Someone else designed the letters you are reading.
I chose the words, but I did not design the typeface. I did not develop the software that arranges the paragraphs, the equipment that stores the text or the technology that makes it visible on your screen.
A machine can correct my spelling. Another can suggest alternative phrasing. I can rearrange entire chapters without copying a single page by hand.
For most of my life, I have accepted these things as ordinary parts of writing.
Then I began using artificial intelligence to develop Max Aura Dungeon, and the tools themselves became part of the debate about whether I was creating art.
Apparently, some tools are much more philosophically troublesome than others.
And I can understand why.
Where the work happens
There are forms of writing in which the physical production of the letters is central to the achievement.
Calligraphy is an obvious example. The movement of the hand, the shape of a line and the relationship between marks can become significant artistic qualities.
The writer of a novel usually works differently.
The design of an individual letter matters relatively little compared with the arrangement of words, sentences, scenes and larger narrative structures.
Both activities involve written language, but they place their creative attention in different places.
Photography introduced another kind of complication.
A photograph can depend on the physical world, the properties of a camera and the behaviour of light. The photographer may make decisions about timing, framing, focus, exposure and what deserves attention.
The machine contributes something substantial to the production of the result.
That does not mean every photograph involves the same degree of artistic intention or achievement.
A casual snapshot and a carefully composed photograph can emerge through similar technology while representing very different kinds of work.
The difference becomes even more complicated in film, where writers, actors, directors, editors, composers and many other participants may contribute to a single finished work.
We have long been comfortable with art that involves several people and technologies.
What we have not always agreed upon is how those contributions should be understood.
The object somebody decided to call art
In 1917, Marcel Duchamp submitted an ordinary urinal, signed R. Mutt and titled Fountain, to an exhibition organized by the Society of Independent Artists.
The exhibition was supposed to operate without a jury.
The work was rejected anyway.
I find that detail particularly entertaining.
An institution that presented itself as open to artists still encountered something it was unwilling to recognize.
Duchamp’s work subsequently became one of the most discussed examples of the readymade: an ordinary manufactured object selected and presented in a context that challenged assumptions about artistic creation.
The Philadelphia Museum of Art’s account of the Fountain controversy explains how the original object was lost and how its influence spread through documentation and later replicas.
The object had not been manufactured by Duchamp’s hands.
The intervention involved selection, presentation and the questions that followed.
That is a useful challenge to the assumption that artistic creation must always involve physically producing the thing being presented.
But I would resist making selection a magical ingredient that transforms every choice into art.
Choosing which chair to put in my kitchen is a decision.
Selecting one sentence from several alternatives is a decision too.
Some choices are casual. Others involve considerable thought, knowledge and an understanding of what the result is supposed to accomplish.
The interesting part is not simply that somebody made a choice.
It is what kind of choice they made and what that choice contributes.
A beautifully written instruction manual
Consider a technical manual.
A good manual may require extraordinary knowledge, clarity and discipline.
Its author must understand a complicated subject and anticipate the difficulties of someone encountering it for the first time.
The illustrations might be elegant. The explanations might be memorable. The structure may guide the reader through a demanding task with admirable precision.
We can recognize that achievement without agreeing that every good manual is automatically a work of art.
That does not prevent an instructional text from being appreciated artistically.
It suggests that craftsmanship, purpose and artistic status do not always follow one another in a simple sequence.
The same difficulty appears in advertising.
A commercial can tell an excellent story while attempting to sell something.
A piece of music can be composed for a film, a game or a marketing campaign and still become meaningful independently of its original purpose.
The commercial intention does not necessarily eliminate the artistic qualities.
Neither does the presence of artistic qualities make every commercial admirable.
Ghostwriting creates another useful complication.
Someone may provide their experiences, ideas and direction while another person writes much of the prose.
The resulting book might be excellent.
Its quality, however, does not settle questions about credit, authorship or whether the arrangement was described honestly to readers.
We are capable of recognizing several kinds of contribution when human collaborators are involved.
AI asks us to examine those distinctions under quite different circumstances.
The work that happens between people
I encountered a different kind of collaborative storytelling long before I began writing novels.
As a young Dungeon Master, I created places, situations and possibilities for other people to explore.
I could prepare an encounter, imagine how its inhabitants might respond and establish the rules governing the world.
Then the players arrived.
They made decisions I had not anticipated, misunderstood things I thought were obvious and occasionally pursued possibilities that were much more interesting than the ones I had planned.
The resulting story depended on the material I provided and on what they did with it.
My role was substantial, but it did not give me complete control over every event.
That experience helped make me comfortable with creative work in which the person initiating something does not personally determine every detail.
It also taught me that a surprising contribution can be wonderful or disastrous.
Sometimes an unexpected action makes the entire story more interesting.
Sometimes it destroys an encounter that required several hours of preparation.
The distinction tends to become obvious remarkably quickly.
When a tool begins making suggestions
AI-assisted writing introduces a different relationship.
A language model can contribute words, dialogue, scene structures and ideas. It can produce passages that appear polished enough to publish almost immediately.
That capacity is precisely why I do not think it makes sense to describe the technology as though it were merely a sophisticated spellchecker.
It is doing something more substantial.
But the fact that it produces substantial material does not tell us how the finished work is developed.
When working on Max Aura Dungeon, I may ask for a scene involving established characters and receive dialogue that sounds perfectly reasonable.
Unfortunately, sounding reasonable is not always what a character needs to do.
A frightened person may become unusually articulate. Someone whose behaviour should reflect a complicated history may respond like a familiar fictional archetype.
An uncomfortable situation can become neatly resolved because the generated prose seems determined to make everyone behave sensibly.
Sometimes a passage contains an unexpected idea worth keeping.
At other times, the most useful thing it does is show me what I do not want.
I then have something tangible to reconsider.
The direction may change. A character may need a different response. An entire scene may be discarded because the words are convincing in isolation but wrong for the story.
These are decisions I make while working with material that the system has helped produce.
They are not the same as having written every original sentence myself.
I see no reason to pretend otherwise.
Making a choice is not the whole achievement
The relationship between a creator and a tool can involve many different arrangements of responsibility and control.
A person might select an untouched AI-generated story and publish it.
Another might use generated prose as an initial draft, rewrite extensively and construct a larger work through numerous deliberate choices.
A third might use AI only to discuss possibilities while writing the text independently.
Calling all three processes AI-assisted writing is accurate enough at a distance.
It tells us remarkably little about what actually happened.
I think that is part of why arguments about AI and artistic creation become so unsatisfying.
One person is imagining a machine producing an entire work without significant human direction.
Another is describing years of worldbuilding, sustained revision and deliberate editorial choices.
They may disagree intensely while discussing processes that have little in common beyond the technology involved.
That does not mean every form of human involvement constitutes an equally significant artistic contribution.
Nor does the history of photography, printing or readymades prove that AI must be treated exactly like earlier tools.
The differences deserve examination.
But so do the actual creative decisions people make with them.
When I choose between two passages, I am not merely voting for the one that sounds nicest.
I am considering the character, the scene, the larger story and the experience I want the words to make possible.
Sometimes I choose neither.
Sometimes I discover that the scene itself needs to become something else.
The interesting work often begins at the point where a perfectly usable sentence turns out to be the wrong one.