# Promotions for Team Plan B, Google Lunar XPRIZE and Innovation and its Enemies ![[resources/images/bryant-robot-dog-landscape-1600.jpg]] [[collections/Livestreams Collection|Livestreams Collection]] | [[collections/Livestream Lessons|Livestream Lessons]] | [[collections/Simple Reminders|Simple Reminders]] *Livestreams compilation dated April 25, 2016, standing for the promotional broadcasts on Team Plan B, the Google Lunar XPRIZE, and Calestous Juma's* Innovation and Its Enemies. > [!info] Note > Calestous Juma's book opens on artificial intelligence and then moves outward from it. Its first sentence is “The rise of artificial intelligence has rekindled a long-standing debate regarding the impact of technology on employment,” and the paragraph that follows treats that debate as one instance among many: exponential advance signalling hope and fear at once, controversies framed as risks to moral values, human health, and environmental safety, and beneath those framings a set of socioeconomic considerations that go unacknowledged — who captures the benefit, who absorbs the risk, and whose cultural identity is altered. Juma's conclusion is that societies carrying great economic and political inequities are the ones most likely to experience heightened technological controversy. *(Calestous Juma,* Innovation and Its Enemies: Why People Resist New Technologies, *2016, jacket copy.)* > > ![[resources/images/2016-07-06-innovation-and-its-enemies-galley-proof.jpg]] > > Artificial intelligence was the entry point rather than the subject. It rekindled an argument that has been running since the printing press and the mechanical loom, and it rekindled that argument across a whole host of technologies at once — genomics, agriculture, energy, transport, synthetic biology, off-world manufacturing — which is what made it the case through which the general mechanism could be examined. The lessons below follow the book's own order of approach, which is why they return to artificial intelligence repeatedly as the live instance while the mechanisms they name apply to any technology that disturbs a settlement. > > This workbook is a compilation of livestreams rather than a single recorded broadcast. Across the spring of 2016 I promoted and discussed: Team Plan B and its Google Lunar XPRIZE campaign, for which I served as advisor to Adobri Solutions Ltd; Calestous Juma's *Innovation and Its Enemies*, which I was presenting with him before publication; and the off-world manufacturing argument that joined them. April 25, 2016 marks the beginning of active promotion on Periscope and Meerkat broadcasts. ## Four Kinds of Opposition Deserve Four Different Answers *Promoting Calestous Juma's book meant explaining, over and over, why opposition to a new technology is normal rather than pathological, and the part people found most useful was that the single word resistance conceals four different things. The mechanism underneath is this.* +Opposition to a new technology arrives in at least four forms, and treating them as one object guarantees that three of them are answered wrongly. Some opposition identifies a real failure, which makes it the cheapest engineering information available. Some expresses a transitional loss, where the objection is true, the technology is still correct, and what is owed is a route through rather than an argument. Some protects an incumbent position and speaks in whichever public language the culture currently respects. And some is only the ordinary request that an unfamiliar thing become familiar before it is trusted. {{Resistance is not one phenomenon with one answer; it is four different signals arriving through the same channel, and the first task is always to sort the mail.}} Dismissing the whole category as ignorance forfeits the information inside it. Accepting the whole category as a verdict forfeits the future. +Innovation attracts opposition because it changes a settlement rather than a device. Every durable technology sits inside an arrangement that determines who knows, who earns, who leads, which skills confer standing, and which institutions remain necessary — and a new instrument rearranges all of that before anyone has agreed to it. This is why the intensity of a controversy tracks the size of the disturbed settlement rather than the magnitude of the technical risk, and why two technologies of identical danger can meet wildly different receptions. Artificial intelligence draws disproportionate heat for exactly this reason: it touches who knows, who judges, and who is believed, which is a wider settlement than a faster engine or a cheaper material has ever disturbed. {{The object of a technological fight is almost never the machine; it is the settlement the machine is about to renegotiate.}} Reading the settlement first explains the fight. Reading the device first explains almost nothing. The six-century pattern behind this, and its application to artificial intelligence, is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## A Rule Written as Universal Virtue Can Be a Moat *This was the part of the book that surprised audiences most while I was promoting it, because the historical examples are so plain once they are laid side by side. The mechanism underneath is this.* +An incumbent defending a market position almost never announces that it is defending a market position. It speaks instead in whichever language the surrounding culture already honors — safety, quality, tradition, employment, the protection of children or the vulnerable — and a portion of what it says will be true, which is exactly what makes the argument difficult rather than easy. The historical record is long and unembarrassed about it: margarine, electricity, refrigeration, and mechanized farming each met organized opposition dressed as public concern. {{The most effective defense of a market position has always been a rule presented as a universal virtue.}} The record shows safety concern and competitive interest travelling together routinely, each fully operative, which is why the sorting has to be done on structure rather than on sincerity. +Three diagnostic questions separate a standard from a moat, and all three can be asked without accusing anybody of bad faith. Who benefits from the proposed restriction, measured by position rather than by stated motive. Whether the same standard is applied to the incumbent system or only to the entrant. And whether compliance can be satisfied only by organizations already large enough to maintain a compliance function. {{Compliance is a fixed cost before it is a moral achievement, and a fixed cost is a filter whether or not anyone designed it as one.}} Applied to artificial intelligence the test becomes productive immediately: a compute threshold, a mandatory evaluation regime, or a licensing requirement each impose a fixed cost that the largest laboratories already carry as ordinary overhead and that a university group or a four-person company cannot. A rule that passes all three questions is probably a standard. A rule that fails the third while invoking the first is a moat, and it will tend to produce the concentration it then cites as proof that concentration was inevitable. The historical cases behind this test are developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Technology Enters a Symbolic World Before a Market *I spent a good deal of the promotional run on this one, because it is the part that engineers find least intuitive and that anyone who has tried to change a community's practices recognizes immediately. The mechanism underneath is this.* +A new technology arrives in a symbolic world well before it arrives in a market. Food carries memory, work carries dignity, family carries obligation, and ritual carries continuity with the dead — so a technology touching any of those is negotiating with meaning rather than with preference, and the negotiation runs on terms no price signal expresses. This is why innovations in nutrition, childbirth, burial, education, courtship, and the authorship of art reliably meet resistance of a different order than innovations in logistics, at identical levels of measurable risk. Artificial intelligence now sits on the last of those, which is why generated prose and generated images provoke a reaction that no account of measurable harm alone will explain. {{A technology that disturbs a practice through which a culture recognizes itself is not facing a consumer objection; it is facing a question about whether the culture survives its own adoption.}} Treating that question as noise produces technically excellent systems nobody consents to inhabit. +Cultural translation is the working method, and it begins with a question rather than a campaign: what human value was the inherited practice preserving? Once that is named, the new instrument can usually carry the same value forward in another form, which converts the choice from one between heritage and efficiency into one between two ways of honoring the same thing. A people will exchange a tool far more readily than they will exchange belonging. {{Ask a culture to abandon its meanings and it will refuse the instrument; show it how its deepest commitments survive inside the instrument and it will carry the thing further than you could.