Chapter 8 Chapter 7 Prospects for Superhuman Intelligence

Of course, if my "ego" is just a bundle of intuitions about known numbers and exact dimensions, then let me bundle it up neatly and make the best use of it.But if this covert personality, with its unusual and satisfying desire, and style of resurgence, strife, and permanence, is not a machine with faulty gears and limited maximum horsepower, but a A living thing, infinitely variable, always readjusting according to circumstances, capable of immeasurable achievements, and capable of inducing pathetic meanness, a living thing in a sense whose destiny dictates; if If its free run is not an illusion, and its possibility of supernatural experience is not a lie, we must, then, prevent ourselves from succumbing to the age-old errors of the mechanistic materialists.

We have a spiritual life because our thinking is a dynamic Darwinian process, so that we can create - re-create ourselves every day.That mental life (a disorder described at the outset of this book) may now be imagined as a Darwinian process of a high order, reaching almost to the top of those levels of hierarchical stability, capable of realizing the sense of self.Such depth and multifaceted abilities may arise from the continual replication of brain codes, competition with other brain codes, and generation of new mutations. It was not a computer, at least not in the sense of the trustworthy machine that would faithfully repeat its actions.For most people, it was something new in the realm of mechanism, with no good analogues at all—except for other known Darwinian processes.But you get a sense of what it might be; looking down at the surface of the cortex (actually the flattened surface) sees what appears to be a mosaic—a mosaic quilt, neither of which is at rest.On closer inspection, each piece looks like a repeating wallpaper pattern, but the pattern of each unit is dynamic, a flickering space-time pattern, rather than a traditional static pattern.The boundaries between adjacent pieces of quilt are sometimes steady and sometimes advance like a battle line.Sometimes a cell pattern recedes from an area, and the triangular array no longer synchronizes like points; after a while another cell pattern may quickly occupy the disorganized territory without encountering resistance.

The prevailing winner of this copying competition, that is, the model with the most harmonies to capture the attention of the output pathway, seems like a good candidate for what we call consciousness.Our shifting focus may mean that another type of copying model comes to the fore.Our subconscious may be patterns of other activities that are not dominant at the time.No particular part of the cortex acts as a "conscious center" for long, before another area takes over. A changing mosaic also seems to provide a good candidate for intelligence.Instructions for new movements are contained in formable spatial patterns.Whereas spatio-temporal patterns can change to find new resonances, this mosaic of changes can discover Barlow's new order.Given the ability to make judgments about a motor-programmed brain code based on the resonance of long-term memory and current sensory input, they can then simulate actions in the real world in a crooked manner.There are situations where it is not obvious what to do next, and they have Piagetian characteristics in their handling of these situations.

These tessellations have an uninhibited side to our mental activity, and we can create new levels of complexity, crossword-like composite symbols that embody new levels of meaning (as is the case when writing poetry).Because brain codes can represent not only sensory and motor schemas, but also thoughts, we can even imagine how high-quality metaphors emerge, and when we enter the imaginary realm of a novel, we can imagine Coleridge ( CO-leridge)'s "intentional suspension of unbelief" can occur. The brain code and the Darwinian process are exactly what I mentioned at the beginning of the book when I said that by the end of the book it is possible for the reader to imagine a process that leads to consciousness that works Fast enough to develop a quick intellect, and will be better than guessing.This final chapter discusses the implications of augmenting our brains and building artificial approximations to them.But first, let's diverge a bit and look at the competing nature of interpretation itself.

The golden rule of measuring explanation—to which all science aspires—is abstraction and mathematical description.It is certainly impressive when one can present a series of ongoing inferences as a set of definitions and axioms.Starting with Plato's Republic, both Descartes and Kant sought to understand how the mind works mathematically.We finally seem to be on the threshold of answering some of these questions. But there has long been a challenge to the entire scientific community, and it will once again manifest itself strongly when science attempts to explain the human mind.Mystical and irrational views of truth derive from spiritual revelation, not deduction; these views hold that scientific truth is inferior and quick-witted compared with that attained by pure contemplation.The second challenge comes from dogma.Galileo was in trouble not because of his astronomical studies, but because his constantly challenging and revising scientific method threatened the very truths that religion had used to build its seemingly eternal and internally consistent worldview.There is also the challenge of what the literary critic George Steiner has called a "romantic existential theological argument"—such as Nietzsche's "preference for instinctual wisdom over futile deduction or Blake's "A criticism of Newton's optical analysis of the rainbow.A fourth challenge is that hidden motives are ubiquitous, or that truth is attached to political views.

