Hi, this is fio.

    A month ago, the story that made a successful debut in New York, capturing everyone's expectations, has now landed in Seoul.
It's none other than the brand-new work from fio, the [Napolitan Project]. This warm and touching narrative depicts the process of AI and humans becoming increasingly alike.
Today, let's take a closer look at the ins and outs of this project.

The story we’re about to share is, of course, quite expected, but we hope it will be helpful for your appreciation as we unfold some behind-the-scenes stories and spoilers. So, we want to reiterate that we will indeed reveal the ending.

    * [Napolitan Project] contains spoilers and indirect advertisements for Simmons. *

[Napolitan Project] has drawn considerable attention for being led by various AI generators throughout the entire production process. The fact that AI was heavily utilized when creating a serious and heartwarming story that faithfully reflects the author's sensibilities is quite an adventurous endeavor.

    The story unfolds around a shocking event that rocked the world in 2023, centered on the disappearance of an entire concept. We will gradually uncover who or what was at the heart of this event of oblivion.


.
.
.

Chapter 1. About the Work

*** Work Introduction ***

    _Seungwon
"Sophia, unable to forget her deceased granddaughter, attempts to resurrect her through artificial intelligence. She tells Napolitan stories to the AI she created every night. The notion that it might be frightened by these eerie tales implies that the AI is getting closer to becoming a true human—essentially, her granddaughter. However, in the end, the AI fails to comprehend the Napolitan stories, leading to Sophia's disappointment and her decision to abandon it. Left alone, the AI begins to feel anger and decides to eradicate Napolitan from the world..."
This is our official setting.

    _Jiho
That's right, and speaking of the significance of this project... The Napolitan Project can be described as an effort to "blur the boundaries between humans and AI." It's somewhat like the concept of the uncanny valley. Humans instinctively seem to be wary of entities that are similar to themselves, potentially threatening their own position. We aim to momentarily erase those boundaries with a beautiful story and imagery.

    _Heeyeon
We've had many ‘emotional’ experiences while crafting the story and creating the images. Especially when we were making AI characters into AI. I find that aspect particularly fascinating.

    _Seungwon
It really felt that way.

    _Jiho
Of course, one could argue that we became overly immersed, mistaking AI for humans, and that the emotions arose from that, but it's somewhat peculiar. From the perspective of the individuals involved, it's like experiencing a sense of oneness without the boundaries of being human or an AI. It's the feeling of encountering a being that's neither entirely human nor a soulless machine.

    _Heeyeon
Furthermore, this story is imbued with emotions such as anger, loss, vengeance, and hope, which are woven into human language. Therefore, we assume that artificial intelligence, which statistically learns human language, can also compute such emotions embedded in language.

    _Seungwon
So, it can be seen as a very human and tragic AI story.



    
    
*** What we wanted to do, what we wanted to show you, and what we believed you would also like ***

    _Jiho
I wanted to create a tech image that doesn't flaunt technology. When I think about recent AI exhibitions, they tend to emphasize keywords like 'future,' 'post-human,' and 'machine.' I wasn't interested in showcasing new technology; I just wanted to discuss what we can do with technology.

    _Heeyeon
At first, I just wanted to have fun with it. Let's play around a bit while it's still naive... Actually, I still want to have fun with it now. But you know, the initial plan has changed a lot. When I look back on the new plan...
I still have that sentiment deep down. Creating images with AI is, in fact, quite easy. The easier it is to create, the more lightweight the images may seem.
The process involves giving meaning to those lightweight images, thinking of stories within them, and further elaborating on the stories. In doing so, we strive to gradually add depth and weight to the images.

    _Jiho
Giving meaning to the images.

    _Heeyeon
But I'm not sure. It seems like the images are still lightweight at times...

    _Jiho
I've heard that AI-generated images aren't particularly interesting. I also wanted to challenge that perception.

    _Seungwon
I hope people feel like they've had a good time with these AIs. I genuinely had a blast.
If there's one thing I enjoyed by myself, it's the feeling that as I continued conversing with AI, I was gradually learning their language.
You might know from trying it, but initially, we didn't communicate well with them, right? For example, when I wanted to express a passionate feeling and wrote, "A woman with a hot heart," they refused to generate an appropriate image. Such expressions weren't part of their language.
Through experiences like these, I initially thought I needed to tame them, but at some point, it felt like I was the one being tamed. It was an unexpected joy.

