Text | Outside the Surface, Author | He Jinyi, Zhou Xiao, Tan Jiuyun, editor | Cao Binling, Fu Xiaoling, Mu Mu
When the CyberCab was released, Musk set the venue at Warner Studios, perhaps hoping that the scene of himself appearing in an unmanned car without a steering wheel would be as classic as a Hollywood movie.
But after the press conference, Tesla's stock price plummeted by 8 points.
No wonder investors are dissatisfied, when Musk set the flag for mass production in 2026, San Francisco citizens were already accustomed to Waymo's Robotaxi.
On the other side of the ocean, in China, Radish Express has not only released the world's only front-end mass-produced L4 unmanned vehicle, but also the footprint of autonomous vehicles has spread to more than ten major cities.
For example, recently, it was approved for Hong Kong's first autonomous vehicle pilot license, and has a springboard to enter the right-hand drive and left-hand drive area - this travel market accounts for 1/3 of the global share.
According to the agency's forecast, by 2030, the global Robotaxi market will reach 834.9 billion yuan, which is roughly equivalent to the volume of 6 Uber (travel business revenue in 2023) and 4 Didi.
Chinese players have been eyeing this sweet cake for a long time.
"End-to-end" one foot of the accelerator has blasted autonomous driving into the L4 era
Looking at the bustling streets of Japan, gray-haired elderly taxi drivers abound, and their trembling driving skills need to use a magnifying glass to look at the map, which makes foreign tourists sweat.
Behind the "old age that should be in retirement, but coming out to break through" is the increasing demand for autonomous driving.
It can be seen that the cost of taking off with the age of the driver is also the taxi fee: the cost of an hour's drive is about 20,000+ yen, which is calculated according to the minimum hourly wage of 1,000 yen in Japan, which means that it will take three days to pay for a taxi.
And this is not an isolated case, as you can see, compared to the $0.35/km, on-call taxi service in Chinese mainland, Hong Kong, China charges about $1.28 per kilometer, and even if you successfully get on the bus, you may be stuck in traffic.
Therefore, people in these areas are actually more eager for "self-driving taxis" than in Chinese mainland, after all, driverless autonomous driving is exactly what these pain points are. At the same time, self-driving technology can also intelligently allocate routes to avoid congested road sections.
However, the strong demand overseas is encountered in the local technology "saline-alkali land".
As we all know, as an upgraded application of the Internet, "software + chip/consumer electronics + machinery manufacturing" is the key element of the implementation of intelligent driving technology.
However, in the 1990 PC Internet wave, the technology rookies of Silicon Valley - Yahoo, Google, Facebook, Amazon, swept Europe, Japan and South Korea, suppressed the development of local Internet companies, resulting in a computer talent gap that continues to this day.
In the era of smart phones led by the iPhone, European feature phones such as Nokia and Ericsson, which had eaten 80% of the global mobile phone market, were almost uprooted, and the local consumer electronics industry collapsed, and the electronic supply chain such as processors, memory, and chips declined.
The only advantage of machinery manufacturing has now been hit by a blow.
Just a week before the announcement of Carrot Run's going to sea, Northvolt, a battery company that is the "hope of the whole village" in Europe and has the mission of "surpassing CATL and seizing the dominance of the new energy vehicle wave", submitted a bankruptcy protection application to the court.
The reason is that the local front-line workers to the board of directors lack experience in battery production, resulting in too low a yield rate and a delay in delivery. It is reported that the only workers in Sweden's local factories who are really skilled and capable of working are more than 100 outsourced workers from China and South Korea.
Under various restrictions, the development of new energy vehicles in Europe is slow. As shown in the figure below, compared with China's 40%+ new energy penetration rate, Europe is only about 20%, and it has declined for a time.
The intelligent driving it is equipped with is progressing in the same way. When the domestic volume is not in the country, NOA Kaecheng, and the technology is infinitely close to L3, Europe is difficult to control even the L2 level.
Mercedes-Benz L2+ level assisted driving solution, the maximum speed is only 64.4km/h, and lane changes are not possible; At the same level, the ideal L9 can drive 120km/h and automatically change lanes.
Now, with the introduction of end-to-end technology, autonomous driving has ushered in the "ChatGPT moment", which can be infinitely close to real people, and has a faster computing speed, which can hit the L4 era.
This time, the leaders of technology and application are obviously China and the United States, where autonomous driving giants gather.
In this new era, the competition of various car companies has fallen on the implementation of operations. As Robin Li said, behind the end-to-end autonomous driving technology is the large model technology.
Specifically, through landing applications, it covers more driving scenarios and various abnormal conditions and extreme conditions, so that the end-to-end system can learn and optimize, achieve higher versatility and stability, and finally form a technology-application rotation.
As a result, the expansion of autonomous driving companies has become a top priority, and this "coincides" with the needs of the above-mentioned markets such as Europe and some of Asia (Japan, Singapore, and Thailand).
In addition, these areas are highly urbanized, and the road facilities are relatively complete overall.
For example, the German ATM system has been automatically controlled; Sweden has announced that it will build the world's first permanent electrified road by 2025 to charge a moving car, which will facilitate the visual recognition and charging process of autonomous driving.