Fundamentally, these are challenges from outside the scientific tradition; their modern advocates will certainly seize upon the chaos of our everyday science and try to exploit them in the same way that Christian Chianism attacks evolutionary biology itself.These types of explanations have long competed with science, and while there have been a few short-term victories (such as Ramantli's exile), they have mostly ended in long-term defeats.The threads of all these various challenges can be seen today in the movements of the laggards of the Age of Reason. We must therefore strive to articulate our scientific explanations without attempting to establish spurious opposites—like the hypothetical dispute over whether evolution is due to genetic mutation or natural selection.That meaningless debate lasted for decades, and it was not until the 1940s that it was eliminated by "modern synthesis". We must avoid using mathematical concepts to make people dizzy, but to enlighten people.We must get rid of "the proof that the imagination requires," and we must also get rid of self-importance or impatience when we conclude that there is no other possibility than the answer we have found.In particular we must be careful when setting our theories at the proper level of mechanistic explanation.

Thus, the neuron-level descriptions of the modern fashionable images of the brain and the mind are merely shadows of the deeper layers of the cytoskeleton—it is at this deeper level that we must seek the physical basis of the mind! I'm sure some physicist or Ecclesian neuroscientist who studies consciousness will say that there must still be a ghost in this machine that jumps through many intermediate levels of hierarchical stability to guide the mysterious Quantum mechanics down to the microtubules of the neuron's cytoskeleton, where the immaterial mind can exchange information with the biological machinery of the brain.In fact, such theorists usually avoid the word "spiritual" and talk about "nose field theory". I will be happy to adopt Dennis's definition as a compromise on the issue of "mystery"; Phenomena to think about. All physicists who study consciousness do is substitute one mystery for another. So far, nothing in their explanations can be combined to explain anything else. just a few words.

Even if they improve on their combination, any effects produced by synchronizing other microtubules provide only another candidate for the overall nature of our conscious experience.Such an explanation would have to be combined with explanations at other levels to refine its mechanistic details, and would have to compete for space with other possible explanations.So far, Darwinian processes seem to be well placed to explain the successes and failures of some important aspects of consciousness. I think we will continue to see, in those tiresome debates over whether machines can really know everything; The opponent is cornered (or at least exhausted; trying to push the opponent through a corner).Unfortunately, even if all scientists and philosophers admitted that the mind is produced by the brain, the complexity of the topic still leads most people to apply some simpler concept (such as "soul") to this complexity. abstraction.Perhaps feeling like a book reviewer, he asks, "Are digital computers just a simpler version of the human brain, as many theorists believe? If so, the implications are dire."

Is it scary?Personally, I find ignorance to be really scary.In this regard, it can be said to have a long history, from using "obsessed" to "explain" mental illness to wizards' pretending to be ghosts.We desperately need a metaphor more useful than the mysteries of nose mechanics; we need a metaphor that successfully bridges the gap between our perceived mental activity and its underlying neural mechanisms. So far, we actually need two kinds of metaphors: one is a top-down metaphor, which maps thoughts onto populations of neurons, and the other is a bottom-up metaphor, which explains how thoughts are generated by what appears to be are produced by disorganized clusters of neurons.But the neocortex's Darwin machine could do both well -- if it were indeed the creative mechanism in it.

To me, the Darwinian machine theory of the neocortex seems to be at the right level of explanation; it does not descend to the level of synapses or the cytoskeleton, but ascends to the dynamics of tens of thousands of neurons involved Levels, the spatiotemporal patterns they generate are the preconditions for movement, that is, for behavior in the world outside the brain.Moreover, the theory is consistent with many phenomena in brain research over a century and could also be tested with some improvement in the spatial and temporal resolution of brain activity imaging techniques and microelectrode array techniques.

