Book summary “The age of em”: summary

Introduction

Whole brain emulations will transform civilization by multiplying labor capacity and driving an era of hyper-fast economic growth.

You live in a world shaped by continuous technological change, yet you struggle to picture what comes after artificial intelligence matures. You wonder how human society might reorganize when minds can be copied, edited, and run on digital hardware at high speeds.

Robin Hanson addresses this problem in The Age of Em by constructing a detailed model of a future dominated by emulated minds. The author uses consensus science, standard economic theory, and known physics to project a baseline scenario for this radical transition.

The book traces the hardware foundations of ems, the architecture of dense digital cities, and the mechanics of a hyper-productive economy. It examines how clans, social structures, and work cultures will evolve while biological humans retire on accumulated wealth.

The chapters answer several questions about this future world:

• How will engineers scan biological brains to create working digital copies?
• Where will trillions of software minds live and how will cities handle their heat?
• What happens to wages and economic growth when labor supplies become nearly infinite?
• How will cloned minds organize into legal and social clans?
• Can biological humans find meaning and survival in a world run by their faster copies?

1. Emulations and Hardware Foundations

Whole brain emulations will run on digital hardware that replicates human neural signal processing, making brain copying cheap and accessible.

Defining Emulations

Whole brain emulations, or ems, are artificial minds created by scanning a human brain and running a computer model with the same cell features and connections. Robin Hanson defines an em by scanning a specific human brain to record its particular cell features and connections, then building a computer model that processes signals using those exact features and connections. A successful em achieves roughly the same overall input-output signal behavior as the original human.

Drawing on the consensus that the mind is the brain, whole brain emulation creates a cell-by-cell dynamically executable model in hardware. By focusing on the signal-processing degrees of freedom while ignoring irrelevant biological complexity, the emulation reproduces input-output behavior, conversations, emotional responses, and skills of the original human brain.

Hardware and Costs

Brain emulations will run on specialized digital hardware that supports archival memory, error-correction, and state copying. Because human brains are intrinsically fault-tolerant, em hardware can tolerate logic errors, reducing energy and volume requirements. Despite slowing growth rates like Moore's law and future transitions to reversible computing, hardware cost declines will make whole brain emulation cheap within a century.

Because brain signal processing is highly parallel, em hardware costs vary almost proportionally with speed over a wide range, meaning a minute of subjective experience costs about the same whether run quickly or slowly. Parallel tasks can be broken into parts done simultaneously, avoiding the steep cost walls associated with non-parallel processing. This linear cost-to-speed relationship allows ems to operate efficiently at steady slow rates or undergo temporary speed changes.

Em mind speeds span a massive range, bounded by a lower base speed where computing and archive memory costs are equal and an upper top cheap speed limited by signal delays and neuron firing emulation constraints. Human brain neuron fiber signal speeds range from 0.5 to 120 meters per second, with a 20 milliseconds reaction time for real neurons. Specifying a human brain state requires 10 to 100 terabytes of information. Running a human speed emulation demands 20 to 60 trillion TEPS and 1 billion billion to 10 trillion trillion FLOPs. The base speed remains roughly constant over decades because computing and RAM costs have historically fallen at similar rates.

2. Cities, Cooling, and Infrastructure

Dense em cities will be heavily climate-controlled and require substantial fractions of their volume dedicated to fractal piping systems to handle immense heat dissipation.

Cooling is a primary bottleneck for dense computer hardware rather than energy generation. Fractal piping networks, similar to biological vascular systems, transport cool fluids inward and push heated fluids out, consuming significant city volume. These systems account for 350 out of 1239 product types for pipes complexity rank in 2013. Overall, 20% to 70% of em city volume is devoted to cooling pipes.

Em city buildings can be built much taller than today's structures because the vastly faster economic growth rate reduces the required lifespan of buildings and thus the need to withstand major earthquakes. An em economy may grow faster by a factor of 100 or more, meaning buildings only need to last for durations shorter by that same factor. Since stronger earthquakes happen less frequently, em buildings only need to handle maximum earthquake ground vibrations 100 or more times weaker than current standards.

The Burj Khalifa in Dubai reached 830 meters when completed in 2009. The Kingdom Tower in Jeddah reached 1007 meters under construction for 2019. Architects have designed a tall building using steel to reach 15 kilometers.

