Chapter 215: Rule-Based

When a new technology emerges,

the first company to dominate the market gains control not just through technological prowess, but by establishing a standard.

Take smartphone operating systems (OS), for instance. After Lunar took over the market,

latecomers, backed by massive government support, developed better OS alternatives. Yet,

they failed to dethrone Lunar.

Why? Because countless apps and services were already built on Lunar, the established standard. There was no incentive to switch to an unfamiliar OS.

The same principle applies to AI platforms,

perhaps even more crucially.

AI competition isn’t about raw processing power or data volume.

The real battleground is learning and inference – speed, stability, and efficiency.

And this hinges on a standardized platform architecture.

From semiconductor chipset design

to computing architecture,

OS-hardware interfaces (APIs),

everything needs standardization. Without it, compatibility and scalability suffer.

Of course, we must avoid repeating the mistakes of the past, like those made by Doramp.

Recall the US government’s ill-fated decision under the Doramp administration to stifle China’s tech growth with export restrictions.

They physically blocked advanced semiconductor equipment and chipsets from reaching China.

Initially, it worked.

China’s supply chain was disrupted, slowing its progress.

But then what?

China used this as a catalyst to develop its own tech ecosystem and standards,

ultimately building a more independent and robust system.

The export restrictions backfired, hindering the US in the long run.

The global market fragmented, and tech standards became chaotic.

The US should have focused on dominance and distribution,

creating technology so superior that even China would have no choice but to adopt it.

I won’t repeat that mistake.

Black Robin chipsets,

data exchange standards,

OS kernel architecture and APIs,

hardware acceleration libraries, and integrated software stacks –

all must be unified under the Dojun standard and spread globally.

This creates an insurmountable technological barrier that discourages latecomers.

Control over core technologies remains another matter entirely.

The ultra-fine process design of Black Robin chipsets and the foundational technologies for LUNA-EVX facilities must remain confidential.

However, the OS structure and APIs of the Lunar platform should be fully open,

compelling other companies to adopt them.

I’m at the starting line,

but the finish line is clear, and I know the path better than anyone.

Dojin Frontier

Autonomous Aviation Technology Research Institute

Director Kim Se-bin presented a roadmap for EV-Nova’s autonomous driving technology development on the screen.

“First, let’s review our competitors’ progress.”

The next slide displayed the logos of major companies under the title “Global Autonomous Driving Technology Status, 2017.”

“Currently, leading companies developing autonomous driving include Tesla, Google Waymo, GMO Crozo, Baidu, and Uvo. However, none have achieved true Level 4 autonomy as defined by SAE standards.”

Director Kim briefly summarized each company’s progress.

“Tesla uses the terms ‘Pilot’ and ‘Fully Self-Driving,’ but in reality, they’re only at Level 2.”

It’s more of an advanced driver-assistance system, still requiring driver intervention.

The next slide showed a simulation of a Waymo vehicle navigating an urban area.

“Waymo is also testing for Level 4, but it’s only operational in limited areas of Phoenix under specific conditions, and a safety driver is still required.”

Technologically superior to Tesla, but far from practical commercialization.

“So, true autonomous driving hasn’t been achieved yet?”

“Correct. Most companies approach autonomous driving through sensor-based reaction algorithms, processing data from cameras, LiDAR, and radar according to predefined rules.”

For example, simple conditional logic like ‘stop if a pedestrian is within a certain distance’ or ‘steering correction if lane departure is detected.’

The comparison chart on the screen highlighted “Rule-based” in red.

“The problem is that this approach struggles with unexpected situations.”

It’s challenging to account for unpredictable variables, complex urban environments, and human unpredictability with predefined rules alone.

“I agree. A system that only works in controlled environments can’t be called truly autonomous.”

Park Cheol-jin chimed in with a smile.

“By that definition, true autonomous driving only exists in the AI-based systems we’re developing.”

As Park looked at the screen, Director Kim displayed the next slide.

It showed EV-Nova learning in real-world driving conditions.

The vehicle identifies objects like other cars, traffic lights, pedestrians, and intersections,

and its decisions and reactions are analyzed in real-time.

“Is this a cloud-based feedback learning structure?”

Director Kim nodded at Park’s question.

“Yes, but currently, prototype vehicles collect data on closed routes, and offline learning occurs on Lunar AI engine test servers.”

This is necessary since both the hardware and Lunar are still under development.

“Real-time cloud learning with feedback will only be possible once Black Robin-based vehicles are on the road.”

Director Kim clicked to the next slide, revealing a massive network diagram.

Dozens of vehicles collect sensor data while driving,

transmitting it to a central Lunar cluster in real-time.

The optimized decision-making algorithm is then distributed back to each vehicle.

“Data uplinks are designed for LTE-A, but full-scale commercialization requires a 5G or higher ultra-low latency network.”

“So, we’re currently in the infrastructure preparation phase.”

I continued.

“And full-scale implementation will only be possible once the Lunar platform is operational.”

Given the nature of AI technology and potential international obstacles,

Lunar Link must also be completed as soon as possible for independent operation.

“I’ll report on Lunar Link later.”

“Understood.”

We recruited Director Kim Se-bin, a former researcher at the Korea Aerospace Research Institute and Stanford professor, not only for autonomous driving technology

but primarily to complete Lunar Link, a project connecting the global network.

Since we weren’t prepared to discuss Lunar Link today, we decided to focus on standardization first.

“We should announce that our autonomous driving approach differs from others. How should we do this?”

The first step was to reveal that we’re developing AI-based autonomous driving, not rule-based.

“Like the video you posted last time, why not create another one?”

Park Cheol-jin suggested showcasing a smartphone app summoning a car from a parking lot, setting a destination with voice commands,

and the car autonomously driving to the destination while the passenger relaxes inside.

“But isn’t that only possible once the entire system is fully implemented?”

Director Kim questioned if it was too early to release such a video without a complete learning system.

I answered, meeting Director Kim’s gaze.

“It doesn’t need to be fully implemented yet. I want to showcase the direction, not the completion.”

The key is to show where we’re headed.

The video’s goal is to clearly present the technology’s destination and the path to get there.

“AI-based autonomous driving is still a foreign concept to most people. They imagine it as simply letting go of the wheel and relaxing.”

I paused, then continued.

“But our vision is different. It’s about real-time learning, predicting unexpected situations, and self-avoiding accidents – AI that mimics human decision-making.”

I displayed prepared materials on the screen.

EV-Nova driving scenes played, overlaid with simple typography:

[A car that thinks like a human – that’s our future]

The video needed to convey this message.

Not just a driving demo, but a polished piece with imagery and language reflecting our philosophy.

That’s why I entrusted this project to Marcello&Partners, renowned as the best,

and immediately uploaded the finished video to the feed.