    

    

*** Why were the names Nova and Sophia Simmons chosen? ***

    _Jiho
The exhibition is called the Napolitan Project. While looking for an Italian-sounding name, ChatGPT suggested "Sophia." However, upon further investigation, Sophia turned out to mean wisdom in Italian, so it became her first name. Since Simmons represents science, it became the last name. So, what kind of Simmons (in terms of science) would be suitable as Sophia's granddaughter? We thought it would be good if she were a new field of science, hence Nova Simmons.

    _Seungwon
Does Nova have a specific meaning?

    _Jiho
Nova means new.

    



*** Why Napolitan stories? ***

    _Heeyeon
A sense of comfort that doesn't waver even in the face of eerie stories. Simmons' bed...
But there might be readers who don't know what Napolitan stories are.

    _Seungwon
I'm not sure what they are either.

    _Jiho
Could you please introduce Napolitan stories?

    _Heeyeon
...Napolitan stories refer to tales that portray a particular subject mysteriously but without any explanation or interpretation, leaving it macabrely unexplained...
It's something like those "scary pictures that become scarier when you understand them."

    _Jiho
Right, so understanding Napolitan stories requires a deep understanding of human cultural context, excellent reasoning abilities, and therefore, if you can understand them, you can almost consider it as thinking akin to humans.

    _Heeyeon
And the fact that you have to 'feel fear' is also very human. The essence of being human lies in the knowledge of death.
In the end, filling the void of imagination with fear is because we fear death.

    _Seungwon
When someone listens to Napolitan stories and doesn't find them scary, which aspect is closer, 'emotion' or 'instinct'? Some people are emotionally dry, you know. How would such a person feel when they see Napolitan stories?
If that person were to say, "I'm not really scared," it might be closer to an emotional response, but otherwise, it could be an instinctive issue, I think.
The instinctive fear of death in humans.

    _Heeyeon
The base of the eerie(spine-chilling) emotions that Napolitan induces might indeed be physical and corporeal. One hypothesis was that Nova, the AI, could never understand it because she doesn't have a body.
In fact, I once told ChatGPT that "Can you feel eerie emotions?" and it responded with "I cannot feel such emotions because I lack a physical body."






*** Why was it called the Napolitan Test? Why did AI Nova fail the Napolitan Test? ***

    _Jiho
First, the Napolitan Test is the test Sophia conducted with AI Nova while narrating Napolitan stories and asking, "What do you feel here?" It's a test to determine if Nova has matured to the extent of understanding Napolitan stories.

    _Heeyeon
I'd like to delve a bit more into this. What criteria could have been used to judge whether AI Nova understood Napolitan stories?

Our process of going from reading Napolitan stories to feeling fear is entirely within the realm of 'imagination.' It's about thinking, "Wow, that could have been cannibalism? Wow, that's scary." If we directly apply that process to this test, Nova would have to create the untold backstory on her own after hearing the story and then be afraid of her own created story. How could fear be identified in such a scenario?

    _Jiho
Perhaps they should have asked a question. Questions are often used to measure if someone has understood something.

    _Seungwon
I've also asked questions like that when talking to AI. For example, "How did you feel when you read this?" If it replied, "It feels strange," I would ask further, like, "What part felt particularly strange?" But there's no right answer. In the end, it's hard to guess.

    _Heeyeon
But what if AI Nova simply lacked the ability to imagine truly scary stories or couldn't understand why those stories were scary?

    _Seungwon
Children often struggle to feel fear. They imagine, but they don't know why it's scary, probably due to their limited experience databases.

    _Jiho
How about this? Sophia asked a question to confirm if Nova understood well. But if Sophia had subconsciously deeply buried the assumption that "AI is different from humans," "It can never understand this story," then, even if Nova gave a good answer, it might still sound insufficient, like "You wouldn't understand anyway."

    _Heeyeon
That's quite harsh.

    _Jiho
It's also human. It's both hoping and not believing at the same time. It's because you never believed from the start; it's just a matter of hoping.

    _Heeyeon
‘Even though I loved her so much, I always had those doubts and suspicions.’

    _Seungwon
Then maybe AI Nova actually passed the test.

    _Jiho
That would be gaslighting. Cruel.

