For most of human history, making culture was difficult. A musician had to learn an instrument. A painter had to acquire technique. A writer needed time, language, persistence, and usually a very tolerant relationship with revision. Film required actors, cameras, editors, sets, money, and an army of people whose names appeared in tiny letters at the end. Even relatively simple creative work carried friction. Creation was constrained by skill, labor, cost, access, and time.

Artificial intelligence is rapidly dismantling that scarcity.

Today, a person can describe a song and hear something resembling a finished recording minutes later. An image can move from idea to polished illustration almost instantly. Text, animation, advertising, voice, video, design, and increasingly complex combinations of them can be generated by systems that require far less technical expertise from the person directing them. Those systems will improve. Their costs will fall. Their outputs will multiply.

This raises an obvious debate: if AI creates the song, picture, film, or story, is it still art?

I increasingly think that may be the wrong question.

The more important distinction may not be between human-created culture and AI-created culture.

It may be between intentional culture and autonomous synthetic culture.

That distinction became unusually clear to me while creating an album with AI.

The album follows the history of artificial intelligence, from humanity's early dream of thinking machines through symbolic AI, neural networks, generative systems, AI agents, possible AGI, and whatever may lie beyond. I cannot play the instruments that appear on these tracks. I could not personally record the singers. I am using AI extensively. I begin with ideas, themes, historical questions, musical instincts, and reactions. My AI collaborator helps elaborate those fragments into structured concepts, genre descriptions, lyrics, and arrangements. A generative music system then renders them into finished tracks.

By the traditional definition, this is heavily AI-generated art.

But the process has been anything but automatic.

One track originally drifted too far from the folk sound I wanted, so we changed direction. We experimented with two male singers and discovered that the music system tended to make them sound like one singer layering his own voice. We tested the same idea with two female singers and found essentially the same limitation. That failure became part of our creative vocabulary. One evening, listening to the album after a few glasses of wine, I suddenly wondered what would happen if the next track contained no vocals and no electronics at all, just Appalachian banjo, acoustic guitar, fiddle, and strings. That sideways idea became an instrumental track representing the transformer concept of “attention,” with instruments listening to one another, answering one another, and passing musical motifs back and forth.

Another track about AI agents became frantic roots rock driven by an almost runaway train rhythm. Because that song was so fast and kinetic, the following track about corporate and governmental control over artificial intelligence deliberately slowed to a heavy, spacious Southern folk-soul groove. That track became unexpectedly powerful, which led us to make the next song, about automation and labor, warmer and more communal.

None of those choices existed in the original album plan.

They emerged through recursion.

An idea produced an output. The output produced a human reaction. That reaction changed the next AI response. The AI response generated another possibility. The music itself introduced surprises and limitations. Those discoveries changed subsequent decisions. The album gradually developed a history.

That history matters.

It is difficult to reduce the process to “human created” or “AI created.” The finished work exists because of a chain of intentions, reactions, judgments, mistakes, discoveries, and decisions that neither participant possessed at the beginning.

I have sometimes called the product of this recursive collaboration a “third mind.” I do not mean another conscious being mysteriously appearing between human and machine. I mean the product itself: the accumulated creative state that neither participant could produce independently. As the collaboration develops, that product acquires themes, constraints, references, aesthetic expectations, and history. Later decisions are shaped by earlier ones.

There is still someone in the process who cares about what the work means.

That may become increasingly important.

Because the same generative technologies that make this kind of collaboration possible also make something very different possible.

Imagine an autonomous cultural production system.

It monitors listening behavior, viewing patterns, social trends, search data, audience demographics, emotional responses, and engagement metrics. It generates ten thousand songs. Another AI evaluates them. The strongest hundred are released to test audiences. Completion rates, skips, replays, shares, and emotional reactions are measured. The system produces variations of the winners. Those variants are tested again. The most successful characteristics are fed back into the next generation.

No songwriter has to decide that a particular song needs to exist.

No musician has to care about what it says.

Eventually, no human necessarily needs to hear most of the generated material at all.

The system could simply search an enormous cultural possibility space until it discovers artifacts statistically likely to hold human attention.

Now extend that model beyond music.

Stories optimized from reading behavior.

Images continuously evolved against click-through rates.

Short-form videos generated from engagement data.

Personalized films assembled for individual viewers.

Synthetic influencers whose personalities adapt algorithmically.

Children's entertainment dynamically tuned to hold a particular child's attention.

Political messages generated, tested, mutated, and personalized faster than any campaign staff could review them.

Advertising that stops being created in campaigns and instead becomes a continuous evolutionary process.

At that point, we have moved beyond AI-assisted creativity.

We have entered autonomous synthetic culture.

The cultural loop changes.

