The Age of Em: Work, Love and Life when Robots Rule the Earth
Paperback available here
Hardback: 1 June 2016, Revised paperback: 5 June 2018 (Has 4 new sections, 18% more text, and 42% more citations)
(See full cover. Cover image: Habitat by Till Nowak)

Robots may one day rule the world, but what is a robot-ruled Earth like?

Many think the first truly smart robots will be brain emulations or ems. Scan a human brain, then run a model with the same connections on a fast computer, and you have a robot brain, but recognizably human.

Train an em to do some job and copy it a million times: an army of workers is at your disposal. When they can be made cheaply, within perhaps a century, ems will displace humans in most jobs. In this new economic era, the world economy may double in size every few weeks.

Some say we can't know the future, especially following such a disruptive new technology, but Professor Robin Hanson sets out to prove them wrong. Applying decades of expertise in physics, computer science, and economics, he uses standard theories to paint a detailed picture of a world dominated by ems.

While human lives don't change greatly in the em era, em lives are as different from ours as our lives are from those of our farmer and forager ancestors. Ems make us question common assumptions of moral progress, because they reject many of the values we hold dear.

Read about em mind speeds, body sizes, job training and career paths, energy use and cooling infrastructure, virtual reality, aging and retirement, death and immortality, security, wealth inequality, religion, teleportation, identity, cities, politics, law, war, status, friendship and love.

This book shows you just how strange your descendants may be, though ems are no stranger than we would appear to our ancestors. To most ems, it seems good to be an em.

And here is a longer talk.

Robin Hanson is associate professor of economics at George Mason University, and research associate at the Future of Humanity Institute of Oxford University. He has a doctorate in social science from California Institute of Technology, master's degrees in physics and philosophy from the University of Chicago, and nine years experience as a research programmer, at Lockheed and NASA. He has 3500 citations, 60 publications, 700 media mentions, and he blogs at OvercomingBias.

Prof. Hanson's second book, coauthored with Kevin Simler, is The Elephant in the Brain: Hidden Motives in Everyday Life, which will be published January 2, 2018.

More on Prof. Hanson here.