At its core, the Darwinian process is, at least among biologists, widely understood as a creative mechanism.Random mutations took thousands of years to form traits, and it took more than a century for us to realize how powerful this copying competition is.In recent decades, we've been able to see the same process operating on timescales of days or weeks, for example in the immune response to form a more suitable antibody.This Darwinian machine of the neocortex operating on the millisecond to minute scale is just changing the time scale again; we should be able to extend our knowledge of what this process can accomplish from evolutionary biology and immunology to thought and The time scale over which the behavior is achieved. It seems to me that we are long overdue to accept James's view of our mental activity.However, many people, including scientists, still hold an unreal view of Darwinism, arguing that it is simply selective survival.On this point, Darwin also played a role in fueling the flames a little.He named his theory the fifth of the six elements, natural selection.What I hope I have done in this book is to bring together all six elements of the Darwinian process and those aspects of its acceleration, and then describe a specific neural mechanism.As a mechanism rather than a metaphor for improvement, the strongest support for my neocortical Darwinian right at this point is cortical neuroanatomy and the principle of dragging oscillators with those 6 elements and acceleration factors of the Darwinian process It fits well. Whether this is the most important process, or whether there is another process that dominates consciousness and speculation, is hard to say; it may be without precedent in biology or computer science—without first discovering some intermediate metaphor , we cannot yet imagine such a process.Seriously, I suspect that "managing" the kind of process that reproduces competition will require an intermediate level of its own description in order to avoid restlessness or stagnation, and I don't use the word "management" in the usual sense of the term Rather, it's a bit like the way global climate models are affected by jet streams or El Niño events.In psychological terms, this "management" might be what Raven calls "a covert personality, with its unusual, satisfying wishes, relapses, and struggles." The complex brain code formed by the Darwinian competition for reproduction can explain many things about our mental activity.It suggests why humans are able to perform far more novel behaviors than other animals (we have non-mainline evolution of non-standard locomotion patterns); it also suggests how we can reason by analogy of all kinds (relationships themselves can have codes that can compete).Because the brain code can be formed from small pieces, you can imagine a unicorn and form a memory of it (the imprint can reactivate the corresponding space-time code of the unicorn).In particular, the Darwinian process is a metaphor-generating machine: you can encode relations among relations and shape them into properties. Such an account of intellectual sexual awareness offers insight into metaphor and action in the imaginary realm.It must tell us about the homogeneity between thought and other mental activities.In the explanation I propose, ballistics and music are intrinsically related to thought and language.We have seen that the emphasis on novel sequences has the potential to generate non-linguistic natural selection that benefits language and vice versa.The overlap between the mouth-face sequence and the hand-arm sequence suggests that both employ the same neural architecture. An important auxiliary use of the Darwin machine may be planning impending motion (rather than ballistic motion) on time scales of seconds, hours, days, or lifetimes.It will be possible to try various synthesis methods, determine what's wrong with them, make them more precise, etc.Those who are good at this can be called smart people. Any explanation of intelligence must also shed light on the ways in which we can gain insight into intelligence other than life on Earth.In short, it must shed light on artificial intelligence (AI), on augmenting animal and human intelligence, which might help detect signals from extraterrestrial intelligence.Not much can be said about the problem of "intelligence outside the earth", but let me try to provide a little perspective from the perspective of behavior, which may help us think about artificial intelligence and intelligence enhancement. An intelligence freed from the necessity of foraging and avoiding enemies, like artificial intelligence, may not need locomotion.Because of this, there is a lack of thinking about what is going to happen next that human intelligence possesses.We first solve the