Essay: The State as a Computational System

1. Abstract

Recently, there's been an explosion in public discourse around AI systems. One strand of this argues it for taking lessons learned from computer architecture and predictive modeling/machine learning and applying them elsewhere. Most notably, these lessons have entered the studies of the human mind - psychology and neuroscience. It is becoming common to use the components of predictive modeling as a toy description for what is happening in the brain. In this narrative, human perception and experience are described as a continuously modeled hallucination on top of our senses, calculated with a brain-based neural network. While in these contexts it is unknown whether it will be more useful to use the “computational framework” in a metaphorical or literal sense, the former allows for intriguing opportunities in its application to a wide ranging set of problems.

In this essay, I suggest we consider the state - that is, the organizational structure that people use to govern themselves and their societies, alongside its ruleset, enforcement, etc - as a notional computational/modeling architecture that consists of the following functions; (a) compute, (b) read/write, (c) data, and (d) objective function. I will make the case that this architecture framework can more clearly focus political debates towards productively addressing the merits and deficiencies of real-world implementations of states. In the application section, I will offer an analysis of popular political divides as characterized within this framework.

2. The State

Let us restate the above informal definition of a state: an organizational structure that people use to govern themselves and their societies. In this paper, we will discuss states, societies, and economies as largely overlapping and inseparable. While there is ample opportunity for interesting discussions about the state, society, or economy each in a narrow sense, in practice the boundaries between these are not clear enough for effective separation. Furthermore, when we consider the impacts of globalization, it may no longer be justifiable to suggest that the footprints of states, societies, and economies come close to lining up. Constructs such as the EU exemplify that states and economies may not be the same “size”, and that an economy may include multiple states, or vice versa. For this essay, we will simplify things by considering the state in the way such things are understood in common discussion, as an entity of government, a people, a society, an economy, etc, all wrapped up together.  

3. Computational Architecture and the State

Computer architecture is a very practical study, but it is possible to describe an abstract representation of computers defined by the functions that comprise them. 

1. Compute

Computer systems include a central and general-purpose compute function, think a CPU in a personal computer, which allows for mathematical and logical calculations and directs the other functions of the system. The compute function allows the computer to take as an input an arbitrary program of instructions, process these instructions, and output results. For our purposes, consider the metaphor of the CPU loosely, and also include similar accessory components such as GPUs, etc, that help with specialized cases of computation but are mostly similar. The compute function is judged by the speed at which single tasks are accomplished alone, and the span of parallel tasks that can be accomplished at one time.  

States likewise possess a compute function defined by the nature and locus of decision-making. Any state includes some processes by which and bodies in which decisions are made, and generally this corresponds to the legislative and regulatory functions. The core government of the state may also share these functions with society at large, as shown by decision-making performed by individuals and organizations. 

2. Read/write

Computation relies on a read/write function that allows the computer to read data into the compute engine and then write the outputs once they are complete. In a personal computer, this corresponds to the ensemble of RAM and storage. More abstractly, the read/write function is the ability for information to flow into the compute engine and for information to flow out of it. If we loosen this definition a bit, we can include peripheral sensors (such as cameras or microphones) or outputs (such as a printer).  Thus, the read/write function allows the computer to sense and interact with itself and the wider world. Read/write capabilities are typically judged by the speed at which these tasks are performed, and the faster the better. 

States interact with the world in myriad ways depending on their implementation. Most read information from the world through security and intelligence organs, perhaps through elections or opinion polls, or through socioeconomic metrics. They write to the world through taxation, military action and policing, and policy-making. 

3. Data

Particularly in a modeling setting, a computer system must have access to a volume of quality data. Data is much-hyped as a buzzword at the time of this writing, but it's true that greater quantities and qualities of data lead to more insightful and accurate results. Data abstractly is a representation of information that the computer can recognize and work with to produce something useful, such as a decision, a prediction, or additional data. Judging data quality is more nuanced and situation-dependent, but typically data that is large in terms of number of observation instances (big data) and data that is wide-ranging in terms of the breadth of included possible scenarios (variance) is considered “good”. Furthermore, data is judged on its cleanliness, the degree to which it is relatively free of nonsense and noise. Relevance and timeliness are further common-sense metrics of data quality.