}} Contempt for the receiving culture is therefore not merely impolite. It is a design failure with a measurable adoption cost. The cultural mechanism behind technology acceptance is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Stable Without Becoming Stationary *Institutional lag was the part of the argument that policymakers in the room always wanted to talk about, and I kept trying to separate the slowness that is wisdom from the slowness that is incapacity. The mechanism underneath is this.* +Institutions learn slowly because they were built to preserve continuity, and that slowness is sometimes the most valuable property they have. A society should not rewrite its law at the tempo of a product launch; courts, licensing boards, professional bodies, and curricula exist precisely to resist the enthusiasm of any particular decade. The difficulty appears when deliberate slowness hardens into an inability to learn at all, because then the institution is no longer preserving continuity — it is only preserving itself. {{Slowness is a feature of an institution until it becomes the institution's only remaining capability.}} The distinction is observable: an institution that is deliberately slow can still tell you what evidence would change its mind, and an institution that has become stationary cannot. +Closing the distance between artificial intelligence and the institutions that must govern it requires standing capacity rather than better intentions. A court, a licensing board, or a ministry that intends to govern these systems needs technical cadres who use them continuously, bounded fields in which limited experiments are permitted, instrumentation that produces evidence instead of anecdote, published findings, and a scheduled revision date written into the rule itself. None of that makes an institution impulsive, and all of it is ordinary practice in competent engineering. {{An institution governs a moving technology by acquiring the habit of revision, not by acquiring a faster opinion.}} The objective is a body that remains stable without becoming stationary, which is a harder design problem than either stability or speed considered alone. The governance argument behind institutional lag is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## The Incumbent System Should Also Appear in Court *Audiences grasped this one the moment it was stated as a comparison problem rather than a psychology problem, so I began presenting it that way. The mechanism underneath is this.* +Unfamiliar risks are judged against an idealized memory of the arrangement they would replace, and the asymmetry is structural rather than irrational. A new machine's failure has a name, a date, and often a plaintiff; the failures of the incumbent system are absorbed into custom and counted by nobody — the misdiagnoses inside normal practice, the applications never read, the waiting lists, the exclusions, the decisions made at the end of a long shift. People know how to fear a novel failure because it is legible. They do not know how to fear a familiar one because it has been reclassified as the weather. {{A familiar system hides its casualties inside custom, which is why it always appears safer than it is.}} +An honest comparison between a new technology and the arrangement it would replace requires both of them to produce numbers. An artificial intelligence system properly carries the burden of proof, and the arrangement it would replace should still be made to disclose its own: error rate, exclusion rate, latency, cost, and the distribution of who absorbs each. A diagnostic model is held to an accuracy standard the clinic it would assist has never once been asked to publish, which is the comparison quietly being avoided. Judging an innovation against a remembered golden age while exempting the present arrangement from measurement is the single most common analytical failure in technological debate, and it survives because nobody has standing to compel the incumbent's disclosure. {{The new technology belongs in the dock, and the arrangement it would replace belongs in the dock beside it.}} Honest comparison frequently vindicates the incumbent. It should at least be made to appear. The comparison problem underneath perceived risk is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## A Frightened Imagination Supplies the Missing Motive *This was the mechanism I most wanted the policy audience to take away, because the usual remedy proposed for it does not work and the one that does is unglamorous. The mechanism underneath is this.* +When people can feel the consequences of an artificial intelligence system but cannot locate its purpose, its owner, its limits, or its route of correction, they will supply the missing explanation themselves, and a frightened imagination is rarely economical in what it supplies. The resulting narrative is not a failure of public intelligence; it is the predictable output of a consequential machine with an unexplained interior. Concentrated benefit accelerates the process, because a visible asymmetry of gain gives the invented motive something to attach to. {{Secrecy does not leave a blank space in the public mind; it leaves a space that the public will fill, and it will not be filled conservatively.}} +Public comprehension is the remedy usually proposed for a system people cannot see into, and it is the wrong target. No amount of explanation renders the interior of a large neural network legible to a non-specialist, and promising otherwise sets up a failure that deepens the distrust. What actually reduces the vacuum is operational legibility: identifiable ownership, named responsibility, stated objectives, disclosed decision paths, observable performance, and a lawful route by which a competent outsider can enter, test, and benefit from the machinery. {{People become resistant to fantasy when they are given standing inside a system, not when they are given reassurance about it.}} Standing is the variable. Transparency about parameters is frequently neither possible nor desirable, and it was never the thing that closed the gap. The relationship between opacity and public narrative is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## A Posture Is Not a Policy *By the time I was promoting this book the public argument had already collapsed into two camps, and the most useful thing the historical material does is dissolve both of them. The mechanism underneath is this.* +Being for artificial intelligence and being against it are postures rather than positions, and neither one survives contact with a specific decision. A posture answers every question the same way, which means it carries no information about the case in front of it; enthusiasm without discernment ends in capture, and refusal without comprehension ends in irrelevance. What replaces the binary is a gradient, which is the instrument people lack when they have only categories: evaluate a given deployment by consequence, reversibility, exposure, and the degree of dependence it creates, and the single question of whether one is for or against artificial intelligence dissolves into a set of answerable questions about particular uses. {{A permanent emotional identity toward technology is a way of not having to think about any technology in particular.}} +The useful distinction is between understanding artificial intelligence well enough to direct it and surrendering judgment to a stance. A mature technological culture can accelerate what enlarges human and national capability, constrain a demonstrated misuse, preserve the research required to improve the thing it is constraining, and revise the whole arrangement when evidence arrives — and it can do all four simultaneously without accusing itself of inconsistency. {{Consistency of posture is the cheapest kind of integrity and the most expensive kind of error.}} The binary persists because it is restful. Gradients require a judgment each time, which is precisely what makes them a form of competence rather than a form of identity. The argument against pro- and anti-technology postures is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Dignity Is a Credible Place in What Comes Next *Of everything in the book, this was the part I pressed hardest in the promotional conversations, because it is where the humane argument and the strategic argument turn out to be the same argument. The mechanism underneath is this.