Chapter 27 5.5 Collaboration without friendship or vision

The trouble with Gaia, for most skeptics, is seeing a nonliving planet as a "smart" machine.Our attempts to engineer lifeless computers into artificial learning machines have met with setbacks.So the prospect of messy artificial learning on a planetary scale seems absurd. But in fact, we overestimate learning and regard it as a difficult task. This has something to do with our chauvinistic plot-taking learning as a unique human ability.In this book, I want to make a strong point that evolution itself is a kind of learning.Hence, wherever there is evolution, even artificial evolution, there will be learning.

Bringing learned behavior down from the altar is one of the most exciting intellectual frontiers we are crossing.In a virtual cyclotron, learning is being smashed into elementary particles.Scientists are cataloging the building blocks of adaptation, induction, intelligence, evolution, co-evolution, etc., into a periodic table of elements for life.The particles needed for learning are hidden in all dull media, waiting to be assembled (and often self-assembled) into a surging and dynamic thing. Coevolution is learning in many forms.Stewart Brand writes in Coevolution Quarterly: "Yes, an ecosystem is a complete system, and coevolution is a complete system in the sense of time. It is normally forward, systematic Coevolutionary self-education, and nourishment from constant correction of mistakes. If ecosystems are maintained, co-evolution is learning.”

The co-evolutionary behavior of organisms might be described in a better term—co-learning, or co-teaching, since the co-evolving parties teach each other as they learn from each other. (We don't have the right words for teaching and being taught at the same time, but our schools would improve if teaching and learning were achieved.) Giving and receiving in a coevolutionary relationship—teaching and being taught at the same time—has led many scientists to think about playing games.Simple children's games like "Which hand has the penny?" have recursive logic like "Chameleon on the mirror".The person who hides the steel 镚'er enters into such an endless process: "I just hid the steel 镚'er in my right hand, so people who guess now will think it is in my left hand, so I will move it to my right hand. But She also knew that I knew what she would think, so I still kept it in my left hand."

Since the thinking process of the guesser is also the same, the two parties constitute a game of mutual prediction of the other party's intentions. The puzzle "Which hand has the hammer" is related to the puzzle "What color is the chameleon on the mirror".The infinite complexity that can be derived from such simple rules intrigued John von Neumann.In the early 1940s, the mathematician developed programmable logic for computers and, together with Wiener and Bateson, opened up the new field of cybernetics. Von Neumann invented the mathematical theory related to games.He defines a game as a conflict of interest in which each party attempts to predict the actions of the other party and take a series of steps to resolve the conflict. In 1944, he co-authored a book, Game Theory and Economic Behavior, with economist Oscar Morgenstern.He sensed that the economy is highly co-evolutionary and game-like, and he wanted to explain it in terms of simple game dynamics.For example, the price of an egg depends on the seller and buyer's mutual guesswork—how much should I bid for him, how much does he think I will pay, how much lower should my bid be than I can afford?To von Neumann's amazement, this endless recursion of mutual deceit, mutual deception, imitation, mirroring, and "game" can generally be settled at a definite price, rather than endlessly entangled.Even in the stock market, when there are thousands of agents playing a game of mutual prediction, parties with conflicting interests can quickly agree on a reasonably stable price.

Von Neumann was most interested in seeing if he could figure out the optimal strategy for such interactive games, which at first glance seem almost theoretically unsolvable.So he proposed game theory as an answer.The RAND Corporation, based in Santa Monica, California, is a U.S. government-funded think tank.Researchers there developed von Neumann's work, culminating in a list of four basic variants of the mutual guessing game.Each variant has a different reward structure for wins, losses or draws.These four simple games are collectively referred to as "social dilemmas" in the technical literature, but can be seen as four building blocks for constructing complex co-evolutionary games.The four basic variants are: the chicken game, the stag hunt, deadlock, and the prisoner's dilemma.

"Grass Chicken Game" is a game for reckless teenagers.Both cars raced towards the edge of the cliff; the driver who fell last was the winner. "Deer hunting" is a problem faced by a group of hunters. They must cooperate to kill the deer. If no one cooperates, it is better to go off and chase the rabbits individually.Are they betting on cooperation (high payoff) or defection (low, but sure payoff)? Deadlock is a pretty boring game, and betraying each other pays off the most.The last one, the Prisoner's Dilemma, was the most instructive and became the test model for more than two hundred social psychology experiments in the late 1960s.

The "Prisoner's Dilemma" was devised by Merrill Flood of the RAND Corporation in 1950.In the game, two separate prisoners must independently decide whether to deny or confess a crime.If both plead guilty, then both will be punished.If both men deny it, they will both be acquitted.But if only one pleads guilty, then he is rewarded and the other is punished.Cooperation pays off, but so does defection if the strategy works.What should you do? If you are only playing once, betraying your opponent is the most logical choice.But when two "prisoners" play over and over again, learning from each other -- aka "repeated prisoner's dilemma" -- the game's progression changes.You cannot ignore the presence of the opposing player; he must be taken seriously, whether as a mandatory adversary or as an accomplice.This closely linked common destiny is very similar to the co-evolutionary relationship between political enemies, business rivals, or ecological symbionts.As the study of this simple game progresses further, the question becomes: what strategy should be adopted in the face of the "repeated prisoner's dilemma" in order to achieve high scores in the long run?Also, what strategies are more likely to be successful when playing against ruthless or friendly types of players?