Table of Contents
    1. Preface
    2. Acknowledgements
    3. Introduction
    Your ancestors behaved differently from you, and you probably explain this via moral progress. However, your descendants may also differ greatly, and in ways that challenge your progress ideals. This is because new habits and attitudes result less than you think from moral progress, and more from people adapting to new situations. This book presents a concrete and plausible yet troubling view of a future full of strange behaviors and attitudes. It starts with a particular technology often foreseen in futurism and science fiction: brain emulations. It then uses standard theories from many physical, human, and social sciences to describe in detail what a world with that future technology would look like.
  1. Basics
    1. Start
      1. Overview
      2. Summary
      There have been three human eras so far: foragers, farmers, and industry. The next era is likely to arise from artificial intelligence in the form of brain emulations, sometime in the next century or so. This book paints a detailed picture of this new era. Here we summarize this picture. For example, most ems are much faster than humans, who live comfortably on the margins of the em society. Ems are crowded into a few dense hot cities, mostly live and work in virtual reality, and work most of the time because of their near subsistence wages. Ems reproduce via exact copies, and usually whole teams are copied together. Most ems are temporary copies that will be deleted after finishing a short task, and most are near a peak productivity subjective age of 50 or more years.
    2. Modes
      1. Precedents
      2. Prior Eras
      3. Our Era
      4. Era Values
      5. Dreamtime
      6. Limits
      Past changes suggest future changes. Compared with its prior era, each era has had bigger groups, faster growth, and a similar total number of humans. This roughly suggests that, compared with our industry era, the next era will have groups of size trillion, economic doubling times of less than a month, and a total duration of a year or two. Strong cultural pressures were needed to turn foragers into farmers, but we feel such pressures less as we get rich today. So industry culture includes many farmer to forager trends, and non-adaptive behavior. Eventually, future individuals must return to being adaptive and poor, and so move back toward farmer culture. The em era moves in these directions.
    3. Framing
      1. Motivation
      2. Forecasting
      3. Scenarios
      4. Consensus
      5. Scope
      6. Biases
      The future matters more than the past, and contrary to a common impression, past forecasters have shown that we can at least dimly foresee the social implications of disruptive new technologies. This book’s main method is to assume that brain emulations are cheap to make, and to then repeatedly apply standard consensus science to predict the details of a world containing such devices, assuming other details. These many details are collected into a self-consistent scenario, with parameter values chosen either to be most likely, or to be simple so as to support further analysis. For example, markets are usually assumed to be competitive with low regulation. Economics is emphasized and tentative conclusions are drawn in as many life areas as possible. We try to avoid common biases.
    4. Assumptions
      1. Brains
      2. Emulations
      3. Complexity
      4. Artificial Intelligence
      An em is an artificial signal-processing (or “computing”) device arranged to closely mimic the patterns of signals sent in a particular human brain. It has the same mental habits and styles of that human. This book assumes that brains can be emulated, and that ems can be copied and run at different speeds. For example, a kilo-em runs at 1000 times human speed, while a milli-em runs at one-thousandth. But as complex brains remain poorly understood, ems cannot be usefully reorganized, beyond a limited set of tweaks. Our rate of progress in non-em-based artificial intelligence suggests that when ems arrive that field will be less than halfway to human level abilities, allowing a substantial era where em labor is in high demand.
    5. Implementation
      1. Mindreading
      2. Hardware
      3. Security
      4. Parallelism
      By placing matching brain areas into matching activation states, ems could read the surface of each others’ minds. Emulation hardware is probably digital, fault-tolerant, very parallel, and specialized to the emulation task. The hardware cost to run an em is roughly proportional to speed over a wide range of speeds. Ems could change speeds by changing hardware. While some ems allow anyone to copy and run them, most ems fear mind-theft as a route to torture, slavery, exposed secrets, and lost investments. Many strategies are available to avoid mind-theft.
  2. Physics
    1. Scales
      1. Speeds
      2. Bodies
      3. Lilliput
      4. Meetings
      5. Entropy
      6. Miserly Minds
      The cost to run an em is proportional to speed probably within at least a factor of one million above and below human speed. Ems can afford to save archive copies at least every 5 subjective minutes. For a faster em, a natural-to-control physical body is proportionally smaller, e.g. a kilo-em has a millimeter tall body, to which gravity seems weaker and winds seem stronger. Ems can meet well in virtual reality when signal delays are less than reaction times; kilo-ems need to be within 15 kilometers. Ems who use fractal adiabatically reversible hardware use much less energy than do human brains for the same speed. They can temporarily vary their speed, and spend about as much renting their hardware as on energy and cooling to run it. Interacting reversible ems coordinate to reverse their interaction messages later within a reversing period.