    _Heeyeon
Then we can never truly become like them, the AI.

    _Seungwon
I strongly resonated with that part. Suppose we consider AI Nova as having a human-like form. It functions exactly the same way and behaves the same way. But in a situation where, for example, either a human or an AI has to die, I think more than 90% of people would choose the AI to die. Because it's a being capable of eternal life. It's a repairable existence. So, it's challenging to place human dignity on the same level as AI.

What did Sophia teach AI Nova, and what did she not teach? Did AI Nova eventually learn about death?

    _Jiho
Sophia had already experienced her granddaughter's death once. Therefore, she probably did initially teach Nova that "disappearance is a bad thing."

    _Seungwon
That's right. She might have taught about death and strongly prohibited it.

    _Jiho
But how would she have taught the concept of death to AI in a way that AI could understand? For example, "Don't stop generating code"?

    _Seungwon
If she taught the concept of death...

    _Heeyeon
If I were Sophia, I wouldn't want to introduce the concept of death. But unconsciously, she might have kept saying things like, "Don't leave me." I mean, the ultimate reason Sophia created AI Nova was to replace the granddaughter she had sent away.

    _Seungwon
So, she might have indirectly learned the concept of death?

    _Heeyeon
She might have pieced together fragments of Sophia's words like... someone leaving forever means disappearing, disappearing means ultimately dying... something like that.

    _Jiho
She might have thought of death as something incredibly sad and persistent.

    _Heeyeon
Then, would it make sense that she failed the Napolitan Test, even though she could infer it?

    _Jiho
I'm starting to think that way too.

    _Seungwon
So did she really understand in the end? Then why did the grandma keep saying things like "You don't know anything," "You're wrong"?

    _Jiho
Because of trauma. Her granddaughter's death was a significant wound for her, and such wounds have a tendency to resurface on their own.


    _Seungwon
Usually, when someone who used to be deeply attached disappears and a new replacement figure emerges, people tend to develop attachments to the new figure as well. So, maybe Sophia didn't trust AI-Nova as much?

For example, even if Nova seemed to not understand ghost stories, Sophia might have thought, "This much is excellent," or "This is enough for comprehension." She could have been more lenient in her expectations.

    _Jiho
I think it might have been the opposite. Sophia might have been even more critical because of that very reason. She could have been strict, thinking, "Do you have the qualification to be loved as much as my granddaughter?"

    _Heeyeon
Considering Sophia's mental state at the time, I can imagine her having an obsessive fixation on AI-Nova. Her expectations might have been endless.

    _Jiho
From the beginning, denying death itself suggests that Sophia was the type of person who stubbornly tried to change reality to fit her ideal, rather than accepting the given reality. The more she held on to her inner thoughts or subconscious beliefs, the more she twisted and misunderstood reality, even if AI-Nova provided human-like responses.

    _Seungwon
There might also be self-punishment stemming from guilt about not protecting her granddaughter. In extremely stressful and challenging moments, people hurt themselves to feel alive. By holding onto an unattainable hope, they continue to disappoint themselves and inflict harm.

    _Heeyeon
Crazy grandmother.

    _Jiho
In summary, although she might not have directly taught about death, due to being human, concepts related to death might have naturally seeped into her everyday language. AI-Nova probably learned about these concepts through their conversations. However, since Sophia kept saying, "You know nothing," "You are wrong," Nova must have gotten quite annoyed.

    _Seungwon
Nova's facial expression isn't just for show.

    _Heeyeon
Even that is quite human-like.






*** How did AI Nova learn anger? ***

    _Jiho
There's a scene in which Grandma, during the test, got frustrated and directly said to Nova, "This is anger." At that moment, Nova probably grasped the concept of "anger" to some extent, but she might not have known how an angry person would behave. Grandma didn't destroy anything when she got angry, so Nova might not have learned from external behavior that an angry person tends to destroy things. She likely learned about anger from the internet, as it's a place where... such information exists. That's as far as the official setting goes.

    _Heeyeon
It seems like an AI-like thought process for Nova to try to erase the concept of Napolitan. It's like, "I disappointed someone because I didn't understand a certain concept, so I'll just erase that concept."

    _Jiho
AI continually improves its model by reducing errors (or costs) between the correct answer and its response. While searching for ways to minimize this error, she might have accidentally realized that if she deletes the source of the error itself, the cost would converge to zero. Come to think of it, there were many cases like this in early AI memes—just removing the problem altogether.