In 1980, Robert Axelrod, a professor of political science at the University of Michigan, organized a tournament to collect 14 different strategies for the "Prisoner's Dilemma" in a round-robin format to see which strategy wins.In the end, one of the simplest strategies won out, called Tit for Tat, devised by the psychologist Anatole Rapobert. "Tit for Tat" is a reciprocating strategy. Cooperation is rewarded for cooperation, and betrayal is rewarded for defection, often resulting in a cycle of cooperation.Axelrod found that repeated play can produce a "future shadow" effect that one-shot games do not have. Cooperation is a reasonable choice.The flashes of cooperation lead Axelrod to ponder: "What conditions are needed in an egoistic world without central authority for cooperative behavior to emerge?"

In 1651, Thomas Hobbes declared that cooperation can only arise with the help of a well-intentioned centralization.This traditional political reasoning has been revered for centuries.Without top-down governance, Hobbes asserted, there can only be group selfishness.Regardless of the economic system, there must be strong forces to enforce political altruism.However, the gradual establishment of Western democracies after American Independence and the French Revolution showed that well-informed societies can develop mechanisms of cooperation without the strong intervention of centralized powers.Self-interest can also breed cooperation.In post-industrial economies, spontaneous cooperation is common.The widespread adoption of industry standards (both quality and protocol, such as 110 volts, and ASCII), and the rise of the Internet, the world's largest form of anarchy, have led to a greater focus on breeding co-evolution necessary conditions for cooperation.

This collaboration is not New Age spiritualism.Rather, it is, as Axelrod puts it, a “collaboration without friendship and vision”—a cold rule of nature that applies at many levels and gives rise to self-organizing structures.Whether you like it or not, you have to cooperate more or less. Games like "Prisoner's Dilemma" can be played by any adaptive individual, not just humans.Bacteria, armadillos, or semiconductor devices in computers can all make trade-offs between the safe gains in the immediate future and the high risks and high rewards in the future according to various reward mechanisms.When playing this game with the same partners for an extended period of time, you are both gaming and engaging in some type of co-evolution.