* +Resistance to artificial intelligence usually begins in material loss rather than philosophical error, and a transition becomes cruel at the moment the people asked to absorb its cost are treated as obsolete evidence of progress. Their knowledge still exists. Their obligations still arrive on the first of the month. Their communities still depend on the income and the competence organized around the departing work. {{Nobody has ever been argued out of a loss by being shown a chart of somebody else's eventual gain.}} Treating a person's accumulated competence as expired inventory is a claim about the economy that the person has every reason to dispute, and disputing it is where organized resistance begins. +Support designed as conversion produces participants, and support designed as consolation produces resentment — a distinction that is measurable rather than sentimental, because it operates on finite resources. A person has a bounded quantity of time, money, mobility, and remaining cognitive budget at the end of a working day, and a program that spends those on training disconnected from real demand has consumed the exact reserves the transition required. Conversion translates existing experience into an adjacent role, funds learning tied to demonstrated demand, preserves mobility where it is possible, and puts the instruments producing the new productivity into the displaced person's hands. {{Dignity is not a severance payment; it is a credible place in what comes next.}} The transition argument underneath technological backlash is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Rules Written Before Contact Regulate a Metaphor *Policymakers in these conversations wanted to know when the right moment to write a rule is, and the honest answer is that the question is posed wrongly. The mechanism underneath is this.* +A rule written before contact with the technology it governs regulates a metaphor, because the only object available to the drafter is the public image of the thing. Statutes drafted against artificial intelligence as a public image rather than against particular deployed systems are the live instance, and they age badly in a single direction. A rule written after every consequence has matured governs a settled market it can no longer shape. Both failures are routine, and neither is corrected by better drafting, because the problem is the attempt to produce a final rule from a single act of foresight. {{Regulation is not a verdict delivered once; it is an instrument that has to learn, and a rule incapable of learning will govern whichever decade it was written in.}} The fantasy of the complete rule is attractive precisely because it promises to end the argument, which is the one thing a live technology never permits. +Iterative governance is the middle path and it borrows its discipline from engineering rather than from software marketing. Define a bounded operating field, instrument it, enforce clear and real limits inside it, publish what is learned, and write the revision date into the rule itself so that revisiting is a scheduled obligation rather than an admission of failure. Law need not imitate a release cycle, and the continuity people require from law is itself a value worth defending against the enthusiasm for iteration. {{A rule that names the evidence which would change it has already done most of the work of being a good rule.}} Versioning what has been learned is available to law without surrendering what makes law worth having. The case for co-evolving rules is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Inaction Has Casualties That Cannot Be Photographed *This is the asymmetry I kept returning to while promoting the book, because precaution is treated as the neutral option and it is not neutral at all. The mechanism underneath is this.* +Delay has a constituency because its casualties cannot be assembled in a room. The disease not cured, the infrastructure not built, the discovery deferred, the enterprise never formed, and the national capability conceded to a faster competitor do not arrive at a hearing with names, photographs, and counsel; they disappear into the counterfactual, where nobody is accountable for them and no journalist can find them. The harms of action, by contrast, are concrete, datable, and attributable. {{A visible harm and an invisible harm of identical magnitude are not weighed equally by any institution ever built, which is why paralysis can be mistaken for prudence.}} Precaution therefore enjoys an accounting advantage it has never earned on the merits. +Competent precaution prices both sides of the decision and says the prices out loud. State what action might cost, state what inaction will cost, name the time horizon over which each arrives, identify who bears each, and then decide in the open which risks are worthy of a future. This improves precaution rather than weakening it, because an unpriced caution cannot be distinguished from an unexamined one, and the people most committed to safety should want the distinction most. {{A one-sided risk assessment is not caution; it is a decision that has concealed half of its own ledger.}} A civilization can be injured by reckless deployment. It can also be injured by teaching every builder that permission arrives only after somebody else has demonstrated the future. Artificial intelligence presents both failure modes at once, and only one of them can produce a photograph. The two-sided ledger behind precaution is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## People Resist Being Treated as Terrain *This was the closing mechanism in the book's argument about imposition, and the one with the most direct application to how artificial intelligence should be introduced anywhere. The mechanism underneath is this.* +A distant institution that arrives with a finished system, a vocabulary of inevitability, and no local route into ownership has already assigned the receiving community its role, and communities read that assignment accurately. The role is terrain — a surface upon which something is done rather than a party by whom something is done — and resistance to it is not a failure of education. It is an accurate response to a correctly perceived status. {{A system delivered finished tells the people receiving it exactly what part they have been given, and no amount of consultation afterward revises the casting.}} This is why identical technologies succeed in one region and are rejected in another at the same price, the same risk, and the same demonstrated benefit. +Real participation changes a community's assigned role rather than the language used to describe it. Local builders, operators, teachers, clinicians, and enterprises can adapt an artificial intelligence system to conditions no distant designer has ever observed, and in doing so they keep part of the competence and part of the value where the implementation actually happens. The test of an introduction is therefore what it leaves behind after the pilot funding ends. {{Adoption lasts when the arrangement departs having created more local agency than it consumed, and it decays whenever the reverse is true.}} An institution that wants durable uptake has to be willing to lose a measure of control over the thing it is deploying, which is the price of the only kind of adoption that survives. The argument against top-down imposition is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## A Worker Cannot Cross on a Bridge in the Future Tense *Employment was the first thing the room wanted to talk about whenever I presented this material, and the honest answer requires holding two things that sound opposed. The mechanism underneath is this.* +Employment is where artificial intelligence enters a household, and the household is the scale at which the usual reassurances stop working. A displaced worker cannot be answered with a chart showing aggregate productivity rising in a later decade; the mortgage arrives monthly and the retraining brochure is written in the future tense. {{A person is being asked to cross a river on a bridge that exists only in an economist's forecast, and the request is reasonable exactly to the degree that something has actually been built underfoot.}} The difficulty is that the opposite policy is worse: preserving every present task also preserves every present limitation, including the ones that make the work dangerous, exhausting, or poorly paid. +The humane response and the strategic response converge on the same instruction, which is to move capability toward the person rather than moving the person toward whatever is left. Teach the new instruments where people already are, distribute productive access on terms an ordinary household can afford, shorten the distance between learning something and earning from it, and build the transition system before displacement arrives rather than after it has become humiliation. {{Automation should remove drudgery from human life without removing the human being from the economy, and the difference between those two outcomes is a policy rather than a property of the machine.