The perfect story that is unsatisfactory

Author | Wei Qiyang

Source | Insight New Research Society

Great Wall Motors has a strong ability to manufacture the topic. The most recent one is that Wei Ruifeng, CEO of Wei brand 2 months ago, "touch porcelain" Huawei on Weibo, shouting Yu Chengdong across the air, and frying the "leftovers" of "backward process". After.

Image source: Li Ruifeng Weibo Screenshot

Today, the relay rod has passed to the hands of Zhixing.

Not long ago, at the HAOMO AI Day, Zhang Kai, chairman of Zhixing, and CEO Gu Weiyi, high -profile official announcement, and stabilized the "China's first place in mass production of autonomous driving".

Prior to this, Zhixing also announced that he won multiple "firsts", including and not limited to:

China's first autonomous driving data intelligent system, Mana;

China's first large -scale mass -produced city -assisted driving product -the cities NOH;

Global's highest computing power -producing autonomous driving computing platform -Little Magic Box 3.0;

China's first 100,000 yuan terminal logistics automatic delivery vehicle -Small Magic Camel 2.0;

Native

For a startup company that has just been established 1,000 days, it is indeed a great achievement in R & D and product landing. What is "great".

On the one hand, many "firsts" of Zhixing are implemented in the L2 -level auxiliary driving, not the autonomous driving of the L4 level that does not need to be manipulated by the L4 level. On the other hand, in the long march of autonomous driving, "starting" The leading lead does not mean that it will rush through the end.

Regarding the current stage of autonomous driving, Wu Gansha, CEO of Yushu Technology, has an accurate exposition. "Autonomous driving is far from entering the second half. Large -scale commercialization phase. "

At present, Zhixing threw a set of "successful methodology" of a self -contained system. The story told is also perfect, but it can be deepened. The story of Zhixing seems to be able to withstand too much scrutiny.

01 Data Assembly is controversial

So far, Zhixing has launched three generations of passenger vehicle auxiliary driving products HPILOT, and has upgraded 6 OTA within two years. The reason why technology and products can quickly iterate are data acquisition and utilization.

Back to the Great Wall, more than ten Great Wall brand models are equipped with the slightest HPILOT products to mass production. In addition, there are Mocha DHT-PHEV lidar version, Euler Lightning Cat, Euler Ballet cat, new generation of Great Wall guns and other models are currently. Settle in one after another.

Based on the advantages of HPILOT on the car, the user's accumulated user assisted driving mileage is close to 17 million kilometers.

At the end, it is obviously the sweetness of the progress of data -driven technology, thus becoming a firm supporter of the progressive route of autonomous driving. Zhang Kai even concluded: "Auxiliary driving is the only way to autonomous driving."

For this reason, a "windmill strategy" has been proposed. The rwitar, low -speed Uonjo -vehicle ecological platforms and intelligent hardware represent the three blades of windmills. The leaves are flying, and the data is closed through the data intelligent system.

There are two places where there are controversy in this set of theory.

The first is the quality of data.

At the end of the Joint HPILOT, there are three versions of iterations, in turn, HWA high -speed intelligent driving systems, NOH smart pilot assist driving systems, and recently released the upcoming city NOH pilot driving system.

There is a detail. There are no laser radar deployed in the first two versions. The integration and perception of the pilot assistance system is to be observed.

On the other hand, the HWA high -speed intelligent driving system and NOH smart pilot assist driving system are all facing high -speed and urban fast road scenarios. Compared with the city market scene, the data structure is relatively single. NoH's algorithm training of NOH pilot assistance system is not known.

The second is the dimension application of data.

According to the windmill strategy, the mass production data of Zhixing must not only promote technology from L2 -level auxiliary driving to a higher -level autonomous driving evolution, but also applies across the field to cooperate with the L2 -level auxiliary driving data with the low -speed unmanned vehicle of the end distribution. Essence

The idea of ​​"Windmill Strategy" is very similar to the "flywheel" L4 concept proposed by Momenta two years ago. Cao Xudong, Momenta CEO, is interpreted: the continuous accumulation and accumulation and accumulation of closed -loop automation by mass production data, data -driven algorithms, and closed -loop automation of the two are the continuous accumulation and accumulation of closed -loop automation. Iterations to promote the "two legs" of mass production of autonomous driving (MPILOT) and completely driverless (MSD).

At that time, many big guys in the industry put forward different views on this technical route.

