1. Research Question

Experiment 13 took the research into regulation.

The more I examined the growing effort to regulate artificial intelligence, the more another question began to emerge.

Governments are developing frameworks. Institutions are establishing standards. Companies are building safety systems. Researchers are testing models. Regulators are considering how increasingly capable systems should be constrained. Capital is allocating.

An extraordinary amount of human effort is being directed towards governing something that, only a relatively short time ago, did not exist in its current form.

That made me step back from the question of how AI should be governed and ask a more fundamental question.

Why does AI exist in the first place?

What human problem are we trying to solve? What limitation are we trying to overcome? And if we are investing so much intellectual, institutional and material capacity into creating increasingly capable machines, what is the civilisational purpose of that effort?

The question is not whether AI is useful. Clearly, it can be.

The deeper question is whether capability itself is sufficient justification for continued development.

Humans have always built interventions to overcome limitations.

Writing emerged as societies became too complex to rely on memory alone. Machines extended physical capacity beyond the limits of the human body. Computers extended our capacity to calculate and process information. AI appears to extend the intervention further, into cognition, prediction, decision-making and increasingly execution.

Each intervention solved a problem.

Each also changed the system around it.

If AI is another intervention in the long human effort to overcome our limitations, what are we actually trying to achieve?

And once we begin overcoming those limitations, what are we trying to preserve as humans?

2. Research Context

The history of technology can be understood as a history of interventions against human limitation.

The first intervention considered in this research is Mesopotamia.

As agricultural settlements developed into increasingly complex cities, the scale of economic and administrative activity began to exceed the capacity of human memory. Goods had to be counted. Labour had to be organised. Resources had to be allocated. Transactions had to be recorded.

The earliest surviving writing systems emerged within this administrative environment. At Uruk, proto-cuneiform records were used to document commodities, labour and economic transactions. Earlier accounting systems used physical tokens to represent goods and quantities. Writing allowed information to be externalised, retained and administered across increasingly complex institutions.

Writing was therefore not initially an intervention designed primarily for literature or abstract thought.

It was an intervention against a constraint.

Human memory was no longer sufficient to support the scale of coordination that the emerging city required.

The record extended memory beyond the individual.

Once information could be recorded, however, it could also be accumulated, compared, classified and administered. The solution to one limitation created new institutional capabilities and new forms of authority.

The same pattern appeared with mechanical reproduction.

For centuries, knowledge could be recorded but remained constrained by the labour required to reproduce it. The printing press changed that relationship by reducing the cost and increasing the scale of reproduction.

The intervention solved a problem of distribution.

Knowledge could move further and faster. Texts could be reproduced with greater consistency. Scientific observations could circulate. Political, religious and philosophical ideas could reach populations that previously had limited access to them.

The intervention therefore changed more than the speed of printing.

It changed the information environment in which societies operated.

Industrialisation introduced another intervention.

Human physical capacity limited the scale and speed of production. Machines and steam power allowed energy to be concentrated and applied beyond the strength and endurance of individual human bodies.

The result was a dramatic increase in productive capacity.

But the intervention also reorganised society. Factories emerged. Labour became increasingly divided and measured. Production became a system rather than a collection of individual tasks.

The machine extended human physical capacity while simultaneously changing the position of human labour within the system.

The twentieth century introduced another constraint.

The problem increasingly became computation.

Scientific, military and administrative systems generated calculations at a scale that humans could not perform quickly enough. Electronic computing extended the capacity to calculate and process information.

ENIAC provides a clear example. Developed during the Second World War to calculate artillery firing tables, it responded to a growing computational backlog that human computers using mechanical calculators could not process quickly enough.

Computing subsequently became general-purpose infrastructure.

The machine was no longer simply a device for performing a particular calculation. It became a system capable of executing different instructions across different domains.

Artificial intelligence extends this trajectory again.

The field emerged from an ambition to reproduce or simulate aspects of intelligence through machines. The 1955 Dartmouth proposal framed the problem around whether aspects of learning and intelligence could be sufficiently precisely described for machines to simulate them.

The intervention therefore moved beyond physical capacity and calculation.

It entered cognition.

AI systems can increasingly process language, identify patterns, make predictions, generate content, support decisions and plan actions. When connected to external systems, they can increasingly execute those actions.

The historical sequence is therefore significant.

Writing extended memory.

Industrial machinery extended physical capacity.

Computing extended calculation.

AI extends elements of cognition and increasingly connects those capabilities to execution.

Each intervention overcame a limitation.

Each intervention also changed the system around it.

3. Research Method

This experiment treats artificial intelligence as part of a longer history of technological intervention rather than as an isolated technological phenomenon.

The research traces a sequence from Mesopotamian administration through writing, mechanical reproduction, industrialisation, computing and artificial intelligence.

Each intervention is examined in relation to the limitation it was intended to overcome, the capability it created and the institutional changes and dependencies that followed.

The research then asks what remained important after each intervention.

This method is designed to separate capability from purpose.

A technology can solve a genuine problem without establishing that every application of the technology is desirable.

The historical comparison is therefore used to examine what is distinctive about AI.

Earlier technologies primarily extended memory, physical capacity or calculation.

AI increasingly participates in cognition, decision-making and execution.

The experiment therefore asks what happens when the intervention no longer merely extends what humans can do, but increasingly participates in deciding and acting within the systems that organise human life.