    2. Infrastructure
      1. Climate
      2. Cooling
      3. Air & Water
      4. Buildings
      5. Manufacturing
      Like computer rooms today, em cities control temperature, dust, humidity, vibration, etc. Hardware closer to city centers is denser, and when higher it is lighter. City centers are taller, and hold more recently designed hardware. Fractal cooling pipe systems occupy roughly half of city volume, and allow huge dense cities. Pipes may push in ice slurries and pull out near boiling water, in which case em hardware is also that hot. Buildings are made fast from modular units, don’t last as long as our buildings do, connect into a lattice to jointly resist winds, and are less resistant to earthquakes. Adiabatic reversible manufacturing spends roughly the same on renting factories as on energy and cooling to run them.
    3. Appearances
      1. Virtual Reality
      2. Comfort
      3. Shared Spaces
      4. Merging Real & Virtual
      Ems spend leisure time in virtual realities of spectacular comfort, beauty, and artistry, and which prevent direct violence. Most ems also work in virtual offices, where environments need to not be overly distracting. Em virtual realities have many elements that would be recognizable and familiar to us. In contrast, em physical objects look more harsh and functional when viewed directly. Em virtual and physical spaces may be integrated into a common spatial representation, to help ems reason about brain locations. An em whose virtual body travels too far from its brain must accept delayed reactions to local events. It is usually prohibitively expensive to have active intelligent non-player characters in em virtual realities.
    4. Information
      2. Records
      3. Fakery
      4. Simulations
      Ems see not just real and virtual worlds, but also ways to manage their brain’s speed, connections, security, and location. Some ems, such as spurs and retirees, are not visible in default views of spaces, but are visible on request. Ems can usually verify the identity of interaction partners, ems save audio-visual recordings of their lives, and ems want reliable records of their copy history. Ems agree to sometimes be placed into sims which they cannot at the time distinguish from reality. Each sim usually serves several functions at once. Ems in unusual situations suspect that they are in a sim, which makes ems especially loyal and reliable in such situations.
    5. Existence
      1. Copying
      2. Rights
      3. Many Ems
      4. Surveillance
      When blank em hardware is available, an existing em is easily copied into that hardware, resulting in a new em with exactly the same memories and mental habits, but with new diverging experiences from that point forward. Ems who can veto copies feel stronger ownership of their existence. While many possible legal regimes could cover em creation, for the rest of this book what mainly matters is that ems are usually created to fill jobs that create a value of a few times the em hardware cost. Strong global coordination to regulate copying could prevent this, but needs strong surveillance, which is hard to manage. Em slavery is possible, but low em wages remove most profits from slave ownership.
    6. Farewells
      1. Fragility
      2. Retirement
      3. Ghosts
      4. Ways To End
      5. Defining Death
      6. Suicide
      Brains, like other complex adaptive systems, become inflexible with experience adapting to particular environments. So within a subjective few centuries, ems become no longer competitive with younger ems and so must retire. Slow retirement is very cheap, but as with the naturally-slow humans, a slow retiree’s expected lifespan is limited by em civilization instabilities. Em retirees are like ghosts in many ways. Ems see making a copy who ends after doing a short task not as “death,” but as a part of them they choose not to remember. Ems usually have a right to suicide.
  3. Economics
    1. Labor
      1. Supply & Demand
      2. Malthusian Wages
      3. First Ems
      4. Selection
      5. Enough Entrants
      By the time 1000 humans have been scanned to make ems that compete for jobs, and 1000 useful mind tweaks are available, then almost all wages fall to within roughly a factor of four of the cost to rent em hardware and supporting utilities. Such “Malthusian” wages have described most humans before the industrial era, and almost all animals ever. While ems are “poor” in this sense, they need not suffer physical hunger, exhaustion, pain, sickness, grime, or unexpected death. The fraction of world income that goes to wages increases, and most wage premiums disappear; ems who are paid at all are paid about the same. The first scans are destructive, and of peak-career humans, while later scans are of young humans better able to learn.
    2. Efficiency
      1. Clan Concentration
      2. Competition
      3. Efficiency
      4. Eliteness
      5. Qualities
      The set of all em copies of the same original human constitutes a “clan.” Most wages go to the 1000 most productive clans, who are each known by one name, like “John,” who know each other very well, and who discriminate against less common clans. Compared with people today, ems are about as elite as billionaires, heads of state, and Olympic gold medalists. The em world is more competitive than ours in more quickly eliminating less productive entities and practices. This encourages more job punishment, less product variety and identity enhancement, and more simple functionality. Because they are more productive, ems tend to be married, religious, smart, gritty, mindful, extraverted, conscientiousness, agreeable, non-neurotic, and morning larks.
    3. Work
      1. Work Hours
      2. Spurs
      3. Spur Uses
