Since 2023, Waymo, the leading autonomous driving company, has laid off more than 200 employees. Embark Trucks, an autonomous driving startup that once had a market value of more than 5 billion US dollars, has now put the disbandment of its business among the options, relying on Volkswagen Argo AI with Ford also announced its closure last year.

In China, some companies that insist on L4 autonomous driving are not as high-profile as before. On the contrary, there are more and more news about designated cooperation with mass-produced car companies. And even those new brands that hold high the banner of intelligent driving seem to be falling behind in the market involution.

If autonomous driving was making rapid progress a year or two ago, now the industry seems to be reaching a contradictory crossroads.

On the one hand, there is the "eve of commercialization" that has been shouted out for a long time, and there are videos of NOAs in various cities flooding the screen. It seems that cars are not worthy of selling cars without some self-driving capabilities.

On the other hand, no company dares to show the optional installation rate of high-end assisted driving in the end market. However, BYD, which has always been considered weak in intelligent assisted driving capabilities, is killing everyone.

Indiscriminate flowers are becoming more and more attractive. What kind of trends lie behind the complex industry changes?

At the 2023 China Electric Vehicles 100-person Conference, Horizon founder and CEO Yu Kai made such a statement. After the industry experienced the peak of rapid technological development, it began to need to return to its business essence. And what is the essence of business? In fact, it means that companies create value, deliver value and obtain value, that is, someone is willing to pay for the product. Return to the business essence: Advanced assisted driving creates value

Yu Kai believes that autonomous driving should return to the business essence, and the value of advanced assisted driving is being recognized by more and more users.

A set of data shows that at this stage, consumers’ core demand for smart driving technology is to relieve driving fatigue, and the most important factor in users’ recognition of high-speed pilot-assisted driving is “making driving easier.”

Not long ago, Horizon discussed a topic at the Journey 5 launch NOA test drive meeting: For users, what is the perception brought about by technological iterations such as computing power and chips?

The ideal car owner "Big D has an attitude" said: "The most intuitive perception is that the display interface is clearer, which can tell me how much things around the car can sense. On the other hand, it is when dealing with unexpected problems. , Active lane changes, avoidance and other strategies are better."

The former gives users more confidence, while the latter makes users more willing to use assisted driving functions.

Yu Kai also talked about his own experience: “Last year, I returned to Beijing from Shandong and queued for 5 hours on the road. At that time, I opened an Ideal ONE, and the whole process did not require me to continue. It’s very easy to step on the accelerator. This is different from the previous experience without assisted driving.”

Therefore, from a business perspective, the core significance of autonomous driving at the current stage is to make driving easier for users. Only by making it easier to meet users' value needs can we obtain business returns.

This is exactly what Horizon is working on.

"Automobile Company" not long ago experienced the high-speed NOA function of the ideal L8 equipped with the Horizon Journey 5 chip. It can be said responsibly that compared with the early pilot assist system, both The overall driving efficiency, lane changing logic, and the more subjective level of confidence and security given to the driver have all been greatly improved.

According to data released by Xpeng Motors, the proportion of high-speed NGP mileage in the total mileage of users during the Spring Festival this year reached 3.25%, which was significantly higher than the 1.38% during the National Day four months ago. promote. Many new power brand users have also expressed their recognition of advanced assisted driving functions on social media.

Compared to Japan and Europe, assisted driving functions are more driven by safety regulations. Functions such as AEB, ACC, lane keeping, emergency automatic braking, etc. all play an active safety role.

The Chinese market’s expectations for assisted driving are based on safety and driven by better user experience and user value. Smart electric vehicles are increasingly replacing traditional BBA luxury cars in the market, and high-speed NOA is rapidly entering mass production, which is proof of this.

In Yu Kai’s view, this is an opportunity for the Chinese automobile industry to change lanes and overtake. No need to worry, now we are making high-speed NOA

How should the autonomous driving industry respond to such opportunities and demands?

Yu Kai believes that last year was the first year of mass production of L2+ high-speed NOA, and this year some leading companies will start to launch urban NOA. However, from a technical perspective, it will take about three years to have a comparison. Good progress.

In fact, it seems that the current advanced assisted driving arms race is in full swing, and the improvement of user experience is not directly proportional to the accumulation of hardware. Just like for most users, assisted driving that relies on camera perception does not seem to be much different from a system supported by lidar.

Therefore, Yu Kai called on the industry not to be too anxious about autonomous driving, because the industry is not developing that fast. By 2025, high-speed NOA must achieve a silky-smooth experience at a reasonable price/performance ratio. At the same time, considerable investment must be made to make urban NOA usable.

In fact, the development of technology and the improvement of user experience have always been complementary and mutually reinforcing, and the same is true for the development of autonomous driving.

When Xiaopeng first promoted NGP, a R&D staff member once said that a very important consideration in the process of developing the system was "allowing more users to use it."

