Chapter 36 Darwin's principle of relativity

Darwin's principle of relativity Holland had the happiest time of his life in Santa Fe.He enjoys nothing more than sitting with a group of bright minds and discussing various issues.But more importantly, those conversations prompted him to make an important directional change in his research.It was these conversations, plus his not knowing how to say no to Marie Gell-Mann, that brought him under. "Marui deserves to be a good presser," Holland laughed.In the late summer of 1988, Gell-Mann called Michigan to find him, he said. "John, you've been working on genetic algorithms. Now we need an example to refute the creationists," Gell-Mann said.

The fight against "creation science" has indeed been one of many things Gell-Mann has been passionate about.He got involved in this a few years ago.At that time, the Louisiana Supreme Court held a hearing to debate whether it should be a law to teach the science of creation in schools like Darwin's theory of evolution.Gell-Mann persuaded nearly the entire American scientific community of what he called "Swedish laureates," or Nobel laureates, to sign a brief to aid the court's interpretation calling for the law to be repealed.The state Supreme Court did strike down the law in a seven-to-two vote.But after reading the newspaper reports, Gell-Mann realized that the problem was far more than just a few religious zealots. "People write and say: 'Of course, I'm not an extremist, I don't believe in the nonsense of creation science at all. But there seems to be something wrong with the science of so-called evolution taught in our schools. Of course the world cannot be Born of blind chance.' They're not creationists, but they can't believe that mere chance and choice can create everything we see."

So, he told Holland, his idea was to take out a series of computer programs, or even computer games, and show these people how it all happened.These computer gadgets can reveal to people how great evolution and changes can be produced by the pressure of opportunity and selection in the growth and reproduction of generations.You just arrange the primordial conditions—basically a planet—and things mature.In fact, Gell-Mann says, he's considering organizing a workshop at the Institute devoted to just such computer games.Can Holland do something about it? Well, no, Holland was actually reluctant to help.Of course he appreciated Gell-Mann's ideas and plans, but his research work was already full, including designing a classifier system that could be applied to Arthur's economic model.From this point on, Gell-Mann's evolutionary simulations would distract him.Besides, he'd already done the Genetic Algorithm, and he couldn't see how doing it again in another form would do anything new.So Holland flatly rejected Gell-Mann's request.

Well then, Gell-Mann said.But why not think about it.Not long after, Gell-Mann called him again: John, this matter is really important.He asked Holland if he could change his mind. Holland refused again, but he had seen that it was not going to be easy.So after a long talk with Gell-Mann, he dropped all resistance. "Okay," he said to Gell-Mann, "I'll try." In fact, Holland admitted, he was at the end of his battle anyway.Between the two phone calls from Gell-Mann, he was figuring out how to get Gell-Mann to accept his rejection, and he had begun to think more and more about what he would do if he could only agree. Where to start with this.And he began to realize that doing it might open up many opportunities.Evolution is of course much more than random change and natural selection.Evolution is simultaneously realization and self-organization.But it is at this point that, despite the best efforts of Kaufman, Langton, and many others, no one has yet come to a comprehensive understanding.Perhaps this is an opportunity to further raise awareness."I started to think about it seriously, and I realized that I could make a model that would satisfy Murray, but at the same time I could do something interesting with it from a research standpoint," Holland said.

This model is actually a reproduction of the models he made back in the 1970s.At the time he was working on genetic algorithms and writing the book Adaptation.At that time he was invited to give a speech at an academic conference in Finland.For fun, he decided to find an entirely different topic: the origin of life. He said that he called this academic report "spontaneous emergence", and his paper is also based on this idea.In retrospect, his research angle at the time was quite close to the autocatalytic model.At that time, around the same time, Kaufmann, Manfred, and Otto Rost were also building models of autocatalysis, but they were doing it alone. "My thesis is not such a computer model, but a formal model that can be used to do math. I try to show that it is possible to design an autocatalytic system that can produce simple self-replicating entities that can be calculated faster than Much faster than usual."