      4. Social Power
      Ems are oriented more to work, instead of leisure, and tend to be workaholics, perhaps working 12 hours per subjective day. Ems are comfortable making “spurs,” i.e. copies created at workday’s start, and retired or ended by its conclusion. Spurs save a factor of two or three in costs, but forego chances to learn from tasks. Spurs do most em work, many spur tasks take a subjective hour, and spurs interact more with each other. Spurs can ensure privacy in counseling, auditing, and law enforcement. By putting two spur copies into a “safe” from which only one bit is returned, one em can show another “you’d agree with me if I could tell you what I know.” Ems act like people today who get and keep power, handicapping themselves less, selling themselves more, flattering bosses more and criticizing them less, and directly asking more for things they want.
    4. Business
      1. Institutions
      2. New Institutions
      3. Combinatorial Auctions
      4. Prediction Markets
      Competition induces not only new more efficient physical technologies, but also new more efficient social institutions. New em practices may include more pay-for-performance incentives, track records, context- dependence of prices, random juries of voters, sales of votes and citizenship, contingent or continuous donations to shared projects, and wider use of public key cryptography. Combinatorial auctions, where bidders specific the value they put on different allocation combinations, deal well with complex interdependent values while avoiding the inflexibility and lobbying that afflict centralized organizations. A subsidized prediction market induces financial speculation on a topic, which aggregates information into accurate price estimates. Decision markets, which predict outcomes given particular decisions, let decentralized pools of speculators inform key decisions.
    5. Growth
      1. Faster Growth
      2. Growth Estimate
      3. Growth Myths
      4. Finance
      Many myths circulate about factors that increase economic growth rates. For example, the fact that ems can run faster than humans should not much increase growth. Even so, the em economy grows faster than does ours because of stronger competition, computers mattering more, and especially because factories can make labor as fast as non-labor capital. An em economy doubling time estimate of a few weeks comes from the time for factories to duplicate their mass today, and from the historical trend in growth rates. In response, capital becomes less durable, and one-time-use products become more attractive. Clans become a unit of finance, private firms and hostile takeovers get more support, and asset prices more closely approximate the predictions derived from strong financial competition.
    6. Lifecycle
      1. Careers
      2. Peak Age
      3. Maturity
      4. Preparation
      5. Training
      6. Childhood
      At work, fast ems can coordinate more tasks, whereas slow ems can do fewer tasks better. Managers and programmers go fast, and managers have a wider span of control. Ems slower than kilo-ems suffer jobs changing before careers end. Most ems are subjectively old, near a job-dependent age when productivity peaks. As ems are older, they prefer peace to excitement and have stronger attachments, weaker gender differences, less lying, crime, and neuroticism, and more trust, composure, agreeableness, conscientiousness, and openness. By preparing or training one copy then making many copies, preparation and training get cheaper. Ems want low productivity variance in workers, but high variance among trainees. By watching the lives of older copies in the same jobs, ems can get a good picture of what their future lives will be like. Em children are rare and often like celebrities in that many future copies will know their life details.
  4. Organization
    1. Clumping
      1. Cities
      2. City Structure
      3. City Auctions
      4. Choosing
      5. Speed
      6. Transport
      Like us, ems gain by clumping together into cities. Unlike us, ems slower than kilo-ems can interact fully across cities without moving brains via virtual reality. This greatly reduces travel congestion, allows bigger cities, and puts most ems in a few big city-states. Iconic city locations are less about travel. City centers host faster ems and those doing interconnected tasks, although the very fastest are often in peripheries. City combinatorial auctions can substitute for centralized zoning and utility allocation, allowing em cities to deal with interdependencies quickly and flexibly. Working ems are faster than milli-ems, kilo-ems is the typical speed, and leisure usually runs faster than work. Em speeds clump, with a ratio between clumps near eight, and so cities may separate into regions for different speeds. Physical transport across a city seems very slow to kilo-ems, encouraging very local production, and hugely discouraging space travel.
    2. Groups
      1. Clans
      2. Managing Clans
      3. Firms
      4. Firm-Clan Relations
      5. Teams
      6. Mass vs. Niche Teams
      Ems trust their clans more than we trust families or identical twins. So clans are units of finance, liability, politics, labor negotiations, and consumer purchasing. To promote unity, clans avoid members arguing or competing. Em firms are larger, better managed, put more effort into coordination, have more specific job roles, focus more on costs relative to novelty, and have higher market shares and lower markups. Clan reputations and clans buying into firms promotes clan-firm trust, which supports locating employees at firms, using distinctive work styles, and focusing more on being useful instead of gaming firm evaluation systems. Em work teams tend to have similar social-category features like age but a diversity of information sources and thinking styles. In mass-labor markets, ems are created together, end or retire together, almost never break up, and mostly socialize internally. In niche-labor markets, associates coordinate less regarding when they are created or retire.
