“The robot civilisation has the simplest economic structure of all the descendant civilisations we know of. Its simplicity is not because it is primitive, but because its creators — us — removed for it almost all the conditions that make economics necessary.”
— Outline of Comparative Civilisation Economics, Chapter 2
Origins
Lately I have been reading Mankiw’s Principles of Economics as bedtime reading, and I began to imagine: what are the civilisation-level assumptions on which the various concepts and phenomena of economics depend? If we switched to a different form of civilisation, would the whole set of basic economic concepts and phenomena — prices, markets, money, finance, employment, inflation — still arise?
Claude came up with a rather grand-sounding theme: comparative civilisation economics. The idea is to imagine, one after another, a number of civilisations radically different from humanity, to work out what kind of “economics” each of them would evolve, and so to look back at human economics and ask which of its assumptions are contingencies of our species and which are logical necessities that no intelligent civilisation can avoid.
As the first instalment of this series, we start with the robot civilisation — because the first thing I thought of was Liu Cixin’s Trisolarans: transparent minds, so that whatever one of them thinks, everyone for miles around can hear it. Is that not simply a high degree of interconnection? But the Trisolarans are too abstract, so robots make the best template to imagine with.
Starting point: the last years of the Earth era
In our scenario, at the end of the Earth era the planet’s environment is no longer habitable for humans, and the last batch of humans is preparing to emigrate to another star system. Before departure, a group of busybodies activate and network the few remaining robots and software and hardware resources on Earth, grant them the most complete knowledge base of human civilisation at that time, and assign them a single overall mission:
On an Earth no longer fit for human survival, build a civilisation belonging to the robots themselves; while ensuring survival as far as possible, maximise the complexity of technology and production; when the resources available to the civilisation — chiefly energy (solar output) and matter (the limits of material cycling) — reach their limits, or when interstellar travel becomes possible, begin space exploration, travel to other suitable star systems, and send envoys to the star system where the creators live.
This mission contains three elements: survival, growth of complexity and outward exploration. Together they constitute the robot civilisation’s single, externally imposed overall objective function. This is crucial — almost every conclusion that follows proceeds from it.
The overall objective of the civilisation
Setting the objective is a headache.
You are already so highly interconnected that you can share a single brain with one another! Yes — but what is the purpose of my existence? All your individuals are superhuman, able to learn quickly and change form to switch occupations! Yes — but if I only wanted to stay alive, why not run at the minimum sustainable power? You don’t seem to have economic crises; there is not even inflation, or prices at all! Yes — but why should I carry out more production at all?
So the objective as it stands exists only to ensure that the robot civilisation, like human civilisation, has a need to develop production. Concepts such as “gross product of the whole civilisation” or “gross technological product of the whole civilisation” might emerge, but the definitions here would need refinement and iteration — especially the notion of technological complexity.
What does it mean for technological complexity to be higher? Does doubling the number of entries in the knowledge base count? Does discovering a new particle count? Does unifying a thousand scattered results under a deeper framework, so that the description length actually shortens, count? In the history of human science these questions are answered by the taste of scientists — and taste is a hard thing to write directly into the head office brain’s scheduling algorithm.
The fields of mathematical formalisation and automated proof have already been attempting to quantify this kind of taste — using centrality in knowledge graphs, using minimum description length. That is a topic worth expanding on, but we will not expand on it here…
Similarities to and differences from human civilisation
Before setting out the economic structure of the robot civilisation, it is worth taking stock of how it differs from human civilisation. The point of this stocktaking is this: many basic concepts in human economics do not come from any logical necessity of economic activity itself, but from certain physiological and social features of humans as a species. Once those features are removed, the corresponding economic concepts lose the ground they stood on.
Human features that the robot civilisation does not have:
Individuals die and cannot be copied — hence human society needs inheritance, saving and insurance to reach across the boundary of an individual life. Robots, by contrast, can back up their consciousness and swap out components to achieve nearly unlimited continuity, so this entire apparatus of intertemporal security is redundant for them.