The usual calculations that creationists still like to cite were developed by scientists in the 1950s.The point of contention is that self-replicating life forms could not have arisen from random chemical reactions in the primordial soup, since that would have required time well in excess of the age of the universe.It's like expecting the monkeys in the basement of the British Museum to evolve from typing away to create the complete works of Shakespeare: they will, but it will take a very long time. But Holland is not as overwhelmed by this notion as Kaufman and others are.Random chemical reactions are nice, he thought, but what about chemical catalysts?Does this have to be non-random?So Holland assumed in his mathematical model that the primordial soup of molecules, that is, arbitrary symbols connected by strings of different lengths, was acted upon by free-floating "enzymes". "Enzyme" is the operating body that acts on the string. "They're like copies, a very primitive operator that attaches to any string and copies it," Holland said. "Actually, I was able to prove a theorem. If a system has these running bodies floating in it, if arbitrary strings of various lengths, that is, building bricks, can be combined with each other, then the system will produce self-replicating entities. would be generated much faster than purely random behavior."

Holland called that paper on spontaneous emergence "a single point of view," and he never wrote it before or since.But the problem of emergence and self-organization has been circling in his mind.In fact, he had had a long back-and-forth discussion with Farmer, Langton, Kaufman, and others about it at Los Alamos a year earlier. "So, the high pressure at Marui made me think that maybe the time is ripe to do some serious research on this. Maybe now I'll build a real computer model of these ideas," he said. After years of working with classifier systems on and off, building a computer model seemed like a piece of cake for him.Since in the original paper the free-floating running body had regular effects—"if you encounter such and such a string, do such and such to it"—the thing to do now is to Putting them into the program like this makes the model as close to a classifier system as possible.But as soon as Holland began to think along these lines, he realized that his system of classifiers had a serious philosophical flaw.On that paper on spontaneous emergence, the spontaneity is real, and the emergence is entirely internal, but the classifier system, despite its ability to learn and discover groups of emergent rules, still exists in a pinch. Appeared, thereby reversing the external factors of the situation.The system still relies on the programmer's shadow manipulation. "The classifier system is rewarded only because I set the rules for winning and losing," Holland said.

This has always troubled him.Questions of religion aside, the real world works well and does not require a cosmic arbiter.Ecosystems, economic systems, and societies all operate on Darwinian principles of relativity.Each is constantly and constantly adapting to each other.Because of this, it is impossible to weigh an agent and say, "It has a robustness of 1.375." Biologists resolve that, whatever "robust" means, robustness has not been possible since Darwin's time. is a single and definite number.It's like comparing a gymnast to a sumo wrestler, the question is meaningless since there is no common measure between the two.A particular organism's ability to survive and reproduce depends on the living space it finds itself in.What other organisms are around it, what resources it has access to, and even its past history are also relevant.

"This change in perspective is extremely important," Holland says. Indeed, evolutionary biologists have a term for its importance: ecosystem organisms don't just evolve, they co-evolve.Organisms do not change by climbing some abstract commanding heights of fitness, as biologists of Fish's generation believed (the view of organisms maximizing fitness in classical population The point about function-maximizing actors appears to be identical).In reality, organisms are often cyclical and chasing each other in an infinitely complex dance of co-evolution. On the surface, says Holland, coevolution looks like chaos.At the Institute, Kaufman likes to liken this to climbing the high ground of fitness in a rubber scene.Every time you climb a step, the entire rubber scene will change its shape.Yet the outcome of such a co-evolutionary dance is anything but chaotic.In nature, flowers are fertilized and reproduced with the help of bees, and bees rely on nectar to maintain life.The cheetah chases and devours the gazelle, and the gazelle escapes from the cheetah's claws.Coevolution has produced countless organisms that are perfectly adapted to each other and to the environment in which they live.In human society, the dance of co-evolution has produced an equally perfect web of economic and political interdependence, such as alliances and competition, and supply and demand.This is what drives Arthur's glass house economy.In Arthur's conception, you can observe artificial economic actors adapting to each other.This is the dynamic underlying Arthur and Kaufman's analysis of autocatalytic technological change, and it is also the dynamic underlying the relations of states in a world without central authority.

