The Age of Em: What If the Future Belongs to Copies of Us?
Reading The Singularity Is Nearer recently reminded me of another book that left a strong impression on me years ago: Robin Hanson’s The Age of Em: Work, Love, and Life when Robots Rule the Earth.
Both books imagine intelligence transforming civilisation, but the contrast is not artificial intelligence versus humanity. Hanson’s emulations would also be artificial minds: computational systems designed to reproduce the functional patterns of particular human brains.
The real difference is where the intelligence comes from. Kurzweil imagines increasingly capable AI merging with and expanding human intelligence. Hanson asks what might happen if the first machines capable of replacing most human workers are not designed from scratch. They begin as copies of us.
That change in origin matters. A civilisation of emulations would not be ruled by incomprehensible alien minds. It would be built from familiar human memories, habits and ambitions made reproducible at industrial scale.
A Future Built from an Assumption
Hanson’s central concept is whole-brain emulation. A particular human brain would be preserved and scanned in sufficient detail, its relevant structures interpreted and a computational model built that reproduced its patterns of signal processing. A successful em would retain the memories, skills, dispositions and habits of the scanned person closely enough to reason and work in recognisably human ways.
Every step in that description contains unresolved problems. We do not know exactly which features of a living brain must be captured to reproduce an individual mind. A map of neurons and synapses may be insufficient if the relevant state also depends on changing synaptic strengths, molecular processes, neurotransmitter systems, glial activity or interactions between the brain and the rest of the body.
The influential whole-brain emulation roadmap developed by Anders Sandberg and Nick Bostrom treats the project as a sequence of distinct challenges involving preservation, scanning, interpretation, modelling and simulation. Progress in one area does not establish that the complete chain will work.
Recent connectomics research makes both the progress and the scale visible. In 2024, researchers published a wiring diagram of an entire adult fruit-fly brain containing 139,255 neurons and 54.5 million chemical synapses. In 2025, the MICrONS project reconstructed a cubic millimetre of mouse visual cortex containing more than 200,000 cells and approximately half a billion synapses.
These are extraordinary achievements, but neither is a functioning emulation of the animal concerned. The mouse dataset covers a small region rather than a whole brain, while a human em would require not only tracing vastly more cells and connections but identifying which biological states must be reproduced for one particular mind to persist.
The Age of Em does not claim to solve those problems. Hanson deliberately places them inside the premise. He assumes that emulations become technically possible, cheap to copy and available before more flexible forms of human-level AI. He then asks what ordinary economics and social science might imply under those conditions.
The book is therefore best read as a conditional exercise: if this particular technology arrives first, and if institutions respond roughly as Hanson assumes, what kind of civilisation follows?
Human Minds Made Reproducible
An em would be artificial in implementation but human-derived in cognitive architecture. It would not begin as an alien intelligence learning to imitate human behaviour from the outside. It would begin with the memories and mental habits of a particular person.
Hanson assumes that ems remain within the broad range of human psychology even when their material existence changes radically. They would still care about status, security, friendship, recognition, useful work and control over their own lives. They would also inherit competitiveness, tribal loyalty, self-deception and the ability to rationalise systems from which they benefit.
This is what makes Hanson’s scenario more unsettling than many stories about machine intelligence. Its inhabitants do not need strange goals to create an alien civilisation. Familiar motives may be enough once minds can be copied, accelerated, paused and deleted.
The future would not become less human in its psychology. It would place human psychology under economic conditions no biological society has experienced.
When One Worker Becomes a Labour Market
The most consequential feature of emulation is not that a mind runs on a computer. It is that the mind can be copied.
A biological worker is scarce. Training a skilled engineer, researcher or manager takes years, and that person can perform only one full-time job at once. Once a useful em template exists, additional instances could be created whenever sufficient computing capacity was available.
One successful worker could become a team, then a department, then a substantial portion of a profession. The copies would initially share memories and skills, but they would begin diverging as soon as their experiences differed. One might learn from a failure the others never encountered; another might become exhausted, disillusioned or hostile to the organisation that instantiated it.
Copying creates scale without preserving one permanent collective person. Each branch becomes a separate worker whose shared past does not eliminate its distinct present interests.
Under Hanson’s baseline assumptions—cheap duplication, competitive markets and relatively little regulation—employers create copies while the value of another worker exceeds the cost of running one. Competition then pushes wages towards the price of computation, electricity, cooling and supporting infrastructure.
This result is sometimes described as labour collapsing to the cost of running a mind. It is better understood as the interaction between a technology and a particular institutional settlement. Restrictions on duplication, collective ownership of templates, licensing systems, strong labour protections or an enforceable right to refuse copying could produce different outcomes.
The economics are not independent of the politics. A mind is cheap to reproduce only after society has decided who may reproduce it, on what terms and with which obligations towards the resulting person.
Hanson also expects a relatively small number of highly productive human templates to generate enormous populations of copies. Once a particular mind performs slightly better under em conditions, replication amplifies the advantage. Groups of related copies could form “clans” that act as units of trust, finance, reputation, political organisation and labour negotiation.
This resembles automation, but with a different concentration of power. Conventional automation replaces workers with tools. Emulation could replace diverse workers with repeated instances of a small number of workers.
