Una Mirada Humana

Why generating content and initiating a creative act are not the same thing.

Filmmaker Christopher Nolan has argued that artificial intelligence cannot create anything. I bring this up because I’ve noticed a growing number of people saying they’re tired of running into AI-generated articles.

On that point, I have to disagree.

AI systems like Claude and ChatGPT, now used by millions of people to write, research, program, and design, cannot wake up one day and decide to write an article about a topic of their choosing. They have no capacity to make that decision without explicit human involvement.

And yet, people often blame the model as though it had not only decided to write the text, but had also chosen Substack as the place to publish it.

I think there’s an important distinction that needs to be made.

If someone says:

“AI cannot create anything.”

That claim is difficult to defend. Every day we see images, music, essays, software, and designs that did not exist before and were generated with the help of AI. In practical terms, something genuinely new has come into existence.

If, instead, someone says:

“AI cannot create on its own initiative.”

Then I completely agree.

AI has no desires, intuitions, ambitions, or inner impulse that drives it to write a novel, develop a theory, or launch a creative project. It requires context, instructions, or a conversation to set the process in motion.

In fact, the way I work with AI illustrates this distinction quite well.

I often begin with nothing more than a vague intuition, a question, or even a contradiction. As the conversation unfolds, connections emerge that neither of us had explicitly formulated at the beginning. I contribute experience, judgment, and the final decision about what deserves to remain. The model contributes speed, breadth of exploration, and the ability to uncover relationships that might otherwise go unnoticed.

For that reason, I believe the discussion should no longer be whether AI creates or does not create. A far more interesting question is this:

Which part of the creative process belongs to the model, and which part belongs to the human being who guides, selects, edits, and ultimately gives meaning to the result?

That question strikes me as far more profound than a simple yes-or-no answer.

If we start from the premise that no AI can initiate a creative process on its own, then everything it generates has an external trigger: a question, an instruction, a context, or an objective defined by someone else.

In that sense, the origin of the process always lies outside the AI.

But it does not follow that every element of the final content comes entirely from the user.

Imagine commissioning a sculptor to carve a statue of Apollo. The client initiates the project, yet no one would claim that the client created the sculpture.

The comparison serves only to distinguish the origin of the commission from the origin of the work—not to equate a sculptor with an AI system.

Something similar happens with AI, though with one fundamental difference: the sculptor possesses intention. AI does not.

The user provides the initial intent and the judgment needed to accept or reject what emerges. During the interaction, the model generates combinations, relationships, and formulations that were not literally contained in the original prompt.

That is why it helps to distinguish between two different things.

The first is the origin of the creative act, which, in the case of AI, always comes from an external agent.

The second is the generation of content, which emerges through the interaction between the user and the model.

Making this distinction avoids two equally simplistic mistakes.

The first is claiming that AI creates entirely on its own.

The second is reducing it to nothing more than a sophisticated copying machine.

Neither description accurately reflects what is actually happening.

As long as AI lacks initiative of its own, every creative act in which it participates will always have an external origin, even if the content itself emerges during the interaction.

That distinction also changes the way we think about responsibility.

If someone publishes AI-generated text without reading it, verifying it, or accepting it as their own, responsibility for that publication rests primarily with the person who chose to publish it. The AI did not decide to make that content public, nor can it be held accountable for doing so.

That does not mean the developers of these systems bear no responsibility. They certainly do—in the design of the model, its safeguards, and efforts to reduce foreseeable errors. But that responsibility is fundamentally different from authorship or editorial judgment.

In other words, AI can generate content, but it cannot decide to publish it.

That final decision remains human.

This distinction matters because responsibility does not arise at the moment content is generated. It arises at the moment someone decides that the content represents what they are willing to stand behind.

A model can produce thousands of possible drafts. It cannot decide which one deserves to become public. It cannot sign it, defend it, revise it after criticism, or answer for its mistakes.

That threshold—the moment an idea becomes something a person is willing to claim as their own—remains, at least today, entirely human.

For centuries, authors have been responsible for the works they sign. Artificial intelligence profoundly changes the way a first draft can be produced, but it does not eliminate the responsibility of the person who ultimately says:

“This represents what I want to communicate.”

At least for now, that decision remains exclusively human.

But what happens when, during the training of a model, internal structures emerge that no engineer explicitly designed? The discovery of what researchers have begun calling “J Space” may seem, at first glance, to challenge everything I’ve argued so far.

It doesn’t.

The fact that functional internal structures can emerge during training without being explicitly programmed does not mean the model decided to create them. It simply means that the optimization process discovered an effective solution. This is a phenomenon of emergence—not an act of intention.

One thing is how a model develops its internal representations during training, where unexpected patterns may arise.

Quite another is who initiates a specific creative act.

J Space belongs to the first category.

It is fascinating because it suggests that models can develop internal representations that no one explicitly coded line by line. But that still does not imply that, on a random Tuesday morning, the model decides:

“Today I’m going to write an essay about Cervantes and publish it on Substack.”

Perhaps the real surprise is not that AI generates content, but that, during training, it discovers internal structures that even its creators did not anticipate. That opens profound scientific questions about how reasoning itself emerges. But, at least for now, it does not alter one fundamental fact: the act of initiating a creation still originates outside the model.

Perhaps the real mistake has been trying to answer a philosophical question with a binary response.

AI does not create in the way human beings do.

But neither is it merely a photocopier of information.

It generates novelty without agency.

And as long as that agency remains uniquely human, the ultimate origin of every creative act will continue to lie outside the model.