The data that is available to states varies widely across time and space. Today, some societies are highly digitized and some states have high visibility through surveillance of their societies, while others are dominated by black markets and secret or illicit activity. Some states have privacy or search/seizure laws and others do not. Abstractly, states desire data on their constituent populace, economy, and organizations as well as the same for foreign allies and rivals. It is with this data that states make laws, green-light operations, or pursue policies. Note that data can be both digital and analog, although digital data is now generally preferred due to computers. 

4. Objective function

Also more suited to the modeling setting, a computer or model must have an objective function. The objective function provides the computer instructions and goals for what to compute and why. We may lump together the notion of a computer program that instructs the steps a computer must follow, and that of an ML modeling objective function, which provides a goal without necessarily laying out the steps to follow. These can be changed by the “user.” Judging the quality of an objective function is difficult and open to interpretation. In the modeling sphere, a model attempts to be “accurate” in its predictions compared to reality, but measurements of this notion of accuracy abound with competing pros and cons. But determining if an objective function has a “correct” or “desirable” goal is up to the perspective of the user. We could take a more global view of objective functions and consider what side-effects its pursuit might entail, such as the paperclip AI example. 

States are often described as driven by desire for maintaining stability, increasing its power, and/or maximizing its people’s welfare, among other things. Furthermore, states are imbued with competing or dominant ideologies that color its goals and methods. Unlike with computers, there is no apparent “user” that controls the state to perform some task, but rather this user either doesn’t exist or arguably the users are the citizens and residents subject to the state’s sovereignty or the individuals occupying positions in the state's bodies. Many mechanisms exist to arrive at an objective function for the state, such as elections, revolutions, etc. Most states have an overarching ideology and goal, but to varying degrees of consensus. 

4. Application 

I believe this framework helps to discuss the pros and cons of competing systems. For example, capitalism was implemented in earnest in the 18th and 19th centuries as individuals accumulated unprecedented private wealth through industrialization in the West. While it is common to this day to debate whether capitalism leads to the most desirable outcome, it is at least undeniable that it has achieved the greatest increases in productivity and consumption in human history. By that metric at least, it was successful from the beginning. Writers and philosophers such as Adam Smith extolled the “invisible hand” of the market, the emergent phenomenon that the collection of individuals making their own self-interested decisions leads to greater productivity and wealth overall. This line of thinking, combined with its visible successes, has been inherited by successive generations into the present and is now married to a fundamental libertarian and individualistic attitude. People in these systems feel that it is morally right for decision-making to be decentralized, and in order to protect this individual decision-making power, property and privacy rights have emerged through formal and informal means. States that were absolutist imperialist monarchies at the start of this transition to capitalism gradually moved towards representative democracies coupled with strong corporations, representing a clear decentralization of decision-making. 

Consider the world in which capitalism emerged - by today’s standards a large and opaque place. Almost all individuals within a state and their actions were invisible, and the fog of distance obscured events with long time delays. Continuing from the earliest human civilizations, the largest human endeavors were taxation, war, and religion, all of which usually performed by the state. Max Weber’s original definition of a state - a monopoly of the legitimate use of violence in a region - reflects this. A small court or aristocracy made decisions for the masses, and typically a hierarchy of decision-making lords shared authority among themselves.

In our computing framework, this world represents one with very little available analog data and no digital data. As is the case today, people had access to analog personal data on their little worlds - their local region, their family, their trade. Larger scale activities were feudally organized. First, local decision makers could access and process the available analog personal data locally. Then up the chain, decision-makers would progressively consolidate the data and decision-making up to the apex ruler. These decision-makers constitute a compute system in a world with very little available compute power. Human brains, particularly uneducated ones, have many strengths, but in terms of raw intellectual calculation power they leave a lot to be desired. Such brains cannot access and process huge quantities of intellectually-useful data. Even outside of the state hierarchy, the populace all privately computed their way through their lives with little outside help. Overall, computation was highly modular. In order for each compute node, and the system as a whole, to function, each node was required to have enough raw power and data to model their specific business or industry or life and to make local predictions. The market and pricing framework acted as the principle read/write function for supply and demand of goods in those models.

