Why is Jiang Haipeng anxious
Jiang Haipeng has been very anxious in the past few days. He said that I had no more than three hours off a day. As he said this, he stood in a packed conference room, facing the media, and honestly confessed, "So I'm sleepy now." ”
There was an immediate sound of good-natured laughter.
Jiang Haipeng continued: "This incident was very sudden - the boss told me that I wanted to try your car. ”
"Yes. He replied.
"I'm going to go live. The boss continued.
This is what makes Jiang Haipeng feel "devastated": as the senior director of Great Wall Motor's intelligent driving, he has just completed the initial development of the software for assisted driving in urban areas, and is on the eve of large-scale road tests. Faced with the boss's "surprise inspection", he worries that the city NOA is not yet mature enough.
But Wei Jianjun is very resolute: "I just let everyone put forward opinions, open the door, let everyone know what Great Wall Motors is doing, and hope that the outside world can see our changes." ”
So, the live broadcast time was set. Wei Jianjun posted a short video, inviting netizens to witness the latest progress of Great Wall Motors in the urban NOA on the morning of April 15. Great Wall Motors, which has always been immersed in technology and research and development, is embracing new changes.
The live broadcast test location is in Baoding, which is also the base camp of Great Wall Motors. Taking an assessment on your own turf, some people joked, should be as easy as an open-book exam.
Jiang Haipeng doesn't think so. Although it is only 140 kilometers away from Beijing, Baoding has a more complex traffic environment: traffic, people and vehicles are mixed, traffic is not standardized, and road construction occurs from time to time. In addition, there is no high-precision map here, and the actual measurement is also a verification of the no-map scheme, "Baoding road conditions are very typical, if Baoding is opened thoroughly, it will get 80% of the scenes in the country." ”
Rather than an open-book exam, the actual test in Baoding is more like answering an additional question.
At 10:30 a.m., Wei Jianjun, Wu Huixiao, and Jiang Haipeng set off from the Great Wall Haval Technology Center to test the situation of the city's NOA together. Driving through the city street, the old town, and then to the narrow street, the intelligent driving completes the navigation lane change, merge in/out ramps, avoid vehicles, pedestrians and detours, etc., and the city NOA also passes smoothly on the main road where the lane line is intermittent.
In the process of turning at the fork in the road, the intelligent driving vehicle encountered the oncoming vehicle on the left. The city NOA chose to avoid it. From one point of view, Great Wall Motor's plan adopts a conservative game strategy.
When driving to the most complex "Liudaokou", which is intersected by six main roads, the intelligent driving system also passed smoothly and accurately identified the cones and pedestrians in the process of passing.
Wei Jianjun felt that during the whole process, Zhijia achieved a silky response like a human old driver in the start and stop, and "the response was also very decisive".
In this one-hour live test, the total mileage of intelligent driving is 16.6 kilometers, and the average speed is 24.7 kilometers per hour. Zero mistakes throughout. The whole live broadcast was watched by more than 10 million people.
From "machine" to "brain"
The whole live broadcast went smoothly, and it was Great Wall Motor's latest generation of intelligent driving system Coffee Pilot Ultra (CP Ultra) that played a role.
This system is based on Orin-X's high-computing domain control platform, with 1 lidar, 3 millimeter-wave radars, 11 high-definition cameras, and 12 ultrasonic radars to form an all-weather perception capability. Achieve 100% high-definition maps and full-scene urban NOA driving experience.
According to Great Wall Motor, the CP Ultra is equipped with dozens of intelligent safety control capabilities such as AEB, LKA, ELK, etc., in addition to regulatory and standard scenarios, it also adds AEB functions to recognize and respond to a variety of general obstacles and pedestrians, further improving the safety of complex scenarios. In the state of intelligent driving, it can brake the vehicle at 120 km/h and the pedestrian at 100 km/h.
It is worth mentioning that Great Wall Motors has launched the SEE integrated intelligent driving model. According to Wu Huixiao, the name comes from three English words, Safety, Efficiency, and Experienced, which represent the triple guarantee provided by the model in the process of intelligent driving:
The first is active safety protection, the second is that the complexity of road conditions to which the vehicle responds has been greatly improved, and the decision-making behavior of braking, turning, and moving forward is closer to that of humans, and the third is to achieve a smooth experience of using all scenarios and driving on the road.
"SEE itself also means 'seeing' in English, which means that this system can see things, see the road clearly, and understand people. Wu Huixiao revealed the deeper expectations in this naming.
In March this year, Tesla pushed the V12.3 version of FSD in North America, bringing autonomous driving to the next stage: no rule code, only neural networks.