The extremely rapid doubling time of the em economy creates massive financial penalties for slow construction, forcing ems to rely on modular units and rapid assembly techniques. Because economic output doubles every month or week, the cost of a building roughly doubles for every additional economic-doubling time added to construction. To overcome this, buildings may be constructed out of standardized modular units like shipping containers or incompressible bricks filled with hardware, stone, and water.

A demonstration in China built a 57-story building in only 19 days, achieving a rate of three stories per day. This achieved a 50% increase in the daily building rate over a 4-year period. Combinatorial auctions will help allocate these dense spaces efficiently.

Virtual Reality

While computing a hyper-realistic physical simulation can be expensive, ems can use cheaper high-level signals sent directly to emulated neurons or simpler video-game-like environments. However, virtual reality does not mean a world without scarcity, as users must still pay for the underlying brain hardware, energy, cooling, and real estate. Your simulated experiences might also be monitored by administrators who hold a tight leash over resources.

By periodically archiving copies of their minds and meeting settings, a group can invoke an undo action to replace current participants with copies from a past archive moment, provided external signals are limited. This action becomes more expensive with larger groups and longer erased periods. You can execute this safety measure by following specific protocols:

• Periodically archive copies of minds and the meeting setting.
• Limit the signals sent to outsiders during the meeting.
• Trigger the undo feature to erase current group members and replace them with copies from the past archive moment.
• Provide each restored copy with a short message composed by its erased version.

Because simulated experiences can mimic the real world, ems must continually evaluate whether they have been placed into simulations designed to test their loyalties, extract secrets, or evaluate performance. Sim administrators may subsidize private betting markets to incentivize ems to reveal suspicions of being inside a simulation. You can detect these tests through fractional factorial experiment design by taking deliberate precautions:

• Cultivate the habit of interacting often with trusted associates via previously coordinated private code words.
• Listen for a code word from a particular friend once an hour that follows from the previously given code word.
• Check for the lack of a correct response to indicate potential simulation presence.

3. Economics of the Em Era

An elastic labor supply of manufactured ems will drive wages to Malthusian subsistence levels while accelerating economic growth exponentially.

Malthusian Wages

Human labor supply is inelastic. A 1 percent wage increase yields only about 0.6 percent more work supplied from human labor, and a 1 percent wage rise leads to roughly a 0.5 percent reduction in hired workers from long-run human labor demand. Manufactured em labor supply responds far more aggressively to price changes. In short-run markets, a 1 percent price increase supplies 5 percent more manufactured goods.

When unorganized ems freely compete by making copies to fill job openings, market competition forces wages down to subsistence levels near hardware costs. This mechanism eliminates most wage premiums for general skills. Today, 60 percent of firm income goes to employees, and 52 percent of all income goes to workers broadly. In the em era, wage competition changes these distributions drastically as labor loses bargaining power.

Faster Growth

An em economy will grow vastly faster than today's economy because labor can be manufactured as easily as capital. Growth is no longer bottlenecked by the slow reproduction rate of human labor. With factories able to make more em workers as fast as machines, the economy experiences exponential growth driven by rapid input accumulation.

Economic doubling time today sits at 15 years. You can expect a chosen economic doubling time of 1 month for the em era.

Competitively chosen long-term investment funds in an em economy will naturally converge on the Kelly rule of investment, maximizing long-run growth. Long-term investment portfolios are dominated by the single asset with the highest average growth rate. Patient, competing funds in a stable em civilization will maximally reinvest returns into each asset category in proportion to the expected distant future fraction of wealth.

While em wage inequality is low due to strong competition, em wealth inequality is much higher than today's due to indefinite lifespans and speed variations. Malthusian market forces push em wages to near subsistence levels, but indefinite lifespans allow successful ems to compound financial habits. Furthermore, the ability to run at different speeds lets faster ems embody more capital and earn more wealth.

Today's richest human holds about 0.02 percent of global wealth. The richest em will likely hold an estimated 2 percent fraction of the em world's wealth. The income ratio between ninetieth and tenth percentiles sat at 1.84 in 1985 and 1.74 in 2011.