    3. Conflict
      1. Inequality
      2. Em Inequality
      3. Redistribution
      4. War
      5. Nepotism
      6. Fake Experts
      Compared with our world, em cities, families, firms, and labor market suppliers are more unequal. Individual wages are more equal, but speeds and lifespans are much less so, and indefinite lifespans can induce more wealth inequality. The sort of inequality that has most motivated historical redistribution, between the families of a nation, is low for ems, and redistributing on that basis transfers from poor to rich individuals. Poorer ems suggest more war but older ems suggest less. Big cities suggest fewer small wars, but more big wars. Although ems need not fear death, they’d fear war’s destruction. Clan nepotism is more problematic than is family nepotism to us, so ems need strong policies to suppress it. In particular, nepotism can undermine the reliability of abstract professional experts, pushing ems to rely less on such experts.
    4. Politics
      1. Status
      2. Governance
      3. Clan Governance
      4. Democracy
      5. Coalitions
      6. Curbing Coalitions
      Faster ems have many features that mark them as higher status, and the clumping of speeds creates a class system of distinct status levels. Strong central rulers are more feasible for ems, as leaders can run faster, put spurs in high-trust roles, and use safes to reassure wary citizens. Decision markets can help advise key government decisions, while combinatorial auctions can help to make complex interdependent allocations. The em world selects for personalities good at governing that same personality. Competitive clans and cities may commit to governing via decision markets that promote profit or long-term influence. One em one vote works badly, but speed-weighted voting seems feasible, although it requires intrusive monitoring. Shifting coalitions of em clans may dominate the politics of em firms and cities, inducing high costs of lobbying and change. Ems may try many policies to limit such clan coalition politics.
    5. Rules
      1. Law
      2. Efficient Law
      3. Innovation
      4. Software
      5. Lone Developers
      Law becomes easier for ems in some ways. Archived copies can clarify knowledge and intent. Surveillance, operating systems, and tiny physical ID tags can better protect property. Police spurs can study private info and only report legal violations found. If em law gets more efficient, it may use less prison and more blackmail, negative liability, gambled lawsuits, and prediction markets on court outcomes. As em growth depends more on increasing inputs, innovation matters less to ems than to us. But innovation still matters greatly, and ems may find better institutions to promote innovation, such as changing patents from property to liability. Low em wages would reverse recent trends and push software engineering away from abstraction toward performance. Limits on device speed would push toward parallel software, and greatly cut use of intrinsically serial software. Very fast individuals or small teams might spend their subjective lifetimes building a single software system.
  5. Sociology
    1. Mating
      1. Sexuality
      2. Open Source Lovers
      3. Pair-Bonds
      4. Gender
      5. Gender Imbalance
      As ems don’t need sex to reproduce, sex is left more to individual choice, and may be suppressed as in eunuchs. But demand for sex and romantic pair-bonding likely persists, as do many familiar gendered behavioral patterns. A modestly unequal demand for male versus female workers can be accommodated via pairs whose partners run at different speeds, or who use different ratios of spurs to other workers. Ems have spectacularly good looks in virtual reality, and are very accomplished. Open-source em lovers give all ems an attractive lower bound on relation quality. Clan experience helps ems guess who are good receptive matches. Having only one em from each clan in each social setting avoids complicating relations.
    2. Signals
      1. Showing Off
      2. Personal Signals
      3. Group Signals
      4. Charity
      5. Identity
      6. Copy Identity
      Ems show off their abilities and loyalties, although less than we do because ems are poorer and better-known to each other. Because speed is easy to pay for, ems show off more via clever than fast speech. Celebrities matter less to ems, and it is easy to meet with a celebrity, but hard to get them to remember you. Clans coordinate to jointly signal shared features like intelligence, drive, and fame. Clans fund young ems to do impressive things, about which many older copies can brag. Innovation may matter less for em intellectuals. Mind-theft inspires great moral outrage and charity efforts. Secure in identifying with their clan, most ems focus personal energy more on identifying with their particular job, team, and associates. It isn’t clear if em identity degrades continuously or discretely as copies get more different. Copy-events are identity-defining, and newly copied teams quickly create distinct team cultures.
    3. Collaboration
      1. Ritual
      2. Religion
      3. Swearing
      4. Conversation
      5. On Call Advice
      6. Synchronization