Historically, the rough pattern was:

human experience ? artistic intention ? cultural artifact ? audience

In an autonomous system, the loop could become:

model ? generated artifact ? audience behavior ? optimization data ? model ? more effective artifact

Humans remain inside the loop, but increasingly as the environment being measured.

Culture starts learning how to grow around our attention.

This does not necessarily mean the results will be bad.

That is part of what makes the problem difficult.

The autonomous song may be beautiful.

The synthetic film may be hilarious.

The generated novel may be moving.

The personalized entertainment might understand exactly what kind of story you want on a particular night.

There is no law saying that something produced without human artistic intention must be aesthetically inferior. Indeed, sufficiently advanced systems may eventually produce cultural artifacts that are technically extraordinary.

So the defense of intentional culture cannot simply be, “Human-made things are better.”

Sometimes they will not be.

Nor should we retreat into definitions of authenticity that insist a musician must physically play every instrument or a painter must personally place every mark. Humans have always created through tools. Cameras did not destroy photography by doing some of the mechanical work. Synthesizers did not invalidate music. Digital editing did not make cinema unreal. AI may become another extraordinarily powerful extension of human creative capacity.

The relevant question is not how much machinery touched the work.

It is whether someone meant something by making it.

That is why I think AI-assisted art and autonomous synthetic culture need to be separated conceptually.

A human may use AI to create nearly every technical component of a piece while still exercising intention throughout the process. The person can choose the subject, establish the values, respond to unexpected results, reject possibilities, change direction, recognize meaning, and decide when the work is finished.

That is very different from a system whose primary intention is an optimization target.

Maximize watch time.

Increase retention.

Reduce churn.

Increase purchases.

Improve emotional engagement.

Generate another variation.

The artifact may still contain what looks like emotion, conflict, vulnerability, longing, humor, grief, rebellion, spirituality, love, or rage. But those qualities may be there because a system learned that particular combinations of them perform well.

And this leads to a question I find increasingly difficult to shake:

Was the work created to express an intention, or was the intention manufactured to produce the work?

That may become one of the central cultural questions of the AI era.

We are accustomed to worrying about information overload. Generative AI may produce something more profound: meaning overload.

There could eventually be more songs than anyone could hear, more images than anyone could see, more stories than anyone could read, more videos than anyone could watch, and more virtual worlds than anyone could inhabit.

Creation itself becomes abundant.

When abundance arrives, value tends to migrate toward whatever remains scarce.

If generating competent cultural artifacts becomes nearly free, then the scarce resources may become human attention, trust, provenance, judgment, curation, community, and intention.

The future problem of art may not be scarcity.

It may be significance.

Why this song?

Why this story?

Why now?

Why did someone decide it needed to exist?

Who stands behind it?

What experience produced it?

What conversation did it enter?

What does it ask of us?

Those questions do not disappear merely because machines become extraordinarily good at producing artifacts. They may become more important precisely because production itself becomes so easy.

This also suggests that the distinction between “human art” and “AI art” may eventually become almost useless.

Most culture may exist somewhere on a spectrum.

At one end:

human intention ? AI assistance ? human judgment ? artifact

At the other:

optimization objective ? autonomous generation ? automated evaluation ? artifact

There will be countless arrangements between them.

The important questions will concern agency.

Who initiated the process?

Who made consequential choices?

Who rejected alternatives?

Who decided what the work meant?

Who decided it was finished?

Who is accountable for it?

And perhaps most importantly:

Did anyone care whether this particular thing existed?

There is deep irony in writing this argument with artificial intelligence.

I am not standing outside AI-generated culture warning everyone to put down the machines. I am creating with them. I find the collaboration exhilarating. It allows me to make things I simply could not make alone.

That experience is precisely why I think the distinction matters.

AI can dramatically expand human creative agency.

It can also remove human intention from cultural production almost entirely.

Those are not the same future.

One is a world in which humans gain new instruments.

The other is a world in which culture becomes an autonomous optimization system constantly generating increasingly effective material for human consumption.

The technology underlying both futures may be almost identical.

The difference is the relationship we choose to build around it.

As generative capability becomes abundant, we may need new cultural norms around provenance, authorship, curation, disclosure, intentionality, and perhaps even spaces deliberately reserved for slower forms of creation. Not because machine-generated culture is inherently corrupt, but because societies should understand when culture is communicating someone's intention and when culture is being algorithmically grown around their behavior.

For thousands of years, we largely assumed that behind every song, picture, story, sculpture, or performance there was someone who, for whatever complicated reason, wanted to make it.

That assumption may soon stop being safe.

A cultural artifact might exist because someone had something to say.

It might exist because a human and an AI discovered something together.

Or it might exist simply because a system calculated that we were likely to keep watching.

Those things may look remarkably similar from the outside.

They are not the same.

When creation becomes unlimited, intention becomes scarce.

And the future problem of art may not be who made it.

It may be whether anyone meant it.