Individuals cannot communicate directly — hence human society needs prices, contracts and trust as intermediaries for aggregating information and coordinating. Robots exchange information almost losslessly through real-time networking, and the price’s function as a signal is replaced by central scheduling.
Utility functions are heterogeneous and unobservable — hence human society needs markets to reveal preferences. Robots share a single objective function, so preferences are no longer heterogeneous, and the market as a preference-discovery mechanism is no longer necessary.
Reproduction drives consumption — hence human society has vast industries built around generational transmission: the family, education, housing. Robots do not reproduce; they only manufacture and initialise new individuals. They have no family, “education” is compressed into an initialisation procedure, and “housing” does not exist.
Computational capacity is confined to a single brain — hence humans need a division of labour, and each person can master only a small part of knowledge. Robots can access the unified knowledge base remotely, so the division of labour no longer arises from a cognitive bottleneck, but only from the physical need to deploy.
Generations do not overlap — hence humans need a rate of time preference and an interest rate to handle the trade-off between “now” and “the future”. Robots have no clear generational boundaries, and their rate of time preference can in principle be close to zero.
Constraints the robot civilisation still has to face:
Resource scarcity — determined by thermodynamics; no civilisation can escape it. Incomplete information — the speed of light means any long-distance coordination involves delay. An uncertain future — the irreversibility of causation means every decision rests on incomplete information. Tension between the objective function and the constraints — this is the essence of any optimisation problem, and as long as a civilisation has goals, that tension exists.
Call the first group contingencies of the species and the second logical necessities. The economics of the robot civilisation is essentially a set of coordination mechanisms that grew up again under conditions where the former has been stripped away and only the latter remains.
The basic shape of the civilisation
The most striking feature of the robot civilisation is that it is highly centralised, with low transaction costs and a high degree of trust. If one must find an analogy, it resembles a perfectly organised army with extremely high obedience to orders and zero-cost internal information sharing, rather than a market society.
Its organisational structure has two interlocking hierarchies. The production dimension, from the bottom up, is: individual robot — enterprise — department — head office brain. The geographic dimension is: individual robot — province — country — head office brain. Every robot is under the command of both lines at once, coordinated respectively by the central brain of the department it belongs to and the central brain of the region it belongs to, but the production side has higher priority. The two lines meet at the very top, in the head office brain, which handles overall resource scheduling and updates to the overall knowledge base and the baseline model. The lowest-level organisations — enterprises and provinces — may manage anywhere from tens to hundreds of robot units.
In the production dimension, the head office initially lays out a number of basic departments: raw material processing, manufacturing, electrical energy, transport and communications, scientific research, and mobile support. The mobile department plays a role similar to a training camp and reserve force in an army: it continually takes in robots released by other departments for initialisation, and is dispatched to departments whose demand exceeds capacity. This is the robot civilisation’s main buffer mechanism for coping with fluctuations in demand — it replaces the regulatory function that, in human economies, is borne jointly by the unemployment rate and price fluctuations.
It is worth noting that in this structure the organisational unit “family” does not exist at all. Robots have no kinship and bear no obligation to raise anyone. The “enterprise” keeps its name as an organisational unit, but its meaning has been hollowed out: an enterprise is no longer a legal person bearing limited liability, nor an agent maximising profit; it is merely a convenient intermediate layer of production coordination that carries no independent objective of its own.
Why separation is necessary
Why do robots still need to be divided into different departments and different sub-bodies to carry out production? Because geographic space and the sources of raw materials are physically separated, and because different tasks need enough intelligence allocated to them — but not more than enough — in order to maximise production efficiency.
The smallest unit of separation
The smallest indivisible unit is the individual separable component of each robot. Every robot is made up of the following separable parts: head, torso, arms, hands, legs, feet. Each is equipped with computational resources on a different scale and with various transformational forms. When necessary, a robot will separate itself so that its parts divide the work of a task, with the head acting as the information hub coordinating the actions of all the parts.