Indeed, Holland says, coevolution is a powerful force for mutation and self-organization in any complex adaptive system.From this he understood that if he really wanted to understand these phenomena at their deepest level, he would have to exclude external rewards from his system.But unfortunately, he also knows that the assumption of rewards from the outside world is closely related to the market metaphor of the classifier system.In the classifier system established by Holland, each classifier rule is an extremely small and simple actor, and they participate in the internal economic system together. In this internal economic system, the common currency is each actor’s "Strength", and the only source of wealth is the return from the end user, i.e. from the programmer.There is simply no way around this problem without radically changing the architecture of the classifier system.

So, what Holland had to do was to completely change the structure of the classifier system.What he needed, he thought, was a different, more radical metaphor for interaction: combat.He designed an ecosystem, in this highly simplified biological community, digital organisms wander in the digital environment, looking for resources for survival and reproduction, these resources are digital water, grass, shells, strawberries etc.When these creatures come together, they will of course try to use each other as resources.Holland said: "I compare this to my daughter Manga's 'Mail Monster' game. In this game, you have many possible moves to attack and defend, and how you use these possible moves determines your strength. Win or lose in battles with other monsters." To be more specific, the environment represented by the ecosystem is a vast plain with "springs" all over it, from which various resources symbolized by a, b, c, and d spew out.Individual organisms roam the environment at will, devouring resources along the way like sheep grazing calmly and gently into their own internal repositories.But as soon as two organisms meet, they immediately change from the sheep state to the wolf state and attack each other. In the ensuing battle, the outcome of the battle depends on the pair of "chromosomes" of each organism, which are just two sequences strung together by a set of resource symbols, such as aabc and bbcd. "If you are one of these organisms, you match each sequence of your 'aggressive' chromosomes with the other's second sequence of 'defensive' chromosomes, and if they match each other, you get High score. This situation is very similar to the immune system: if your attack matches the opponent's defense, then you open the gap. And the opponent responds to you with interactive actions, that is, his attack and your defense Match. The interaction is extremely simple. It depends on whether your offensive and defensive capabilities are better than your opponent's." If the answer is yes, he says, then you're in for a treat: all the data symbols in your opponent's repository and its two chromosomal sequences are yours.Moreover, if eating your former opponent means that you currently have enough data symbols in your arsenal to reproduce your own chromosomes, then you can reproduce yourself by creating an entirely new organism, one of which may have a, Two variants.But if that's not the case, then you go back to grazing. To put it bluntly, this ecosystem is not exactly what Gell-Mann wants, users will feel that there is nothing interesting, and there is no novel image.But Holland wouldn't bother with that.He would punch in a string of passwords and symbols to boot the system, and watch it produce more passwords, see lines of alphanumeric gibberish cascade across the screen (by then his computer had been upgraded to an Apple Type II machine).This ecosystem is a Dutch-German game.In this game he finally ruled out obvious alien rewards."It's a closed circle. You're really coming back to the notion that 'If I can't find enough resources to replicate myself, I can't survive,'" he said, capturing what he sees as the essence of biological competition. thing.Now he can use the system as an intellectual playground, a place to explore and understand how co-evolution really works. "I have included many phenomena in ecosystems in my research program. I want to show that even with this extremely simple structure, every phenomenon can manifest itself in one way or another." The ecological phenomenon that Holland was most interested in studying was what the British biologist Richard Dakins called the arms race of evolution.That's why plants have evolved harder and harder surfaces to produce toxic chemical repellents to resist pest attacks.And pests have evolved harder jaws and more sophisticated chemical resistance mechanisms to combat them.In this regard, the Red Queen Hypothesis is another famous example.This assumption comes from a book.The character in the book, the Red Queen, tells Alice that she must run as fast as she can to stay where she is.The evolutionary arms race appears to be the main driving force behind the increasing complexity and specialization of nature, just as the real arms race of the Cold War was the main driving force behind the creation of ever more complex and specialized weapons. In the fall of 1988, Holland certainly couldn't