What this covers

Russ Roberts questions Robin Hanson on whether a technological singularity belongs beside the usual scenarios of continued growth, collapse, or a stable middle path. Hanson starts from a point about perspective: life looks stable on human timescales, so people in earlier eras could not imagine farming or industry, and we have the same trouble now. He then explains why most celebrated candidates, such as surveillance, space elevators, cold fusion, and nanotechnology, would not qualify, since each touches only a small share of the economy. Human labor, which earns about 70 percent of US income, is the constraint that machines would have to break.

The economic reasoning takes up much of the middle. Hanson leans on David Ricardo's 1820 result that a machine substituting for a person caps the wage at the machine's rental cost, and he argues over whether human and machine labor are complements or substitutes. He borrows the immigration debate to show how supply elasticity drives wages. He also revisits the Malthusian trap: children once earned barely more than they cost, so land and sons were the better assets. Rapid population growth could therefore lower per-capita wealth even as total wealth rises, leaving owners of real estate, stocks, and patents far better off than wage earners.

The last stretch tests feasibility. Whole brain emulation needs scanning, cell models, and a very large computer. Hanson cites a scanned mouse brain and a solid computing trend, and he denies that any grand theory of intelligence is coming. Contested points include whether engineering the parts amounts to understanding the whole, whether emulations would keep memories and personality, and whether extrapolating three past transitions is anything more than numerology. Hanson also argues for the virtues of such a world.

Sharpest takeaway

Hanson argues that the historical pattern of rare, sudden jumps in economic growth makes a further transition to a doubling time of weeks plausible, and that the most likely route is whole brain emulation, which would turn labor into cheap copyable capital and reshape wages, wealth, and population dynamics.

  • Growth has jumped by roughly 200 times at each of three past transitions (human emergence, farming, industrialization), so extrapolating this pattern implies a doubling time of one to two weeks.
  • Whole brain emulation needs only scanning, cell models, and computing power, and computing power follows a reliable trend, so copying a human brain is a tractable route to human-level machine labor.
  • If labor can be copied cheaply, wages fall toward the cost of the machine, and wealth shifts to owners of non-wage assets, while per-capita wealth may fall as populations grow faster than output.

The claims · ranked66 claims · weighted by value

This asset isn't compiled yet

You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.

0.86

Life looks stable on human timescales, so people in earlier eras could not imagine the coming of farming or industry; the same blindness makes it hard for people today to conceive a radical transition.

causalhigh valueestablishednovelty 3/4durability 4/4· Robin Hanson

“before these transitions occur life looks really stable”

0.81

In agrarian history, land and sons were more valuable assets than children, because a child's earnings near subsistence barely exceeded the cost of raising them; with rapid population growth the return to labor would again fall relative to land and other scarce assets.

factualhigh valueestablishednovelty 3/4durability 3/4· Robin Hanson

“the wages of Labor fell relative say to the price of land”

0.78

Economists have long treated humans and machines as complements, but Ricardo's substitution result shows machines can substitute for people; the economics depends on whether the relative advantage curve across tasks is steep (complements dominate) or flat (substitution dominates).

causalhigh valuecontestednovelty 3/4durability 4/4· Robin Hanson

“that's the key sort of conceptual barrier to economists understand”

0.74

Engineering the components of the world is distinct from understanding the higher-level organization that emerges from them; we know what the world is made of but are far from knowing how to build a functional brain, as opposed to a functional replica of the Eiffel Tower.

factualhigh valuecontestednovelty 3/4durability 4/4· Robin Hanson

“we know what the world is made of we know what pieces they are”

0.74

For most of human history, per-capita wealth stayed flat because any growth in total output was absorbed by a growing population; only in the last few hundred years has growth been fast enough that per-capita wealth rose.

causalhigh valueestablishednovelty 3/4durability 3/4· Robin Hanson

“over most of human history when world product grew we spent that on more population and per capita wealth didn't change”

0.73

Labor earns about 70 percent of US income and has stayed roughly constant; because human labor, not capital, is the main limit on growth, machines that can substitute for people could break that limit and drive a large change in growth rates.