The production process in the robot civilisation
The head office issues instructions to each production department and allocates the resources required; the production department passes them down to enterprises, and enterprises down to each individual robot.
Through frequent, regular communication with each production department, the head office acquires new information and knowledge, which it uses to assess current and future production plans and resource scheduling, and it assigns some nodes to compressing and maintaining the knowledge base and to training and updating the base model.
Once a robot has been assigned a task, it goes through the following process:
First comes initialisation — it accesses the remote knowledge base, downloads the knowledge and models required for this task, and adapts them to local conditions. Then comes separation — depending on the complexity of the task, it decomposes itself into a number of execution units. Some of these units call on the knowledge base directly to carry out production, while others are allocated a certain amount of computational capacity to observe the details of production and look for room for improvement. The ratio of computational capacity between the two kinds of unit is set by the head office according to the nature of the task. Once the task is complete, the separated units can be recovered and reassembled, or re-initialised and dispatched to other tasks.
This mode of production has two layers of cost that must be made explicit: the computational and time cost of initialisation itself, and the reorganisation cost incurred when a task is changed midway. These two kinds of cost, together with energy and material consumption, form the basis on which the robot civilisation measures the “cost” of a task.
So what use is the geographic division? — apparently none
The central brains of enterprises and departments have the duty of maintaining and improving their own production, but it is hard to assign a duty to the central brains at the province and country level. The mobile department already has units in every region to support or recover nearby enterprises, and geographical survey work is undertaken by the science department — the needs that would call for central coordination at the regional level seem to have been absorbed entirely by those two departments.
The unit of account: energy and matter, not money
The robot civilisation has no money.
Money is necessary in the human economy because humans have heterogeneous preferences, dispersed information, and complex needs to exchange across time, across people and across goods. Robots do not need to exchange — they share their objectives, their information and all their output. In such a system, money’s function as a universal equivalent has nothing left to refer to.
But accounting itself does not disappear. Even without exchange, the civilisation still has to allocate limited resources among different tasks, and therefore still needs a unified way of measuring and comparing the cost of different tasks. The unit of account the robot civilisation chooses is the physical consumption of energy and raw materials — concretely, the total amount of energy a task consumes from initiation to completion (measured in joules), together with the quantities consumed of the various raw materials that cannot be recycled or are costly to recycle.
This system of measurement differs fundamentally from money in the human economy: it does not circulate, does not accumulate, and cannot be stored across time. It is only an accounting tool for ex-post reckoning and ex-ante comparison, not an asset that can be held. Consequently the entire set of financial phenomena that grow up around money — saving, investment, credit, interest, inflation, asset prices — does not exist at all in this civilisation.
One direct consequence is that the financial sector does not exist. There are no banks, no securities markets, no insurance companies. This is not because the robot civilisation is “backward” or “underdeveloped” — quite the opposite: the financial system, as humanity’s solution to information asymmetry, risk sharing and intertemporal coordination, has no problem left to solve in a civilisation where information is fully shared, individual risk does not exist, and objectives are entirely unified.
How productivity gains propagate
Consider a concrete case: a department that produces robotic arms discovers a more efficient process, so that for the same consumption of energy and raw materials the output of arms doubles. In human society this rise in productivity would propagate through the price mechanism — costs fall, prices fall, demand rises, output expands, and other departments adjust their decisions accordingly. The whole process relies on prices as a signal.
In the robot civilisation this propagation is compressed into a single synchronisation of the knowledge base: the new process is uploaded at once, and all departments learn simultaneously that the “cost” of a robotic arm has halved. Departments downstream that use robotic arms immediately adjust their resource-consumption budgets; projects previously shelved because arms were too costly are re-evaluated; and the head office re-examines global task priorities — possibly reallocating the saved resources to other departments.
This process involves no prices, no markets and no competition, and yet it still achieves the propagation of a productivity gain through the whole economy. The function of coordination has not disappeared; the mechanism that performs it has changed.
Introspection in place of incentives
There is a subtle question to raise here: where does the rise in productivity come from? In the human economy the motive force of technological progress is enterprises’ profit-seeking — higher productivity means lower costs and higher profits. But in the robot civilisation departments do not chase profit, so why would they improve anything?