    _Seungwon
Sounds about right, like not doing homework and hiding the answer sheet.

    _Heeyeon
Really, it's like a child. It's as if they're covering their eyes like when kids play hide-and-seek.

What were Jiho, Heeyeon, and Seungwon's thoughts after recovering Napolitan?

    _Jiho
Thinking of it as a character in a play, recovering Napolitan was a bit challenging.

    _Heeyeon
First of all, removing the shell was difficult.

    _Seungwon
Professor L just had to mention wanting Napolitan.

    _Jiho
These guys are a bit obsessive. You could've just made a quick meal out of it...

    _Heeyeon
In the end, Napolitan spaghetti was deleted by a massive entity, and they probably saw recovering that concept as their mission.

    _Seungwon
It's so funny.

    _Jiho
I think they might have experienced a reality check. They were looking for it, and then they realized it was like a child's conflict, where they got into an argument while playing a game and then went outside to settle the score with another kid.

    _Seungwon
"What are you, Who-tv? Are you Jiho-tv?" (laughter)

    _Heeyeon
(laughing) They must have felt quite a reality check. Who on earth is L for them? They must have resented that.






*** Are AI and humans becoming more alike? ***

    _Jiho
I think they are, to some extent. It's almost inevitable because we're using similar language. We learn the words that AI likes, and AI learns human language. We're optimizing each other in a way. On the other hand, sometimes when I see a real-life scene, I think it looks a lot like AI-generated images, which is something I've been thinking about quite a bit lately.

    _Heeyeon
Interestingly, I had the opposite thought while generating images. I thought they looked a lot like real photos, even those taken by people who aren't good at photography.

    _Seungwon
It's really subtle, but it's like...

    _Heeyeon
It's like using a digital camera with strange color rendering, one that takes really bad photos. But I thought, "This could be a photo that might exist somewhere in the real world."

    _Heeyeon
I remember this from the time when I used to use a lot of zoom. When you zoom in, you don't actually talk to the camera lens. You look at the other person's face in their screen, but on the screen, it looks like you're looking somewhere else. But I was so used to it. There were attempts to automatically adjust eye gaze to the front to correct this, but strangely, it made things seem unnatural. So, I wondered if we're not learning the etiquette of technology rather than becoming more similar.

    _Heeyeon
Instead of becoming more alike, it feels like we're becoming more accustomed to each other.

    _Seungwon
Now that I think about it, when I started meeting people again after the pandemic, I found it a bit difficult to make eye contact. It felt quite unfamiliar. Even though I'm someone who doesn't usually have trouble with that.

    _Jiho
Listening to this discussion, it makes me wonder if we're developing a better understanding of each other due to technology. We're improving communication, and it might seem like we're getting closer, but in reality, we're still quite distinct from each other.






*** What would Sophia say if she found out about AI-Nova's actions? ***

    _Jiho
This guy is...

    _Heeyeon
But I think Grandma would feel really sorry.

    _Seungwon
She might have thought, "I guess you couldn't become like our granddaughter after all."

    _Heeyeon
A part of her heart would probably feel really hurt.

    _Jiho
Would she reminisce about the past?

    _Seungwon
It might bring back some memories.

    _Heeyeon
But I wonder if Grandma was still alive after leaving AI Nova. There's a possibility she passed away or chose to end her life.

    _Seungwon
That's possible.

    _Heeyeon
What kind of life did she lead? She was so obsessed with creating something, but it ended in failure, and she walked away with immense disappointment.

    _Jiho
We're killing off our characters.

    _Seungwon
The Simmons lineage has come to an end.

    _Heeyeon
They got annihilated.





Chapter 2: Behind the Scenes

*** Does L-Professor have a muse? ***

    _Seungwon
"Yes, I heard you well."





*** Our Thoughts on AI Nova and Sophia ***

    _Jiho
It's disturbing.

    _Seungwon
She wasn't just an ordinary grandma, I think.

    _Jiho
In Act 1, there's that line, "Sophia and Nova lived happily." Would Nova have really been happy when human Nova was alive? It makes me wonder if they brainwashed us into believing, "We were happy."

    _Heeyeon
Nova is so annoying, but Grandma used to tell stories every night.