In addition, Europe, Japan and other regions have the foundation of the automotive and electronic parts industry chain, which is convenient for localized cooperative procurement; The regulations on the important factors for the implementation of autonomous driving are also relatively complete in these regions.
With such conditions for going to sea in front of us, the field of autonomous driving is naturally in turmoil. It can be seen that since the second half of the year, China and the United States have set off a "Kaesong War", and it has been applied in many regions and countries outside the local area.
However, even if you are prepared, it is never easy to open up in a blank market.
Robotaxi出海的风大,浪也大
If the vehicle is not adapted, and the translation goes out to sea to the right-hand drive and left-hand area, there may be problems such as abnormal shaking of the brake pedal and stalling in the first few years on the road.
This is because the right-hand drive and left-hand countries and regions, most of which belong to the former Commonwealth countries and regions, are located in humid and warm coastal areas, and the wet and salty air blown by the sea breeze will form salt spray on the surface of electronic components, corrode sensors and other components, and accelerate the speed of vehicle wear.
Based on this, overseas enterprises often need to develop their own cleaning systems to adapt to the complex local climate and road conditions.
It can be seen that the completely different driving environments at home and abroad always test the players' ability to localize.
However, for China's Robotaxi backed by "industrial Cthulhu", the hardware problem is not tricky - at present, the domestic autonomous driving industry chain is becoming mature, and some parts and components have led the world, which can quickly respond to the lean production needs of going overseas.
The real challenge is the customization of the software system.
The driving habits of overseas users are different from those in China, for example, domestic car owners want to have strong safety warning signals, but European car owners are used to driving quietly, and frequent prompts will be regarded as noise. In addition, autonomous driving systems also need to learn and adapt to traffic regulations in different regions.
But understanding is only the first step, and if you want to do a good user experience, you have to continue to hone your technology.
For example, Europeans like to live in the suburbs, and there will be many high-speed scenes on the commute to the city center, and it is often a high-speed without a speed limit.
The faster the vehicle speed, the more the autonomous driving system needs to have sufficient perception, planning, and control capabilities to "see the road clearly, pay attention to the road conditions, and operate quickly".
To put it bluntly, it is necessary to have a more high-definition camera, more powerful computing power, and process more images and information per second.
For example, Germany is one of the first countries to legislate on autonomous driving, but the driver is required to sit in the seat and be ready to take over the vehicle at any time.
However, today's L4 level of autonomous driving may bring new changes.
Take the domestic Radish Express as an example, it provides almost 100% of the fully driverless ride-hailing service in Wuhan, you know, Wuhan is also called "a city full of superhero drivers" on the Internet, and the traffic accident rate ranks among the top in the country all year round.
Not only that, the radish run has also gone out of Wuhan and landed in more than ten cities such as Beijing, Shanghai, Guangzhou, and Shenzhen.
This can be achieved because L4 autonomous driving has strong safety capabilities and generalization capabilities.
After the full evolution of multiple technologies such as sensor sensitivity and system algorithm capabilities, L4 unmanned vehicles can be fully autonomous driving without human supervision, and Tesla and Baidu's unmanned vehicles have eliminated or hidden the steering wheel.
In addition, L4 also achieves greater safety redundancy (safe for extreme environments and redundant for normal environments), which makes vehicles better able to cope with stricter traffic regulations and more congested and disorderly road conditions even when they go to sea.
Furthermore, with the support of the end-to-end autonomous driving model, the system can also rely on continuous learning to accelerate the understanding of traffic regulations in different countries and regions, and continuously improve the algorithm capabilities.
This also determines that the longer the autonomous driving mileage accumulated by players and the more cities covered, the more obvious technical advantages they will have, just as Baidu's Apollo autonomous driving open platform was upgraded to version 10.0 and released to global users on December 4.
Of course, all this has to be based on the premise of getting access qualifications, otherwise it is impossible to accumulate experience and train large models, which is why Baidu's stock price rose after Radish Express drove into Hong Kong.
In addition, Robotaxi going overseas on a large scale will also "empty the wallet", after all, the new energy infrastructure in most overseas areas is underdeveloped, taking Hong Kong as an example, the number of electric vehicles accounts for only 11.2%, and the supporting facilities naturally cannot keep up.
In this way, overseas enterprises must have the ability to reduce the mass production and operating costs of vehicles, otherwise it will be difficult for the money-burning game to continue.
At present, players at home and abroad have given their own solutions, and the mass production cost of unmanned vehicles has reached the level of 200,000 yuan.
From this perspective, Robotaxi is also ready to fight a tough battle, and the next year or two will be the time to decide the fate.
brief summary
In 2013, when Robin Li completed Baidu's Silicon Valley R&D center, he looked forward to AI and said: According to Moore's Law, if you continue to do it for ten or twenty years, the AI brain is likely to be smarter than the human brain, and then qualitative change will occur.
However, ten years of persistence and hundreds of billions of R&D expenses have been invested in exchange for a cold bench. It's just that on Wall Street, people don't like the "long-termism" in the uncertain track of technological innovation.
Now, the boots have finally landed, the intelligent transformation of automobiles has fully entered the L4 stage, and the time has come for "long-termism" to be realized, and with the carrot run into Hong Kong and the world, Chinese players will also have the opportunity to rush and take the lead in this blue ocean market, and grab the qualification to define the track.