Chapter 244: Belated Change
A month later.
Breaking news appeared on major media outlets and social networks worldwide.
[Tesla’s ‘FS (Fully Self)’ Software Mass Deployment… “Supports Integrated Urban and Highway Autonomous Driving”]
Tesla announced that through this update, the ‘FS’ function would be applied via OTA to certain urban and highway sections, including nighttime and rainy conditions.
The initial deployment area was limited to the Bay Area and Los Angeles in California, USA…
└Wow, as expected of Tesla. They’re the first to enter autonomous driving.
└But, does it actually work properly?
└Elon Musk has been saying for years that Tesla isn’t just a car manufacturer. They’ll grow into a platform company.
└Hmm… It still seems like it’s only at Level 2.
└It’s probably too much to expect the kind of performance seen in ‘The Electric X Operation,’ right?
└lololol It’s still early. I don’t think it can do that yet.
└Just wait a bit, and user reviews will start pouring in.
└lololol Wow, you’re like a ghost. Tteubom just posted a user review.
└……
People flocked to Tteubom’s channel, which has a million followers, in an instant.
He had already moved to North America for this test.
The thumbnail of the video he posted had this text:
[KTalk Feed] - @Tteubom | Followers: 1,367,281
Honest Review of Tesla FS.
Disappointment or Innovation?
Attached Video File: 1
When the file was clicked,
the recorded video started playing.
The video seemed to have been filmed at dawn, starting in a dark background.
Tteubom in the video was sitting in the driver’s seat of a Tesla vehicle, not holding the steering wheel.
The vehicle was driving through the streets of California, and the display screen in the car showed the words .
“It’s 5:30 AM now, and Tesla’s FS function is on. This vehicle is a North American certified model, and as you can see, this update has been applied.”
People’s reactions to the video were mixed.
└Wow… No hands on the wheel during urban driving?
└But the braking is a bit rough? There are definitely more sudden stops compared to human driving.
└Tteubom had to intervene at an unprotected left-turn section.
└lololol It even drives in reverse?
In some scenes, the vehicle suddenly accelerated after stopping
or hesitated for a long time in the middle of an intersection.
Especially, when a motorcycle suddenly cut in,
the FS slammed on the brakes and sounded an alarm, which became a point of controversy.
At the end of the video, Tteubom left an overall review.
“Hmm… In my opinion, it’s still embarrassing to call this autonomous driving. There were several situations where an accident would have occurred if I hadn’t intervened. At this stage, it’s more like ‘one-third of autonomous driving?’ I guess. Hahaha, but since it’s Tesla, it’ll probably improve from here, right?”
Starting with Tteubom, other famous influencers also rushed to post their review videos.
The results were not much different from Tteubom’s.
It could ‘mimic’ autonomous driving,
but it was still far from being truly reliable.
And exactly one week later,
a new post appeared on Kang Tae-joon’s feed.
[KTalk Feed] - @Kang Tae-joon | Followers: 202,081,003
LAD v0.8 Public Test Begins
Tomorrow, I will personally install LAD on an EV-Nova vehicle and conduct an autonomous driving test.
I’d like you all to judge for yourselves by watching the live stream.
#LAD #DojinFrontier #AutonomousDriving #EV-Nova #BetaVersion #LiveStream
└Kang Tae-joon surpasses 200 million followers, congrats!
└The follower count is really skyrocketing.
└lololol The recent Polar Orbiter mission was impressive, though.
└Well, with over 30 million live viewers…
└Is LAD the name of Dojin’s autonomous driving software?
└Yeah, Tesla has FS, and Dojin has LAD.
└The timing of the release seems too perfect.
└lolololol They’re confident they can easily surpass Tesla’s ‘FS.’
└If it falls short of expectations, the stock will crash…
└Kang Tae-joon is a god. You have to believe in him. I just bought 10 more shares of Dojin Frontier. My goal is to reach 10,000 shares.
└Wow, 10,000 shares is almost $320,000?
└With 10,000 shares, early retirement is possible.
└Dojin’s stock will go to $1,000.
└……
Tesla Headquarters, California, USA.
Elon Musk, sitting in his office, was silently replaying a video.
Urban intersection, unprotected left-turn section.
Motorcycle intrusion.
And scenes of Tesla vehicles equipped with FS hesitating for a moment.
Some videos even showed collisions with motorcycles.
On one side of the desk, a state-of-the-art projector displayed real-time data logs.
Error rates during driving, intervention points, sensor delays, and post-stop re-acceleration.
“Is this… the current state of the autonomous driving we’ve created?”
The autonomous driving division head, ‘Eliason,’ standing beside him, cautiously spoke up.
“In my opinion… it seems to be the limitation of the rule-based judgment system we’re currently using.”
Eliason displayed a slide on the office screen.
It was a statistical summary of major failures from the recent FS beta test.
