On July 20th, Industrial Securities and 36Kr held the "Smart Driving Investment Closed Door Meeting" in Beijing. With the theme of "Overheating and Immaturity is an Investment Risk, or the Starting Point for Development", the conference invited more than 10 listed companies and star startups, as well as senior industry experts and policy experts, to review the development process of intelligent driving and discuss the opportunities and future of intelligent driving from the perspective of development.
The main theme of the meeting was "Is there a bubble in the valuation of autonomous driving companies?" "How are the autonomous driving grading standards defined and implemented?" "What is the prospect of autonomous driving?" And other three major issues to discuss, hoping to achieve the first and second market linkage in the field of intelligent driving, as well as to answer questions for investors and industry practitioners.
Hou Jun, COO of Zhixing, delivered a speech with the theme of "Exploration and Landing of Autonomous Driving Solutions for "Feasible, Reliable and Commercial"". The following is a transcript of the contents, compiled by 36Kr:
CoO Hou Jun
<h3>Self-driving track: China and the United States</h3>
From 2018 to 2030, from closed scenes, low-speed logistics to gradual opening, really to the future Ofotaxi, this trend is becoming more and more obvious, including today's closed scenes have begun to land, and the time for commercial scenes to land is becoming clearer and clearer.
This track is the fastest running in two countries, that is, the United States and the United States, in the United States we mainly look at three players, the first player is Google's Waymo, which focuses on pure L4 unmanned driving, and the technical construction is one step in place. The second player is Tesla, a gradual evolution from current assisted driving to the future. The third player is Cruise, which is gradually evolving from mass production passenger cars from assisted driving to autonomous driving, while taking into account the logistics industry.
These three routes have their own advantages, but from the current state of development, the latter two seem to be in a better situation after 2019, and the Waymo model is sought after by everyone before 2019. The other two models, Tesla and Cruise, are a combination of car-making and technology companies.
The benchmarking mode of The Ultimate Wisdom Walk is Cruise. Cruise is in high-speed manned assistance, high-speed manned complete to low-speed cargo complete, what is the logic of these three quadrants? We have such a view, in the development process of unmanned driving, progressive we think it is easier to land, first of all, low speed to high speed, first loaded after manned, first commercial and then civilian, there will be a crossover in the middle, from our point of view, this progressive is easier to achieve closed loop.
In addition to the three laws just said, we also think that there will be three stages in automatic driving, and the name of these three stages is the name I reported today, called feasible, reliable, commercial, feasible is to say that your perception ability, your intelligent algorithm, your vehicle control can be used, can run, called feasible.
The second is reliable, automatic driving has a big premise is the vehicle regulation level, we call safety, vehicle regulation level, all day long, all terrain, the real automatic driving vehicle is not only in the experimental field, is to run in the north of the south sea. The purpose of automatic driving is not only for scientific research, but ultimately for commercial use, and if it is commercial, it is in the four quadrants of automatic driving where your final scene demand is located, whether it is terminal logistics, urban distribution, or manning, which is a huge difference. Cost is also a key factor, and when autonomous driving comes out, the ultimate customer and consumer is unwilling to pay for the additional cost of autonomous driving.
The last very important factor is the policy and regulation, after the automatic driving comes out, the current traffic rules, and now the supporting related insurance will change dramatically, who is the main body of responsibility, who is the beneficiary will have a very big difference.
Therefore, these three stages, feasible, reliable, and commercial are necessary experiences, and the entire mochi we have followed in the commercialization process, including ourselves and our partners.
<h3>Windmill strategy: autonomous driving from passenger cars, low-speed logistics vehicles and intelligent hardware layouts</h3>
The three elements of data intelligence, data collection, processing to the application is our core, these three blades quickly rotate to form our windmill strategy.
There are two stages of data intelligence, one is from perceptual intelligence to cognitive intelligence, cognition we will see from the map to cause and effect to reinforcement learning, perception is vision, hearing, touch, of which there are two technologies that we believe are the core key.
The first technology we are based on natural language processing, it is called Transformer, it is the most advanced technology of visual processing, it will be much faster and more capable of visual processing than other technologies, but it has great requirements for data volume, which is suitable for companies with a large number of data acquisition capabilities, such as Tesla, which uses similar technology to carry out a lot of visual processing.
New technology Transformer
The second technology is CSS, that is, cognitive intelligence collaborative scene security, which is developed and invented by us, why do we say so? In order to ensure true safety, we will actually determine the core axiom, and the premise of this axiom is to ensure that people and autonomous driving ensure safety first in a certain scenario.
New technology CSS
The three blades of our windmill strategy, namely the small magic box, the small magic disk and the small magic camel, this is the whole of the passenger car we from L2, L3, L4, from assisted driving to autonomous driving of the four stages, L2 began commercialization in September this year, L3 commercialization at the end of this year, at least more than fifty thousand cars this year, we predict more than one million vehicles within three years, within five years 2025 predict million vehicles, may be to four or five million vehicles loaded, This creates great convenience for our data intelligent data collection.
The core product of intelligent hardware is a small magic disk, which is a cost-effective low-speed chassis, which is mainly the current larger partners, including Meituan, SMIC Data, etc. to provide the underlying capabilities to help them. In the whole low-speed intelligent hardware, we emphasize that it is such a concept, we are the ecological helper of the low-speed platform, we have no scene in this platform, it is our partners who have a scene, we call Meituan, JD.com, SF, and Postal They have a scene, we are divided into five layers from top to bottom in this scene, from factory services, supply chain and hardware services, system tool services, vehicle services, transportation services, and make full use of our supporting services with the Great Wall to help our partners.
The entire logistics industry in recent years will have a huge change, that is, the business and order volume surge, the number of labor force is declining, and the average per capita wage of the industry is rising, the employees of terminal logistics may be 10 to 20 million people, and it is difficult to recruit new people to help everyone deliver express delivery today.
In fact, we predict that from 2019, the cost of low-speed logistics vehicles will be about one million, this year may be about 250,000, the next three years of vehicles will drop to about 100,000 in 3-5 years, this year is the first year of commercial use of low-speed logistics vehicles, which also includes and many partners to start large-scale commercialization, but also based on the rapid decline in cost, but also based on feasible, reliable, commercialable can be deployed stage. Partners can see that including Great Wall Motor, Meituan, Wumart, etc., the current investors are Shougang Fund, Meituan, Hillhouse Venture Capital made a Pre-A round of investment.
The second core capability is product leadership, and the core ability of product leadership lies in the commercialization ability of scale landing, including small magic box, small magic disk and small mo camel; the third core ability lies in data-driven, our data-driven lies in the scale and processing capacity, including the transformer just mentioned; the last ability is industrial collaboration, whether we are on the 5S ecology of low-speed cars. Or for the ability of the entire supply chain, make full use of the capabilities of large OEMs to help our entire ecosystem.