It had been a while since Chun Min-jung last visited the Vice Chairman's office.
From the moment she stepped in, she wore a bright smile, like a puppy waiting to be praised.
"You must have completed a project."
"How did you know? We've finished the Smart Set-Top Box project!"
Taewoo Electronics had already launched Pixi, a set-top box equipped with artificial intelligence.
However, it was still in its early stages and could only perform basic functions, so it couldn't
truly be called a smart set-top box.
"Sounds like Pixi's performance has seen a major upgrade."
"We expanded its internal search functions, so now it can respond to most user questions. It can also understand every regional dialect in the country, so no matter what dialect a user speaks, it functions without any inconvenience. We even trained it to recognise North Korean dialects."
The most common complaint Pixi users had was related to speech recognition.
It worked fine for frequently used commands, but struggled with anything even slightly more
complex.
"If it's trained in all dialects, that should drastically reduce recognition errors."
"It also supports English, Chinese, Japanese, and Spanish. It can search foreign-language content quickly and even offers basic interpretation features."
"Interpretation? That's unexpected."
I had already believed Pixi's performance was ten years ahead of its time.
But with an interpretation function, it felt like we had leapt not ten, but twenty years into the
future.
Before my regression, most households owned smart set-top boxes.
But none of them had interpretation features—I had never even heard of such a function, let alone used one.
"It also includes a simple home security feature. If you're away for an extended period, you can activate a security mode. A micro CCTV installed in the box streams real-time footage to your phone, and if the mode isn't deactivated, it automatically sends an alert."
"That could lead to malfunctions or false alarms."
"To prevent that, it won't contact the police directly. For apartments, it alerts the front desk, and for houses, it contacts a pre-designated person. That can all be configured."
This was exactly why I had invested so heavily in AI.
Once properly developed, artificial intelligence could generate hundreds of trillions of won in
added value.
Of course, that was only possible thanks to Chun Min-jung's ability to turn ideas into reality.
"When will it be ready for release?"
"We could launch it immediately if we wanted to."
"The president of Taewoo Telecom is going to love this."
"And… another project has also been completed."
Wait, not just one?
Curious about what else she had brought, I scooted my chair closer.
"What kind of tech is it this time?"
"Well, it's not something to get overly excited about. Do you remember when you instructed us to develop an AI that could learn to play Go?"
"Of course I remember. Are you saying you've already completed an AI capable of playing Go?"
"It's still an early version, but it's strong enough to beat most amateur players. It's not quite ready to defeat professionals yet, but with a few more tweaks to the algorithm and performance, it'll be possible."
Artificial intelligence was still a new frontier.
To attract global attention, a major event or issue would be needed.
Before my regression, it was the showdown between AI and top-tier professional Go players that created a worldwide buzz.
But if I recalled correctly, that kind of event didn't happen until ten years later.
And yet here we were—ten years early—with Chun Min-jung and her team already developing an AI capable of beating human players at Go.
"If it's still at the amateur level, then it's not quite ready for a public reveal. How long until it can beat a top-tier professional player?"
"Honestly, not long. It's already learning from thousands of game records every day. And since we're feeding it records from players ranked at least 6-dan, it should be able to match
professionals in about three months."
The difference between humans and computers was simple.
Humans get tired. Computers, as long as they have a stable power supply, can run 24/7.
"Is it possible to train an AI on games too? I'd love to see it take on professional gamers in realtime strategy games similar to Go."
"That's absolutely doable. In fact, it might be easier than Go. Professional gamers do have a wide range of strategies and tactics, but in reality, there are still recognizable patterns."
Chun Min-jung clearly had a deep understanding of gaming.
Then again, I hadn't met many people in the IT industry who didn't like games.
"Before we release the Go AI, I'd like to debut the gaming AI first—with an event match against professional gamers."
"Actually, I've been developing one on my own for fun."
"You were developing a Starcraft AI?"
"Well, yes. I was just playing for fun, but someone I matched with kept insulting me for being bad. So, in a fit of frustration, I started building a Starcraft AI."
Chun Min-jung really had a one-of-a-kind way of thinking.
I told her to study English—she built a translation system.
She got flamed while gaming—so she built a gaming AI out of spite.
"What's its win rate?"
"My AI is currently ranked second on the ladder. It wins almost every match against amateurs,
and it has about a 50% win rate against professional gamers."
"Can that win rate be improved?"
"At the moment, the AI is running on my personal computer, so it can't perform at full capacity.
If we move it to hardware optimized for AI, it should be able to achieve an 80% or higher win rate even against professional gamers."
Chun Min-jung was confidently projecting a very high win rate.