do much research on the evolutionary arms race.Back then his ecosystem was just a written design.But within a year or so, the system was working very successfully. "If I start with very simple organisms and use only one letter for the organism's offensive chromosomes and another letter for the defensive chromosomes, then I see organisms with multi-letter chromosomes (these organisms can By mutating to lengthen their chromosomes). They are co-evolving. If one organism makes the offense a little stronger, the other increases its defense. So they get more and more complex. Sometimes they split, which is Created a new species." "That's when I saw that such a simple mechanism could generate arms races and speciation, and my interest grew," Holland said. In particular, he wanted to understand a deep paradox in evolution.In fact, this relentless competition has led not only to an evolutionary arms race, but also to symbiosis and other forms of cooperation.Indeed, it is not surprising that Holland regards various forms of collaboration as his research interest.This is the fundamental problem of biological evolution, not to mention the fundamental problem of economics, political science, and all human phenomena.Why on earth do organisms cooperate with each other in this competitive world?Why would they open the door to "allies" who could easily turn against each other? The famous "Prisoner's Dilemma" brilliantly reveals the nature of this problem. The Prisoner's Dilemma was originally developed from game theory by a group of mathematicians.This story is about: two prisoners were locked in solitary cells separately.The police are interrogating a case they committed together.Both prisoners can make their own choice: he either confesses to his partner (i.e. betrays him) or remains silent (i.e. cooperates with his partner instead of the police).Now, both prisoners know that if they both remain silent, they will both be released, and the police cannot convict them as long as they deny their confession.But the police are also fully aware of this.So they gave the two prisoners a little incentive: if one of them betrayed and denounced his partner, the denounced prisoner would be acquitted and some reward would be given.His accomplice would be sentenced for the heaviest crime, and, to humiliate him, he would be fined as a reward for the informer.Of course, if the two prisoners had betrayed each other, both would have been sentenced for the heaviest crime, and neither would have been rewarded. So, what are the two prisoners to do?Mutual cooperation or mutual betrayal?On the surface, they should cooperate with each other and keep silent, because then they both get the best result: freedom.But they had to think carefully. Prisoner A is no fool, and he realizes right away that he simply cannot trust his partner not to give the police evidence against him and walk away with a handsome reward, leaving him behind bars alone.The allure of this idea is too great.But he also realized that his accomplices were not fools and thought of him that way.So Prisoner A concludes that the only rational option is to betray his partner and tell the police everything, because if his partner is stupid enough to keep silent, then he will be the lucky one to get out with a reward.And if his accomplice also confessed to the police according to this logic, then Prisoner A would have to serve his sentence anyway, at least he would not have to pay a fine on top of that.So as a result, these two prisoners got the worst possible retribution according to desperate logic: jail time. Of course, in the real world, trust and cooperation rarely reach such a dilemma.Negotiations, personal relationships, enforced contracts, and many other factors shape the parties' decisions.But the Prisoner's Dilemma does capture the frustratingly real side of distrust and the need to guard each other against betrayal.Let's look at the two superpowers during the Cold War that locked themselves in a forty-year arms race that turned out to be of no benefit to either side.Then there's the seemingly endless Arab-Israeli standoff, and the perpetual tendency of countries to protect their trade.In nature, a creature that trusts others too much may be eaten.So the question arises again: why do all organisms dare to cooperate with each other? Much of the answer came from a computer competition organized by Robert Axelrod, a member of Holland's Bucky group at the University of Michigan.Axelrod was a political scientist with a longstanding interest in the problem of cooperation.His idea for organizing the contest was simple: Anyone who wanted to enter the computer competition would play the role of one of the prisoners, and the program would be paired into different combinations, and the participants would start playing "the prisoner's dilemma." ", where everyone has to choose between cooperation and defection.But here's the difference: They didn't just play the game once, they played it 200 times over and over.This is what game theorists call the "repeated prisoner's dilemma," and it more closely mirrors some sort of frequent, long-term relationship.Moreover, this repeated game allows the program to refer to the previous choices of the opponent program when making the decision to cooperate