4. Research Findings

The first finding is that AI belongs to a much older civilisational pattern.

Human societies repeatedly develop technologies to overcome biological, informational and organisational limitations.

AI is the latest expression of that impulse.

It exists because there are things humans want to know, produce, predict, coordinate and accomplish that exceed the capacity of unaided human cognition.

The second finding is that capability does not establish purpose.

The fact that a system can perform a task does not establish that the task should be automated. The ability to make a decision does not establish that a machine should make it. The ability to execute an action does not establish that authority to execute should be transferred.

The third finding is that every major technological intervention creates new dependencies.

Writing created dependence on records and administrative systems. Industrialisation created dependence on machinery, energy and production infrastructure. Computing created dependence on digital systems and computational infrastructure.

AI increasingly creates dependence on models, data, compute, networks and execution infrastructure.

The fourth finding is that the governance of AI cannot be separated from the governance of infrastructure.

AI does not operate independently. It depends upon energy, compute, data, telecommunications, identity systems, payment systems, software and physical infrastructure.

As these systems become more deeply embedded in social and economic life, control over them increasingly becomes a question of social consequence.

The final finding concerns preservation.

The object of preservation cannot simply be existing human tasks.

Human history has repeatedly involved the removal of tasks from human labour through technological intervention.

The deeper question is what should remain possible for humans after the task has been removed.

The research identifies agency, dignity and continuity as central preservation conditions.

Agency concerns the continuing ability to choose, participate, refuse, redirect, intervene and act.

Dignity concerns the recognition that human value cannot be reduced entirely to productive usefulness.

Continuity concerns the ability of institutions and societies to remain functional when technologies fail, incentives change or critical infrastructure becomes contested.

The object of preservation is therefore not human limitation.

It is meaningful human agency within systems that are increasingly capable of overcoming human limitation.

5. Research Implications

This finding changes the governance question.

If the objective of technological development is to overcome human limitations, governance should not attempt simply to preserve those limitations.

It should not preserve inefficient work merely because humans once performed it. It should not prevent useful automation simply to maintain existing arrangements. It should not treat human involvement as valuable merely because it is human involvement.

The question is more precise.

What must remain humanly possible as technological capability increases?

Humans must remain capable of participating in consequential systems. They must be able to understand consequential decisions to a meaningful degree. They must retain the ability to refuse where refusal is legitimate. They must retain mechanisms for intervention. They must be able to contest consequential outcomes. They must retain meaningful control over critical resources and infrastructure.

Human value must also not become entirely dependent upon whether a machine can perform the task previously associated with that value.

This changes how human oversight should be understood.

A human can remain technically present while possessing no meaningful authority. A human can approve a decision without understanding it. A human can possess an override mechanism that cannot realistically be exercised. A human can remain formally accountable while practical authority has migrated elsewhere.

Human presence is therefore not sufficient.

What matters is whether humans retain meaningful capacity to shape consequence.

The implication for governance is that increasingly capable AI requires attention not only to what systems can do, but to how authority is allocated around their capabilities.

The governance question moves from restricting technology in the abstract to organising the relationship between human agency, machine capability and the infrastructure through which execution occurs.

6. Conclusion

The research began by asking why AI exists.

The historical investigation produced a straightforward answer.

AI belongs to the long human effort to overcome limitation.

Human societies have repeatedly externalised capacities that constrained them. They externalised memory through writing. They mechanised physical production. They externalised calculation through computing. They are now increasingly externalising elements of cognition, decision-making and execution.

There is nothing inherently wrong with this progression.

Overcoming human limitations is one of the defining characteristics of civilisation.

But every intervention changes the system around it.

The critical question is therefore not whether we should preserve the world as it was before AI.

We cannot.

Nor should we necessarily try.

The question is whether, in overcoming our limitations, we also surrender the capacities that make us able to govern the consequences.

We are not trying to preserve human limitation.

We are trying to preserve meaningful human agency through the process of overcoming it.

This is where the research moves beyond AI capability.

When intelligence becomes connected to execution, capability becomes authority.

A system that can act creates a different governance problem from a system that can merely calculate.

The question is no longer simply what AI should be permitted to do.

It is who, or what, has the authority to execute the consequence.

Who authorised the system to act? What can it access? What is it permitted to do? Where does its authority stop? Who can intervene? Who can override it? Who retains control over the infrastructure through which the action occurs?

The research therefore moves from the question of what we are trying to preserve to the question of how authority should be organised when execution is no longer exclusively human.

This is the point at which the research enters Execution Authority Governance.

7. References

Schmandt-Besserat, D. From Accounting to Writing. University of Texas at Austin.

University of Pennsylvania Museum. Digital Corpus of Cuneiform Lexical Texts: Archaic Administrative Corpus.

Metropolitan Museum of Art. “The Origins of Writing.”

Computer History Museum. Birth of the Computer.

Computer History Museum. The Minds Behind ENIAC.

McCarthy, J., Minsky, M. L., Rochester, N. and Shannon, C. E. (1955). A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence.

OECD. (2022). OECD Framework for the Classification of AI Systems. OECD Publishing.

UNESCO. (2021). Recommendation on the Ethics of Artificial Intelligence.

Santoni de Sio, F. and van den Hoven, J. (2018). “Meaningful Human Control over Autonomous Systems: A Philosophical Account.” Frontiers in Robotics and AI, 5, Article 15.