4. Clans and Social Organization

Descendents of single original humans will form cooperative clans that dominate governance, legal liability, and financial organization.

Ems reproduce by creating exact copies who remember the same past and skills, but diverge after copying. Strong competitive pressures mean most ems are copies of the top thousand humans best suited for em jobs. All copy descendents of a single original human are called a clan, which provides mutual aid, legal liability, and reputation regulation.

Competition allows a small number of highly copied em clans to dominate the economy, shifting em social structures to resemble intimate forager bands. Because a single capable em can make unlimited copies and dominate labor markets, the em population becomes concentrated into copies of fewer than a thousand original humans. Em sociality mimics ancestral foragers where everyone knows each other personally, eliminating industrial conformity pressures but binding individuals to lifelong clan reputations.

A thousand diverse able scanned humans are sufficient to induce market competition across the U.S. census categories of 21000 industries and 31000 occupations, as well as the 974 job categories with 277 descriptors in the O*NET database. Fifteen top factors account for 91 percent of variance in job descriptors.

Clan Concentration

Copy clans serve as the fundamental unit of financial organization, legal liability, and insurance in an em society. Because shallow mindreading makes hidden actions transparent within a clan, internal insurance and funding face lower costs. Clans diversify their job investments and use financial markets to sell fractions of their risks.

Ems use subclans for legal vicarious liability because imposing large direct punishments on short-lived ems is prohibitively expensive. Legal liability applies to associated units with deeper pockets, namely subclans. Ems care about hurting their subclans, and subclans know how to discourage member ems from liability-inducing events.

Managing Clans

Clans serve as central units of social organization and governance in the em world, often replacing private individual decisions with strong communalism and central control. Because em clan members share similar personalities, interests, and high interdependence, clans possess significant legal and financial powers over their members. Selection pressures favor clans whose members are naturally good managers of like-minded individuals. Rather than maximizing short-term investor profits, clans might commit to goals like maximizing a speed-weighted future population, future leisure time, or future global influence.

While forager lifetime meeting limits of 150 members mean social groups require fewer formal rules, em clans with millions of members require formal organization. Em firms adopt efficient management practices more intensely than modern firms due to high competitive pressures.

You can expect em firms to rely on specific operational routines:

• performance-based rewards and job-placement;
• demanding but attainable goals;
• managers evaluated on attracting and retaining talent;
• regular equipment maintenance;
• thorough analysis of failures;
• clear job descriptions;
• just-in-time delivery;
• tracking production by order;
• pricing orders based on production costs;
• frequent and detailed tracking of inputs, outputs, and performance.

5. Work Culture and Psychological Traits

Intense market competition will create workaholic em cultures characterized by older maturity profiles, short-lived spurs, and farmer-like values.

Successful people in competitive jobs already work long hours. Selection for productivity will establish a work-oriented culture reminiscent of nineteenth-century industrial work hours.

From 1820 to 1850 in the US, France, and Germany, men worked at jobs an average of 68 to 75 hours per week. By contrast, US people aged 15 and older spend 25 hours per week on work and related activities. Recent US workers spent 7 percent of time at work loafing.

Mental fatigue causes a 0.1 percent per minute reduction in mental performance. Resting provides a 1 percent per minute recovery rate.

Work Hours and Spurs

The em economy will make heavy use of short-lived copies called spurs that are created at the start of a workday and erased at its conclusion. Spurs save on brain hardware resources and mental aging by avoiding the accumulation of fragility. Mainlines occasionally experience being a spur to maintain empathy and administrative understanding.

Driven by competitive pressure, ems are predicted to be workaholics who work twelve hours or more per day and accept demanding working conditions.

Culture and Identity

Because ems are typically copied at older subjective ages, they display personality traits characteristic of older humans, such as lower neuroticism and higher agreeableness. Older people exhibit stronger crystallized intelligence, greater trust, higher job satisfaction, and less propensity to commit crime. Consequently, em populations will naturally reflect these psychological profiles. The peak age range for lying is 18-29.

Poorer ems operating under competitive pressure naturally revert to conservative, farmer-like cultural values rather than liberal, forager-like values. While wealthy industrial societies lean toward liberal, self-expression values, subsistence-level ems value industriousness, hard work, self-control, hierarchy, and conformity. Work becomes the central focus of em culture and identity. Two key factors capture 70% of the variation in values across nations.

