The approach and extent of dealing with issues related to autonomous driving vary depending on the position. It is undeniable that Elon Musk of Tesla has contributed to raising the stock price by announcing that he will unveil a fully autonomous driving car one year after a year since 2016. Those who support Tesla promote it as a giant technology company like Apple. On the other hand, companies like Waymo from Google, GM Cruise, Ford Blue Cruise, Hyundai Group’s Motional, Amazon Zoox, Baidu from China, and Pony.ai are approaching the issue more conservatively compared to Tesla. It is still unknown which side is right. However, the position difference between Tesla, led by a few genius engineers, and the multitude of opinions is clear. Amid this, Elon Musk declared that he will release FSD version 12 at the end of this year, not a beta version. Of course, it is for the robo-taxi business. Last week, he demonstrated a vehicle equipped with an internally developed system by driving it himself. Before that, there was an anti-robo-taxi movement in San Francisco, USA. Let’s summarize the two incidents.
Written by Young Seok Chae (Director, Global Auto News)
Tesla has unveiled FSD (Full Self-Driving) beta version 12. Elon Musk tested the new version by live streaming while riding in the vehicle. The video, which has already surpassed 11.4 million views, had very poor image quality, possibly due to using an old smartphone. Although it may not be directly related to autonomous driving, as an action by a giant technology company, it can undermine trust. It can also be pointed out that Elon Musk was seen manipulating a mobile phone with his hand while in the driver’s seat, violating current traffic laws.
The test took place mainly in Palo Alto, a relatively quiet area in San Francisco, where Tesla’s engineering headquarters are located. The 45-minute livestream was conducted with Elon Musk and engineers riding together. The maximum speed shown in the video was 43 miles per hour on the instrument panel. When San Francisco granted permission for autonomous driving by Mercedes-Benz in June, the speed was 40 miles per hour. 43 miles per hour could be a mechanical error.
It is worth noting that the final version of Mercedes-Benz Drive Pilot, which received permission to operate in San Francisco as the second location after Nevada, aims to achieve level 3 autonomous driving at a maximum speed of 80 miles per hour. Like Nevada, Drive Pilot can be used on certain high-traffic highway sections in California.
The version tested this time is referred to as version 12, not a beta version. Although it has not entirely moved beyond the beta stage yet, Elon Musk emphasized that he can keep his promise to release the full version at the end of this year.
“Neural net responds to all situations without internet connection”
The most prominent feature of version 12 is that vehicle control is operated by a neural net rather than engineers directly inputting data into the code. It is still in the internal testing phase. Elon Musk has been repeating the phrase “one year later” since its initial introduction in 2016, but this time, it is firmly set for the end of the year.
Although it is impossible to assess all situations through the video accurately, there seems to be some improvement in intersection navigation. In beta version 10.69, there were many instances where it did not work properly, requiring the driver to take control. In this video, most of the 45-minute duration had no issues. It also appeared to have improved its response at roundabouts, known as “rotaries” in the Western region. It also looked good when it automatically parked in the appropriate position upon reaching the first destination set through navigation.
However, as Elon Musk mentioned, there were situations where the vehicle initially drove in the wrong lane. The most attention-grabbing moment was when the car detected a green traffic light for another lane and attempted to drive when the light was red, prompting Musk to disengage the system. This indicates that malfunctions can occur in other areas as well. If someone like Elon Musk does not pay attention to the vehicle’s situation and sits in the back seat, it can lead to accidents.
Tesla’s FSD is equipped with 8 cameras based on Hardware 3.0. It operated at a maximum of 36 frames per second input, and Elon Musk stated that it can process calculations with pure AI faster. He also added that the theoretical maximum of the software is about 50 frames per second. Still, for FSD V12 software to function correctly, it only needs to provide 24 FPS under road conditions.
It is emphasized that FSD 12 can be operated without an internet connection. Elon Musk also mentioned that FSD AI now receives an enormous amount of video to identify tasks that must be performed in various situations, rather than individually coding each road element or recognizing each person riding a scooter, like a rotary. The recognition of bicycle lanes on the right side of the lane was clear, and it emphasized that it responds more conservatively to pedestrians and people riding bicycles simultaneously.
Through this, FSD V12 has removed hundreds of thousands of lines of code, making it lighter and more agile, and it can continue to operate even in unfamiliar terrain without a data connection. Musk also mentioned that Tesla has FSD V12 beta testers worldwide, including New Zealand, Norway, Thailand, and Japan.
During the test drive, the engineer explained that the current priority when it rains is to drive slowly, as the development is ongoing. Overcoming that robo-taxis will have to stop operating in severe weather seems to be a difficult task.
Even if Tesla presents FSD as a completed version, it cannot capture all the countless phenomena and teach them to artificial intelligence through videos. The roads of the human world are filled with innumerable variables at every moment. That’s why traditional automakers use the term ADAS more often than autonomous driving. No one knows which side will conclude.
Although the video has already surpassed 11.4 million views on X, it is considered a failure to demonstrate the performance of FSD v12 adequately.
Nevertheless, Tesla takes a significantly different technological approach than other autonomous driving companies. While companies like Waymo, Cruise, and Baidu rely on a foundation of 360-degree LiDAR, Tesla’s Full Self-Driving (FSD) system uses cameras and neural networks. In videos, Elon Musk has repeatedly emphasized that accumulating visual data through cameras is more important. He believes that training the AI solely on that visual data is sufficient. On the other hand, companies utilizing LiDAR technology assume the need to integrate with different infrastructures, including V2X, to ensure safety.