}} Nothing in the architecture of these systems selects between them. The full argument about displacement and distribution is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Observation Is Infrastructure and Participation Is the Question *This is where I most wanted to separate two things the public debate keeps fusing, because fusing them produces a conclusion that serves nobody. The mechanism underneath is this.* +Pervasive sensing, analysis by artificial intelligence, and institutional memory are infrastructure, and infrastructure is what makes a country competent. A state that can see will route resources well, detect failures before they cascade, prevent fraud, protect the grid and the water and the ports, and coordinate work at scales no unaided bureaucracy has ever managed. {{Intelligence is manufactured out of telemetry, and a nation that cannot observe itself cannot govern, defend, or improve itself.}} The capability is therefore worth building rather than worth apologizing for, and the people building it are doing serious national work. +The question that decides whether such a system earns legitimacy is a question about participation rather than about observation. A citizenry treated only as a source of data and never as participants, builders, beneficiaries, or owners of the resulting capability will experience a competent system as something performed upon them, and will resist it accordingly — which is the asymmetry Juma's framework predicts generates the most durable opposition. {{Legibility should run in both directions far enough that people can understand the systems they inhabit, enter them with skill, and help decide what national intelligence is for.}} Opposition concentrates where benefit concentrates. Broadening the entry points is cheaper than managing the resentment that follows from not broadening them. The distributional argument underneath this is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## A Future Becomes Credible When Part of It Is Delivered Early *Audiences would accept the long-run argument and then ask what to say to somebody carrying the short-run cost, which is the right question and the one the long-run argument cannot answer. The mechanism underneath is this.* +Disruption is felt immediately and improvement is promised distantly, and that mismatch rather than ignorance is what produces technological skepticism. A person facing present displacement is being asked to extend credit to an institution on the strength of a return scheduled for a later decade, and credit is extended on the basis of demonstrated performance rather than stated intention. {{A promise of future abundance is a currency that cannot be spent in the month it is offered.}} Telling a skeptical public that artificial intelligence will benefit their grandchildren is accurate and does almost no work, because the grandchildren are not the party being asked to absorb the transition. +Legitimacy during a long transition is purchased with something present and verifiable. Retraining has to terminate in demand that exists, access has to be real rather than announced, and the productivity the new systems generate should become visible as cheaper services, lower bills, stronger infrastructure, shorter waits, and new categories of work before the public is asked for further patience. {{Deliver a fraction of the promised future early and the rest of it becomes believable; deliver none of it and the whole of it becomes a slogan.}} Progress asks people to release arrangements they understand, and competent leadership gives them something solid to stand on while they let go. The argument about timing and credibility is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Follow the Incentive, Then Return to the Principle *The book's central analytical move is this one, and it is the part that makes people uncomfortable in both directions at once. The mechanism underneath is this.* +Moral language is frequently the acceptable surface of a material dispute, and recognizing that is an analytical instrument rather than an accusation. A threatened profession speaks of dignity. An incumbent speaks of safety. A faction excluded from new wealth speaks of human nature. In the artificial intelligence debate all three speak at once, which is why the argument sounds moral and moves on money. Each of those claims may carry real moral information, and each also has an economic structure that can be read: the revenue at risk, the status in question, the proposed distribution of control, and who would be licensed to operate if the argument prevailed. {{An ethical claim and an economic interest can occupy the same sentence without either one cancelling the other, which is why following the incentive is a first step rather than a verdict.}} +The discipline has two halves and most people perform only one of them. Follow the incentives, the threatened status, the displaced revenue, and the proposed allocation of authority; then return to the ethical question with a cleaner view and answer it on the merits. {{A society governs technology badly when it mistakes every interest for a principle, and just as badly when it assumes every principle is an interest wearing a disguise.}} The first error produces capture dressed as virtue. The second produces a cynicism that cannot recognize a real moral objection when one finally arrives, which is the more dangerous condition because it disarms the only instrument that would have identified a genuine harm. The method behind reading a technological controversy is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Work Carries a Story a Person Can Tell About Their Place *This was the mechanism that surprised the business audiences most, because they had been treating work as compensation and the resistance made no sense to them under that accounting. The mechanism underneath is this.* +Work is never only income. It is rhythm, status, competence, inheritance, belonging, and a story a person can tell about why their position in the world makes sense — and artificial intelligence disturbs every one of those before it touches the wage. This is why occupational displacement produces grief rather than merely budgeting, and why the grief is so frequently misread as technophobia by people who have never organized an identity around a craft. {{A technology that alters a trade is renegotiating the social identity built around the trade, and the identity was never listed on the invoice.}} +Adoption improves wherever a transformation preserves honorable continuity while opening a larger future. People can learn new instruments without being informed that everything they mastered was foolish, and the practitioner's judgment, taste, accountability, and understanding of what the work is for can survive into the new arrangement when a builder takes the trouble to carry them across. {{The wise builder does not ask a culture to become rootless; the builder shows how its deepest capacities survive in another form.}} Telling people their mastery was merely inefficiency is the single most reliable way to convert a manageable transition into organized opposition, and it costs nothing to avoid. The argument about identity and technological change is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## A System Acquires What Its Designers Do Not Know *Participation is the part everyone agrees with and almost nobody implements on a schedule that matters, so I kept restating it as an engineering argument rather than a civic one. The mechanism underneath is this.* +Participation is an information channel before it is a courtesy, and that is the reason to build it rather than the reason to praise it. Workers understand the job as performed rather than as specified. Communities understand the local constraint that appears in no requirements document. Operators understand the failure mode that precedes the incident. Citizens understand what a proposed change will actually ask them to surrender. {{Every one of those is knowledge the designers do not possess and cannot generate, which makes participation a sensing apparatus rather than a gesture of respect.}} A system built without it is not merely less popular; it is less informed, and it will discover what it lacked through failure instead of through conversation. +A system acquires outside knowledge only through instruments built to carry it, and goodwill supplies none of them. Field observation where the work is actually performed, structured operator debriefs, incident channels that reach engineering rather than public relations, adversarial testing conducted by the population that will be scored or sorted by the system, and a route by which a correction can be demanded and tracked are the mechanisms, and each one is a piece of sensing hardware for a class of knowledge no requirements document contains. {{The knowledge a designer most needs is held by people who will never be in the room, which makes a channel to them a component of the system rather than a courtesy extended to them.