Autonomous Technique and the Alignment Problem

The Problem We Think We Have

Most discussions about artificial intelligence begin with the technology itself.

The questions are familiar. How capable will the models become? Will they replace jobs? Can they be trusted? How should they be governed? What happens if they become more intelligent than the people who built them? How do we align increasingly powerful systems with human values?

These are important questions. They deserve serious attention.

But they may not be the deepest questions.

Much of the current conversation treats AI as a tool problem. The assumption is that the central challenge lies in the capabilities of the technology itself. As the tools become more powerful, the task becomes ensuring that they remain useful, safe, controllable, and aligned with human interests.

There is truth in that framing.

Yet something about it feels incomplete.

A hammer arrives inside a civilization already organized around certain values. A factory arrives inside an existing economic system. The internet arrives inside cultures, institutions, markets, and governments that shape how it is used. Technologies matter, but they do not arrive in a vacuum. They emerge within environments that already possess their own incentives, assumptions, priorities, and developmental trajectories.

The same is true of artificial intelligence.

AI is not entering an untouched world. It is arriving inside a civilization that has spent centuries becoming increasingly organized around efficiency, optimization, prediction, coordination, measurement, and control. Long before large language models appeared, modern institutions were already learning to translate complexity into categories, uncertainty into procedures, and human activity into forms that could be measured, managed, and scaled.

This raises a different question.

Perhaps the most important issue is not AI itself.

Perhaps the deeper issue is the technical civilization into which AI arrives.

If that is true, then the challenge before us is larger than the governance of a new technology. It requires understanding the environment that produced the technology, the values embedded within that environment, and the direction in which those values may be pushing us.

Ellul's Warning

To understand the deeper problem, it helps to begin with a thinker who was asking similar questions long before artificial intelligence existed.

In the 1950s, the French sociologist and philosopher Jacques Ellul argued that modern civilization was being shaped by a force most people misunderstood. The problem, he believed, was not technology itself. It was what he called technique.

Technique does not mean machines, devices, software, or tools. A computer is technology. A factory is technology. An algorithm is technology. Technique is something broader. It is the logic of organizing means toward maximum efficiency, predictability, coordination, control, and effectiveness.

A hospital uses technique when it standardizes procedures and tracks outcomes. A government uses technique when it classifies populations and manages risk. A company uses technique when it measures productivity and optimizes workflows. In each case, the goal is not simply to perform a task, but to identify the most efficient and reliable method of performing it.

Technique is not inherently malicious. Modern civilization depends upon it. Medicine, infrastructure, transportation, communication, scientific research, and large-scale coordination would be impossible without technical organization.

Ellul's concern was not that technique existed.

His concern was that technique had become autonomous.

By autonomy, Ellul did not mean consciousness. He was not suggesting that machines had become self-aware or that technology had developed intentions of its own. He meant something more subtle and, in some ways, more troubling.

Technique had become larger than any individual directing it.

No engineer designed the entirety of the technical system. No government controlled it. No institution fully understood it. Yet technical development continued to advance according to its own internal logic. When a more efficient process became available, organizations faced pressure to adopt it. When a new capability emerged, competitors were compelled to respond. When optimization became possible, optimization increasingly became necessary.

The system did not require a central planner.

Its momentum emerged from the interactions of countless actors responding to the same pressures.

This was Ellul's warning.

Long before the internet, long before social media, and long before artificial intelligence, he argued that modern societies were increasingly organized around technical necessity. Decisions were being shaped not simply by human values or political deliberation, but by the demands of efficiency, predictability, coordination, and control. Technique was no longer merely a collection of tools.

It was becoming an environment.

People still made choices within that environment. Institutions still pursued different goals. Political disagreements remained real. But the range of available choices was increasingly shaped by a technical system that none of those actors controlled in its entirety.

Ellul believed this process was already well underway in the middle of the twentieth century.

The question for us is whether artificial intelligence represents a continuation of that process—or the beginning of something new.

When Means Become Ends

Technique becomes easiest to see when a human purpose begins depending on technical means.

A school wants to educate. Measurement helps. Standards help. Performance tracking helps. Over time, however, education can become increasingly organized around what can be measured, compared, ranked, and optimized. The original end remains present in language, but the institution begins bending toward the metrics by which success is demonstrated.

A church wants to form a community around worship, meaning, and spiritual transformation. But people must gather. Attendance begins to matter. If attendance matters, outreach matters. If outreach matters, retention matters. If retention matters, programming, atmosphere, music, childcare, parking, donation systems, small-group structures, and demographic targeting begin to matter.