In the previous V11 version, Tesla's engineers also needed to manually write 300,000 lines of C++ code to input concepts such as traffic light signals, pedestrians, and roads, while in V12, the neural network model will directly output decision-making commands such as steering, acceleration, and braking after input images, and no longer rely on rule judgment. The code is only 2000 lines. The level of involvement of manual rules is significantly reduced.
The rules of the game have been rewritten, and end-to-end has become a new battlefield for intelligent competition. Starting from this idea, Great Wall Motors has introduced end-to-end technology to ensure experience, and realized the link integration of perception and decision-making. By feeding data for joint training, the SEE model can also learn "how to drive" directly like a human driver, rather than a rule-defined driving behavior. This means that as long as the data is continuously fed, SEE has unlimited growth possibilities and lays the technical foundation for the evolution from L2 to L4.
"In the past, the concept of intelligent driving was to teach the machine how to drive, and to break down complex driving tasks into single tasks that the machine could understand. The emergence of AI models provides us with new ideas. If we regard a large model with complete cognitive ability as an 'adult' and realize the transformation from 'teaching machines to drive' to 'teaching AI to drive', is this the future of intelligent driving based on large models?" Wu Huixiao believes that with the rule-driven shift to data-driven, the SEE integrated large model has also completed the transformation from "machine" to "brain".
"A lot of cars either don't go around when they encounter obstacles, or they steer very stiffly, and the car drives like it's breaking, and I want our cars to come out very, very silky, waltzing instead of mechanical dancing. Wu Huixiao said.
At present, Great Wall Motors has taken out 200 cars for generalization tests, and in addition to the original four cities of Baoding, Shenzhen, Chengdu and Beijing, the top 15 cities of Great Wall Motors sales are also expected to take the lead in opening the city. According to the previous rhythm arrangement, 100 cities will be opened in 2024.
Wu Huixiao also mentioned in a previous interview with Cheng Li, the founder of Cheyun, that intelligence will cover the whole line from oil vehicles to new energy vehicles.
Continue to invest in intelligence
In the face of the wave of intelligence, this hard-working car company that believes in the iron law of technology does not want to miss the key node of industry change. In fact, Great Wall Motors has invested in relevant research for a long time to promote Great Wall Motors to accelerate its transformation into a "global intelligent technology company".
In 2009, Great Wall Motor has successively carried out research and development on key technologies such as 5G and autonomous driving. A R&D team for driver assistance systems has been established in-house;
In 2017, through its subsidiary in the United States, Great Wall Motor plans to spend $3.51 million to establish an R&D center in Detroit and build an autonomous driving team.
In 2019, Great Wall Motors separated the Intelligent Driving Foresight Department and established Momo Zhixing. It has also become Great Wall Motor's intelligent driving supplier, and so far, its mass-produced intelligent driving product HPilot has been equipped with more than 20 models of Great Wall Motors;
In 2020, Great Wall Motor released the "Coffee Intelligence" vehicle intelligent brand, covering intelligent products in the whole domain such as intelligent cockpit, intelligent driving, and intelligent electronic and electrical architecture.
With chatGPT becoming popular all over the world, the new paradigm is also refreshing the competitive landscape of the automotive industry. At last year's Shanghai Auto Show, AI intelligent transformation became the consensus of many car companies. Great Wall Motors is also involved in this. In September, the AI Lab was officially established to build a technology middle platform from the R&D end to the vehicle end for coffee intelligence, which is also a cutting-edge technology R&D center at the group level, providing general large-scale model solutions for the Great Wall in the future.
At the beginning of this year, Great Wall Motors elevated AI to an enterprise-level strategic dimension, and the AI Lab reported directly to the head Wei Jianjun, with a fixed weekly meeting. The interior has been described as a green light for intelligence.
With the advent of the era of AI-defined automobiles, computing power has become the primary productive force. Great Wall Motors has built its own Kyushu Supercomputing Center, which has completed the data closed loop of large models, big data and large computing power. According to data from Great Wall Motors, the computing power scale of the Kyushu Supercomputing Center has reached 1.64EFLOPS, achieving a high-performance storage of 5 T/s and a high-performance network with a communication bandwidth of 3.7 TB/s.
On the basis of a certain amount of computing power, how to squeeze the computing power of the supercomputing center to the most efficient and improve the efficiency of data training also requires a large amount of capital investment. In 2021, Wu Huixiao saw the amount of money spent on data and computing power, and was surprised and sighed that it was "a lot of money", but within a few days, "I stood in front of the boss with the people who did data and computing power and said, I want more money." This is enough to see that Great Wall Motors dares and is willing to make enough investment in this forward-looking technology.
With the increasing weight of intelligence in the automobile competition, Great Wall has also been increasing the relevant "chips", with 10 R&D centers around the world, and more than 4,000 R&D personnel related to intelligence.
