南式的遺傳演算法，根據這些規則表現的好壞來進行演化。結果，他發現最後找到的規則正好都密布於邊緣地帶。一九八八年，派卡德把結果發表成一篇叫朝向混沌邊緣的適應（Adaptation to the Edge of Chaos）的論文，這是混沌邊緣的名稱首次出現在學術期刊（當時蘭頓仍然稱之為混沌初始）。 考夫曼聽到了十分震驚。我曾經想到過在相變中可以產生複雜的計算，但是我從來沒有想過天擇也能產生同樣的結果，我壓根兒沒有這樣的想法。 一旦有了這層領悟，自我組織和天擇的老問題也就豁然開朗：過去二十五年來，考夫曼一直聲稱自我