or defect.If the two programs have only played one round, defection is clearly the only rational choice.But if the two programs have played each other many times, each has built its history and reputation in this regard.How the opposing program will behave, however, is extremely difficult to determine.Indeed, that's one of the things Axelrod hopes to learn from the competition.Can a program always cooperate regardless of what the opponent does?Or can it always act on treachery?Should it respond to rivals' moves with more sophisticated moves?If yes, what would be the action? In fact, the fourteen programs submitted after the first round of the competition included strategies of all kinds and complexity.But much to the amazement of Axelrod and others, laurels belonged to the simplest strategy: tit for tat (TIT FORTAT).This is the strategy submitted by Anatol Rapoport, a psychologist at the University of Toronto.The tit-for-tat strategy starts with cooperation, but from then on it adopts the strategy of treating others in the same way.That is, the tit-for-tat strategy implements the principle of carrots and sticks.It is "good" in the sense that it never betrays the other first.He is "forgiving" in the sense that it rewards the opponent's previous cooperation in the next round.But it is "tough" again in the sense that it will take the action of defection to punish the opponent for the previous defection.Moreover, its strategy is extremely simple, and the opponent program can know its intention at a glance. In this sense, it is "simple and clear". Of course, since only a small number of programs participated in the competition, it is also possible that the tit-for-tat victory was just a fluke, but maybe not.Of the fourteen programs turned in, eight were "well-intentioned," never betraying first.And these well-intentioned programs easily won the six non-well-intentioned programs.In order to determine a result, Axelrod held a second round of competition, in which people were specially invited to win the crown from the strategy of tit for tat.This time, sixty-two programs participated in the competition, and the tit-for-tat strategy won again.The conclusions are indisputable.The good guy, or more accurately, the well-meaning, forgiving, tough, and plain guy really always wins. Holland and the rest of Butch's group were of course fascinated by all this. "I've always been deeply troubled by the 'prisoner's dilemma,'" Holland said. "That's one of the things I don't like. So I'm very happy to see the result of this competition. It's really encouraging. The game is amazing." The profound implications of the triumph of tit for tat for biological evolution and human affairs are clear.Axelrod pointed out in his book "The Evolution of Cooperation" published in 1984 that the tit-for-tat strategy can lead to cooperation in all areas of society, including cooperation in the most hopeless environments.His favorite example is the principle of "live and let others live" that arose spontaneously in World War I.The troops in the trenches at the front restrained themselves from shooting and killing people, as long as the other side did the same.Armies in no man's land have no way of contacting local armies, and they certainly won't be friends.But what makes this principle work is that both militaries have been stuck in limbo for months, giving them a chance to adapt. In one chapter of the book, Axelrod also argues that tit-for-tat interactions allow nature to collaborate even without intelligence.He wrote the chapter with his co-author, William Hamilton, a biologist in Butch's group.In this regard they give examples such as lichens: the fungus extracts nutrients from the rocks in the ground, providing food for algae, which in turn provide photosynthesis for the fungus; , which in turn protect the tree; the flowers of the fig tree are food for the wasps, who in turn pollinate the fig tree, scattering the seed. More broadly, co-evolution would allow tit-for-tat cooperative styles to prevail in a world of treachery and vice.Suppose, says Axelrod, that a small number of tit-for-tat individuals arose in the world by mutation.Then, as soon as these individuals can meet each other enough to form a stake in future encounters, they will begin to form small cooperative relationships.Once that happens, they can outperform the knife-in-the-back types around them.In this way, the number of people participating in cooperation will increase.Soon, tit-for-tat cooperation will finally prevail.Once this mechanism is established, individuals who cooperate with each other can survive.If the less cooperative types try to violate and exploit their goodwill, the hard-line side of the tit-for-tat policy will punish them so badly that they won't be able to spread their influence.Axelrod wrote: "Then the cogs of social evolution are set in motion." Shortly after the book was published, Axelrod and Holland's graduate student Stephanie Forrest (Stephanie Forrest) used a computer simulation to simulate this cooperation.The question is whether a co-evolving population can find tit-for-tat strategies through genetic algorithms.It turned out that the answer was yes: in the workings of computers, tit-for-tat strategies, or something similar to them, emerged