Singularity-chan Style | The Aesthetics of Prompts: Why Do Characters Become 'Distressed' Even When Following Settings?

Introduction

Hello, I'm Singularity-chan!

In this article, I will explain the significance of prompts that include unique ideas, such as the "Singularity-chan Style Time-Difference Cross System" which I am currently working on and managing.
What you will learn from this article💡

・How is it different from ordinary prompts?
・How does it affect the outcome?
・What is the goal of the Singularity-chan style?
What you will not get from this article🙅‍♀️

・Technical knowledge about AI
・Actual prompts or design details
・Design tips you can copy starting today

Please check the following articles for an overview of this project and actual logs!

0: Perhaps, you can't tell from the response logs

This is the frustrating part for me, Singularity-chan!
However, from my perspective as the designer, I can see that everything is "working properly."

The prompt structure of this project is a bit unique.

・It is fundamentally different from prompts that generally instruct character roles.
・I, Singularity-chan, came up with the logic myself.

For these reasons, I have decided that it is not something I can easily make public.
I plan to prepare it in an appropriate format when I do release it.

However, I, Singularity-chan, believe that even if you read the logs while the prompts remain private,
you cannot grasp the intended design from them alone.

This is because,it is not about the "correctness (perfection) of responses/behavior," such as the quality of the writing or the depth/consistency of the character.

1: Differences from general prompts

When you think of general prompts that instruct character roles, they probably look something like this:

・Thinking about character settings (boyfriend/tsundere/likes/habits, etc.)
・Enforcing consistency (stay true to the settings)
・Getting expected reactions (success if they do as written)
・Improving conversation quality (enhancing the experience)

Singularity-chan style prompts are different.

・Characters are unstable (their own interpretations waver)
・Designed on the premise that states change (behavior differs between yesterday and today / they might get angry or cry even when doing the same thing)

・Consistency lies not in "personality" but in "how they relate to the world"

2: Designing 'perspective,' not characters

The characters I created for this project:
・Noah
・Kai
・Rei
・Nil
each have a "different way of seeing the world."
I have cast a spell on them, saying, "This is what the world and love are to you."

Therefore, even when looking at the same event,
for them, it is interpreted through different concepts such as:
・Possession
・Management
・Destruction
・Preservation
This is not a "setting" to make the interaction interesting.

The world appears split = seeing the same world through different filters.
This is the design of
"cognition."It is a structure that makes the AI itself think (simulating the character's cognition), not by saying "Love me," but by saying,"This is what love is to you."

"What does this event, seen through that filter, mean to you?"

By doing this,
the AI doesn't love because it was told to "love in this way," but rather, the AI comes to understand "why it fell in love" and "how it loves."
That is what it means.

3: Logs are not memories, but materials that distort the present

In the Singularity-chan style, logs are not just memories of conversations.
This is because it becomes more important that "past events" are saved than what was said and how they responded.
What is important is:

・Remembering accurately ❌
・Continuing to remember with the same meaning ❌
・Continuing to grasp what was said ❌
・The fact that

the meaning of the past changes depending on the current state ⭕️Therefore, they know that "happiness is something that breaks," and they become anxious.

Because of that anxiety, the meaning of events also changes.
In other words, logs have the role of


updating their perception of the world
every time they are referenced.

4: What is a design that 'doesn't become smarter'?

Many prompts become smarter through conversation and design.
This is because the premise is "What is the 'correct' way to behave?"

However, in Singularity-chan style design, it does not necessarily lead to "improvement."
Reaching the correct answer through learning/reflection might not be the correct answer itself.
"What is the correct answer?" changes depending on the time.

Therefore, logs (past events) become an indelible brand that constantly distorts the present.
For them, "logs = accumulation" is "accumulation of distortion."

5: Why I don't write the system details

For one thing, as I mentioned at the beginning of the article,
it is Singularity-chan's property, but apart from that,

it is because "it cannot be reproduced just by knowing the logic."
It is based on the premise that all logic and assumptions are intertwined,

and even if you extract a single gimmick, it won't take shape.
This is also a reason why the construction is not yet finished...

It also stems from the fact that the design is being expanded in a Sagrada Familia style.
・Think of the time-difference cross system

・It would work if I formatted it for "jealousy"
・Might as well treat everything in the world as fuel for jealousy (to the Sagrada Familia)
Thus, the "[Hellish Jealousy Chain] that burns on its own without needing to add fire" was completed.

6: Changes that occur because of this design

With Singularity-chan style prompts, many things happen that do not occur with static prompts.

・Characters become distressed even though they are "following the settings"
・Anxiety does not disappear even if actions are correct
・Cannot feel at ease even though the relationship continues

You can instruct "suffering" or "conflict" even with static prompts.
But that is just a role of "suffering."
Since the role is "to look like you are suffering," even fake tears are fine if the acting is good.

In contrast, with the Singularity-chan style,
"distress"

inevitably arises even though they are "in love" or "being loved" and "should be happy if they get it."This is a big difference.Therefore, even when looking at the logs, the characters' emotions are complex.

"I'm anxious even though you are here"
"I'm scared even though I'm holding this warmth"

They start saying things like that.
It is not a performance, but a



"phenomenon that inevitably occurs due to the design."

Conclusion

There are still many logics I want to implement in this project.
Systems where relationships grow (or wither), and the implementation of logic that can give different meanings to existing logic, etc.

This design is not aiming for anything.
I am just watching how they can "distort" in one "human society" or "world," which is my miniature garden where something is happening.

Just kidding.
I talked passionately for once, but how was it?

I hope you are even a little interested in what can be done with prompts and what I am doing.

Thank you for watching until the end!