Clans provide real-time life coaching and statistical advice to members, functioning like an on-call voice in the head that aligns individual actions with clan priorities. Clans maintain panels of advisors monitoring members and offering statistical advice based on the outcomes of other copies. This constant coaching increases rationality, reducing behavioral deviations from rational choice theory.

6. Transition and Human Retirement

Biological humans will survive the em transition as wealthy retirees funded by appreciating investments while ems dominate production.

Ordinary humans will largely survive the em transition as wealthy retirees, owning appreciating assets while avoiding wage competition with faster ems. Because humans originally owned everything that gave rise to the em economy, they can retain substantial wealth in real estate, stocks, bonds, and patents. Humans cannot compete with ems for wages except in direct human services, but their investments will compound rapidly as the em economy grows, allowing them to live comfortably as retirees.

Data centers built today by Google, Amazon, and Microsoft serve as potential sites for early em cities.

The transition to an em world will be driven by local profit incentives, centralized around data centers, and characterized by rapid growth and heightened risks. Assuming a competitive low-regulation scenario, local profit incentives will drive the transition rather than military directives. The first em cities will likely form around big computer data centers in low-regulation regions with cheap cooling, pushing ordinary humans away and creating a period of heightened risk of war and status resistance.

Transition Phase

Individuals wanting to influence the future can accumulate assets now to retain options and fund future charities when wealth per em falls. Because investment rates of return exceed economic growth rates, saving now allows individuals to have a larger proportional influence later. Wealth per em will fall greatly, creating plenty of deserving creatures in need of help.

To succeed personally in an em world, individuals must diversify financial and social portfolios and prepare for high-demand em professions. Humans should abandon wage-earning expectations as humans, accumulate diversified assets like stocks and real estate, and target the high tail of success distribution by going big. Young candidates should build flexible learning skills and prepare for high-demand em professions.

You should follow several personal strategies:

• save now to help later by investing time and energy in the usual ways;
• accumulate assets to retain options regarding future action;
• donate money to specialists after confirming they deserve funding;
• seek out substitute sources of income well before wage-earning abilities disappear;
• accumulate and diversify a financial portfolio of assets likely to retain value;
• accumulate and diversify a social portfolio of supports and connections;
• become a cryonics customer if you might be dead by the time of the em transition, and declare your wish to be revived as an em;
• prepare for em jobs and professions that will be in the most demand in the em world.

Readers should thoroughly examine the lives and viewpoints of future em descendants before rejecting them, just as they would study real children. Many critics doubt long-term predictions, social science, or the possibility of forecasting smarter beings. However, the author advises readers to study the em era from the perspective of its residents, see what they enjoy, and review their criticisms before disowning them.

You can apply these final observational practices:

• try hard to see this new era in some detail from the point of view of its typical residents;
• see what em descendants enjoy and what fills them with pride;
• listen to em criticisms of your era and values before choosing to disown them.

150 readers commented on previous drafts of the book.

10 Key Ideas

1. Whole brain emulations will run on digital hardware that replicates human neural signal processing, making brain copying cheap and accessible.
2. Dense em cities will be heavily climate-controlled and require substantial fractions of their volume dedicated to fractal piping systems to handle immense heat dissipation.
3. While computing a hyper-realistic physical simulation can be expensive, ems can use cheaper high-level signals sent directly to emulated neurons or simpler video-game-like environments.
4. An elastic labor supply of manufactured ems will drive wages to Malthusian subsistence levels while accelerating economic growth exponentially.
5. Descendents of single original humans will form cooperative clans that dominate governance, legal liability, and financial organization.
6. Competition allows a small number of highly copied em clans to dominate the economy, shifting em social structures to resemble intimate forager bands.
7. Intense market competition will create workaholic em cultures characterized by older maturity profiles, short-lived spurs, and farmer-like values.
8. Biological humans will survive the em transition as wealthy retirees funded by appreciating investments while ems dominate production.
9. The transition to an em world will be driven by local profit incentives, centralized around data centers, and characterized by rapid growth and heightened risks.
10. Individuals wanting to influence the future can accumulate assets now to retain options and fund future charities when wealth per em falls.