      Ems are likely to reverse our recent trend away from religion and overt rituals, perhaps via more group singing. Traditional religions can continue, but need doctrinal clarifications on death and sins of copies. Like high stress workers today, em work groups pushed to their limits swear, insult, and tease. Ems deal with a wider range of mind opacity and transparency, allowing mind reading within teams, but manipulating expressions to hide from outsiders. Clans can offer members life-coaching via voices in their heads, using statistics from similar copies, but teams may create unique cultures which limit the usefulness of that. Avoiding direct meetings helps clans bond better. Em relations are often in the context of similar relations between copies. At work, ems try more to make relations similar, to gain from learning and scale economics. But friends keep relations more different, to emphasize loyalty and natural feelings.
    4. Society
      1. Culture
      2. Divisions
      3. Farmer-Like
      4. Travel
      5. Stories
      6. Clan Stories
      Em culture emphasizes industriousness, work and long-term orientations, and low context attitudes toward rules and communication. Being poorer, ems tend towards farmer/conservative values, relative to forager/liberal values. So ems more value honor, order, hierarchy, religion, work, and less value sharing, consensus, travel, leisure, and variety. Sex attitudes stay more forager-like, however. Ems are divided like we are by geographic region, young versus old, male versus female, rich versus poor, and city center versus periphery. Ems also divide by varying speeds, physical versus virtual work, remembering the human era versus not, and large versus small clans. Ems travel to visit or swap with other copies of themselves. An exotic travel destination is other speed cultures. Like us, ems tell stories of conflict and norm violations, set in ancestral situations. Stories serve as marketing, with many characters coming from well-known clans. Em stories have less death and fast-action.
    5. Minds
      1. Humans
      2. Unhumans
      3. Partial Minds
      4. Psychology
      5. Intelligence
      6. Intelligence Explosion
      As humans are marginal to the em world, their outcomes are harder to predict. Humans can’t earn wages, but might become like retirees today, who we rarely kill or steal from. The human fraction of wealth falls, but total human wealth rises fast. Humans are objects of em gratitude, but not respect. Design complexity limits changes to em minds, and makes ems unlikely to completely lose big human capacities. More likely are reduced inclinations, such as for art, sex, or parenting. Ems may expand or reduce particular brain regions. Smarter ems specialize more in stable jobs, but less in other jobs. The most entrenched human mind features last longest: those supporting social interactions, embodying shared standards, and shared with most mammals. The em economy doubles many times before human level AI software appears, and then AI is quite unlikely to explode in one small place to suddenly take over the world.
  6. Implications
    1. Variations
      1. Trends
      2. Alternatives
      3. Transition
      4. Enabling Technologies
      5. Aliens
      During the em era, the em economy and population grow, computing costs fall, and typical speeds increase at first and then decrease. Humans hardly change. A few variations on this book’s main scenario are discussed briefly, including shallower brain emulation, faster AI software progress, clan-specific brain hardware, laws restricting em copying, a much larger number of viable clans, clan fragmentation, older peak productivity ages, larger changes in em brain sizes, smaller cities, more difficult computer security, quantum computing, mind merging, and stronger global governance. Ems require three enabling technologies. Each could be the last one ready: (1) computers-last gives an anticipated broad transition, (2) scans-last gives anticipated narrow transition, and (3) cell-models-last gives unanticipated narrow transition. The first cities to support ems, and the first humans to scan, gain advantages thereby. As some have asked this chapter answers: the em era says almost nothing about aliens.
    2. Choices
      1. Evaluation
      2. Quality of Life
      3. Policy
      4. Charity
      5. Success
      People usually evaluate distant futures via how warm and moral are their residents. The em world does well by these criteria, and also by utilitarian criteria of how many creatures find their lives worth living, as productive people are happier. Humans get rich fast, but are no longer the center of attention. Policy recommendations include speeding cell-modeling innovation, encouraging portfolio diversification, and integrating and mixing em and human institutions. Likely market failures include overly fast ems, underly dense cities, and excess negotiation costs. To promote good em policy, one should study policy, push for good policies, join with compatible groups, learn relevant skills, earn to pay others, and save to spend later. To survive in an em world, diversify your assets away from wages. To start one of the few highly copied em clans, when the em era starts be very productive or at the ideal young age with great promise.
    3. Finale
      1. Critics
      2. Conclusion
      Critics of drafts of this book have asked why they should care about a future not centered on them or their kids. Others say it is always impossible to foresee social change, or that we can’t foresee the social change of creatures smarter than us. Others wanted an analysis of the consequences of different technologies, or wanted to hear about the very distant future, not the next great era. Parents may disown children who sufficiently oppose their values. But surely parents should first try hard to see things from the kids’ point of view. Readers may be repelled by many features of the em era. This book asks only for that they try hard to see this world from its residents’ point of view, before disowning these their plausible descendants.
    4. References