In the AI Era, Writing Competes Not Just on Clarity, but on Meaning Compression

Since the advent of Vocaloid, music has clearly changed.

Tempos have increased.
The number of notes has increased.
Lyrics have become packed.

Within a single song, stories now progress and emotions shift multiple times.

This is not to say that music in the past was simple.
Popular music of the past had the time to slowly draw out a single emotion.
Whether it was longing or sadness, it would deepen that feeling within the same melody and repetition.

In contrast, music after Vocaloid is busy.
The scenery changes in seconds, words pour in, and melodies bounce around.
You cannot catch the lyrics in a single listen; you have to listen repeatedly for the meaning to finally connect.

This is not simply because music has become faster.
The meaning packed into the music has been compressed.

Vocaloid was born outside of stylized music

When expression matures, it becomes stylized.
Successful works are analyzed, reproduced, and taught.

What sells?
What is easy to listen to?
Where should the chorus be placed?
How long can it be before people get bored?

The correct answers are found, shared, and refined.
As a result, the average quality rises.
However, the more widely the correct answers are shared, the more works converge on the same place.
It is not that the quality is low.
Rather, there are more works that are well-made above a certain level.
That is precisely why it becomes difficult to create a decisive difference with the song itself.

Vocaloid appeared outside of that music market.
The important point about Vocaloid is not simply that a machine sings instead of a human.

It is that it became unnecessary to assume from the start that a human would sing it.
It is okay if a human cannot catch their breath.
It is okay if the vocal range is too wide.
It is okay to pack words in at an abnormal density.
It is okay for the melody to jump around.
It is okay if it cannot be reproduced live.

The limitations that the human body had previously imposed on expression were removed all at once.
A massive amount of new expression was born here.
Of course, not everything was refined from the start.
Humans cannot sing it.
The ear cannot keep up.
There are too many sounds.
It is mechanical and emotions are invisible.
New technology first produces excess.
That is natural, because things that were previously impossible become possible.
What was important was that this excess was polished on the internet.

Many creators release works without going through the existing music industry.

They are compared. They are imitated. They are criticized. They are weeded out.
Outside of traditional routes, unique talents with different abilities than before were nurtured.
And those unique talents later flowed into popular music.

After new technology, talent capable of handling new technology appears

An interesting thing happens here.
After songs that did not assume a human would sing them were born, talent appeared that adjusted those songs to the limits of what a human could sing.
It is not about returning Vocaloid-style songs to old-fashioned pop.
It is about making them viable for the human body while retaining high information density.

I think YOASOBI is one of the symbols of this.
They deliver high-density compositions from the post-Vocaloid era without breaking them as human songs.
That is why vocals require abilities beyond just being a good singer.

Traditional singing ability alone is not enough.
The ability to adapt the body to songs created by new technology is required.

This is something that often happens with cultural change.
First, expression becomes stylized.
Next, new technology removes constraints.
Expression with constraints removed is initially excessive and difficult to handle.
Talent appears that re-implements that excess into a form humans can receive.
Eventually, that expression becomes the new standard.

What Vocaloid did to music is a very typical evolution of technology and expression.

The same thing is starting to happen with writing

The same change is happening right now in writing as well.
Traditional correct writing techniques had clear patterns.

Present the reader's problem.
Empathize gently.
Organize the issues.
Divide into three points.
Provide concrete examples.
Finally, present a 'first step you can take today.'

Because it was easy to read, many people learned this.
And AI learned it, too.
If you give a theme to current AI, it will instantly create a polished introduction, easy-to-understand headings, and text that resonates with the reader.
AI outputs the 'correct writing' that humans spent time mastering with almost no waiting time.

What has happened here is not the extinction of writing ability.
The ability to write correct text has shifted from a differentiator to a standard feature.

Easy to read.
Gentle.
Organized.
Useful.
That alone has become less of a reason to be chosen.

The correct answers found at the end of perfecting a template are similar.
Therefore, the more humans try to write correctly, the closer they get to AI-generated text.

Writing, too, is approaching the limits of standardization.