}} An artificial intelligence system deployed without those channels still learns what it was missing, through the failures of the people it was pointed at, which is the most expensive sensing apparatus ever assembled. The case for inclusive innovation is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## An Ethical Word Acquires Teeth When It Names an Operating Arrangement *The artificial intelligence ethics vocabulary was already accumulating in 2016 and I was trying to say what would make it operational rather than ornamental. The mechanism underneath is this.* +Words such as fairness, safety, dignity, and accountability become useful at the moment they are attached to an operating model, and remain decorative until then. Each of them resolves into four answerable questions: who is permitted to act, who may appeal a decision, who carries the cost when the system is wrong, and who receives the productive gain. {{An ethical principle that cannot name who acts, who appeals, who pays, and who gains has not yet become a principle; it is a mood with a vocabulary.}} Specify those four and the language acquires enforcement, auditability, and a shape that survives staff turnover. Leave them unspecified and the same words will be recited indefinitely while the economic machinery continues unchanged underneath them. +The practical question in every such debate is therefore architectural rather than rhetorical, and it can be asked in one sentence: what arrangement of access, responsibility, correction, and benefit makes this stated value durable after the conference has ended? That question is answerable, and answering it is ordinary institutional engineering of the kind professions have performed for two centuries. {{Ethics becomes a fog whenever every participant is performing concern and nobody is specifying an arrangement.}} The fog is not produced by insincerity. It is produced by stopping one step before the work begins, which is why so many well-intentioned frameworks leave no trace in the artificial intelligence systems they were written about. The argument about ethical vocabulary and operating structure is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Scale Becomes Leverage or Predation Depending on the Appetite Directing It *I wanted the business audience to hear this precisely, because the easy version of the argument is wrong and the easy opposite is also wrong. The mechanism underneath is this.* +Corporate scale is not the problem and capture is, which is a distinction worth holding firmly because collapsing it produces bad policy in both directions. Large organizations can assemble laboratories, compute, capital, manufacturing, logistics, and patience in combinations that small groups cannot sustain, and that assembly is how certain classes of problem get solved at all — medicine, materials, energy systems, national infrastructure, and the training of frontier artificial intelligence, which no small group has yet accomplished alone. {{The question about a large organization is never its size but the appetite directing what its size has assembled.}} Pointed at manufactured need and engineered dependency, scale becomes predation. Pointed at medicine, energy, education, infrastructure, national strength, and instruments that enlarge what an individual can do, the same scale becomes civilizational leverage. +The remedy for concentrated competence is broader participation rather than compulsory incompetence, and the two are routinely confused by people who mean well. Breaking capability into pieces too small to accomplish anything does not distribute power; it distributes incapacity and leaves the field to whoever is willing to reassemble it elsewhere. {{Concentrated competence is answered by widening who can own, operate, and build with artificial intelligence, never by arranging for nobody to have it.}} Ownership of productive assets, open access where risk permits, public compute, procurement that rewards contestability, and enough technical literacy in the population to use what exists are the working instruments. Character among the people directing the assembly is the other half, and it cannot be legislated. The argument about scale, capture, and distribution is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Choosing the Aim Is the Work That Does Not Get Cheaper *The identity question was already visible in 2016 and I was asked about it constantly, usually by someone whose profession was about to become inexpensive. The mechanism underneath is this.* +Writing, drawing, diagnosis, calculation, and strategic judgment carried identity because access to the skill was scarce, and artificial intelligence reproducing the performance changes the scarcity without settling the value of the person. The first crisis is therefore psychological rather than economic: an activity used as proof of being special stops functioning as proof, and a great deal of inherited self-definition becomes unavailable in a very short period. {{A capability that makes an old distinction cheap has not made the person cheap; it has moved the level at which a person becomes distinguishable.}} That relocation is disorienting precisely because the previous level had been reliable for a working lifetime. +The human contribution that survives a repriced skill sits one level above the performance, and it asks more of a person rather than less. Choosing the aim. Recognizing which problem actually matters. Joining disciplines that do not normally meet. Carrying responsibility for an outcome somebody can be held to. Establishing taste, which is the compressed judgment of many prior decisions. Deciding what the instruments should serve. {{The durable human contribution moves upward when the performance moves downward, and the upward position asks more of a person rather than less.}} This is a demanding future rather than a diminished one, and it rewards the people who were already working at that level before the instruments arrived. The argument about identity under repriced skill is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Distance Produces Mythology and Contact Produces Proportion *Regulators in these conversations wanted to know how to govern something still arriving, and the answer turns on proximity rather than on philosophy. The mechanism underneath is this.* +Governance becomes theatrical whenever it regulates an imagined object instead of an operating system, and the imagined object is the only one available to a body that has never used the thing. A working artificial intelligence system reveals users, failure rates, dependencies, incentives, error distributions, and consequences that speculation cannot supply at any level of intelligence or good faith. {{A regulator without contact governs the public image of artificial intelligence, which is the one version of it that nobody actually operates.}} This is why rules written at maximum distance from implementation age in months and why their authors are plainly surprised when they do. +Proximity to an operating system can be required without waiting passively for harm, and the instruments that deliver it are well understood. Test environments, staged deployment, instrumented pilots, measurement, professional standards, incident reporting, and rules that change as evidence arrives all deliver proximity without surrendering authority. The regulator should be close enough to the machinery to understand what is being governed and independent enough to say where its use becomes unacceptable. {{Distance produces mythology and contact produces proportion, and only one of those can be written into a statute that will still make sense in five years.}} Both halves of that posture are required, and most failures come from having only one. The argument about regulatory proximity is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Restraint Became a Geopolitical Position *This is the part of the stagnation argument that was easy to miss in 2016 and is now the dominant consideration, so it is worth stating as it was already forming. The mechanism underneath is this.* +Artificial intelligence contributes directly to national strength, which converts domestic restraint into an international position whether or not the people deciding intend it that way. Scientific instrumentation, industrial capacity, medical research, logistics, energy management, and defense all compound from the same underlying competence, which means a decision to slow development is simultaneously a decision about relative standing. {{Restraint in a strategically consequential technology is never only an ethical posture; it is also a transfer of position to whoever declines to exercise it.