This principle still exists today, just like the beginning of the "flywheel effect". ?

The development of urban NOA still has a long way to go, but it has important value in data collection, scene collection, and feedback research and development. The continuous expansion of user usage scenarios can also allow more users to Have the opportunity and conditions to use advanced assisted driving functions.

The purpose of making high-speed NOA is to bring the user experience to the extreme, so that users can establish trust and dependence on advanced assisted driving, and at the same time better understand the value and usage boundaries of advanced assisted driving. Horizon: Use algorithm optimization to approach the upper limit of user experience

In Yu Kai’s view, what is currently achieved from dozens of Tops to hundreds of Tops or even 1,000 Tops are all high-speed NOA, and the difference is not too big, so The industry still has a lot of work to do.

This is also the direction Horizon is working towards.

Horizon believes that it is necessary to continuously optimize software algorithms based on given computing power, use more data, and constantly approach the upper limit of user experience. "I believe that in a few years, we will reach this level, about a few hundred Tops, and the NOA in urban areas can also be good." Yu Kai said.

At the world's top artificial intelligence computer vision CVPR conference not long ago, Horizon as the first author proposed an end-to-end autonomous driving algorithm framework based on Transformer, and this paper has more than 9,000 stood out among the submitted papers and was selected into the TOP12 best papers.

This is Horizon’s forward-looking research and development in algorithms. For the first time, a neural network is used end-to-end in monitoring, tracking and prediction, and arrow trajectory prediction. It is all completed with a simple architecture from front to back. The traditional approach is to package them, split them into many modules, and then make them independent.

This also means that Horizon may be able to use end-to-end large-scale data to train the entire autonomous driving system.

Based on the understanding of algorithms, Horizon has injected forward-looking research on algorithms into the design and development of chip architecture. This is also the core computing architecture developed by Horizon - BPU (Brain Processing Unit). ).

The third-generation BPU architecture - Bayesian architecture is used in the Journey 5 chip, which can efficiently support Transformer calculations. Compared with competing chips, Horizon can complete more tasks with lower power consumption. Efficient computing.

Higher computing power and efficiency allow Horizon to use larger data and larger models to unsupervisedly learn human driving attempts, “just like you learn from a large number of unsupervised, It’s the same as learning from a vast amount of natural text without annotations,” Yu Kai said.

Specifically, by given the current traffic environment, navigation map, and a driver's driving behavior history, machine learning is used to predict his next driving action.

Experimental data also proves that when the parameter scale of the model continues to increase, the test loss will decrease accordingly, that is, the more parameters, the smarter the system.

Relying on a strong ecosystem, Horizon can help OEMs achieve their dreams. Horizon has also been committed to building a software and hardware ecosystem for intelligent driving and robotic computing. In the industry, Qingzhou Zhihang’s urban NOA solution based on a single Journey 5 chip has already won a mass production project from a well-known car manufacturer, and autonomous driving companies such as Pony.ai and WeRide are also developing on Horizon chips Software solutions.

The time window of the chip industry is very narrow, and the window period does not choose a certain chip technology pioneer company, but chooses an ecosystem that develops the fastest, just as Android and Apple have survived, and windows phone failed. Therefore, the purpose of Horizon's ecology is to involve more partners and make the pie bigger together. Through the collaborative technology platform of software and hardware, and the flexible and open business cooperation model, we can work with upstream and downstream partners. , to realize the run-through, upgrading and prosperity of the entire industrial chain.

The commercial progress made by Horizon is also visible to the naked eye. Currently, more than 50 models have been mass-produced, more than 120 models are pre-installed, and nearly 3 million car-grade autonomous driving chips have been shipped. , the Journey 5 chip has also become the only mass-produced chip in the industry with more than 100Tops computing power, and has been installed and delivered on the ideal L8. It has also achieved strategic cooperation with the Volkswagen Group. Viewpoints from major automobile companies

In Yu Kai’s view, the development of autonomous driving in the next 10 years will be like “people riding horses”. Autonomous driving can be achieved, but humans and vehicles also cooperate, and the ultimate responsibility is Attribution still belongs to the driver.

In fact, although some car companies have already implemented L3 level autonomous driving functions, we can also see that its use conditions are very strict, which is naturally related to "let more people use it" contradiction.

But no matter what the trade-offs and choices are, people-oriented should be the core.

Human beings are the ultimate beneficiaries of all technological developments, and they are also the payers of these technological developments. The path forward for autonomous driving will not be achieved overnight. In this process, how to balance the relationship between technology and market demand, and how to make the two promote each other, may be a topic that the entire industry should constantly think about.

At present, focusing on user value may be a more pragmatic and feasible choice for the autonomous driving industry.

This article comes from the author of Yichehao Automobile Company. The copyright belongs to the author. Please contact the author for any form of reprinting.

The content only represents the author's opinion and has nothing to do with Bitauto