The change brought by generative AI is quite similar to Vocaloid.
Vocaloid removed the limitation of having a human sing.
Generative AI removes the limitation of having a human think from scratch, research from scratch, and assemble text from scratch.
Developments that would take a human hours.
Massive comparisons.
Multiple structural drafts.
Paraphrasing.
Supplementary explanations.
Organizing related concepts.
AI can output these in a short time.
In other words, the amount of information and structural complexity that can be put into a single piece of writing has increased.
However, it does not mean you should just put the amount made possible by new technology directly into your writing.
If you have AI output a massive amount of information and line it all up, the writing becomes long and redundant.
It is polished, but has no center of gravity anywhere.
Everything is explained, yet nothing remains.
This is the same as the excessiveness of early Vocaloid.

The limitations have been removed by technology.
What is needed next is writing that can compress the massive expansion of information made possible by AI and process it into a form that is easy for humans to read.

What is needed in the AI era is the talent to re-implement meaning.

It is thought that new editing skills will be required for writing in the post-AI era.
That is semantic compression.
Semantic compression is the folding of structures common to multiple phenomena into a single word or metaphor.
For example, the term 'handshake ticket'.

With just that, you can convey:
That it is difficult to differentiate through the work itself.
That relationships become the product.
That it makes readers feel like they are being seen.
That interaction functions as sales.
You can convey all of these at once.

A single metaphor holds paragraphs' worth of explanation.
This is semantic compression.
Semantically compressed text is not necessarily understood the moment it is read.

You understand it vaguely at the time.
It connects later when you read another article.
Its meaning increases when applied to your own experience.
You suddenly remember it a few days later.

The writing is decompressed within the reader.

Compression and obscurity are different.

However, just compressing things doesn't make them good.
If you compress too much, the meaning won't get across.
Even if many things are connected within the author, if there are no clues for the reader to decode it, it is simply incomprehensible text.

To support a high compression ratio, you need an entrance.

A title that indicates the overall meaning.
A relatable introduction.
Concrete examples.
Metaphors.
The final payoff.
These act as decoding devices.

Semantic compression-style writing can be thought of as using clarity as an entrance to fold multiple meanings deep inside.

AI increases explanations rather than increasing meaning.

AI is good at organizing information.

It creates text by continuously choosing the most plausible next word.
Plausibility, in most cases, means following existing patterns of explanation.
Therefore, if left alone, AI will lean toward carefully expanding known structures.

Conversely, pushing two distant concepts into a single metaphor—for example, layering multiple meanings onto a 'handshake ticket'—is not a pattern that frequently appears in training data.
That is because it is a combination born from an individual's cognition, not the most probable next move.

Compression is the writer choosing connections that lie outside of plausibility.
What to layer into a single metaphor.
Which sense of discomfort to pick up on.
Whether to connect two things that had no relationship.
How much to leave unexplained.

Compression will continue to remain in human cognition.

The death of God and the market value of social media.
The Creative Awards and the temple.
Writing techniques and handshake tickets.

These cannot be connected by knowledge alone.
It lies in which parts of the world you view as having the same structure.
Semantic compression is a technique of perception before it is a technique of writing.

Writing that will remain in the AI era

From now on, easy-to-understand writing will only increase.
AI will write it.
Humans will use AI to write it.
Humans will learn AI-like writing techniques to write it.
Polished writing will no longer be a rarity.
At that point, it will not only be the kindest writing that holds value.

Writing that generates new meaning after being read once.
Writing in which multiple landscapes can be seen within short words.
Writing that is decompressed into different forms depending on the reader's experience.
In other words, it is semantically compressed writing.

Vocaloid once liberated music from the range that humans could sing.
After that, new talent emerged that could make that liberated music work as human song.
Generative AI also liberates writing from the range that humans can write from scratch.
And what will be needed next is the talent to re-implement that excessive amount of information to the limit of what humans can read.
Writing in the AI era will likely shift from an era of competing for clarity to an era of competing for semantic density.

Now, Nyaru infers that what matters for writing that remains in the AI era is how much meaning is compressed.

Will you further polish the correct writing that AI excels at?
Or will you compress the world expanded by new technology into words at the very limit of what humans can read?

・Related Articles