}} The question of whether to slow down stays open; what changes is the set of quantities being weighed when it is asked. +Protecting incumbents is the second cost and the quieter one. Broad prohibitions and heavy permissioning regimes are absorbed most easily by the organizations already large enough to maintain compliance functions, so a pause imposed in the name of caution frequently arrives as a competitive advantage for whoever is already established. {{A rule that only the largest participants can satisfy has decided who the future belongs to while appearing to decide only how fast it arrives.}} Competent precaution therefore prices the surrendered capability, the protected incumbent, and the delayed science alongside the risk of moving, and states all four in the open rather than pricing only the risk of moving. The full argument about the costs of restraint is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## A Profession Meets the Instrument Through Its Customers First *Of all the lag problems in the book's framework, this is the one I expected to arrive fastest, because it depends on the route of first contact rather than on anybody's deliberation speed. The mechanism underneath is this.* +A profession now encounters artificial intelligence through the people it serves rather than through its own governing body, and that reversal rewrites who sets the terms. Clients arrive having already used the thing. Students submit work produced with it. Patients bring printouts. Opposing counsel files briefs drafted by it. The licensing board, meanwhile, has not yet agreed on a vocabulary, which means the practitioner is improvising a standard in front of the person they are serving. {{A technology that reaches the customer before it reaches the regulator has already set the terms the profession will be arguing about.}} Reactive prohibition is the predictable output of a body asked to rule on something it has never operated. +Systems literacy distributed through a profession closes that kind of gap, and the level required is lower than the panic around it suggests. The public servant, judge, teacher, physician, and executive have no need to become model engineers; they need enough systems understanding to recognize what class of decision has entered the room, what the instrument can and cannot be relied upon to do, and which questions to ask before accepting an output. {{A profession keeps authority over its own work by understanding its instruments well enough to govern them, and loses that authority to whoever understands them better.}} Standing technical capacity beats committees convened after each surprise. The diffusion argument behind this is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Influence Before the Architecture Hardens *Everyone in these rooms endorsed public engagement and almost nobody scheduled it early enough to matter, so I began making the argument about timing rather than about virtue. The mechanism underneath is this.* +Engagement is valuable in exact proportion to how much the design can still change, which makes it a scheduling decision rather than an ethical one. While an architecture is soft, the people affected by it can still embarrass it into becoming better: they know where a process is brittle, which error class cannot be tolerated, what language will be misread, which workflow quietly depends on a judgment the system has removed, and what benefit would make the disruption worth carrying. {{Consultation conducted while the design can still be altered produces influence, and influence is the only thing that produces legitimacy.}} None of that knowledge is a property of the system; it belongs to the situation the system is about to enter. +Engagement conducted after the architecture has hardened requests applause, and the people asked for it recognize the difference immediately and resent being asked. The resentment is not irrational, because they have correctly identified that their participation has been converted into decoration for a decision already taken. {{A process that cannot be changed by what it hears has not consulted anybody; it has staged a hearing.}} The artificial intelligence laboratories and deployers that acquired this knowledge early shipped fewer catastrophes and absorbed less opposition. The ones that waited acquired the same knowledge later through complaint, regulatory inquiry, and litigation, which is the most expensive channel available for information that was free at the beginning. The argument about engagement and legitimacy is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## One Word Describes a Drafting Aid and a Consequential System *Juma's treatment of biosafety favored risk-scaled oversight rather than prohibition, and carrying that into artificial intelligence exposes a vocabulary problem that governs everything downstream. The mechanism underneath is this.* +A single term is being used for a drafting aid and for a system making consequential decisions inside defense, medicine, finance, and critical infrastructure, and governing both through one emotional category is an abdication of judgment rather than an excess of caution. The word names a family of methods, not a level of risk, which means every argument conducted at the level of the word is conducted about nothing in particular. {{A category that contains both a spell-checker and a targeting system is not a risk class; it is a failure to have drawn one.}} Demanding zero risk from such a category would immobilize the harmless half and tell you nothing useful about the consequential half. +Proportion is the instrument that keeps both safety and invention alive, and it scales along four observable variables: the consequence of an error, the reversibility of the outcome, the exposure of the population, and the degree of human or institutional dependence on the system being right. Low-consequence use stays open enough for learning and enterprise. High-consequence use carries evaluation, telemetry, named responsibility, and a demonstrated capacity to intervene before harm compounds. {{Oversight that scales with consequence is what allows a society to be careful about the things that matter without being useless about everything else.}} The alternative is a single rule written for the average case, which protects nothing and prevents a great deal. The case for risk-scaled oversight is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## The Public Supplied the Substrate *Safety dominated these conversations and I kept trying to widen the question, because safety is one governing variable among several and the others decide whether a capability is tolerated at all. The mechanism underneath is this.* +Whether a capability creates value is rarely the live question; what social structure receives the value is almost always the live question. A frontier system rests on a substrate the public supplied — the data, the labor, the capital markets, the electrical grid, the spectrum, the educated population, and the legal order that makes any of it ownable — which makes the terms of return a legitimate subject rather than an ungrateful one. {{A capability built on a public substrate owes the public an account of what it returns, and the account is a condition of legitimacy rather than a gesture of generosity.}} Framing that question as hostility to invention mistakes the people asking it. +Distribution is also far wider than cash, and the narrow reading does real damage to the argument. It includes expertise becoming cheap enough for an ordinary person to buy, public services that work, ownership of productive assets spread beyond a founding class, enterprises formed by people who previously could not have formed one, national systems that function under stress, and the practical ability of an individual to act at a scale once reserved for an institution. {{The measure of a technology's legitimacy is the capability it leaves in ordinary hands, which is a larger and more interesting figure than any dividend.}} Those are the terms on which a transformation is actually judged, and they can be measured. The distributional argument is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Secrecy and Exclusion Are Separable Variables *This distinction was the one I most wanted the policy audience to carry out of the room, because conflating the two produces a demand no serious state can meet. The mechanism underneath is this.* +Secrecy and exclusion travel together out of habit rather than necessity, and separating them resolves most of what looks like an impasse. Defense, security, and competitive research require protected methods, and a confident nation protects them as a matter of ordinary competence. What corrodes legitimacy is secrecy bundled with three other conditions that are each independently removable: no lawful route into participation, no answerable authority, and no public evidence of benefit. {{A protected method is compatible with an open purpose, a named command, published standards, and a visible route in — and the bundle is a choice rather than a consequence.