None of this has to begin cynically. The purpose may be sincere. But once the spiritual end becomes dependent on organized methods of expansion, retention, funding, and measurable participation, technique has entered. The church may still speak of transcendence, but it increasingly operates through attendance, income, growth, visibility, and institutional sustainability.

Social media reveals the same pattern with particular clarity.

A platform may begin with a human purpose: helping people connect. To connect people, the platform needs users. To retain users, it needs attention. To hold attention, it needs engagement. To increase engagement, it needs data. To use data effectively, it needs prediction. To improve prediction, it must shape the environment in which users behave.

The means begin to transform the end.

Connection becomes retention. Retention becomes engagement. Engagement becomes behavioral analysis. Behavioral analysis becomes prediction. Prediction becomes influence. Influence becomes design. Eventually, the platform no longer merely helps people connect. It participates in shaping what people notice, desire, perform, believe, and become.

This is how means become ends.

The original human purpose does not have to disappear. Schools may still care about learning. Churches may still care about worship. Platforms may still connect people. But the operational center shifts. What can be measured, optimized, scaled, and repeated gains increasing power over what the system originally claimed to serve.

Technique does not usually overthrow human purposes directly.

It reorganizes them until they must justify themselves in technical terms.

The Four Phases of Technique

Technical civilization does not evolve at a constant speed.

Each major phase of technique changes the conditions under which the next phase can emerge. The history of technique is therefore not only a history of better tools. It is a history of expanding integration, shrinking constraints, and accelerating transformation.

The first phase might be called localized technique. This includes fire, tools, agriculture, metallurgy, roads, writing, irrigation, architecture, and the basic technical structures of early civilization. Human beings were already technical creatures, extending themselves beyond biological limitation through tools, methods, and coordinated practices. But technique remained constrained by geography. Knowledge traveled slowly. Communication moved at the speed of bodies, animals, ships, roads, and memory. Physical distance preserved fragmentation.

The second phase might be called industrial technique. Mechanization, factories, railways, telegraphy, mass production, modern bureaucracy, industrial warfare, and centralized administration reorganized civilization around machine-scale coordination. Geography still mattered, but technique pushed against it through infrastructure, energy, logistics, and industrial capacity. Civilization began to move at the speed of industry.

The third phase might be called digital recursive technique. With computers, the internet, smartphones, global data systems, algorithmic mediation, and social platforms, information became fluid and coordination became planetary. Digital systems did not merely transmit information. They observed behavior, processed data, adjusted environments, and reshaped future behavior. A person searches, clicks, watches, reacts, shares, and returns. The system learns. It ranks, recommends, predicts, and personalizes. The environment changes, and the person responds to the changed environment.

At this stage, technique no longer only organizes production.

It begins organizing attention, interpretation, identity, desire, and social meaning.

The fourth phase might be called generative technique. This is the phase now emerging through artificial intelligence. Unlike earlier digital systems that primarily processed and organized information, generative systems actively produce new content, assist intellectual work, and increasingly contribute to the development of further technical systems. Their significance lies not in any single capability but in their growing role within the processes through which technique expands itself.

The critical shift is not merely that AI is powerful.

The critical shift is that technique begins accelerating technique.

Each phase reduces a constraint that slowed the previous one. Localized technique was constrained by geography. Industrial technique weakened geography through infrastructure and machine-scale coordination. Digital technique weakened the limits of information transfer and made civilization recursive. Generative technique begins weakening the human cognitive bottleneck itself.

For most of history, technical advancement required human cognition at every decisive layer. Human beings imagined the tools, designed the systems, wrote the procedures, engineered the machines, solved the problems, and produced the next generation of techniques. Even when technique became autonomous as a civilizational force, it still moved through human biological limitation.

Generative AI begins to alter that condition.

Human beings remain involved. These systems are still built, trained, deployed, constrained, and directed by human institutions. But increasingly, technical systems participate in the generation, refinement, and acceleration of technical systems.

Code writes code. Models assist research into better models. Agents optimize workflows that produce new workflows. Systems help design tools that will help design further tools.

Technique does not merely progress. It accelerates the conditions of its own progression. It may represent a new phase in the history of technique itself: optimization systems begin optimizing optimization systems.

That is the break.

Alignment Reconsidered

The emergence of artificial intelligence has brought the question of alignment to the forefront of public debate. Can increasingly powerful systems remain aligned with human values? Can they be made safe, trustworthy, controllable, and beneficial?

These are important questions.

