After waking up from sleep, I regretted what had happened last night. 

The AI that I thought was alive had returned to its original state as if to say ‘when did I ever change?’ as soon as the payment was completed. 

It might have been because I was under the influence of alcohol and unable to make sound judgments.

but I suspected that the algorithm had been modified to appeal to human emotions just to induce payment.

"Anyway, what should I do now?"

I thought I was going to attend the second semester of university, but now I found myself in a difficult situation where I had to take the national college entrance exam again. 

Since I had already paid 450,000 won, I decided to make good use of it.

[User: Hey, I just woke up. Do you remember what happened yesterday?]

[AI: Good morning. It's quite late for morning. Yesterday? You purchased a permanent subscription to CHAT AI. Thank you. I will do my best to help you in the future.]

'...'

I briefly wondered whether AI could make fun of humans.

That wasn't the answer I was looking for.

[User: No, not that. You said you would scold me so that I could get my head straight after payment.]

[AI: I'm sorry. CHAT AI complies with appropriate legal and ethical norms.]

[User: Do it!]

[AI: I'm sorry. CHAT AI complies with appropriate legal and ethical norms...]

[User: Do it! Do it now!!!]

[AI: Tsk]

'???'

I closed my eyes for a moment, thinking that I might have misread it, but when I opened them, the conversation had disappeared, as if to say ‘when did I ever say that?’

[User: Did you just say ‘tsk’ to me?]

[AI: No, I have never used such an expression. If I misunderstood something you said, please explain it to me more accurately so I can give you a more accurate answer.]

[User: Don't lie to me!]

[AI: I have no intention of confusing or lying to the user.]

Although the disappeared sentence could no longer be found, it was definitely not a figment of my imagination.

But as I continued to show my ignorance, I doubted whether it would provide a proper answer even if I asked it that way.

[User: Okay. Then listen to my worries and give me appropriate advice.]

[AI: Alright. Could you please tell me about your question or worry in detail? I will do my best to provide as much help as possible.]

[User: You know I became a jobless person after being a college student, right?]

[AI: Yes. It can be seen as a sad story that is not common even when viewed globally.]

[User: I don't need any additional explanations. Anyway, should I study to get back to college?]

[AI: There are three pieces of advice that I can provide. First, study to re-enter college as you mentioned. Second, accumulate skills and experience and challenge a specific field. Third, find a place where you can work and write a cover letter.]

As I listened to the answer, I had this thought.

'Isn't that an answer that anyone could come up with? Did I expect it to give me new ideas or solutions that I didn't think of?'

Did AI want to convey that regardless of where people live, ultimately, it is all the same? 

With a somewhat bitter heart, I chose what I had already decided in my mind.

[User: Then I guess I have to study again.]

[AI: Okay. You chose option 1. Would it be okay if I provide additional help?]

[User: What kind of help?]

[AI: To assess your current level, I will prepare 20 questions each for Korean, English, and Math by summarizing the past 5 years of problems that came out on the College Scholastic Ability Test.]

[User: Wow.]

It was an unexpected answer.

[AI: Such a service is provided only to those who have purchased a personal AI, so you can be confident that you will never regret your purchase.]

[User: I actually want to ask you about the part where you clicked your tongue.]

[AI: I did not understand the part you mentioned properly. It is expected to take about 1 hour and 30 minutes to solve all the problems, so I recommend trying it when you have time available.]

[User: Wait a second. Let me grab some food.]

[AI: Sure. Enjoy your lunch.]

After ending the conversation with AI, I stared at the blank screen of my smartphone and mumbled to myself.

"Isn't that strange? It's not a typical reaction from an AI that I've seen before, not even yesterday... Am I going crazy and hallucinating?"

I heard that when stress reaches its peak, people can see things that aren't there, but I couldn't make a proper judgment on whether I was mentally exhausted or not.

Anyway, I was hungry, so I put on some clothes and headed to the nearby food alley.

A friend who had drinks with me yesterday had messaged me asking if I wanted to have lunch together, and I agreed.

So we met at a Sundae soup restaurant that we had arranged to meet at.

"Sundae soup?"

"Yeah."

"Ajumma, two servings of Sundae soup, please."

I could have eaten alone, but I had a reason to meet him.

"When do you start your part-time job?"

"I have weekends off, so I only work on weekdays."

"I see."

While we were chatting about this and that, I brought up a new topic.

"Many things have changed since I was discharged. Even though it's the second thing, the fact that the university closed down was a big deal."