}} +Citizens do not require blueprints, and the demand for them is usually a proxy for something else. What a public actually needs in order to extend confidence is narrower and more achievable: assurance that competent people can enter the system on merit, that conduct inside it carries consequences, that the capability belongs to a recognizable national project with a stated purpose, and that somebody can be identified who answers for it. {{Confidence in a secret program is built from the legibility of its purpose, its people, and its accountability, none of which require disclosing how it works.}} A state that supplies those four can keep its methods indefinitely. A state that supplies none of them will be narrated by its own population. The argument separating secrecy from exclusion is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Mood Is Not Administration *This was the closing argument of the whole framework whenever I presented it, because both of the available moods feel like positions and neither one is doing any work. The mechanism underneath is this.* +Worship and panic are both delegations of judgment, which is what makes them restful and what makes them useless. Worship hands the decision to the machine and calls the handover progress. Panic hands the decision to fear and calls the handover prudence. In each case a person has been relieved of the obligation to evaluate a particular system for a particular purpose, and in each case the relief is the entire appeal. {{A mood is attractive precisely because it answers every question identically, which is the same property that makes it carry no information about any question.}} Civilizations fail technological tests by substituting mood for administration, and the substitution is almost always performed by people who believe they have taken a stand. +Administration is the unglamorous alternative and it has a checklist rather than a posture: define the objective, measure actual performance against it, assign authority to somebody who can be named, retain the power to correct what is found, distribute capability widely enough that the system acquires defenders rather than only subjects, and decide in the open which risks are worth carrying for which returns. {{Governing a powerful technology is municipal work — objectives, measurements, authority, correction, and distribution — and it is dull in exactly the way competence is usually dull.}} Treat the capability as infrastructure under construction: powerful, unfinished, political in its distribution, and too consequential to be left to either a sales pitch or a nightmare. The governance argument that closes this framework is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## A Prize That Ends Without a Winner Can Still Succeed *Promoting the Google Lunar XPRIZE meant explaining what a competition of that kind is actually for, which is rarely the purse. The mechanism underneath is this.* +A prize of this class is an instrument for converting an impossibility into a costed engineering problem, and that conversion survives whether or not anybody collects. Before the competition, privately organized lunar access existed as a category error — the sort of thing a serious person did not propose in a meeting. Afterward it existed as a set of known subproblems with known price ranges: launch procurement, lander guidance, surface mobility, thermal survival, communications, and the regulatory path for a non-state actor operating off Earth. {{A competition's real output is the cohort of people who learned that an impossible thing is merely expensive, and that cohort persists after the deadline.}} +A prize of that design purchases distributed competence rather than a demonstration. Teams raise capital, assemble systems engineers, negotiate launch, navigate export control and licensing, and discover which assumptions were decorative — and all of that knowledge disperses into the industry afterward through the people who hold it. {{The failure of every entrant is compatible with the success of the enterprise, because the knowledge was the product and the trophy was the advertisement.}} A field opens when enough people have done the work to know what the work consists of, which is a slower and more durable mechanism than a single victory would have produced. The account of the campaign and its objective is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## A Destination Becomes a Manufacturing Site *The 3D-printing objective was the part of the mission I most wanted audiences to understand, because it is the part that changes the arithmetic rather than the itinerary. The mechanism underneath is this.* +Printing an object on another world converts that world from a destination into a manufacturing site, and the difference is economic rather than symbolic. A destination requires that every tool, fastener, replacement part, and structural element be lifted out of Earth's gravity well and delivered at launch cost. A manufacturing site requires only that the instruction set and the machinery arrive, after which mass is sourced locally. {{Every kilogram manufactured where it is needed is a kilogram that never had to be accelerated out of a gravity well, which is the only cost reduction in spaceflight that compounds.}} +The strategic value therefore sits in the inversion rather than in the artifact. A printed bracket proves nothing interesting about brackets; it demonstrates that the loop from local matter to useful hardware can be closed outside Earth at all, which is the precondition for everything built on top of it — shielding, landing pads, pressure vessels, replacement machinery, habitat. {{A first print is a proof of loop closure rather than a proof of manufacturing, and the loop is the thing the whole architecture waits on.}} The remaining distance is unglamorous and substantial: power, abrasive dust, thermal cycling, autonomy, and the part tolerances that structural use demands. The off-world manufacturing argument is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## What a Small Organization Can Credibly Attempt Is the Real Barrier *This was the part of the project that interested me most as a systems question, because the binding constraint was never the physics. The mechanism underneath is this.* +The constraint on an undertaking of this kind is the credibility threshold rather than the technical threshold, and the two are routinely confused. For decades lunar exploration belonged to governments with multibillion-dollar budgets, which established an unexamined premise that the budget was the requirement. The premise governed who applied for funding, who was taken seriously in a meeting, which engineers considered the field open to them, and which investors would return a call. {{A barrier made of other people's estimates is still a barrier, and it holds long after the engineering reason for it has dissolved.}} +Attempting the thing is what tests the premise, which makes a serious attempt valuable even when it does not succeed. A small organization that assembles a credible lander and rover programme has demonstrated something about the category rather than about itself: that private-sector ingenuity can reach the problem, that relentless technical iteration can substitute for institutional scale, and that a working blueprint exists for whoever comes next. {{The first group to treat an undertaking as survivable changes the probability every subsequent group assigns to it, and capital, talent, and regulatory willingness follow the revised estimate rather than the original one.}} The argument about perception and private space capability is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Decomposition Is What a Serious Attempt Leaves Behind *The moon shot became a metaphor for ambitious work in every field, and the transferable part is not the inspiration. The mechanism underneath is this.* +An undertaking whose probability looks small from outside is converted into a sequence of solvable problems by one operation, which is honest decomposition. Vision supplies the target, engineering discipline supplies the decomposition, and perseverance supplies the time — but the decomposition is the only one of the three that produces a transferable asset. Name an outcome current means cannot reach, break it until each piece is tractable, assign each piece a cost and an owner, and the impossibility has been replaced by a schedule. {{An impossible objective is almost always a correctly stated objective that nobody has yet decomposed.}} +Most of what a serious attempt produces is therefore the map rather than the arrival, and the map survives the attempt's failure. This is why the symbolic moon shot transfers to domains with no connection to spaceflight: the structure holds wherever somebody is willing to state an outcome plainly and then do the unglamorous work of finding its parts. {{A failed attempt that decomposed the problem correctly has delivered most of the value a successful one would have, and it has delivered it to everybody.