When new large-scale activities developed in the industrial revolution, they likewise developed as hierarchical modular computation centers such as corporations or companies separate from the state. This likely occurred out of necessity, for arguably the compute and data needs of these new parallel activities - such as running industrial factory systems - were beyond the capability of a single state computation center to manage. There may have been no way to efficiently process such wide and disconnected data without loss and no way to coordinate the autonomous modular compute nodes of the state’s leaders. It is therefore reasonable to expect a system resembling a capitalist one to emerge in this environment. What is not necessary is for a self-serving philosophy to grow up with this system that reinforces the system with morality and ethics, which did occur and is the dominant philosophy we have in capitalist societies today. This paper makes no comment on the ethical or moral dimensions of capitalism, but rather argues that the system was reasonable given the data and compute available at the time of its development. 

One way to tackle this is to consider the objective function of the system. Arguably the objective function for a capitalist society is profit-seeking. Because computation in capitalism is modular and parallel across nodes of individuals and corporations, this profit-seeking is specifically profit for individuals rather than profit for the whole collective. This tends to increase productivity and consumption, which typically improves the quality of life of producers and consumers, but also tends to increase greed and inequality. Furthermore, there may be blind-spots from the perspective of the individual that are important to the whole - these are typically called negative externalities in economics.   

Now let’s consider the world of the late 19th century and early 20th which had largely adopted capitalism. This is a world that has accumulated complex organized systems that more effectively marshall their computation and data resources. States and corporations were more organized through bureaucracy. However, data availability and computation capabilities  improved since the advent of industrialization due to improved technology. Data was better collected, indexed and processed by organized bureaucratic systems relying on formal paperwork. Data transmission was more frictionlessly and quickly communicated via formal postal systems and new communication technologies like the telegram or telephone. As a result, any compute node had access to more data. Later in this time frame, the first electrical systems and first computers were developed and adopted. Data and compute were becoming more and more centralized due to the simple fact that it was now more possible to do so and that doing so produced greater productivity results and wealth. 

With such marked improvements and increased centralization of the two, it is entirely reasonable to expect people at the time to aspire to test the boundaries of what might be accomplished. This happened with the emergence of socialism as a response to capitalism. Socialism is a system that proposes to eliminate private ownership of capital in favor of collective or often state ownership. Many, but not all, socialists proposed eliminating markets and prices as a method for organizing production and instead sought to centrally-plan economic production, usually at the state level. Socialist thinkers and writers extolled ideals of equity such as “from those according to their ability, to those according to their need.” Just as with capitalism, there are many flavors to socialism, but the most commonly implemented socialist systems correlated with authoritarian or totalitarian states exercising public ownership and economic planning. As with capitalism, this paper does not express an opinion on the ethics and morals of socialism. However, it is possible to claim that the objective-function for socialism is to equitably increase total public welfare via collective ownership. 

Within the computing framework, socialist states do away with modularized compute nodes such as corporations and absorb them into the state’s computation function. This centralized compute node can then access all data feeds across the economy and coordinate these feeds to attempt to produce a single global plan. As of the establishment of the major examples of socialist states, the Soviet Union and People’s Republic of China, the primary computation engine remained human brains organized in efficient bureaucratic grids and data existed as economic statistics that were numeric but analog. Computers existed at the time of these state’s establishment, but remained relegated to specific scientific use-cases for some time yet to come, and their raw power remained puny compared to those of the present day. It follows that, for a centrally-planned socialist state economy to function, that state must have access to sufficient computational power and data to be able to create a faithful global model of the society and economy writ-large and perform accurate predictions using that model. Compared to the local models of corporate or individual actors in capitalism, this socialist global model was clearly more complex and demanding. Due to the strong apex leaders that frequently sat atop these state bureaucracies, the decision-making capabilities of these states often boiled down to the capabilities of the apex leader alone. 