problem of movement, and only later, in terms of system and ontogeny, do we gradually raise more abstract questions, and look forward to the future through speculation about the future. There may be other paths to high intelligence, but upward progression from locomotion is the paradigm we know of.Interestingly, this paradigm is rarely mentioned in psychology or artificial intelligence.While there is a long line of thought in brain studies that emphasizes "motor development upwards," discussions of cognitive function more commonly emphasize a passive observer who uses intelligence to analyze the sensory world.Contemplation of the world still dominates most studies of the mind and can be downright misleading in itself.The exploration of the human world is accompanied by constant speculation and intermittent decisions about what to do next, and these characteristics must be included in the way we intellectually delineate the points of dispute. It is difficult to estimate how frequently advanced intelligence will appear in evolutionary systems on Earth and elsewhere in the universe. The main limitation is that we currently know nothing about how to overcome the blind end in nature (that is, the tendency to be stuck in equilibrium and to be fixed). Limits make most guesswork moot.In addition, there is the requirement of continuity, that a species be stable enough at each step in its evolution not to disintegrate itself, and competitive enough not to be in a contest with an all-in-one specialist defeated. Listing intelligence, if done sufficiently, is probably slightly better than testing other species (or computers) in a way that tests human IQ.But we now understand the physiological mechanisms that help the brain make correct guesses and discover new sequences. We can evaluate two promising species (artificially created or enhanced schemas) to see how many intelligent building blocks they each manage to assemble, or how many stumbling blocks they manage to avoid.My current assessment checklist would emphasize: • An extensive set of movements, concepts (such as words) and other tools.But even with a large vocabulary accumulated by literacy over a long lifetime, advanced intelligence still requires additional units to form novel high-quality synthesis. • Tolerance to creative disorder.This disorder enables individuals to sometimes break free from old categories and create new ones. 6 or more simultaneous workspaces ("windows") per individual, enough to choose among different analogues; but not so many that it would eliminate the tendency to split groups and thereby build new vocabularies . • Ways to create new relationships between these workspace concepts.This relationship is more refined than "is-a" and "is-larger-than" that many animals can grasp.Tree relationships seem to be particularly important for language structures like ours.Our ability to compare two relations (analogy) enables operations to be performed in the metaphorical space. • Capability of offline prototyping before being functional in the real world.It is a forming process projected in some way combined with the 6 elements of Darwin, they can replicate, change and compete, judging by the multi-faceted environment; the more successful models become the model variants of the next round center).This process is also combined with certain acceleration factors (equivalent to reorganization, climate change, islanding), which has a shortcut that allows the Darwinian process to operate at the level of thought rather than movement. • The ability to form long-term and short-term strategies, by taking intermediate steps to help set the stage for future craft.It would be even more helpful to be able to project a calendar and monitor the progress of those strategies and tactics. Chimpanzees and bonobos may be missing several elements, but they do a better job than the current generation of AI programs. Another implication of my Darwinian theory is that even with all the ingredients, we would expect considerable variation in intelligence because there are individual differences in many ways: in taking shortcuts; in finding the right analogy; in the level of abstraction; in the handling of speed; in the spirit of tenacity (not more is better, and boredom may give better model variants a chance to develop). "Yes! In our country," gasped Alice, "if you run as fast as I do now, you'll end up in another place after a while." "Slow country!" said the Chess Queen. "Here, you see, you must run as fast as you can in order to stay where you are. If you want to get anywhere else, you must run twice as fast!" carol, Why don't more species have complex thinking abilities?Of course, comic strips have played up wild tales of silent intelligence in insects, too.But if they had even one-tenth of