Every complex adaptive organization faces fundamental tradeoffs.Creatures must choose between perfecting existing skills and traits (strengthening their legs to run faster) and trying out new traits (wings).It can't do everything at the same time.This daily challenge is the trade-off between exploitation and utilization.Axelrod made an analogy with a hospital: "Generally you can imagine that trying a new drug has a lower return on efficacy than trying to find out what is already available. But if you give all patients the best drug at the moment You will never be able to verify the efficacy of new drugs. From the perspective of individual patients, it is better not to try new drugs. But from the perspective of social aggregates, it is necessary to do experiments.” Development (future benefits) and utilization (current stable What should be the ratio of winning chips) is a game that the hospital has to play.Living organisms make similar trade-offs in deciding how much they should mutate and innovate in order to keep up with changing environments.When a large number of creatures are making similar trade-offs and interacting with each other, a co-evolutionary game is formed. Axelrod's 14-player "Prisoner's Dilemma" round robin tournament is played on a computer. In 1987, Axelrod expanded the computer game by setting a system.In the system, a small group of programmed players execute randomly generated "prisoner's dilemma" strategies.Each random strategy is scored after playing a round against all other running strategies, and the strategy with the highest score has the highest replication rate in the next generation, so the most successful strategy can be multiplied and propagated.Many strategies win by "preying" on other strategies, and thus can only thrive if the prey survive.This leads to a mechanism for the cyclical fluctuations in populations that abound in wildernesses in nature, and how fox and rabbit populations ebb and flow in year-to-year cycles of co-evolution.When the number of rabbits increases, the number of foxes increases; when the number of foxes increases, the number of rabbits dies.But without the rabbit, the fox would starve to death.The fewer foxes, the more rabbits.If there are more rabbits, there will be more foxes, and so on. In 1990, Christian Lindgrey, who worked at the Niels Bohr Institute in Copenhagen, expanded the number of players in this co-evolution experiment to one thousand, and at the same time introduced random interference, and made this artificial co-evolution process multiply to thirty thousand generations after.Lindgrey found that groups of dumb individuals participating in the "prisoner's dilemma" game not only reproduced the ecological fluctuations in fox and rabbit populations, but also produced many other natural phenomena such as parasitism, spontaneously emerging symbiotic symbiosis, and The long-term stable coexistence relationship between species is like a whole ecosystem.Lindgrey's work has some biologists excited because cycle after cycle emerges in his long game of bouts.The duration of each cycle is very long; and within a cycle, the mixture of "species" with different strategies is maintained in a very stable state.These flourishes, however, were interrupted by sudden, short-lived episodes of instability in which old species died out and new ones took root.A new stability was quickly reached among species with new strategies, and continued to develop for thousands of generations.This pattern fits with a common pattern of evolution found in early fossils, known in evolutionary circles as punctuated equilibrium, or "punkeek" for short. These experiments yielded a remarkable result that has attracted the attention of anyone wishing to harness the forces of co-evolution.Here's another law of the gods: In a world adorned with "chameleons on mirrors" of nesting garlands, whatever ingenious strategy you devise or evolve, if you absolutely obey it and serve it, From an evolutionary point of view, this strategy cannot compete with other competitive strategies.That is to say, how to make the rules work for you in the protracted war is a competitive strategy.On the other hand, introducing a small amount of random factors (such as errors, defects) can instead create long-term stability in the world of co-evolution, because in this way, some strategies cannot be easily "copycat", so that they can survive in a relatively long period of time. Dominance over the period.Without disturbances—that is, unexpected or perverse choices—there would not be enough periods of stability to sustain the system, and escalating evolution would lose its chance.Mistakes can keep the coevolutionary relationship from sinking into a self-sinking vortex because of too tight glue, so as to keep the coevolutionary system moving forward.Salute your mistakes. These coevolutionary games played on computers offer additional lessons.The distinction between zero-sum and non-zero-sum games is one of the few game-theoretic ideas that has permeated popular culture.Chess, elections, racetracks, and poker are zero-sum games: the winner takes what the loser loses.The wilderness of nature, the economy, consciousness, and the web are non-zero-sum games: the existence of a bear does not mean that the wolverine will fail.The interlocking and interconnected nature of conflicts in coevolution means that overall gains can benefit (and sometimes hurt) all members.Axelrod told me, “One of the earliest and most important insights from game theory is that the strategic content of non-zero-sum games is very different from that of zero-sum games. Any harm done to others in zero-sum games is Good for you. In a non-zero-sum game, you can both prosper and lose together. I think people often see the world in terms of a zero-sum game when they shouldn't. They often say: 'I'm better than other people Do well, so I should prosper.’ And in a non-zero-sum game, even if you do better than someone else, you can be just as poor as him.” Axelrod noticed that, as a winner, the Tit for Tat strategy never considered exploiting the opponent's strategy—it just fought back in the same way.In a one-on-one duel, this strategy cannot outperform any other strategy; but in a non-zero-sum game, it can achieve the highest cumulative score against many strategies, thus winning the championship.As Axelrod pointed out to the creator of "Prisoner's Dilemma" William Poundstone: "The idea is incredible. How can you win the championship without defeating everyone in chess?" But in common Evolving - changing in response to oneself - winning without beating others.Astute chief executives in the corporate world are now acknowledging that, in the age of networks and alliances, companies can make a lot of money without hurting others.This is the so-called win-win situation. Win-win is the story of life in the co-evolutionary mode. Sitting in an office full of books, Robert Axelrod is still immersed in the understanding and thinking of co-evolution.Then he added: "Hopefully my work on cooperative evolution will help avoid conflict in the world. Have you seen the award from the National Academy of Sciences," he said, pointing to a plaque on the wall, "They think it helps Avoid nuclear war." Although von Neumann was a key figure in the development of the atomic bomb, he did not explicitly apply his theories to the political game of the nuclear arms race.After von Neumann's death in 1957, military strategy think tanks began to use his game theory to analyze the Cold War, in which two rival superpowers smacked of "forced cooperation" in a co-evolutionary relationship.Gorbachev had a fundamental coevolutionary insight.Alcedro said, "He saw that the Soviet Union would be safer by reducing the number of tanks, not by increasing them. He unilaterally eliminated 10,000 tanks, making it harder for the United States and Europe to maintain a large military budget. This started the full-scale process of ending the Cold War." For the "false gods," the most useful lesson to be learned from coevolution is that control and secrecy are counterproductive in a coevolutionary world.You can't control it, and it's better to be open than secretive. "In a zero-sum game you always want to hide your strategy," Axelrod said. "But in a non-zero-sum game, you might make the strategy public, and then the other players have to adapt to it." Gorbachev's strategy worked because he implemented it publicly; if Simply unilaterally cutting arms in secret will accomplish nothing. Chameleon on Mirror is a completely open system.Whether it's lizard or glass, there are no secrets.Gaia's large closed loop is constantly looping because all the smaller loops within it communicate with each other in constant co-evolutionary communication.We know from the collapse of the Soviet command economy that public information keeps an economy stable and growing. Co-evolution can be seen as two parties caught in a web of mutual teaching.Co-evolutionary relationships, from parasitism to alliances, are inherently informational.A steady exchange of information welds them into a single system.At the same time, the exchange of information—whether it’s insults, or favors, or just general news—opens up ground-breaking ground for cooperation, self-organization, and win-win outcomes. In our nascent network age, frequent communication is creating increasingly sophisticated artificial worlds, poised for the emergence of co-evolution, spontaneous self-organization, and win-win cooperation.In this era, the opener wins, the central controller loses, and stability is a permanent state of decline guaranteed by continuous errors.