Yu Yan, CEO of Qingzhou Zhihang, believes: "From the types and computing platforms of sensors, the application scenarios of L2/L3 are very different from L4, and it is difficult to form interoperability with the data of L4. Big challenge. "

The engineer of Xiaoma Zhixing believes: "It is relatively easy to use a high -profile L4 -level autonomous driving architecture to do L2/L3 level. Conversely, low -configuration data is difficult to use for testing and certification of the entire L4 level system. "Tunsen's future chief scientist Wang Naiyan even wrote a special debate in Zhihu." In the intentional or unintentional rendering of some manufacturers and some media, the data is the pirate of the spiritual sanxation. 'Growing' into a L4 system ... "

Wang Yan believes, "If the architecture of a system is designed for L2/L3 (allowing the system failure caused by a single point failure), then even if there are more data, it cannot automatically complete the failure detection and processing, and complete the level of the level. Capital. "A reliable L4 system requires creation of engineering, algorithms, hardware, and vehicles in order to create.

Obviously, the data can be reduced, but it is difficult to increase.

02 terminal delivery order suspicious cloud

In addition to passenger cars, the maximum achievement made by Zhixing in the field of autonomous driving is the bottom line of the end-of-the-end distribution of the unmanned vehicle with the "Small Magic Camel 2.0". Pull to the price of 100,000 yuan.

Following this carefully created Wang -level product, the official announcement, as of September 2022, the order volume of Xiao Mo camel has exceeded 90,000 orders. To this end, it also built a "star ring ring in Baoding. Factory "flexible manufacturing base, with an annual production capacity of 10,000 vehicles.

90,000 delivery orders are indeed a number worthwhile, but how these 90,000 orders are realized, but it is a big question mark.

There is a small detail here, and there is a "front department" in exaggerating marketing.

In June last year, "Cyber ​​Automobile" disclosed that it has declared that its unmanned vehicles have entered the normal operation in "Wumi Multi -Dot Supermarket in Shunyi, Beijing, and the cumulative service of nearly a thousand orders." It is found that this is not the case, but "there will be a staff member with the remote control remote control, and occasionally come to test."

Another direction of thinking, assuming that the 90,000 orders have no water, it is indeed a single accumulation, so the new problem arises. It only takes more than 1 year to complete the order. Is the technical maturity and public acceptance of human delivery so high? Has the era when unmanned delivery has come to the ground?

In fact, after nearly two years of resistance, the value of unmanned vehicles has been released, but their technical capabilities are also recognized by society.

For example, there is no perfect solution to the unmanned vehicle in the face of the upper and lower elevators, and it will be "stuck" when encountering steps or deceleration belts in the community;

For example, the space of the community is narrow, the driving channel is not smooth, and the unmanned vehicle is often "dilemma" when the unmanned road passes through similar lines, becoming a congested node;

For another example, some community road design is twisted, and many unmanned vehicles have insufficient turning radius to be able to do delivery tasks.

During the delivery of Shanghai Jiaotong teachers and students, in terms of food delivery, comparing a variety of distribution tools such as private cars, self -developed unmanned cars, and unmanned small buses with a larger amount of cargo in school automation. The child is the best.

The experience of Shanghai Jiaotong teachers and students is: Sanjiezi can go to the streets and send meals to the closest place to the dormitory building.

It is more regrettable that there is no excessive display of the technical capabilities of the small camel, but the focus of marketing is placed on the ultra -low price.

At the end, Zhixing specifically calculated the cost account: A courier was calculated at a monthly salary of about 10,000 yuan, and it was almost 130,000 yuan a year. The cost of Xiao Mo Camel was basically the same as the annual salary of a courier.

However, the number of unmanned vehicles is not only compared with the cost of purchasing costs with the labor cost of the courier/delivery staff.

For example, in the field of passenger cars, due to different functions of SUV and cars, there will be differences in prices. The price is not the decisive factor that affects the landing of unmanned vehicles. Whether the accumulated data is enough, no matter what price, whether the existing technical structure can support the needs of commercialization.

However, the current reality is that all unmanned delivery projects have autonomous driving engineers to participate in the whole process and are always preparing to solve the problem.

Guan Chaowen, CEO of Yunzhicha in Beijing in Beijing, once talked about this status quo. "At present, an unmanned car barely replaces a courier with a monthly salary of 6,000 yuan, but I have to take an engineer with a monthly salary of 30,000."

As a result, Liu Zhiyong, a real machine intelligent CEO, judged that unmanned driving must land on the logistics full chain. Only revolutionary phenomena will occur. "

According to the above logic, it takes a longer time to win the battle scale of the end logistics intelligent delivery vehicle.

03 Dust the prosperity of the Great Wall is difficult to balance

Finally, take a look at the growth of the slightest Zhixing.