our pre-planning thinking ability, they would pose a serious threat to Africa.The absence of higher intelligence species, I suspect, is due to a barrier that needs to be crossed.It's not just a "brain-sized Rubicon," a body image that allows you to mimic someone else, or a dozen other advances in hominids that came after apes. A little Intelligence can be a dangerous thing—whether it is an alien, artificial, or human intelligence. Intelligence beyond the ape must always be sailing between two twin dangers, like ancient seafaring Home must circle the rocks of Scylla and the vortex of Charvbdis". The vortex of dangerous innovation is the more obvious danger. The dangers posed by the rocks are more subtle: conservative conservatism ignores chess red And then Alice's explanation of the need to run as hard as you can to stay in place. For example, when you're driving in rapids, if you can't keep your speed in the main course, you'll tend to hit rocks. Intellect is also A by-product of racing. Foresight is our particular form of running, a quality so crucial to intelligent navigators that it is necessary for longer-term survival. As evolutionary biologist Gould Admonishment: "Intelligence and development are a glorious accident of evolution.By virtue of its power, we have become stewards of the continuity of life on Earth.We didn't ask for the role, but we couldn't say no to it.We may not be fit for the job, but we're in the job. "When talking about other intelligent species, what about something we could create ourselves? The possibility that a human mind embedded in a silicon chip is a replica of the fine structure of the brain has attracted some attention. ...I don't think such an "immortal machine" that turns a human brain into a software general-purpose computer might not be very feasible. Even if neuroscientists end up as some physicists and computer scientists rashly assume Solve the "readout" problem like you do, I think unless those working lines are well tuned and stay that way, there is a good chance of dementia, psychosis, seizures. Just think of those with paranoid and obsessive-compulsive symptoms Human: When detention in a mental hospital is unlimited, no longer bounded by a person's lifetime, then "falling into an endless cycle" will take on new meaning. Who wants to take such a ridiculous bet?" It would be better, I think, to know the genes and the basic properties that are intended to be reproduced in successive generations.Dawkins sees these replicated relationships clearly in a book.My friend the futurist Thomas F Mandel saw this too.He had lung cancer and his prospects were looking bleak, but in reply to his friend who studied space cybernetics, he said: I had another motivation in starting this project.Seriously, in the 5 months since my cancer diagnosis, this motivation has permeated almost everything I do. I thought, like everyone else, that my physical body itself would not last forever, and I guessed I had less time left than the actuary had allotted, but if I could stretch out and touch every One that I know... I can give all of myself and the idiosyncratic patterns that make up Thomas Mandel so that when my physical body dies, I don't really leave. ...my subject will be here, part of this new space. It's not an original and innovative idea, but it's worth a try anyway.Maybe someday someone can fix all the pieces with some Mandelian element, and I could be the haughty, obsessive, tender, compassionate, and whatever else you all seem to sense that I might be everything. Certain programming of artificial intelligence may also produce intelligent robots.But I think that with the help of principles discovered in neuroscience, we can build a software general-purpose computer that can talk like a human, be as affectionate as our pets, and imagine metaphors, at several different levels of abstraction. think on. Basic software general-purpose computers can at a minimum reason, classify, and understand speech.I imagine that even rudimentary software general-purpose computers will be admittedly "conscious", self-centered like us.I do not mean the less important aspects of consciousness, such as awareness, alertness, sensitivity, excitability.I don't mean self-awareness, that doesn't seem to matter.Egocentric consciousness, I think, would be easy to achieve; making it useful to the intellect would be more difficult. From my point of view, software general-purpose computers that are gradually updated will acquire intellectual awareness such as steerable attention, mental repetition, syntactic language production, imagination, subconscious information processing, planning for the future, and strategic planning. Decision making, especially our human soliloquy in waking or sleeping.Although such software general-purpose