At a Rise of AI Virtual Chat in 2020, economist Robin Hanson discusses his book The Age of Em with host Fabian Westerheide. Hanson explains that the book takes one standard futurist assumption, that individual brains can be scanned and modeled cell by cell into a working emulation, and works out its social and economic consequences in detail rather than debating whether the technology is likely. He argues that a competitive emulation economy would select scans of only the few hundred best humans, make billions of copies of them, and push wages down to subsistence level, while the original biological humans who start out owning the capital remain comparatively rich as the economy doubles roughly every month. Hanson describes how emulations would concentrate into a few very dense cities, form status classes based on running speed, face limited career lengths despite being technically immortal, and likely gain broad practical autonomy because humans running a thousand times slower cannot supervise them closely. He also addresses opposition to the technology, sex and reproduction among emulations, and questions comparing his scenario to AI and to science-fiction depictions of uploaded minds.
Key takeaways
- Hanson assumes ordinary economic and social structures, such as firms, law and nations, continue largely unchanged in the age of emulations.
- The emulation economy would drive wages to subsistence level while the biological humans who initially own the capital stay comparatively rich.
- Because early emulations are unmodifiable "black boxes," operators can only copy, pause, delete or change the running speed, not edit them internally.
- Rights and autonomy for emulations would follow practical efficiency rather than moral consensus, similar to how slavery loses its advantage at subsistence wages.
- Running at roughly a thousand times human speed, emulations would split into high-speed and slow, low-status "retiree" classes.
Detailed notes
The assumption behind whole-brain emulation
Hanson says his book takes one common assumption from futurism and fiction, that a brain can be scanned in fine spatial and chemical detail and reconstructed from generic computer models of each cell type, and builds out its consequences as a proof of concept for doing more detailed futurism. The resulting emulation, or Em, reproduces the original person's input-output behavior when connected to artificial eyes, ears, hands and mouth, but Hanson deliberately sets aside whether it is conscious or "really" the same person; a freshly started emulation remembers being human and experiences itself as a continuous person. He assumes emulations are initially unmodifiable "black boxes": operators can copy, delete, pause or run them faster or slower, but cannot edit their internal, spaghetti-like structure. He also assumes ordinary economic and social institutions, firms, law, nations, continue to apply rather than dissolving into something wholly new.
A subsistence economy of copied minds
The emulation economy is competitive and capitalist, Hanson says, and selects scans from only the few hundred best humans in each profession, then makes billions of cheap copies of them. Because emulations can be copied so easily, competition drives their wages down to subsistence level, similar to pre-industrial societies. Biological humans, who own essentially all the capital at the outset, become the rich, largely non-working class living off returns on that capital, though a growing share of ownership over time shifts toward emulations who are closer to the economic action. Hanson stresses he is forecasting what a competitive market would produce, not what anyone wants, and that the book only covers the "age of Em" itself, an era of change comparable to an entire industrial, farming or foraging era compressed into roughly one or two subjective years, after which something else follows that he declines to predict.
Cities, work and the physical world of emulations
Emulations exist physically as computer processes in server farms, but experience a constructed virtual reality as real as a human's everyday surroundings, switching to a harsher, functional view of server racks, cooling pipes and power plants only for physical-world tasks. Because emulations can interact without physically moving, the usual traffic-congestion limit on city size disappears, so Hanson expects them to concentrate into one or a handful of very large, dense cities rather than spreading out. These cities would likely grow first out of existing data centers near human population centers, later creating conflict as emulation cities expand and displace nearby human communities, while humans remain spread more thinly across the rest of the earth.
Rights, autonomy and the risk of conflict
Hanson argues that rights are not determined by a world's basic structure but by what is efficient and what people with resources want; poor, subsistence-level societies, including his emulation world, mostly do what is efficient. Using slavery as an analogy, he notes that owning slaves offers little advantage once wages are already at subsistence, so emulations are likely to be granted extensive practical freedom because their work is too complex and too fast, roughly a thousand times human speed, for humans to control in detail. Maintaining integration between human and emulation institutions, he says, reduces the risk that emulations conclude they no longer need humans and simply take their assets; war remains possible in this world but not constant.
Status classes, inequality, gender and death
Because emulations can run at different subjective speeds, Hanson expects a class hierarchy in which high-speed emulations hold high status and slower, retired emulations become a low, "ghost"-like class. Individual emulations stay relatively poor, but inequality between clans, descendants of the few hundred original scanned humans, can grow large and compound over time because near-immortal emulations can preserve wealth-building habits that biological death normally resets. Emulations face a limited subjective career length, since minds appear to age independently of the hardware running them, after which they typically retire and run more slowly. Reproduction happens through copying rather than childbirth, which Hanson expects to make emulations more liberal about sex, though gender itself persists since brains are not deliberately modified, creating a "mating problem" because the emulation economy has no reason to preserve equal numbers of each gender.
Q&A highlights
Audience polls showed 66 percent supporting research into executable brain models, about half expecting emulations to gain human-like rights over time, and 70 percent expecting humans and emulations to coexist. Asked about opposition, Hanson points to workers displaced from jobs emulations take over, and to people who would call the initial, destructive brain-scanning process murder or suicide, a conflict he expects to sharpen once the emulation economy wants to scan children's brains to get more adaptable young minds. Asked whether he plans to write fiction, he says his book deliberately has no plot or characters so that everything in it stays logically consistent, unlike stories such as the Amazon show Upload, whose premise breaks down once its economic logic is followed through. On neuromorphic AI, he argues there is no single algorithmic key to intelligence, only many accumulated smaller algorithms, so emulations remain easier to predict than general AI. He also confirms that emulations would knowingly agree to periodic training simulations designed to test their loyalty and composure.
Quotes
The emulation economy is a competitive capitalist economy and it selects from all humans to scan and workers the few best.
Robin Hanson
My book is like science fiction except there's no plot and there's no characters and it all makes sense.
Robin Hanson
The emulations will basically have to be given a lot of autonomy because they're running the emulation world.
Robin Hanson
Questions answered in this talk
What is a whole-brain emulation ("Em") according to Robin Hanson?
According to Robin Hanson at a 2020 Rise of AI Virtual Chat, an Em is a computer model built from a fine-grained scan of one human brain combined with generic models of how each type of brain cell behaves, reproducing the same input-output behavior as the original person without being guaranteed to be conscious or exactly the same identity.
Would emulated minds have rights or be treated like slaves?
Hanson argues at the same talk that emulations would likely gain extensive practical autonomy because the emulation economy runs at subsistence wages, where owning slaves brings little advantage, and because the humans nominally in charge move far too slowly, about a thousand times slower, to supervise emulation work in detail.
How does Robin Hanson's Age of Em scenario differ from general AI?
Hanson explains that emulations are a narrow, more predictable subset of possible artificial minds because they reuse known human psychology, letting him forecast their economy and society in detail, unlike the far larger and less predictable space of general artificial intelligence.