On the bright side, the end is incubated from the forward -looking section of the Great Wall Motor Technology Center Intelligent Driving. At the end of 2019, from the Great Wall independence, although multiple rounds of financing was completed later, the introduction of Meituan, Qualcomm, Gaoma Capital, and the first trip of the way Holdings and other shareholders are currently valued at $ 1 billion, but the absolute major shareholders are still Great Wall Motors.

According to the data of the enterprise, Baoding City of Motorcycle Technology Co., Ltd. (holding 29.14%) and Hebei Xiong'an Great Wall Motor Technology Co., Ltd. (holding 27.75%) two total holdings of 56.89%. Among the many shareholders, in addition to Meituan and Qualcomm's industrial resources, most of the other shareholders are the nature of financial investment. This is caused. Although it is born in the Great Wall, it is very dependent on the situation of the Great Wall.

Gu Weizheng expressed his position very straightforwardly last year: "The Great Wall itself is a very thick leg. It has sold 1 million units for several years, and the strategy of 2025 sold 4 million vehicles (one year). Why don't I serve the Great Wall well? "

If it is a start -up enterprise, it is natural to hold the "thighs" like the Great Wall, but if you have a farther ambition, you will only be tied to an OEM in depth, and the value will be very limited.

The reason is very simple. The Great Wall is prosperous, but the upper limit of the slightest is the upper limit of the Great Wall; but if the Great Wall is declining, the lack of buffer because there is no industrial ecology, and the impact will be higher than the Great Wall.

The end also realized that "cannot hang on a tree", gradually dilutes the "blood" relationship with the Great Wall, and emphasizes its independence. At the brand day event in April this year, a "6P Open Cooperation Mode" facing the field of passenger car driving was launched, intending to develop more cooperative OEMs, but as far as the results shown at present, have not made breakthroughs.

From the perspective of the bystander, the end may never break through.

As mentioned above, the ending of the data -driven strategy, it has indeed achieved prominent results through data driving in R & D and landing. The data from the Great Wall models "feed".

In other words, if there are other brands in cooperation with the slightest, you need to face the problem of data belonging and utilization directly. Therefore, as long as the Great Wall still controls the slightest, any brand has doubts, "Will my data be a wedding dress for the Great Wall, "?

If this "heart knot" cannot be unlocked, it will be difficult to take a step out.

The embarrassment is how to position it, whether it is an independent company that is separated from the Great Wall, or a subsidiary of the Great Wall.

If it is the former, it is necessary to further dilute the Great Wall shares in the subsequent financing process and introduce more industrial investment, but will the Great Wall let go easily?

If it is the latter, there may be more hurdles to cross.

In terms of intelligent transformation, foreign car companies have actually carried out a lot of practice.

Toyota originally had a department named TRI (Toyota Research Institute) specializing in autonomous driving research. However, Toyota still spent 500 million and $ 400 million. Investment (Uber sold the "Advanced Technology Group" (ATG) to Aurora at a price of $ 4 billion in December 2020, and formed an in -depth cooperative relationship with Aurora.

After GM purchased CRUISE, which cost US $ 1 billion to acquire the autonomous driving startup CRUISE in 2016, it successively acquired Zippy.ai, the last mile of automation startup, "chip -level" laser radar technology provider Strobe, etc. Essence

The public has both the AID of the autonomous driving research and development department, and then established the car.software of the automotive high -end auxiliary system department, but still merged AID to ARGO AI at a price of $ 1.6 billion. In addition, it added an additional $ 1 billion in cash investment.

After so many examples, I just want to explain that it is difficult for traditional cars to get autonomous driving based on their own resources and capabilities. Aside from institutional processes and cultural genes, rarely the development costs of the bottomless cave. Essence

Therefore, for the slightest, it is also possible to use the L2 -level auxiliary driving as the base plate to make the Great Wall's intelligent transformation of the Great Wall's intelligent transformation.

04 conclusion

Zhang Kai said on HAOMO AI Day that in 2022, Zhixing faced three major battles: the battle of data intelligent technology, the battle of assisting the driving city market, and the battle of the scale of the smart delivery vehicle at the end of the logistics.

If you don't have an accident, in the near future, you will still hear the full comprehensive war report, of course, it will also reproduce such doubts today.

We do not doubt the great ambitions, we only worry that we have lost ourselves in the intricate changes.

The word "Lao Tzu" is taken from the eternal saying of "Lao Tzu".

How to adhere to the original intention and how to be down -to -earth is the core of the self -introspection. Perhaps on the next Haomo AI Day, more pragmatic is needed, not another self -hi -type carnival.

- END -