Korea was considered the birthplace of esports, and that reputation was largely thanks to the elitelevel performance of its professional gamers.
They would win international tournaments almost by default.
It was often said that qualifying for domestic tournaments was harder than winning global
competitions—so achieving an 80% win rate against Korean pros was no small feat.
"The win rate's even better than with Go."
"That's because in games, there's a limit to how fast a human can physically execute commands. There's a stat called APM—actions per minute—that measures how many commands a player can input in one minute. Even for top-tier pros, it's hard to average over 500 APM, but AI can easily reach two to five times that."
Humans had to move their fingers and physically press keys.
But for a computer, there's no need for a keyboard—it could issue commands directly, making its reaction speed vastly superior.
"Then let's schedule the event for next month."
"Perfect timing! The first half of the Pro League season just ended, so scheduling won't be an
issue. Would it be okay if I selected the participating players? There are a few I really want to
see."
"Work with the planning department and put together an event outline. As for the player lineup—
feel free to pick whoever you like."
"Really?! I've always wanted to meet pro gamers in person—this is a dream come true! Thank
you so much!"
Did she always like esports this much?
Chun Min-jung was practically bouncing like a fan about to meet their favorite idol.
She dashed off to the planning department, clearly too excited to sit still.
The StarCraft exhibition match held in Yongsan, Seoul.
Being known as the sacred ground of esports, the venue always attracted large crowds, but today, unlike usual, many journalists had filled the seats.
"Ladies and gentlemen, we are about to start the unprecedented AI vs. pro gamer showdown!"
The game began with the familiar shout of the caster.
I had played StarCraft a few times but wasn't very skilled, so I didn't fully understand how the match was unfolding.
Perhaps for my sake—or maybe just for her own hobby—Chun Min-jung provided real-time
commentary beside me.
"The player competing in this round has an extremely strong early-game strategy. In his debut match, he swept the ELO number one player with a rush build. Although his ranking is low, I specifically added him to the lineup because his games are so exciting."
"A rush build? What's that?"
"To put it simply, it means a 'sharp build,' but actually it's short for a 'rush build'—a strategy where you overwhelm your opponent before they can even respond."
I kind of got it.
They had brought in a player strong in the early game, probably to see how the AI would respond.
"If it's an early-game strategy, the AI might be at a disadvantage."
"That's a common misconception. Actually, the AI is stronger in the early game. This is true both in games and Go AI as well. Just watch!"
The player using the early rush built structures inside the AI's territory and started producing
units secretly.
Though the AI noticed late, with dazzling worker control, it quickly destroyed all the enemy
units.
"That was stopped way too easily."
"Trying to win the early control fight against an AI is a completely wrong idea."
"Will the next match end this boringly, too?"
"Don't worry. This player wasn't even in the top 10, but next up are players ranked in the top 3."
The first match ended with the AI victorious.
The early-rush player wore a sulky expression, and the crowd cheered "It's okay" at his pure,
honest reaction.
Meanwhile, the second match was ready.
A player with the looks of a celebrity took his seat.
"He could be a celebrity, not just a pro gamer."
"His skills match his looks—extremely flashy control and stable management. Though he's now ranked second, last season he was the reigning champion."
Was I reading something into it?
I slightly bowed my head to check, but Chun Min-jung wasn't holding anything in her hands.
She really was a hardcore esports fan, memorizing every player's traits perfectly.
"Starting match two~!"
The game began again with the caster's loud voice.
Maybe because the previous player was crushed by an early rush strategy, this time the opening phase passed without much action.
"They're both using full late-game builds. They focus on solid resource gathering early and plan for a big battle in the late game."
"So who's in the better position?"
"Actually, the AI has the edge in build order. Not by much, but about 60-40 in favor of the AI."
So the result might come quickly.
With a much higher APM than a human and overwhelming multitasking ability, the AI looked very strong.
I relaxed a little and watched the screen.
Fifteen minutes in, battles erupted everywhere.
Both sides were desperately looking for weaknesses to exploit and using skillful unit control to gain advantages.
"Vice President! They've mined all the resources. Now they have to fight with what's left."
"So it's come down to a unit-control battle. The AI should be favored then."
"Oh? The pro gamer just lost two units!"
The pro player made a mistake.
In a situation where every unit was precious, he lost two units near the enemy base.
Of course, the AI rushed in immediately.
But then — just like it had been waiting — the pro moved his main army to ambush the AI's
forces.
"Wow! It wasn't losing units, it was setting a trap. The AI's win rate just dropped to 20%!"
He's going to win this?
In that brief moment, the pro gamer spotted and exploited a weakness in the AI.
This must be why Korea is called the birthplace of esports.