}} The undertakings that leave nothing behind are the ones that were never decomposed, only announced. The symbolic argument behind the moon shot is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## The First Rung Matters Because of the Ladder Above It *People asked why a single print on the Moon deserved the language I was using about it, and the answer is entirely about what sits above it. The mechanism underneath is this.* +A first demonstration earns its significance from the ladder it is the bottom of rather than from its own magnitude. Additive manufacturing off Earth is the first rung of a structure whose upper rungs are autonomous robotic systems, self-maintaining industry, universal constructors, and machines carrying human organization beyond the biological reach of the people who designed them. Judged alone, the rung is a modest technical result. Judged as the entry point to that ladder, it is the step the rest of the architecture waits on. {{A first rung is worth the attention given to the ladder, and the error is to appraise it as though it were the summit or as though it were nothing.}} +Honest appraisal requires keeping the levels separate, which the enthusiasm around such projects rarely does. A printed component is demonstrated capability. A rover that manufactures its own replacement parts is architectural possibility, with the pieces existing at different levels of maturity. A machine ecology that expands without resupply is a scenario with identified unsolved problems. {{Distinguishing what has been demonstrated from what has been architected from what has been imagined is the whole of credibility in a long programme, and collapsing them is how a serious objective acquires a reputation for fantasy.}} The strategic reading of the mission is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## A Seed Factory Carries Organization Rather Than Matter *Explaining von Neumann's question to an audience encountering it for the first time was one of my favorite parts of this presentation, and the key is to say what is actually being transported. The mechanism underneath is this.* +Von Neumann asked whether a machine could be programmed to build an exact copy of itself and whether such a system could construct other machines, and the answer that matters is about what gets carried rather than what gets built. A universal constructor is a robotic printer and assembly plant able to manufacture a wide range of objects including the parts required to reproduce its own productive machinery. Placed on a spacecraft it becomes a probe that travels, gathers local matter, uses available energy, and builds productive capacity from which further machines emerge. {{A seed factory transports organization rather than material, which is why its payload mass bears no relation to the industry it can become.}} +The scaling resembles biological reproduction and carries the same five conditions, each of which is a real engineering problem rather than a formality. One probe becomes two, then four, then eight, provided the system can obtain energy, extract useful material, manufacture components, preserve its instruction set across copies, and repair the machinery the cycle depends upon. {{Nothing in the architecture asks for a machine that creates something from nothing; it asks for a system carrying enough organization to turn matter already present into an expanding industrial ecology.}} Of those five conditions, the manufacture of complex components from local material remains the hard boundary. The explanation of constructors and probes is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Closure Is a Gradient Rather Than a Threshold *Productive closure is the term that does the most work in this architecture, and treating it as a switch is what makes the whole programme sound like science fiction. The mechanism underneath is this.* +Productive closure describes the condition in which a robotic ecology can maintain, repair, and expand itself without a physical supply chain from Earth, and it is a gradient rather than a line that gets crossed. Early systems achieve partial closure by manufacturing bulky, simple, tolerance-forgiving structures locally — shielding, pads, pressure vessels, structural members — while importing complex electronics. Later systems move the boundary outward as each capability matures. {{Closure is measured as the fraction of a system's own mass, by component class, that it can produce from local material, which turns an argument into a number.}} +Posing closure as a gradient rather than a threshold changes what can be planned, which is the practical gain. Asking whether a system is self-sufficient produces a debate with no resolution and no intermediate deliverables. Asking what percentage of its mass it can presently make, and which component class is the next cheapest to localize, produces a roadmap with measurable milestones and an order of operations. {{A threshold question yields an opinion and a gradient question yields a schedule, which is the difference between a vision and a programme.}} The instruction set, the seed factory, the extraction of oxygen and metals from regolith, and the maternal systems that produce daughter systems are then each positioned on that gradient rather than argued about in the abstract. The closure argument is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Falling Launch Costs Do Not Answer a Compounding Tax *The economic argument for seed industry is the one most often answered with cheaper rockets, and the answer misses what kind of cost it is. The mechanism underneath is this.* +Lifting industrial mass out of a gravity well is a tax that compounds with ambition rather than a price that can be negotiated down. Reducing the cost per kilogram is valuable and changes what missions are affordable, and it leaves the structure of the problem intact: a civilization whose entire industrial base must be launched pays again for every expansion, every replacement, every repair, and every increment of scale. {{A cheaper rocket reduces the rate of the tax; local manufacture removes the thing being taxed.}} Those are different categories of improvement, and only the second one scales. +Seed industry attacks the compounding, which is why the Moon is the correct proving ground rather than the destination. A maternal system producing daughter systems could build solar arrays, communication relays, replacement machinery, and radiation-shielded habitat before any crew arrived, and lower gravity makes structural printing cheaper there than anywhere a crew would want to live. The Moon is also close enough for teleoperation when autonomy fails and harsh enough to be an honest test of dust, vacuum, and thermal cycling. {{The objective was never the Moon; the objective is closure, and the Moon is where closure can be attempted within a communications delay a human operator can work inside.}} The deep-space argument is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]]. ## Ambition Out of Proportion to Resources Is the Normal Frontier Configuration *Describing the actual size of this team was always the most persuasive part of the presentation, because the ratio is the argument. The mechanism underneath is this.* +The configuration worth studying in a project like this is the ratio of ambition to resources rather than either quantity alone. A Canadian company, a founder, an advisor, and an objective sized for a national space programme is an arrangement that sounds like a category error and is in fact the ordinary shape of frontier work. Frontiers are opened by groups whose ambition exceeds their resources by an amount that would be irresponsible in an established field, because an established field has already priced its problems and a frontier has not. {{A frontier is identifiable by the ratio of what its participants are attempting to what they have, and the ratio always looks indefensible from inside a mature industry.}} +A group attempting something at that ratio reliably produces competence rather than a guaranteed result, and the competence disperses whatever the outcome. Systems engineering under real constraint, regulatory navigation nobody had mapped, capital formation for an objective with no comparables, and the willingness to attempt a thing before anybody had demonstrated it was survivable — all of that leaves with the people who learned it and reappears in the organizations they join next. {{The industry that follows a frontier is largely staffed by people whose own attempt did not work, which is why counting winners understates what an opening cost and what it produced.}} The account of the team and its mission is developed in [[blog/20160523 - Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow|Deep Use of Existing Technology, SpaceX, and the Technologies of Tomorrow]].