Taking the Soviet Union and China as examples, the two failed to achieve long-term increase to public welfare, or to production and consumption. While moral or ethical debates between capitalism and socialism abound with no clear winner, there is a clear winner with respect to economic growth in output. The Soviet Union produced five-year economic plans with quotas that did not match realities on the ground and overly prioritized military expenditure. Due to unilateral decisions with respect to food production by leaders like Stalin and Mao, mass famines occurred. The socialist state in the Soviet Union failed and was replaced with a particularly oligarchic and kleptocratic brand of capitalism. I would argue that the Soviet Union failed because its golden age existed and ended at a time, prior to the 1990s, with insufficient compute power and data collection capabilities. A centrally planned economy state was not reasonable or feasible. 

Westerners observed the end of the Soviet Union with a perception of their own system's superiority and confirmation that their system is the only one that has ever been successful, but more importantly is the only one that could ever be successful. As mentioned before, the capitalist system developed in a time when no other reasonable option existed, but developed a self-serving philosophy that proposed that no other option existed due to moral and ethical reasons rather than pragmatic ones. The failure of capitalism’s primary alternative seemed to confirm the pragmatic side, but also fed the ideological side to the West’s faith in capitalism, and saw the proclamation of “the end of history.” 

However, China provides an interesting counterpoint. The socialist regime in China was instituted well after the Soviet Union, and the economic conditions on the ground at its establishment were much more backwards, necessitating a long period of growth prior to reaching maturity or parity with most of the rest of the world. Until recently, China has grown mostly through a familiar “catch-up” effect. All of this is to say, critically, that China’s socialist regime came to maturity after the 1990s. As discussed above, for the full lifetime of capitalism and socialism up to this point, the notion of “computation” and “data” had been dominated by human brains and bureaucratic organizations, and analog numeric or descriptive data, all of which is fairly limited. In the 1990s, three trends began that fundamentally shifted this reality. 

First, the computer. Vastly more powerful and cheaper than ever before, personal and commercial computers took off very quickly. Human beings for the first time could now offload cognitive computation tasks from their brain onto a computer. Computers excel at repetitive tasks that humans are not skilled at, and so the pairing of the human brain and the computer expanded the possible capacity of notional computation in society. This expansion continues through to 2020 due to increasing power, decreasing costs, and more human-integrated implementations of computers such as smart-phones. 

Second, the internet. The speed, depth, and breadth of data transfer both literally and notionally in society exploded beyond precedent. Any individual human or computer now had nearly frictionless and lossless access to the world’s data, particularly compared to the flawed communication systems of the past. The usefulness of interconnectivity has compounded as additional aspects of society and the state on-board themselves to the internet. This connects compute nodes, both human and machine, to each other more tightly, and eases the access for these compute nodes to the data in the environment. 

Third, data. The first two points above combine to greatly incentivize the accumulation of all kinds of data, and for that data to be ever more conducive to processing by computers. More and more activities are conducted on or by computers and communicated through the internet, producing ever more flows of data. And as prior sources of data flows are on-boarded to the general computing system, totally computer-native data flows emerge. This data creates a larger and larger footprint to the digital representation of the physical world and its systems as well as a new natively virtual world and its systems.

Now, if we suppose that the Soviet Union and socialism in the 20th century failed due to a lack of compute and data, what does this mean for a world with plentiful compute and data resources such as the world today or of the near-future? China confronts us with this question as the major implementation of a high-compute and data-rich socialist state. The CCP in China hangs over the state and all state-owned and “private” chinese firms to create a clearly centralized system. China’s populace grips their smartphones as firmly as any in the world and the state surveils the virtual and physical worlds to an unprecedented degree. If all activities in a society are surveilled and accessible to a centralized notional computation center, all that remains is a need for sufficient computational capacity to produce a useful model of just the kind previously discussed that could be the engine for a successful integrated socialist state. Is capitalism’s diffuse decision-making, which once fit reality but now is primarily supported ideologically, now not in synch with the compute/data environment we live in today? Tellingly, the so-called Big Tech companies in the West are already erecting the scaffolding for a Chinese-style system to be possible in the West.