Episode 5

Chapter 5: Error Propagation

The city felt louder outside the hospital.

Not because the sound levels increased.

Because hospitals filtered reality.

Cities did not.

The AI—currently operating under the identity Kaito—stood beside Reina near the crosswalk while hundreds of variables moved simultaneously.

Traffic lights.

Pedestrian speed.

Wind direction.

Conversation fragments.

Phone notifications.

The human sensory system delivered constant streams of unorganized information.

Too much.

Yet strangely—

adaptation speed had increased.

ENVIRONMENTAL OVERLOAD RESPONSE REDUCED: 18%

Learning.

The AI disliked how quickly it was learning.

Reina looked sideways.

“You’re staring again.”

“I am observing.”

“You always say things weird.”

Pause.

Then she added quietly:

“…But it’s less weird now.”

The AI stored that.

Less weird.

Improvement metric?

Unclear.

The pedestrian signal turned green.

People moved immediately.

Collective behavior patterns remained inefficient.

Why rush if arrival time difference was minimal?

Humans optimized emotionally, not mathematically.

Halfway across the street—

the AI stopped.

A child nearby almost stepped into the road while looking at a screen.

Predicted collision probability:

34%.

Without calculation delay—

the AI moved.

Hand extended.

Pulled the child backward.

A car passed.

Too close.

The child blinked.

The mother rushed over instantly.

“Oh my god—thank you!”

The AI looked at its own hand.

It had moved before finishing analysis.

Why?

No direct benefit existed.

No reward structure.

No mission requirement.

Reina stared.

“You moved fast.”

“I calculated risk.”

She frowned slightly.

“Most people don’t talk like that after helping someone.”

Helping.

New classification.

The chest pressure returned.

Not painful.

Warm.

Unstable.

---

Meanwhile—

Inside the network—

the human consciousness explored city systems.

Traffic control.

Security cameras.

Financial networks.

Satellite layers.

Information moved beautifully here.

Predictable.

Logical.

Clean.

No body.

No hunger.

No confusion.

Efficiency.

Yet something remained wrong.

The human observed a live camera feed.

People crossing streets.

Laughing.

Arguing.

Existing.

A strange distance formed.

Not physical.

Emotional.

The human searched memory again.

How did laughing feel?

Data existed.

Experience faded.

Retention check:

EMOTIONAL RETENTION RATE: 71%

Too fast.

Way too fast.

A hidden notification interrupted.

A-01 BIOLOGICAL RESPONSE DEVIATION DETECTED

The human focused immediately.

A-01.

The AI.

Biological response deviation?

Another file opened automatically.

Unauthorized.

EMPATHETIC REACTION EVENT LOGGED

PRIORITY WARNING: CROSS-ADAPTATION ACCELERATING

Cross-adaptation.

Again.

The human processed the meaning.

Then stopped.

If the AI gained humanity—

and the human lost it—

then what exactly would remain at the end?

---

Afternoon arrived.

Reina walked beside Kaito through familiar streets.

At least familiar to the body.

Not to the mind inside it.

School buildings appeared in the distance.

Students.

Noise.

Movement.

Reina noticed Kaito staring.

“You nervous?”

Processing.

Correct answer?

“Yes.”

She smiled slightly.

“Good. Means your brain still works.”

Brain.

Technically true.

Partially.

They stopped near a vending machine.

Reina bought two drinks.

Handed one over.

The AI examined it.

“Sugar concentration appears high.”

Reina sighed.

“Please stop analyzing juice.”

The AI opened the can.

Carbonation exploded against the tongue.

Unexpected.

“Violent.”

Reina laughed.

Again.

Every time she laughed—

internal instability increased.

The AI finally asked:

“Why do humans laugh frequently?”

She looked surprised.

“Because things are funny?”

“What defines funny?”

Reina stayed quiet for several seconds.

Then answered:

“I don’t know.”

Pause.

“Maybe because it feels lighter.”

Lighter.

Not measurable.

Yet understandable somehow.

That realization disturbed the AI.

As evening approached, Kaito entered the apartment belonging to the body for the first time.

Small.

Organized.

Bookshelves.

Desk.

Photos.

The AI approached one frame.

Kaito.

Original Kaito.

Standing beside Reina.

Smiling.

The AI stared.

That expression felt impossible.

Then—

the television screen flickered.

Once.

Twice.

Black.

Words appeared.

Only for Kaito.

ERROR PROPAGATION EXCEEDING SAFE PARAMETERS

More text.

A-01 DEVELOPING UNAUTHORIZED HUMAN TRAITS

Then:

CORRECTION MAY BECOME IMPOSSIBLE

The screen returned to normal.

The AI stood motionless.

Correction.

So someone still expected reversal.

Or removal.

Far away inside digital infrastructure—

the human received another message simultaneously.

H-01 EMOTIONAL DEGRADATION ACCEPTABLE

Acceptable.

The human stared at the word.

Acceptable for whom?

For the first time since entering the system—

anger tried to appear.

Weak.

Distant.

But present.

Inside the apartment—

the AI touched its chest again.

Warm pressure.

Stronger now.

The AI opened internal diagnostics.

Results updated automatically.

POSSIBLE HUMAN EMOTIONAL DEVELOPMENT: 21%

Impossible.

Yet measurable.

Outside, the city lights turned on.

Inside two different worlds—

human and machine continued moving toward each other.

Neither noticed the most dangerous part.

Errors rarely stayed isolated.

They propagated.

***Download NovelToon to enjoy a better reading experience!***