For Wei, he understands the importance of winning technology and R&D opportunities in a limited competitive window. In an internal speech, he expressed his determination to fight, "In China, there is only three or five years of opportunity to overtake with intelligent and electric lane changes." There is only one chance to rush, fast, in order to maintain value, slow is equal to depreciation. It's important to get a head start and access resources. ”
In fact, the desire to become a leader in autonomous driving has driven Great Wall Motors to continue to expand its technological horizons. In the core power system of the car, Great Wall Motors in more than 20 years of technical polishing, finally got rid of the dependence on foreign countries, the first Chinese high-end powertrain 3.0T + 9AT/9HAT "14 horsepower more than the same type of engine in Europe", representing China's current level in powertrain; The layout of the common development of the three major routes of hybrid power and hydrogen energy.
With the incubation or acquisition of companies such as Honeycomb Energy, Momo Zhixing, Xianlin Intelligence, Jiayu Intelligence, MIND Electronics, Seiko Motors, etc., Great Wall Motors has built an industrial system of the whole industrial chain from vehicle R&D and manufacturing, electronic and electrical architecture, core power system, intelligent application, etc., forming an open and shared ecological map of the whole industry chain. This is the "forest-like ecosystem" established by Great Wall Motors, in which various brands, industrial categories and technologies are scattered, Wei Jianjun concluded, "We have competition, but also an ecological model of symbiosis, sharing and coexistence".
Wei Jianjun has to communicate with netizens regularly
On the evening of March 28, the Xiaomi SU7, which had been brewing for three years, was launched as scheduled. Wei Jianjun, who was sitting in the audience, witnessed the whole process. As the owner of Xiaomi's first car and the largest brand, Wei Jianjun admitted, "From the performance of the whole vehicle to the detailed design, whether it is intelligence or Xiaomi ecology, I am very impressed", and the fiery atmosphere of the whole press conference, as well as the hot search entries that are high before and after the release, may also give him more touch, "The Xiaomi press conference really deserves its name".
In this era when fuel vehicles are turning to new energy vehicles and hardware manufacturing is giving way to software intelligence, Wei Jianjun once again realizes the necessity of change: in the past year, Huawei and Xiaomi have entered the electric vehicle market, making car manufacturing no longer a traditional and silent thing. Amplifying the voice of car companies in the Internet has become part of the competition.
Seeing this, some changes have already taken place. In June last year, a number of senior executives of the Great Wall collectively settled in Weibo to learn ideals and build an Internet communication ecology; on March 26, Wei Jianjun released his first Weibo; on April 12, he previewed the city's NOA measurement plan in the form of a short video.
Two days later, Wei Jianjun appeared in the camera wearing a black motorcycle suit. He said that Great Wall Motors had some traditions in the past, burying its head in promoting urban NOA for 10 years, and rarely disclosed relevant information. But today is a good start. Wei Jianjun decided to further embrace the Internet, he said, "to play an exemplary role and communicate with netizens regularly".
Wei Jianjun, who enjoys the reputation of "Baoding Car God", loves cars, and he once said in public, "If I become a chairman, then I will become a professional racing driver." In 1990, Wei Jianjun took over the debt-ridden Great Wall Motors, and with the first batch of Changan cars, he brought the depot on the verge of life and death back on track.
In the following 30 years of market ups and downs, Wei Jianjun led Great Wall Motors to find opportunities to survive and expand. With the scale growth achieved in the pickup truck market and SUV market, Great Wall Motors has exceeded one million sales for several consecutive years, ranking among the top independent car companies. In 2021, Great Wall Motor's revenue exceeded 130 billion yuan and more than 1.28 million new cars were sold, a record high.
However, surrounded by halo and applause, Wei Jianjun did not forget to reflect, and he proposed at the 30th anniversary of the establishment of Great Wall Motors in 2020 that if we can't see subversive changes, it must be us who are subverted. If we don't dare to break the rules, then the rules will quickly become a prison of creation.
In this anxious competition in the industry, volume technology, volume research and development and volume price are no longer enough, competing for traffic, shouting out their own advantages in the market, has become the unshirkable responsibility of the founders of car companies. Just like in the past, Wei Jianjun often appeared in the factory, and later, he appeared more in the Great Wall Motors Technical Center, and now, Wei Jianjun is determined to devote himself to the Internet and bring Great Wall Motors more into the camera.
He said, "In the competition in the automobile market for more than 30 years, I feel like swimming in the ocean, there is no way back, and only by struggling to swim forward to the other side can I see hope." Therefore, I have never dared to be satisfied, nor have I ever been satisfied. ”