computers operate on principles very similar to those used by our brains, they are carefully designed so that they can be remedied when difficulties arise.I've been able to see one way of doing this design, and that's to take those Darwinian elements and patterns of cortical wiring, resulting in triangular arrays and subsequent hexagonal replication competition among pattern variants.Insofar as these functions operate at much higher speeds than our own brains operate on the order of milliseconds, we have seen an aspect of "superhuman intelligence" derived from "software general-purpose computers."If a software general-purpose computer can reach new levels of organization (meta-metaphor), it may point the way to what educational efforts should be made to help humanity take the same step.However, this is an easy aspect, an extrapolation of existing trends in computer technology, artificial intelligence, neuropsychology, and the neurobiology of the human brain.Distilling wisdom from knowledge certainly takes longer than distilling knowledge from data.However, there are at least three difficulties.The first difficulty is to confirm that this software general-purpose computer is indeed compatible with the ecosystem composed of various animals (such as us humans).Especially with the human ecosystem, that's because the competition is strongest between closely related species, which is why neither of our cousins, Australopithecus or Homo erectus, have ceased to exist, and why there are only two The reason why omnivorous apes have survived to this day (the rest are vegetarians with extremely long intestines to extract calories from large amounts of food).If it weren't for climate change itself, our more immediate ancestors might have weeded out other apes and great apes as competitors.The world of the future will be a more demanding struggle against the limits of our intellect than a soft hammock where we can lie comfortably waiting for our robot minions to serve us. Aldo Leopold said in 1948: "The preservation of every wheel and cog is the chief concern of the wise tinkerer." Introducing a potent new species into an ecosystem is no easy task . It's often beneficial when automated re-layout happens so slowly that no one goes hungry.Previously, everyone gathered or foraged their own food, but in industrialized countries, the development of agricultural technology has gradually reduced the percentage of the agricultural population to about 3%.This frees up many people to do other things.The relative proportion of these "occupations" has changed over time, with many moving from manufacturing to service industries in recent decades.A century ago, the two largest occupational groups in the developed world were farm workers and domestic servants; today they make up only a small fraction. However, software general-purpose computers will even replace some of the more educated workers.The prospects for the poorly educated or below-average intellect are dire than they are now, but humanity would clearly benefit as well.Imagine if there is a superhuman teaching machine as a dynamic teaching machine, it will have a real conversation with the students; it will never tire of practicing with the students, and always remember to provide the students with the necessary variety of things so that they do not feel bored; can Tailor the curriculum to a student's specific needs; be able to routinely scan for signs of a developmental disorder, such as dyslexia or a short attention span, and you'll feel the same way. Silicon superhumans can also apply these talents to teach the next generation of superhumans to evolve more intelligent superhumans only through mutation and selection: eventually, the star silicon human can be copied.Afterwards, each offspring is educated differently.Depending on experience, some may have desirable characteristics and values ​​(such as sociability or concern for human flourishing).We can again select stars among them to copy.Since the copying process involves the memory of time (that is another advantage of silicic intelligence besides its ability to be remedial), experience will accumulate gradually, truly Lamarckian: future generations do not need to repeat the mistakes of their predecessors. The second difficulty is the question of value; what is accepted value?How can they be implemented in silicon chips? A rudimentary software general-purpose computer would be amoral, like our pets or young children, with immature intelligence and speech.They don't even possess some of the innate qualities that keep our pets safe.We humans tend to be seen by our pets as their mother (for cats) or the leader of their pack (for dogs); they obey us.This cognitive confusion about our roles allows us humans to benefit from our innate social