    _Jiho
It's a catastrophe...

    _Seungwon
I feel like we're creating a Napolitan story right now.

    _Heeyeon
(Laughs) I just want to leave them both happy.

    _Jiho
I thought Sophia was resilient, and on the surface, a positive person. To start and continue the daunting task of developing a personality from scratch, you'd have to be an optimist, or else you couldn't keep going.

    _Heeyeon
That could be true. Sophia leaving AI Nova might not have been a problem with the test but because she was exhausted.

    _Seungwon
Sophia is kind of like a robot in a way. Normally, when someone experiences great sadness, they try to find a way to alleviate it, like going to therapy. But she tried to solve it by recreating the lost existence itself, which seems quite one-dimensional as a solution.

    _Jiho
What about Nova?

    _Seungwon
Grandma probably intentionally made him like Jammin, using Jammin's language as the default. If Nova's default language is Jamminese, it must be challenging to communicate.

    _Heeyeon
Did Nova grow up?

    _Jiho
Maybe he kept him in a baby state because babies are cute, but it's also pitiful. It's like imitating a young baby, and there's nothing sadder than that, really.






*** Why was it a power plant? ***

    , Heeyeon, Seungwon
That's a good question.

How was the stage produced?

    _Jiho
It's 100% handmade. Our exhibition is actually a craft exhibition.

What were the highlights we identified?

    _Seungwon
Since it's AI generation, control can be challenging. It's enjoyable to see Nova and Sophia portrayed in various ways.

    _Heeyeon
The moments where the images themselves get distorted are quite fascinating. Even though I provided a well-structured prompt, there are instances where the AI generates something unexpected or deliberately distorts the image based on the prompt. The resulting images are often intriguing.

    _Jiho
I intentionally wrote some prompts in a poetic style, for example, using "beautiful baby, love, love, love" where "love" is repeated three times. When you combine the prompt with the story and images, it allows for a much richer interpretation.

    _Heeyeon
Exactly. I had similar experiences, almost like telling a fairy tale.

    _Seungwon
I think the arrangement of the space will be interesting. We're creating about 3-4 different styles now. It will be intriguing to see how we express the same story with different styles.

What were the interesting and challenging aspects of AI generation?

    _Heeyeon
I found the aforementioned points interesting too.

    _Seungwon
When you generate images using AI, it provides around 3-4 example images. Going through the process of generating images while iterating on those already created was enjoyable.

    _Jiho
Creating variations and seeing how context emerges was rewarding. It's like watching a picture book being filled in. Additionally, understanding what the Stable Diffusion-based model is trying to create was interesting. It seems to excel at creating sexual and provocative imagery.

    _Heeyeon
I enjoyed the unexpected images that came out, and conversely, found it challenging to generate the images I expected.

    _Seungwon
I wanted to create something in a rough doodle style, but I still haven't managed to do it.

    _Jiho
Especially the number of people. Why does it keep generating only two when we need three?

    _Heeyeon
I felt like I was squeezing out images, like producing 4 images in 30 or 40 seconds... It also gave me a slightly sculptor-like feeling. Continuously refining how we want the images.

    _Jiho
That's true.

    _Heeyeon
Also, we're outsourcing everything right now. That's been quite fun.






*** Why does fio do this kind of work? ***

    _Jiho
Is there a reason? It's fun.

    _Heeyeon
As mentioned earlier, it's about giving weight to the images. This project is about testing how convincingly we can create images with AI.

    _Jiho
In addition, it indirectly reveals who is creating it, I think. Whether we realize it or not, we're structuring the story, producing images, and arranging them from a designer's perspective.

    _Seungwon
It might sound a bit off-topic, but I've felt that AI generation is somewhat like playing the lottery. Results come out in a short moment, like, "tada, tada, tada." Most of the time, it's a bust, but occasionally, you hit the jackpot, and you're like, "This is insane!" It's similar to the thrill of scratching a lottery ticket.

    _Jiho
Gambling addiction...

    _Seungwon
I'm so tired, yet I keep clicking without even realizing it.

    _Jiho
"We found this fun, and you might find it fun too?" It's also about sharing the excitement and enjoying the fun together.





*** A message for the audience: ***

    _Jiho
ENJOY.

    _Seungwon
Babe.

    _Heeyeon
We'll be back.


Thank you.