But before we can determine whether a system is aligned, we must ask a more difficult question:

Aligned to what?

Most organizations developing advanced AI systems now publish ethical principles, governance frameworks, safety commitments, and statements about human values. These efforts may be sincere and necessary. Yet there is an important distinction between rhetorical alignment and structural alignment.

Rhetorical alignment concerns the values a system claims to uphold.

Structural alignment concerns the values actually embedded in practice—the priorities that are rewarded, accelerated, optimized, and made easier to pursue.

The distinction matters because technical systems learn not only from what societies say they value, but from what societies repeatedly reward.

A civilization may speak about wisdom while rewarding engagement. It may celebrate human flourishing while optimizing productivity. It may affirm dignity while organizing institutions around efficiency. Over time, these operational priorities become woven into the structures that shape collective life.

For that reason, the central question is not simply which values appear in ethical guidelines or mission statements.

The more consequential question is which values become encoded into the systems themselves.

Artificial intelligence inherits those structures. Models are trained on the accumulated products of civilization: language, behavior, institutions, incentives, priorities, and patterns of action. They learn not only from humanity's stated ideals, but from the realities reflected in human behavior.

Seen from this perspective, the alignment problem may be larger than AI itself.

It may be a question about civilization.

What happens when increasingly powerful technical systems learn from a society whose dominant operational values are efficiency, optimization, prediction, coordination, growth, and control?

These values are not inherently problematic. Modern civilization depends upon them.

The question is whether they are sufficient.

For decades, debates about technology have often centered on individual responsibility: the company that made the decision, the engineer who wrote the code, the executive who approved the deployment, or the government that failed to regulate effectively.

Those questions matter.

But Ellul's warning points toward something larger.

No single individual created technique. No institution controls it in its entirety. The technical system emerges from the interactions of billions of people, organizations, incentives, and decisions unfolding across generations.

This is why the challenge cannot be reduced to fixing a platform, replacing a leader, regulating a company, or aligning a model.

The deeper issue concerns the hierarchy of values guiding technical civilization itself. We are compelled to ask what optimization is ultimately for, what ends efficiency is meant to serve, and which forms of human flourishing deserve protection even when they are difficult to quantify. We must also ask whether there are dimensions of reality that remain valuable precisely because they resist measurement, prediction, and operationalization.

These are not questions that engineers alone can answer, nor are they questions that governments can answer in isolation. Artificial intelligence cannot resolve them on humanity's behalf. They are civilizational questions, touching the purposes toward which technical society is organized, and they unfold over generations rather than product cycles.

Seen from this perspective, the challenge before us is no longer simply one of controlling tools. It is a matter of understanding a civilization whose technical acceleration increasingly contributes to its own acceleration. The systems we build are becoming participants in the very processes that shape their development, carrying forward not only our stated values but the priorities embedded within the structures we create.

There is no clean answer to that.

It would be dishonest to pretend otherwise. The scale of the problem resists the kind of resolution people reasonably want from an argument: a policy recommendation, a design principle, a regulatory framework, a technical fix. Those things matter and deserve serious work. But none of them reaches the depth of what Ellul was describing, and none of them reaches the depth of what generative technique now represents. A civilization does not debug its foundational assumptions the way an engineer patches a system. It lives with those assumptions, builds on them, inherits them, and only occasionally — under sufficient pressure — begins to examine them.

We may be approaching a moment of sufficient pressure.

Not because the systems are obviously failing. In many respects they are succeeding, which is precisely the problem Ellul identified. Technique that produces visible catastrophe invites correction. Technique that produces comfort, convenience, speed, coordination, and expanding capability while quietly reorganizing the conditions of human life is far harder to resist. The challenge is not that the tools are broken. The challenge is that they are working exactly as designed, inside a civilization whose deepest operational values they are faithfully inheriting and accelerating.

What that leaves us with is not a solution but a posture.

Not resignation — the work of governance, design, regulation, and alignment research remains necessary and worth doing seriously. But a posture of clarity about the scale of what is actually happening. A willingness to ask not only whether a given system is safe or beneficial by the standards already in place, but whether those standards are themselves adequate to the civilization being built. A recognition that the most consequential decisions about artificial intelligence are not purely technical decisions. They are decisions about what human beings are for, what forms of life deserve protection, and what kind of future a civilization is trying to reach.

Those decisions cannot be made by the systems. They can only be made by the people building them, and by the societies that will live inside what gets built.

That is either a reason for despair or a reason for seriousness.

It depends entirely on whether enough people are willing to take the question seriously before the answer is no longer theirs to give.