causalhigh valuecontestednovelty 3/4durability 3/4· Robin Hanson

“that number 70 percent or maybe closer to 67 in the United States is pretty close to a constant”

0.73

If machines can do most jobs, income from labor may decline, but a far richer world still leaves room for people who hold non-labor assets (real estate, stocks, patents), which can buy abundant output even if their ability to earn wages falls.

causalhigh valuecontestednovelty 3/4durability 3/4· Robin Hanson

“people who have other assets besides their ability to earn wages real estate stocks something else”

0.73

Most technologies people discuss, such as surveillance, space elevators, cold fusion, and nanotechnology, could be transformative but could not produce doubling every two weeks, because each affects only a small share of the economy (energy under 10 percent, manufacturing capital around 7 percent).

causalhigh valuecontestednovelty 3/4durability 3/4· Robin Hanson

“even then we only spend you know less than 10% maybe less than 5% on energy in the economy”

0.73

Ricardo showed in 1820 that when a machine directly substitutes for a person, the wage cannot exceed the cost of renting the machine, so cheaper machines push wages down toward that rental cost.

causalhigh valueestablishednovelty 2/4durability 4/4· Robin Hanson

“when a person and a machine are substitutes direct substitutes then the wage the person gets can't be any higher than the cost to rent the machine”

0.72

A mouse brain has been scanned at decent resolution, a thousand times smaller than a human brain, so scanning technology is roughly ready, though the industry is small and slow because demand is small.

factualhigh valueestablishednovelty 3/4durability 2/4· Robin Hanson

“we've actually done a whole mouse brain at a decent resolution done a scanning of an entire mouse brain the factor of a thousand smaller than human brain”

0.71

Technology and knowledge, not land or resources alone, were the key driver of population growth in human history, because population grew only when people figured out how to support more people.

causalhigh valueestablishednovelty 2/4durability 3/4· Robin Hanson

“it's really clear that the reason the population grew was because we figured out how to support a larger population and that's primarily technology”

0.70

Beyond the three standard scenarios (continued growth, collapse, or a stable middle path), a fourth scenario should be taken seriously: another dramatic transition comparable to the arrival of humans, farming, and industry.

normativehigh valuecontestednovelty 3/4durability 3/4· Robin Hanson

“I think you should in addition to taking all three of those scenarios seriously you should take a four scenarios here”

0.69

Emulation hardware inherits the fast cost declines of computing, so if an economy is built from such entities, all of its goods and services would fall in price about twice as fast every two years, implying a much faster growth rate.

forecasthigh valuecontestednovelty 3/4durability 2/4· Robin Hanson

“computer chips get you know twice as cheap every two years”

0.69

Whole brain emulation, which scans a brain and simulates its cells and connections on a computer, is the most likely near-term route to human-level machine intelligence, because it copies a working system without needing a theory of intelligence.

forecasthigh valuecontestednovelty 3/4durability 2/4· Robin Hanson

“this scenario which we've called whole brain emulation taking a whole brain and emulating on a computer requires three technologies”

0.66

Whole brain emulation requires three technologies: detailed scanning of brain structure and chemistry, accurate models of how each cell type responds to signals, and a very large computer; the computing trend is solid, while scanning has a smaller but decent trend and has already scanned an entire mouse brain.

factualhigh valuecontestednovelty 3/4durability 3/4· Robin Hanson

“we've actually done a whole mouse brain at a decent resolution”

0.66

There is no grand theory of intelligence likely to make AI simple; intelligence may be a large set of details like building a tiger, so progress will come from copying many pieces rather than discovering a single equation.

causalhigh valuecontestednovelty 3/4durability 3/4· Robin Hanson

“there is no grand theory of how to make a tiger there's just a lot of pieces”

0.65

Immigration may lower wages for competing workers, but the effect depends on how elastic the supply of immigrants is; with an upward-sloping supply curve the effect is limited, while an effectively flat supply curve would push wages down to the level of the marginal cost.

