♪ Take notes ♪The rules of honesty at this station.
I used to spend two months writing a 120,000-word science fiction novel with AI.
It was the first time that I was so immersed in a virtual world that I conceived. I began to think about why people should act like this, what would be the consequences of a rule, where a pen should be buried and when the reader should discover it.
I even came up with a word of my own called the Narrative Engine.
The so-called narrative engine is why the reader continues to look.
A story can’t just keep satisfying readers. You give an answer and you give it new consequences; you solve a problem and you make the reader realize that there is a bigger problem behind it.
Some narrative engines can run through the whole work.
For example, the civilizational crisis of Triangular, the existential crisis of El Niño. Crises and countdowns are good, because the less time, the less space a person can escape. The person must make a choice, which exposes his character, relationship and values.
In writing, I began to consider the reader’s mood map: What he expected now, when he should be satisfied, when to interrupt him, when to make him nervous and when to give him a brief breath.
I didn’t learn it after reading a few writing books, but I realized it when I went into the world and tried to write a story.
But it didn’t work out.
I voted it for the editor, not approved. I’ve tried to write online in AI, and the results are not ideal either.
That’s what made me rethink one thing:
AI, can you write a novel?
AI is able to write a 60-ton novel in a stable way.
If the criteria are simple sentences, complete sequences, people can speak, chapters can move forward, then AI can already write well.
It can even be written more like a novel than many untrained people.
The problem is that novels themselves are a product whose reproduction costs are close to zero.
The cost of copying a million copies of a 100-cent novel and a 60-percent novel is almost the same. Why do readers spend time reading a 60-cent novel when they can choose a really good one?
Many commodities have a value of 60 per cent.
A 60 per cent software could solve practical problems, a 60 per cent report could help people to finish their work, and a 60 per cent poster could be enough for publicity.
But that’s not true.
It is not a functional entry point, but a few hours, or even dozens of hours, in the lives of readers. It must be worth the reader giving it his time.
Thus, AI ’ s ability to produce 60-percent novels on a large scale does not mean that it can create the value of novels on a large scale.
On the contrary, when there is an infinity of 60 cents, what is really scarce becomes those parts that cannot be stabilized: the author ’ s unique experience, judgment, desire, prejudice, and what he wants to say with this story and someone else.
AI cannot escape the “AI taste curse”
A lot of people think that if you change a few common sentences, less dashes and less rows, you can get rid of the AI.
But I don’t think it’s a fixed “smelt.”
It is not a word, nor a sentence, but a natural human alert to a standardized paradigm.
When an expression is mastered by AI on a large scale, humans quickly get tired of it. AI begins to imitate human expression, and humanity continues to seek new expression. AI chases back, humans run ahead.
So there is no hint to permanently eliminate the taste of AI.
It looks like someone wrote it today, and once it’s been repeated 10,000 times, tomorrow it’s going to be an AI.
The deeper question is that the essence of novels is not just language.
The novel is a spiritual exchange between the author and the reader. The author turned himself into a character, a destiny and a world that he could not speak directly; the reader suddenly recognized himself in the story of another person.
It’s not a set of words that readers really want to touch, but the person behind them.
AI can imitate expression, but it is difficult to replace “Why does one have to write these things down?”
But I still think it’s worth writing novels in AI.
Because the value of AI writing novels may lie not in the reader, but in the author.
In writing Rebirth: Review, AI helped me express a lot of things that could only be obscured.
Then I began to judge.
Is that what I want to say?
Why would this person make such a choice?
Isn’t it too early to satisfy the reader?
Can you recycle this pen?
What is the narrative engine that now drives the story forward?
The more AI produces, the more judgment I have to make. It is also in the process of denial, selection, feedback and revision that I begin to truly understand writing.
This 120,000-word novel may not be a successful work, but those two months were not wasted.
For the first time I entered the world I created, and for the first time I started thinking like an author.
I begin to understand why someone is willing to complete a novel for a year, two years, or even longer. Because novels do not quickly produce a story, they take a long time to understand what they do not understand.
AI is more like a writing aid wheel.
A person who has never written a novel, facing a blank document, is probably unable to write a single sentence.
AI can make the world work first.
It can help the author complete the scene, try the structure, simulate the conversation, organize the set, or temporarily express what is still vague in the author ’ s mind. The author does not have to start with an absolute blank, but can judge and modify it in an imperfect version.
It’s like a bicycle wheel.
The auxiliary wheel can’t win for you, but it can make you feel riding for the first time. And when you get to the balance, the most important thing is that one day it can be removed.
I’m still going to use AI to help write.
But if I really want to write a novel of my own, I could take a year, maybe two. I need to write my own scenes, find my own language, live with people, and change them over and over again.
It’s certainly much slower than making AI produce dozens of chapters a night.
But novels may be one such thing: it is only when the author is willing to put his real time in it that the reader is willing to give it to him.
So if I wanted to sum up this experience, I wouldn’t say:
I wrote a novel with AI.
I’d rather say:
I wanted AI to be a writer faster for me, and then I found out that AI just sent me to where the writing really started.
The failure was the value of the novel as a commodity.
Success is that I started being a writer.