behavior.We might want something similar for our intelligent machines, but since they're far more capable of doing bad things than pets, we'll probably need realistic defenses, more so than muzzles, leashes, and fences. something that works. How do we build that abstract defense?Like Isaac Asimov's Laws of Robotics?My guess is that it will require multiple copies of the star silicon man, a bit like domestication of dogs.This gradual evolution over many generations of superhumans may partially replace the biological inheritance that comes with birth, which may minimize the tendency of silicon superhumans to be harmful to society and limit their harmful behavior. If this is the case, it will take decades to evolve from a rudimentary software general-purpose computer with only immature intelligence to a safe superhuman who does not need constant guidance.Early models may have been deft and talkative rather than cautious or tactful, a dangerous synthesis with potentially harmful social implications.They will have top-notch abilities, but they will not be based on well-tested evolutionary predecessors whose abilities have been tested. It reveals the past, diagnoses the present, and predicts the future. A third difficulty is how to moderate the human response to this perceived challenge.Just as your immune system, an overzealous response to a stimulus, can damage you through allergies and autoimmune disease (perhaps kill you due to anaphylactic shock).Humanity's reaction to silicon superhumans can also create enormous tension in our current civilization.Once the software general-purpose computer has played a prominent role in the economy, a serious response to it threatens to destroy the existing social system, even the system in which farmers feed the remaining 97% of the population.Remember that famines kill people because of a malfunctioning distribution system, not because there isn't enough food somewhere in the world. But the Luddites and saboteurs of the twenty-first century will be aided by some fundamental features of human behavior that played little role in nineteenth-century Europe. Groups try to differentiate themselves from other groups, Despite the benefit of a common language, most tribes throughout history have always exaggerated linguistic differences with neighboring tribes, making friend and foe irrelevant. It is certain that the "Turing Test" will be used frequently , people will try to understand if there is a real person on the other end of the phone line.To reduce this anxiety, machines can be asked to speak in special voices, but this is not enough to prevent us from getting nervous with them. Software general-purpose computers and superhumans may also be limited to certain occupations.Their entry into other domains is subject to an evaluation process in which a new model is scrutinized against the background of a sample of real human societies.If the potential for serious side effects was so great, and the speed of introduction likely to be so rapid, we would be advised to take steps with the U.S. Food and Drug Administration (FDA) to test the effectiveness of new drugs, new medical devices,安全性和副作用相似的程序。这与其说减缓了技术的发展，倒不如说减慢了它的广泛使用，并有可能暂作退却，以防止依赖性发展太甚。 软件通用型计算机可以局限于有限的相互作用的范围之内；为了使用交互网络或电话网络，它们可能需要严格的批准。对于只有新手执照的超人来说，对其输出可能会采取“延迟一天”规则，以避免某些“程序交易”的危险。对一些新手我们可能需要某种计算机遏制装置，类似于我们遏制致命病毒的生物危害。 对真理的探寻是掠夺性的，它确确实实是一种狩猎，一种征服。在《共和国》一书中描述了一个典型的时刻，那时苏格拉底”和他的同伴在争论一个抽象的真理。他们叫喊着，就像那些发现追捕猎物的猎人……（即使禁止科学上的探寻）在某个地方，某个时候也总有那些对绝对思想的毒药上了痛的人，或者是单枪匹马，或者是成群结队，在努力试图建立生物组织，确定遗传的特性，在云雾室中产生夸克——的踪迹。并非是为了名声，不是为了人类的利益，不是以社会正义或社会利益的名义，而是因为一种内驱力，这种内驱力比爱更强烈，甚至比恨也更强烈，是它使人类对某些东西感兴趣。因为其本身神秘的缘故，因为它存在在那里。 这就开始提出了以下问题：“什么是这个社会该做的正事呢？”是打碎枷锁，优化培育，使人类变得无所不能？还是制造比人类更高明的计算机？很可能我们能兼而为之（就像那些教师的助手一样）。但是，在我们轻率地蜂拥而至产生超人的过程中，我们需要保护人性，这是一种重要的补救形式。 然而，我们能够导人谨慎从事的途径受各种冲动的制约，正是这些冲动把我们引导到这种智能的转折： 好奇心是我们自身的初始动机——智力是如何产生的？这肯定也是许多计算机科学家的初始动机。但是，即使好奇心因其本身的原因有点破足（就像各种宗教试图做的那样），其他冲动也把我们引导至相同的方向。 象棋红后效应的技术版本。如果我们不改进技术，别的人会这样做。从历史上来看，技术上的竞赛的失败常常意味着被你的竞争者所取代（或消灭）——以国家而言，而不只是公司。从最近几十年在数字计算机中速度和信息量以每18个月增加一倍的增长曲线来看，世界的其他部分不会减慢速度，即使大部分国家决定这样做。就生物技术来说，情况也一样。 对文明的严重的环境威胁要求庞大的计算机资源尽快地发展，因为我们气候通过洋流的重组在约几年之内就“变换排挡”。这样的突然变动（全球变暖似乎使这种变动更可能发生）现在会引发第三次世界大战，因为每个人（不只是欧洲人）都在为生存空间而斗争。为了我们自身的生存，我们的一个迫切任务是学会如何推迟这些气候的变动。为模拟全球气候所需要的大型计算机与模拟脑的工作过程所需的计算机十分相似。 我并不知道有任何设法延迟作出决定的现实的途径，使这种向超人的转折以一种更审慎的速度推进。因此，在今后的几十年中，我们将需要面临超智能机的问题，这不会由于减慢技术进步本身而有所推迟。 确实，我们的文明将在终极的意义上“与造物主比赛”：它正在进化一种比地球上现存的更高明的智能。对于我们来说，有必要变成为一个思虑缜密的造物主，明智地对待世界及其脆弱的本性，敏感地注意到为了站稳脚跟需要做些什么，从而来有效地防止倒退，使我们称之为文明的木屋不至于崩溃。 仅仅在2个世纪之前，我们用纯粹理性能对每一件事都作出各种解释，而现在大多数精心构制的和谐结构都在我们眼前土崩瓦解了。我们说不出话来……我们已经发现公。何提出重要的问题，现在我们紧迫地需要某些答案。我们现在意识到，我们再不能通过探寻我们的精神来做到这一点了，因为在那里并没有足够的东西可供我们搜寻；我们也不能通过揣测或者为我们自己编制故事来发现真理。我们不能停留在我们现在所处的位置，耽于今日的认识水平，我们更不能倒退。我不认为在这方面我们会有任何真正的选择，因为我们能看到的仅有一条路伸向前方。我们需要科学，需要更多、更好的科学，不是为了技术，不是为了消遣，甚至不是为了健康和长寿，而是为了充满睿智的希望，我们的文明要存在下去必须获得这种希望。