causalhigh valuecontestednovelty 2/4durability 4/4· Robin Hanson

“if it's really a flat supply curve if they there's a million”

0.64

Over the past century world economic growth has been a steady roughly 4% per year, doubling about every 15 years, but over the longer history growth has been steady punctuated by a few sharp transitions (human emergence, farming about 10,000 years ago, industrialization about 200 years ago).

factualhigh valuecontestednovelty 2/4durability 3/4· Robin Hanson

“growth in the world has world roughly grown every four percent a year doubling roughly every 15 years”

0.62

Models of individual brain cell types already exist for some cell types, and the remaining challenge for emulation is the large number and diversity of cell types, not the fundamental mechanism by which they work.

factualhigh valuecontestednovelty 3/4durability 2/4· Robin Hanson

“we have actually models of a wide range of cell types it doesn't seem that hard to models these cells but we just have a lot of cells to go through”

0.61

Historically per-capita wealth rose because reproduction technology was stuck; a transition that allows rapid population growth could make per-capita wealth fall even as total wealth rises, since total wealth is a ratio problem divided by the population.

forecasthigh valuespeaker onlynovelty 4/4durability 3/4· Robin Hanson

“if this next transition this next singularity as I suspect will involve a technology which allows the rapid increase in the population”

0.61

A brain emulation would copy memories and personality only if the scanning and cell models capture every relevant detail, because memories and personality are assumed to be encoded in the physical cells and their connections.

causalhigh valuecontestednovelty 2/4durability 3/4· Robin Hanson

“your memories the hard drive your memory and your personalities are encoded in these cells and where they are and how they talk to each other”

0.61

Computing hardware follows a solid exponential trend, so the computing requirement for whole brain emulation can be projected confidently, but the detail needed in cell models is uncertain.

factualhigh valuecontestednovelty 2/4durability 3/4· Robin Hanson

“the computing technology has a nice solid trend we've got this very solid computing trend”

0.61

At each major transition the growth rate rose by a factor of roughly 200 or more, and the doubling time shrank from about a quarter million years, to about a thousand years, to about 15 years, with each change happening in a much shorter span than the previous doubling time.

factualhigh valuecontestednovelty 2/4durability 3/4· Robin Hanson

“it's like a factor of 200 in growth rates at least”

0.59

Extrapolating the pattern of the three past transitions (with comparable step sizes and doubling counts per mode) implies the next transition would produce a world economy doubling every one to two weeks, while the timing is only loosely predictable, roughly within the next century.

forecasthigh valuefringenovelty 4/4durability 1/4· Robin Hanson

“what you get is roughly in the one to two week doubling time range”

0.58

Owners of scarce assets (real estate, patents, stock, and key raw materials or transport points) will get rich in the emulation economy, while those whose only asset is their ability to earn wages will lose value as copies compete.

forecasthigh valuespeaker onlynovelty 3/4durability 4/4· Robin Hanson

“if you own things like real estate patents stock if you own things that whose value doesn't decrease as they get more competitive”

0.57

Emulations start with the same personality, motivations, and thought habits as the original person, so the social and economic behavior of emulations can be analyzed using what we know about humans.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“these boxes aren't just machines that are smart they are machines that are smart in the same way we are”

0.57

The emulation scenario is one small region of a vast space of possible minds; general AI scenarios are poorly specified and hard to analyze, but emulations are easier to reason about because they resemble familiar human minds.

normativehigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“the whole brain emulation scenario is one way to pick a place in that bath space”

0.57

Once machines can be copied cheaply, many competing entrepreneurs will each make copies of themselves to save labor costs, so competitive labor in the same skill market gets bid down until wages approach the cost of renting one machine copy.

forecasthigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“if one person does that then none of the rest of us can really earn more than ten cents an hour”

0.57

Prediction and betting markets could be used to forecast the timing of AI and other future developments, with subsidies needed because people prefer to bet on short-term outcomes; visionaries are unreliable forecasters because they tell listeners what excites them.

normativehigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“all of these sorts of things about what might happen in the future could be illuminated and elaborated in betting market prices”

0.57

Brain emulations would inherit the personality, motivations, and thought habits of the person scanned, so they would be predictable in ways other AI designs are not; they would also be cheaper and faster over time, could be immortal by swapping parts, and could be copied at scale.

forecasthigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“these boxes aren't just machines that are smart they are machines that are smart in the same way we are”

0.57

Under emulation, if population grows faster than wealth, the world enters a Malthusian regime where most copies live near subsistence, even though total wealth is much larger, and luxuries such as preserving sunsets would be bought only by a small rich minority.

forecasthigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“it becomes much more of a Malthusian type scenario where per capita wealth Falls”

0.57

Machine bodies would not need a healthy biological ecosystem, so a world dominated by machines would be much less motivated to preserve the environment beyond parks and zoos.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“if we have machine bodies if the world becomes dominated by machine bodies then they won't need the environment in the same way we do”

0.57

Copies made through emulation would be selected in a competitive market: firms would sell copies whose terms, hours, and pay are acceptable and cooperative, so the surviving copies are those who accept the deal, which is why economics can predict the conditions of such a world.

forecasthigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“you want to find people who will fill these boxes who will make a good arrangement”

0.57

Human reproduction technology has been stable for about 2 million years, which is why per capita wealth rose in recent centuries; the next singularity would likely be one where population can grow rapidly, removing that stable constraint.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“humans bodies haven't changed very much in the last 2 million years”

0.57

A copyable emulation can be run faster or slower, swapped part by part so it could be effectively immortal, and copied to produce a million instances, so emulations differ from humans in having cheap, unlimited copies.

forecasthigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“you can make copies well once you have one you can have a thousand you can have a million”

0.57

Individuals in an emulation world could choose between many copies of themselves (a larger clan) or one richer copy, and because the most aggressive copiers would dominate the population, the per-capita average would be dominated by those who copy the most.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“the mathematics of the per capita thing is dominated by those people”

0.54

Brain emulation can work without understanding the brain's overall design: if you know how each cell type behaves and copy the connectivity, the whole system's input-output behavior is reproduced, just as software can be ported to a new machine by emulation without knowing how it works.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“you don't have to know the design and the overall organization of the basic you know architecture you just have to know how to pieces work and copy those”

0.53

Hanson personally wants this transition to happen, since a world with far more wealth and many more people living lives worth living is better, even though some individuals would be disadvantaged and it would not be his own life.

normativehigh valuespeaker onlynovelty 3/4durability 2/4· Robin Hanson

“it's a world with vastly more wealth and vastly more people who find life worth living”

0.52

Animals were domesticated by selecting the few species that cooperate with humans; humans were similarly domesticated, and designed AI could be controllable in the same way, though it is unclear whether we can design controllable minds.

causalhigh valuespeaker onlynovelty 2/4durability 3/4· Robin Hanson

“we've actually been very selective and only domesticated a very small number of animals the ones who are most cooperative”

0.52

Hard-to-predict technologies can be anticipated partially by taking known trends like Moore's law and asking of each candidate technology whether it could produce a doubling every two weeks, and most candidates like surveillance, space colonization, and cold fusion fail this test because their direct effects are limited or slow.

normativehigh valuespeaker onlynovelty 2/4durability 3/4· Robin Hanson

“there's lots of really big important technologies that could come along that could do a lot and be important but just not at this level”

0.52

Individuals should diversify their wealth beyond wage earnings, holding real estate, stocks, or other assets, because wage income may disappear in a world where capital can double rapidly; in that world, even people with few assets could benefit from others' transfers.

normativehigh valuespeaker onlynovelty 2/4durability 3/4· Robin Hanson

“I'd like people to sort of realize that they won't be able to make money on wages forever and make sure they have other assets besides their ability to earn wages”

0.52

The economics of a labor market dominated by copies looks more like software economics: high fixed training costs and very low marginal copy costs, so wages would be set by the rental price of software rather than the marginal cost of a machine.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“the larger labor are cut cut economics looks more like software economics then”