In response, a prominent American analyst named Dan Ives, who strongly supports Tesla, stated that the next three decades will bring the most significant transformation in 30 years. He predicted that Tesla, like Apple, will generate astronomical profits as a disruptive technology company utilizing advancements from superchargers to batteries and FSD with AI.
However, even now, there are numerous reports of operational errors in Tesla’s FSD in the United States. Occasionally, incidents with a certain degree of probability, such as experiencing the same errors under the same conditions, have also been reported.
In San Francisco, USA, there is also opposition to robotaxi services.
On the other hand, in San Francisco, USA, there is an ongoing movement against the operation of robotaxis. Multiple media outlets, including Forbes, Automotive News, Detroit News, and Electrek, have reported that members of a group called “Safe Street Rebel” have been placing cones on top of robotaxi hoods in San Francisco since July. As reported by these sources, the orange construction cones have become a symbol of resistance against autonomous vehicles.
In San Francisco, there have been instances where autonomous vehicles have hindered emergency response teams and caused traffic accidents. However, the opposition from the San Francisco Department of Transportation, city police, and fire department officials to these concerns has not led to any resolutions.
Last Friday at 5:40 PM, the California DMV demanded that GM’s Cruise robotaxi division reduce its operations in San Francisco by over 50%. This request was made due to an incident where a Cruise vehicle ignored a red light and collided with a San Francisco firetruck. This was not an isolated incident, as the San Francisco Fire Department reported 55 similar incidents.
As a result, Cruise has reduced its operations by 50% during the day and 150% at night. The San Francisco DMV stated that this reduction will continue until corrective actions are taken.
Forbes reports that although there hasn’t been an official investigation in San Francisco, there is a shared opinion that Waymo’s vehicles have performed better and caused fewer problems than Cruise. However, Cruise argues that the higher number of operations is the reason for the discrepancies. It’s important to note that Waymo operated mainly during complex daytime traffic, while Cruise initially received permission to provide passenger services only at night. Both companies, however, have not provided precise figures about the differences in their operations.
Despite the opposition, San Francisco authorities approved the 24/7 operation of driverless, fully autonomous taxis within the city on August 10, 2023. The California Public Utilities Commission (CPUC) announced the expansion of autonomous driving services requested by GM Cruise and Waymo, both of which had applied in December 2022. Three out of four commissioners involved in the decision favored this expansion. In response, Safe Street Rebel organizers stated that they would continue their fight against autonomous vehicles, especially if local transportation authorities formally requested a review of their decision by the state regulatory body.
The recent vote has allowed Cruise to expand its operations in the city and Waymo to initiate commercial services. The fully commercial service is approved for downtown San Francisco in the northern part of the state. The combined fleet size of both companies currently exceeds 500 vehicles, and both have plans to expand further. GM Cruise began offering driverless robotaxi services for a fee in certain city areas as of June 2022, marking an evolution from the initial limited testing phase that had restrictions on usage times and areas.
GM emphasized that competition with ride-hailing giants like Uber Technologies will intensify with expanded daytime operations. Given the concentration of tech companies in California and San Francisco, the region has become a hub for autonomous driving development, attracting tech professionals worldwide. Local governments and authorities are generally open to adopting new technologies.
Anticipated Changes in the Robotaxi Business
Now, with the ability to summon robotaxis 24/7, traditional taxi services may take a hit. Robotaxis, charging only $3 per ride compared to a regular taxi’s $10 for the same distance, will likely disrupt the market. Resistance to such changes is expected due to safety concerns and vested interests within related industries.
Safety is a paramount issue that requires extensive discussion. Accidents involving autonomous vehicles can raise significant concerns and resistance from users accustomed to conventional systems. The ongoing debate about which driving mode, human or autonomous, is safer in the event of an accident adds to the skepticism.
Another aspect involves the clash with traditional taxi businesses. While perceptions may vary by region and country, the disappearance of livelihoods sustained until now could be a significant concern for public transportation stakeholders.
From a technological perspective, new challenges are emerging. Recently, Tesla announced plans to license its Full Self-Driving (FSD) technology, drawing parallels to Apple’s iOS and Google’s Android for smartphones. This brings to mind the past when Microsoft’s Windows Mobile, Amazon’s Fire OS, and BlackBerry OS lost prominence with the rise of iOS and Android’s dominance in the market.
Tesla’s FSD, leading in the autonomous driving field, has the potential to dominate the market, similar to an operating system. As a result, companies like GM’s Cruise, Ford’s BlueCruise, Hyundai’s Motional, and Baidu’s Apollo are investing heavily in high-end autonomous development. The fundamental approach of these companies differs from Tesla’s. Tesla claims cameras, artificial intelligence, and neural networks can handle all tasks. In contrast, the mentioned companies focus on a LiDAR-centric approach and envision systems integrated with V2X communication.
The question of which system will be safer and more user-friendly remains uncertain. However, the possibility of a few dominant autonomous driving OS providers akin to smartphone providers is conceivable. Tesla has already demonstrated a willingness to share its Supercharger network with other companies in the charging market to increase profitability.
Additionally, connectivity and infotainment systems, referred to as software-defined vehicles, are interconnected. Honda, considering independent investment challenging, has invested around 300 billion yen in GM’s Cruise over 30 years. This might signal a desire to generate profitability in other sectors through collaborations like the one with Sony in the entertainment sector.
On the other hand, within the United States, opinions regarding resistance can emerge during the introduction stages of new technologies. Given that emerging technologies can directly threaten local communities and lifestyles, resistance or hesitation toward technology can be seen as rejecting change.
Most Commented