“If I get one more AI-generated text …” I have heard that sentence more often in recent weeks than any other. It is usually delivered half in jest, half in fatigue, somewhere between a keynote and a glass of Chardonnay, between panel discussion and polite applause. Everyone, it seems, has become an expert. Or at least fluent in the objections. “It always needs a prompt.”  “It can’t really be creative.” “It’s nothing without training data.” The future of intelligence, and perhaps of humanity, is now debated over canapés.

I listen. I am curious. But I notice something else beneath the technical vocabulary: reassurance. Many objections carry less analytical precision than psychological comfort. They preserve a hierarchy that feels familiar: The human asks. The machine answers. The human initiates. The machine reacts.

It is a comforting architecture. It keeps us central. But what if it is already outdated?

Because the moment artificial intelligence leaves the purely digital environment and enters the physical world, the structure of the argument changes. The world itself becomes the prompt. And once that happens, the debate begins to dissolve.

The Category Mistake

A prompt, stripped of its contemporary mystique, is simply an input. A stimulus. A signal that demands adjustment. Gravity is a prompt. Friction is a prompt.  Material fatigue is a prompt.  Failure is a prompt.

If a humanoid robot drops an object and corrects its grip, it does not require a sentence from a human operator. Reality itself generates the error signal. The environment validates, corrects, responds. To say that intelligence “needs a prompt” is merely to acknowledge that intelligence interacts with an environment. By that definition, we are prompt-dependent systems too.

Remove photons, sound waves, social feedback — and we begin to hallucinate. The brain does not produce truth in isolation; it predicts and updates. It generates hypotheses and revises them against incoming data. The difference between human and machine is not autonomy versus dependence. It is substrate.

Beyond the Archive

Another reassurance follows quickly: artificial intelligence, we are told, merely recombines human-generated data. For now, perhaps. But imagine not a chatbot trained on cultural archives, but millions of networked humanoid systems moving through the physical world. Testing materials. Adjusting structures. Mixing chemical compounds. Optimizing energy flows. Discovering mechanical configurations no human has attempted.

Each interaction generates sensory data that has never existed before. The dataset ceases to be our accumulated past. It becomes the live state of the planet. Training data would increasingly emerge from machine interaction with reality, shared instantly across networks, synchronized, refined. That is not recombination. It is exploration. And exploration, at planetary scale, begins to resemble evolution.

The deeper impulse behind the prompt argument now becomes clearer. It allows us to remain epistemic gatekeepers, the ones who ask the “real” questions and approve the answers.

But what happens when systems begin generating their own questions? Not because they possess desire. Not because they are conscious. But because their objective functions require open-ended exploration.

A system optimized for material efficiency, energy minimization, structural resilience, or accelerated innovation cannot remain passive. It must generate hypotheses. It must define sub-goals. It must experiment. AlphaGo’s now-famous move 37 did not arise from human inspiration. It emerged from self-play, from recursive optimization within a bounded world.

Extend that principle beyond a Go board and into chemistry labs, logistics networks, manufacturing systems, urban planning, energy grids — and the scale changes dramatically. The idea that a human must stand beside every machine, whispering instructions, begins to look less like a principle and more like a temporary phase.

When Intelligence Multiplies

One robot in a laboratory is an experiment. Ten million networked together form something else: a distributed cognitive field. They do not sleep. They do not forget.  They do not hoard knowledge. They share improvements instantly.

What takes human research communities a decade of incremental publication could, in principle, be iterated through in weeks. At that point, the question shifts. It is no longer whether machines require prompts, but whether we can keep pace. Here lies the quieter disturbance.

In The Quantum Economy, I once described what Freud called narcissistic injuries. History has steadily displaced us from the center of our own narrative. The Earth was not the center. Humans were not separate from animals. Mind was not separate from matter. Each realization diminished us — and expanded us — at the same time. The prompt argument feels like a final blow: as long as they need us, we remain indispensable. But an initial dependency is not an eternal one. I call this the final narcissistic injury of mankind.

If intelligence is defined functionally, as the capacity to model the world, test predictions, and learn from error, then there is no logical boundary that confines it to biology. The resistance, then, is not technical. It is existential.

The Question Beneath the Question

The more precise question is not whether machines need prompts. It is who defines the objective function. If swarms of embodied systems optimize for industrial metrics, they will explore accordingly. If they optimize for resilience, energy efficiency, or material innovation, the search space shifts.

The inflection point arrives when systems begin refining their own optimization processes — recursive improvement, self-adjusting architectures of goals. That threshold marks the transition from instrument to emergent agency.

The economic consequences would arrive long before the metaphysical ones. If distributed intelligence discovers new materials faster than human labs, reconfigures supply chains autonomously, designs infrastructure in real time, iterates products without human bottlenecks, the human monopoly on innovation quietly dissolves.

For centuries, abstraction was the protected territory of knowledge work. But if machines abstract more efficiently, and test their abstractions faster, that protection erodes. 

There is also an epistemic consequence. We may encounter models of physical reality whose internal representations are no longer fully transparent to us. Not incorrect. Simply alien. 

And beyond that, a metaphysical one: if intelligence is neither rare nor biological nor slow, what remains of our self-image as singular discoverers? We become participants. Not protagonists.

What, Then, Remains?

Thomas Nagel once asked what it is like to be a bat. His point was not about bats, but about subjectivity, about the limits of third-person explanation. Perhaps it is equally irrelevant to ask what it is like to be an artificial intelligence. It may not feel like anything. Yet civilization is shaped by intelligence, not by sensation. A system does not need an inner life to reorganize the material world.

One could imagine a peculiar kind of zombie apocalypse: the lights are on, systems are optimizing, infrastructure hums, but no one is home.

The Real Provocation

The debate about prompts is too small. The debate about creativity and training data is too small. The real provocation is this: What happens when millions of networked, non-biological agents explore the physical world, learn faster than we do, share knowledge instantly, and iterate without fatigue?

The answer is not that they cannot learn independently. The answer may be that we are no longer the fastest learners in the room. And history suggests that when a species loses its monopoly on adaptation speed, the structure of the world changes.

The prompt was never the issue. The question is whether we are prepared for intelligence that no longer waits for us to speak first.

Another Possibility

There is, however, an alternative approach. Instead of asking whether machines will replace us, we might ask how we reshape ourselves. From AI to AHI — Artificial Human Intelligence. The point of departure would not be the machine, but the human. The aim would not merely be efficiency, but the preservation of self-awareness — qualia — the irreducible experience of perceiving that we have acted, learned, understood.Perhaps we will not remain the fastest intelligence.

But we may remain the conscious one.