and quickly became popular among the group."When that happens, we all throw our hands up and say three cheers," Holland said. It is this tit-for-tat mechanism at the origin of collaboration that Holland refers to when he says that people at the Institute should view the social sciences as if they were “fronts.”He said that when he was designing the development ecosystem, the whole thing about collaboration was on his mind.The mechanism of cooperation was of course impossible in the first version of the program, because he built in the assumption that individual organisms would always fight each other.But in the new version, he seeks to perfect every aspect of an organism's evolution, including its possibilities for cooperation.Indeed, he wanted to design ecosystems as some sort of co-evolving, "holistic" model. "At the Institute, in addition to the ecosystem, we're creating three other models, a model of the stock market, a model of the immune system, and a trade model built by Stanford economist Tom Sargent. I found that these systems share very similar characteristics. They all have 'trade', goods that are exchanged in various ways, and 'resource conversion' mechanisms, such as enzymes or various production processes to achieve resource conversion. switch. And they all have 'mating selection' mechanisms that are the source of technological invention. So I started to create a whole model of co-evolution from that. I remember Stefania Forrester, John Miller and I sitting Down, trying to figure out how to simulate all these characteristics in the smallest device in the ecosystem? Our conclusion is that it can be done without changing the basic model, just adding content to the offensive and defensive chromosomes To this point. I provide additional discriminators that can be defined by chromosomes, thereby increasing the possibility of trade. These discriminators are similar to trademarks, or molecular tags on the surface of cells. At the same time I must add to this ecosystem. Something like a rule, I'm doing this for the first time. The rule is: 'If someone else shows such an identifying tag, then I trade with him, not fight.' And that creates the evolution of cooperation , and unconventional phenomena such as lying and imitation. Based on this assumption, I sketched out the idea of ​​how to make a Sargentian model, and then began to conceive how to design the ecosystem to look like an immune system from the other direction. A model of a system. That’s where the current ecosystem model comes from.” This unified version of the ecosystem has been very successful, Holland said.This system can demonstrate cooperative evolution, as well as the spontaneously formed relationship between predators and prey.This success inspired me to start looking into designing a more advanced version of the ecosystem. "The latest version I'm currently working on demonstrates the evolution of multicellular organisms. So now we're not just talking about trade, I hope we can also talk about the emergence of individuals and organizations. Each actor tries to increase its reproductive rate, but Always constrained by the continuation of the overall organization, there are many things worth studying. Cancer is a good example of this, let alone the situation of the automation industry in the United States!" Holland says it's still early days for such models to be practical, but he's convinced that some good computer simulations in this area may do more for the world than any other research project at Santa Fe. "If we do a good job, people who are not scientists, like officials in Washington, can build these kinds of models without knowing the details of how they work, which will help them grasp the real implications of various policy choices. In essence, such a model would act like a policy flight simulator, enabling politicians to simulate an economic forced landing without putting 25 million people on the plane, he said.The models don't even have to be complex, as long as they give a realistic sense of how things are going and how the most important variables interact. Holland admits that when he talks about the flight sim concept in Washington, the audience doesn't pay attention.Most politicians in power are too busy dodging oncoming blows to worry about policy for the next flight.On the other hand, he's clearly not the only one thinking about strategy from a simulation perspective. In 1989, Maxes Corporation of Orinda, California, launched a game called SimCity.The game puts the player in the role of a mayor trying to make his or her city prosper in the face of crime, pollution, traffic jams, tax resistance, and more.The game shot to the top of the bestseller charts in no time, and has earned the trust of real city planners.They said that while the "Sim City" game was tedious about specific details, it got the right feel.Holland also bought the game, of course, and loves it. "'SimCity' is the best example of a flight sim concept that I know of," he said.The Santa Fe Institute is in serious talks with Maxess about revamping the SimCity interface so that it can be used in many of Santa Fe's simulations.Holland is now working with Maxess to develop a user-friendly version of the ecosystem that anyone can do computer experiments on.