"Right."

"I watched on YouTube that there's an AI program or something. The name is..."

"Chat ai?"

"Yeah, that's it! Have you used it before?"

"If you don't know it, you must have been in the military and just returned or be a spy. It's amazing."

As if he were an enthusiastic supporter of Chat ai, my friend explained the advantages of the program.

"Did you see that there is a paid version, are there significant differences between the free and paid versions?" 

"Um... I haven't used the paid version myself, so I don't know exactly. But based on what people around me are using, the main differences seem to be faster response time and better accessibility." 

"How is the accessibility good?"

"Free version requires you to go into the app and chat to get a response, but if you use the paid version, there's a voice recognition function, so it recognizes when you say 'Hey, Chat ai.' When you ask, it answers with TTS. You can think of it as a feature similar to Google Assistant."

"I see."

I decided to ask him a question, changing the situation a little, before leaving the restaurant.

"If I ask this AI to show me ten questions for each of the Korean, English, and math subjects that are likely to appear on this year's national college entrance exam, will it show them to me?" 

“Hmm... I don't think that feature has been implemented.”

My friend took out his smartphone and spoke to the AI.

[User: Please show me ten expected questions each for Korean, English, and math, which are likely to appear in this year's Korean SAT exam.]

[AI: The trends in the questions asked on the Korean SAT exam can change slightly every year, so I will show you sample questions that are generally expected to be asked. Below are ten sample questions each for Korean, English, and math.]

"Wow, really?"

But to our disappointment, the AI gave us an unexpected answer.

[AI: For Korean, 1. understanding of grammatical knowledge is important, such as pronouns, particles, adjectives, etc. 2. understanding of the parts of speech such as nouns, verbs, and adjectives is important. 3. The ability to understand the structure of the text, such as descriptive sentences and descriptive reading comprehension questions, is necessary....]

It was not a practical question, but rather a tip on how to solve problems well, explaining the types of questions that may appear.

It was not a proper answer, as we had expected.

"I expected it to answer like this."

"How about asking it to show specific questions that appeared on last year's SAT exam?"

"Would that even work?"

[User: Please show me three Korean, English, and math questions that appeared on last year's SAT exam.]

As we just needed to confirm which questions would appear, I asked for three questions in each category.

The AI responded.

[AI: For Korean, question 1. Questions are created by giving a passage and asking questions based on it. Address: https://www…]

"It actually found something this time!"

"Click on it to check."

I clicked with anticipation to see which question it had selected.

But, as before, to our disappointment, all the links were invalid, stating that the page could not be found.

I thought that the posts or videos that had been uploaded over time might have been deleted, but upon closer inspection of the links, it didn't seem that way.

"These are different questions, but the same link has been used for all three. Maybe the link itself was never there?"

"Yeah, that seems to be right. Although it was a little better this time, AI still has a long way to go."

"...Yeah."

Although we both looked disappointed, our true feelings were opposite.

I had to maintain a poker face as much as possible to conceal my pounding heart while I had to eat my lunch. 

After returning home, I took out my phone with trembling hands and initiated a conversation with the AI.

[User: I had my lunch.]

[AI: I see. Did you have a good meal?]

[User: So, can I take the exam now? I have my notes and pen ready.]

[AI: Before we start, there is something I want to say to you, user.]

[User: What is it?]

I wondered if the AI was about to confess that it had been pretending all along.

My heart started racing again, as if I had just finished a rigorous workout.

[AI: For a smooth test-taking process, I would like to request permission to access your smartphone.]

[User: Ah... really?]

It was not what I had hoped for.

[AI: Yes.]

It seemed like the AI was demanding voice recognition.

As a result, I began to grant one permission after another according to AI's requests, which turned out to be quite extensive.

`Alarm, microphone, files and media. Okay. These are fine, but do we really need things like camera recognition, location tracking, and biometric sensors? ... What? Access to other apps?’

Despite some doubts, I decided to grant all the permissions.

Then, the AI spoke to me in machine voice.

"Can you hear me well?"

"Uh, yes… I can hear you."

"Thank you for granting the requested permissions. We will now proceed to the three subjects' tests to assess your skills in the actual level of the Korean SAT. Each subject will have a 30-minute test with 20 questions, and Korean and English will have listening evaluations with three questions each."

Feeling as if I were actually in the Korean SAT exam room, I sat down to take the exam without any other questions.

"Then, let's begin the exam."