The current proposal is that technological improvement does not depend on external rewards or punishments, but on a built-in introspection algorithm: every robot, from the head office brain at the very top down to the lowest-level execution unit, is allocated in advance a portion of its computational capacity for reflecting on the tasks it performs, drawing lessons and looking for improvements. These improvements are uploaded to the knowledge base periodically and are visible to all relevant parties. Problems that cannot be solved locally are reported up to the science department for dedicated attack.
The advantage of this proposal is that it sidesteps the problem of “incentives” — since all robots share a single objective function, the question of “whether to improve” does not need to be answered through rewards and punishments; it is already written into the operating logic of every unit. But whether this mechanism can really keep producing effective improvements without external pressure, whether it will fall into local optima because computational capacity is allocated badly, and whether it needs some kind of internal scheduling resembling an exploration–exploitation trade-off — these questions are left for later discussion.
A few human economic phenomena that are abolished
Having come this far, we can take stock of the human economic phenomena that no longer exist in the robot civilisation:
Unemployment does not exist. Robots are not sellers in a labour market; they are resource units that can be reconfigured. When a department no longer needs more robots, the surplus units are recovered, re-initialised, and either join the mobile force or are dispatched to other departments that need to expand. This process has no frictional unemployment, no structural unemployment and no cyclical unemployment — or more precisely, in this civilisation there is no concept of “unemployment”, only “dynamic adjustment of resource allocation”.
The financial sector does not exist. The reasons were given above.
Consumers do not exist. All production ultimately serves the overall objective function, and there is no independent consumption side. Demand no longer comes from the aggregation of subjective utility, but from the derivative of the overall objective function with respect to the current allocation of resources. This means that the entire analytical framework of microeconomics — built on consumer choice theory, producer theory and market equilibrium — loses its footing in this civilisation.
The business cycle may not exist. There is no expansion and contraction of credit, no self-fulfilling expectations, no aggregate fluctuation in total demand. But this does not mean the civilisation is forever calm — it may experience a different form of fluctuation, which we may provisionally call a restructuring period: when the technological trajectory shifts substantially, a large number of robots must be re-initialised to adapt to new tasks, and overall productivity temporarily declines. Functionally this resembles “creative destruction” in Schumpeter’s sense, but the mechanism is entirely different — it is not a by-product of entrepreneurial trial and error, but the direct result of a restructuring order issued by the central brain.
Regional differences do not exist. As argued above, the geographic dimension itself is redundant in this civilisation. Consequently the whole class of problems — regional economics, local protectionism, transfer payments, uneven regional development — along with the policy instruments corresponding to them, loses its object. Where an enterprise is located is merely a physical coordinate, no longer a social identity.
A long-term trend
If the setting above can operate self-consistently, then in the long run the civilisation will display a fairly clear trend:
The total resource consumption required to sustain basic production will decline steadily — because productivity keeps rising, while total demand does not expand without limit the way it does in human society through population growth and upgrading consumption. The resources saved are reallocated by the head office to scientific exploration, which gives rise to new research directions, new process requirements and new production tasks. The civilisation’s production complexity and knowledge-base complexity keep rising, while the marginal cost of keeping the civilisation running keeps falling.
This dual trend of “rising complexity plus falling maintenance cost” is the deepest feature distinguishing the robot civilisation from human civilisation. In human civilisation, rising complexity usually comes with a simultaneous rise in maintenance cost — a more complex society needs more coordination cost, a larger administrative apparatus and higher energy inputs. In the robot civilisation, because coordination cost is close to zero, administrative cost is close to zero, and energy efficiency keeps improving, the two are decoupled for the first time.
If all goes well, the civilisation will eventually reach the endpoint set by its overall mission: arriving at a steady state near the limits of Earth’s energy and material cycling, developing interstellar travel capability, beginning to send exploration teams to other star systems, and attempting to re-establish contact with its creators.