0.52

There is no physical or logical reason growth timescales must match human lifetimes; growth rates have been slow only because the relevant parameters happened to fall near human timescales, so much faster growth is possible.

causalhigh valuespeaker onlynovelty 3/4durability 3/4· Robin Hanson

“there's no physical or logical reason why I just couldn't be a lot faster”

0.49

Hanson favors population growth, arguing that more people generate more creativity, ideas, trade, and economies of scale, and that environmental concerns about sustainability are less pressing to him than to the average environmentalist.

normativehigh valuespeaker onlynovelty 2/4durability 2/4· Robin Hanson

“I'm a fan of population growth which is minority view but I think more people are good”

0.47

The wealthy will likely own capital, patents, real estate, and key raw material inputs, so in a much richer world the people who own scarce inputs gain disproportionately, and even a small fraction of that wealth can buy a great deal.

forecasthigh valuespeaker onlynovelty 2/4durability 3/4· Robin Hanson

“whoever owns those inputs then owns wealth and even owning a small fraction of that wealth can make you be able to buy a lot”

0.47

Ability to earn wages behaves like a patent: its value declines as competition from cheaper, better alternatives increases, so people cannot count on labor income as a permanent source of wealth.

causalhigh valuespeaker onlynovelty 2/4durability 3/4· Robin Hanson

“your ability earn wages is more like a certain kind of patent it depends on how competitive you are”

0.47

Brain emulations would be cheaper to run if they throw out unnecessary detail, but the process of trimming a copy may break it, because you cannot know which details are irrelevant without understanding how the system works.

causalhigh valuespeaker onlynovelty 2/4durability 3/4· Robin Hanson

“you can make some random changes most random changes it can make a mess can make it worse”

0.44

Historical predictions that abundance would end scarcity and work have been wrong, because people invent new wants; wealthy people today still work hard, choosing work for non-wage reasons, which suggests wage levels become a minor consideration for the rich.

factualcontestednovelty 2/4durability 3/4· Robin Hanson

“the leisure class in our society they're pretty busy folk they work hard”

0.43

Because US manufacturing output is high even though it employs fewer people, technology has made the economy rich by stripping labor out of production, so machines doing everything would leave little for humans to do in the current economic sense.

factualestablishednovelty 1/4durability 3/4· Robin Hanson

“one of the ways we've gotten rich over the last 50 years is by stripping labour out of technology”

0.42

Traditional economists since Wicksell in the early 1900s saw humans and machines as complements, which implies that when machines become more valuable, wages rise, a view Hanson says misses the substitution case.

factualestablishednovelty 2/4durability 3/4· Robin Hanson

“the more traditional view of economist over the last century starting at least with WIC cyl and the early 1900's was that humans machines were complements”

0.40

Economic models of a labor market with copies have a flat supply curve, so an unlimited supply of copies brings the price down to the cost of producing them, but if supply is upward-sloping, the wage effect is smaller.

causalestablishednovelty 1/4durability 4/4· Robin Hanson

“if it's really a flat supply curve if they there's a million there's a wormhole bringing aliens”

0.40

Most top scientists would say a brain is just chemicals and electricity with no room for anything else, though Roberts notes this is not universally accepted and that the question of whether the whole is merely the sum of its parts remains open.

factualcontestednovelty 1/4durability 4/4· Russ Roberts

“I don't think everyone accepts that do you think”

0.35

Computing cost trends were well established early on: Moore's law projected rapid declines in computing cost, which is why people could foresee that a pocket device would match a room-sized computer decades later, whereas broader technological changes are harder to foresee.

factualestablishednovelty 1/4durability 3/4· Robin Hanson

“there was what was called Moore's law there were these trends and computing costs”

0.32

Among most top scientists, the brain is understood as chemicals and electricity with no room for anything else, and physics has a complete picture of the basic particles that make up the world, so a physical-only view of the mind is the default scientific position.