The original biological person whose brain supplied a valuable template might become wealthy or influential. Later copies could remember the same childhood while possessing none of the original’s property or control over further duplication. The central question is no longer merely whether one unusually capable engineer can be copied a million times. It is who controls the right to instantiate, employ, archive and terminate those million people.
Speed as Material Inequality
Because ems run on hardware, Hanson imagines that they could operate at different speeds. A faster em would think, work and experience more subjective time during each external hour, while requiring greater computational resources. A slower em would be cheaper to maintain but would experience the surrounding civilisation moving rapidly past it.
Subjective and objective time would separate. A biological human could watch a day pass while a fast em experienced years of work and conversation. An archived or retired copy might survive cheaply at a very low speed while barely participating in current events.
Speed would therefore become more than a technical setting. It would function as income, status and access. Executives or specialists might run quickly because accelerated thought allowed them to respond before competitors and coordinate larger organisations. Poorer ems might operate at speeds that made ordinary interaction with faster society difficult.
Hanson’s world does not abolish material inequality by replacing bodies with software. It translates inequality into computing capacity and therefore into unequal access to experienced time itself.
The effect complicates familiar ideas about work and leisure. A slow em may be inexpensive to preserve but effectively excluded from contemporary life. A fast worker may accomplish extraordinary amounts while also experiencing an equally extraordinary duration of effort. A civilisation could contain people living in the same external hour while inhabiting radically different social times.
One Past, Several People
Many discussions of mind uploading begin with questions Hanson largely brackets. Would an em be conscious? Would it be the original person? Would a destructive scan amount to survival, or to death followed by the creation of a convincing successor?
Hanson’s project is mainly social and economic. He assumes that sufficiently accurate emulations possess human-like experiences and asks what kind of civilisation they would build. That assumption allows the analysis to proceed, but it does not resolve the philosophical problem.
This is where Thomas Nagel’s What Is It Like to Be a Bat? remains relevant. A complete objective description of a system does not automatically establish whether there is something it feels like to be that system.
An em might insist that it was conscious. It could remember childhood, describe pain, fear termination and object to being copied without permission. From the outside, those responses might be indistinguishable from those of a biological person. Society would therefore confront a practical moral problem before the metaphysics had been settled: either the em experiences suffering and deserves protection, or we have created an entity displaying every meaningful sign of suffering while remaining uncertain whether the experience exists.
Copying creates a related identity problem. Suppose an em is produced while the biological person remains alive. Both remember the same past and can plausibly claim psychological continuity with the person who entered the scanner. A second em multiplies the claim. None possesses a hidden biographical fact that makes it uniquely authentic.
The copies may all continue aspects of the earlier person, but they cannot remain one person after their experiences diverge. Each becomes a separate centre of memory, preference and responsibility.
The useful question is therefore less “Which copy is really me?” than “What obligations exist among people who remember being the same person?” This is developed more fully in the essay on what happens when one remembered life becomes several people.
Hanson’s world turns those philosophical questions into administrative ones. Employers, courts and governments would require rules for consent to copying, ownership of templates, inherited contracts, liability among diverging copies, archived mental states and the termination of short-lived workers.
A copy created to complete a six-hour task might remember agreeing to the arrangement before the copying event. That shared memory would not automatically settle whether this new person consents to being ended when the task is complete.
The Power and Danger of Conditional Precision
The Age of Em is extraordinary because Hanson follows his assumptions much further than most futurists would dare. He considers cities, cooling systems, workplaces, clans, marriage, religion, law, retirement, sexuality, speed differences and the social lives of people who share memories.
The density of detail makes the world feel grounded. It can also make the opening assumptions disappear from view. Precise conclusions are not the same thing as confident predictions when they depend on a speculative technological sequence and a deliberately simplified institutional baseline.
If flexible human-level AI arrives first, there may be no distinct em era. If copying remains expensive, if embodiment proves central to cognition or if emulations receive strong legal rights from the beginning, the economy may behave very differently. Hanson discusses alternative assumptions, but the baseline scenario remains the structure that gives the book its force.
The reader therefore has to preserve a distinction between internal consistency and probability. The book does not establish that this civilisation will exist. It demonstrates that a few unusual premises, combined with familiar incentives, can generate a future stranger than most science fiction.
Kurzweil and Hanson do not ultimately describe opposite destinations. Both blur the boundary between human and machine. The difference lies in the route and in what remains stable during the transition.
Kurzweil expects technology to expand and transform the capabilities of human minds through AI and integration with machines. Hanson preserves something closer to present human cognitive architecture while changing the conditions under which minds are produced, copied, employed and discarded.
One scenario changes what a mind can become. The other changes what it means for a mind to be scarce.
Hanson’s version may be more disturbing because its inhabitants remain recognisable. They do not need alien values to produce an alien society. Human ambition, loyalty and self-justification may be sufficient once labour, memory and identity become reproducible infrastructure.
Copied minds would still need contracts, electricity, cooling, legal status and protection from unauthorised duplication. They would form friendships and rivalries, remember lives that no longer belonged exclusively to them and discover that technical immortality is compatible with poverty and powerlessness.
The future may indeed be transformed by intelligences that think very differently from us. Hanson’s alternative is that it could be transformed by minds that think almost exactly as we do, multiplied until the human person becomes a scalable economic input.
The machines would be built from us. That would not necessarily make the society they create humane.
Comments
Post a Comment