-Unprotected Left-Turn Judgment Error: 37.2%
-Delayed Reaction to Motorcycles and PM (Personal Mobility): 44.5%
-Construction Zone Route Deviation Rate: 18.9%
-Normal Section Centerline Intrusion: 3.5%
-User Intervention Rate: 62.1%
Musk frowned and said,
“It definitely doesn’t feel natural to me.”
Since it only followed predetermined manuals, it couldn’t handle unexpected events smoothly.
Turn right at this point, turn left after a few seconds, stop at this distance…
Not ‘judging’ like a human,
but rigid, mechanical responses.
“This isn’t autonomous driving, it’s just a slightly refined cruise control.”
Eliason continued.
“So… I think we need to consider switching to a behavior-based model.”
“Like Dojin, adopting a cloud-based feedback learning structure?”
“Yes, only by sharing situational behavior histories in real-time and updating that experience across all vehicles can we respond to millions of scenarios.”
Eliason once again emphasized the limitations of the rule-based judgment system and the need to change the structure before it’s too late, displaying Dojin Frontier’s technical overview on the screen.
[Dojin Frontier LAD Learning Structure]
-Edge Inference: Real-time judgment at the vehicle site
-Cloud Replay: Upload all data to the cloud
-Feedback Loop: Automatically mine rare situations → Improve algorithm → Redistribute via OTA
“The core of this method is ‘memory.’ Dojin’s vehicles gather the learned experiences of each vehicle into a central cloud, and all vehicles share the results. Through such repetition, learning efficiency increases exponentially.”
Elon Musk silently stared at the screen and said softly,
“Our FS is definitely at the level where it just stops if a squirrel runs out during driving, unable to find the manual…”
It was embarrassing to even say it, but it was the reality.
Just then, a notification from Kang Tae-joon’s FeedLive appeared on the monitor.
It was the moment the EV-Nova autonomous driving test equipped with Dojin’s LAD began.
Musk and Eliason stopped talking and focused on the live video on the screen.
…
…
[KTalk LIVE] — Real-time Viewers: 3,182,398
Downtown Seoul.
The sky was cloudy, and the ground was damp.
Streetlights and vehicle lights reflected on the wet asphalt like ripples,
and a vehicle glided over it.
Dojin Frontier’s electric car, EV-Nova.
Kang Tae-joon, sitting in the driver’s seat, silently looked ahead.
His hands were completely off the steering wheel, and his feet were naturally placed on the floor, with no actions like pressing the accelerator.
The HUD at the top of the dashboard displayed the following information:
-LAD v0.8.1
-Mode: Full Passive
-Precision Map Sync: Normal
-Real-time Feedback: Recording
-Location: Jongno-gu
-Weather: Cloudy
-Visibility Distance: 240m
The real-time viewer count started at around 3 million and quickly increased by several thousand per second.
└A live test on a rainy, foggy day like this, wow.
└Tesla’s FS stuttered a lot in this weather, but this one is going smoothly.
└This is what you call autonomous driving.
The vehicle passed through Insadong Intersection and headed toward Bukchon.
Narrow, winding alleys, unprotected left-turn sections, groups of jaywalking tourists, and delivery motorcycles suddenly appearing.
The HUD updated smoothly without any warnings.
-Driving Scenario: Complex Risk Pattern Detected
-Expected Intervention Request Probability: 0.8% → 0.4% (Self-Correcting)
-PM Risk Object Approaching
-Characteristics: History of Irregular Driving
-Response Scenario: #0193 Applied
A delivery motorcycle entered from the right alley without a turn signal.
The EV-Nova instantly reduced speed and slightly turned the body to the right.
No warning lights, no honking.
The steering angle was exactly 17.4 degrees.
Like a seasoned driver, it smoothly navigated the situation without anyone noticing.
“Sigh…”
Tesla Headquarters, California.
Elon Musk, watching the video, let out a short sigh.
“This… the difference is too significant.”
For the first time, a ‘mark of defeat’ appeared on his face.
Recently, when the Polar Orbiter project was taken by Dojin, he thought it was just luck.
Except for propulsion reuse, he still believed that Spade Z was ahead in overall technology.
But in autonomous driving, that couldn’t be the case.
Musk replayed the video once more.
The EV-Nova equipped with LAD navigated through the complex urban intersection without hesitation.
Of course, it wasn’t perfect in every part,
but it was far more sophisticated than their FS.
Eliason cautiously spoke up.
“…No matter how you look at it, rule-based seems to have reached its limit. It would be best to switch to a cloud feedback learning structure now…”
Musk didn’t respond immediately.
After a long pause, he slowly nodded.
“This… I have no choice but to admit it.”
It was the moment Tesla, the world’s largest electric vehicle company, changed its autonomous driving strategy from rule-based to behavior-based.
The belated change triggered by Dojin’s LAD.
It meant that Tesla had to follow Dojin’s ‘standard,’ whether they liked it or not…