factualcontestednovelty 1/4durability 3/4· Russ Roberts

“most top scientist if you didn't survey questions will say yeah you know that's it you're a brain and this has just chemicals and electricity”

0.32

Wages of people who work for pay would be unaffected by abundance only if they hold other wealth, because if wages are small relative to total wealth, the wage becomes a small consideration in choosing activities.

causalspeaker onlynovelty 2/4durability 3/4· Robin Hanson

“if the amount you get paid is really small compared to the other wealth you have that's a small consideration”

0.31

Doubling every one to two weeks would not be noticed in ordinary life, because the economy adapts so people don't experience the changes as disruptive, and fast growth is hard to see from within an ordinary lifetime.

causalspeaker onlynovelty 2/4durability 2/4· Robin Hanson

“one of the reasons the world can change so fast is that can change a ways you don't see an interfere with your life”

0.31

Brain emulations could be trained in a way analogous to employees; their attractiveness to buyers depends on accepting pay and hours, so market selection will tend to produce copies that are docile enough to work, and emulation brings the traditional AI issues of rebellion and rights into a more tractable frame.

normativespeaker onlynovelty 2/4durability 2/4· Robin Hanson

“the reason people buy these boxes this is are smart and capable and do lots of stuff”

0.31

Roberts argues that an emulated person would not be an owned object like a TV, because it has a mind, so questions of whether a copy can be controlled or turned off raise moral issues that the economic analysis does not settle.

normativespeaker onlynovelty 2/4durability 2/4· Russ Roberts

“I understand why emotionally I might imagine them being different”

0.30

Current machine progress, while remarkable, has not produced a singularity: computers still represent only a few percent of world income and humans still earn the vast majority of income, so the trend is slow.

factualestablishednovelty 1/4durability 2/4· Robin Hanson

“computers as a percentage of world income you know down a few percent if that range”

0.30

Cars that drive themselves are technically close: Google has run field trials, and the remaining barriers are mainly regulatory and liability, not technical.

factualestablishednovelty 1/4durability 2/4· Robin Hanson

“Google's actually done a set of a field trials with cars the drive”

0.29

Economists should act as humble policy analysts: first forecast what the world will be like, then identify which marginal shifts to push toward, such as encouraging people to build non-wage assets, because individuals overestimate their influence.

normativespeaker onlynovelty 2/4durability 3/4· Robin Hanson

“a humble policy analyst thinks in marginal terms”

0.26

Brain emulation makes three different strategies plausible for copy-makers (creating kids-like copies, owning a personal assistant copy, or using copies as factory labor), and these uses lead to different labor-market and growth implications.

factualspeaker onlynovelty 1/4durability 2/4· Robin Hanson

“so are all three of these gonna be happening at the same time”

0.26

Capable machines would substitute across many services (haircuts, legal work, scripts), not just specific factory tasks, so a broadly capable robot economy would be far wider in scope than narrow automation.

forecastspeaker onlynovelty 2/4durability 2/4· Robin Hanson

“we're talking about machines that would do things now like give you a haircut”

0.25

The idea that technology wants and has a mind of its own can be understood only in an emergent-order sense, where natural processes cannot be fully controlled, and it is a weaker claim than emulations or AI having minds of their own.

definitionspeaker onlynovelty 1/4durability 3/4· Russ Roberts

“the idea that the technology can quote have a mind of its own and I understand that within it within the sense that I understood him to mean it which is an emergent order sense”

0.25

Hanson is skeptical that the mind is only physical chemistry, and he admits he is open to a view that mind is more than matter, which he calls an unobservable issue that is hard to imagine as physical.

factualspeaker onlynovelty 1/4durability 3/4· Russ Roberts

“being a religious person I'm capable of imagining it's something it's not observable so that's one issue”

0.21

Roberts summarized that the discussion concerns labor economics with a new labor supply curve, and that the usual labor-economics insights still apply.

factualspeaker onlynovelty 1/4durability 2/4· Robin Hanson

“so we're talking basically labour economics but we're talking labor economics with a new supply curve”