Text karakush
A while ago, He Xiaopeng went to San Francisco, USA, and tried Waymo's unmanned driving.
Waymo took a major step towards commercialization by charging for driverless taxi services in San Francisco last August. Although it had already started charging services in Phoenix, San Francisco is more than 10 times more densely populated than Phoenix, which challenges and meanings.
In San Francisco, Waymo has installed 250 white Jaguars, which are open to recruiting users in limited areas and gradually expanding its territory. As of last year, it covered nearly 47 square miles of the city, with more than 100,000 waitlists. The website reads: Redefining the way you move around San Francisco.
Like every user, He Xiaopeng will hear this passage after getting in the car:
The experience may feel futuristic ahead of its time. Please do not touch the steering wheel or pedals during the trip. If you have any questions, you can find information in the Waymo app, such as how we keep it safe or clean. ”
For He Xiaopeng, he may feel more certainties about the future. He said that Waymo can take over hundreds of kilometers in the city, and the whole experience is only a little less than that of people, which is very close. However, can it be used on a larger scale, and can it be done at a cheaper cost? Xiaopeng wants to be this leader.
This is yesterday, at the Xiaopeng New Year Intelligent Driving Collection Conference, he set a new goal: to lead the new decade of autonomous driving and change human travel.
Despite the fact that the modern Chinese intelligent driving circle is very popular and the rhetoric is inflated, few people have risen to the level of "human beings". However, Xiaopeng - I believe that there are many friends who are willing to believe it.
Looking back at the way it has come, you will find that this oriental original Tesla no longer needs to benchmark against any opponents to show its identity, and it has a place in China's smart car track.
Since 2019, G3 has introduced visual parking, seen the surrounding parking spaces, selected parking, and opened up the label of "the strongest parking on the surface", driving a handful of new forces to cut in from the "scenario-based" autonomous driving function;
By 2020, the P7 will launch a full-scene voice intelligent cabin that can be seen and spoken, and carry high-speed NGP under the premise of large-scale mass production and controllable cost, setting the threshold for "smart" electric vehicles;
By 2021, the P5 equipped with lidar will be mass-produced, and the new perception of the education industry and the market will be lidar, and then in 2022, 30TOPS will be used to achieve high-end intelligent driving in urban areas;
Then the first to introduce AI visual large model, turn to light map + visual large model, and by 2023, the new XNGP of the Fuyao platform will allow Beijing to enter the era of intelligent driving for the first time;
In January of this year, Xpeng carried out a full push of XNGP without maps in 243 cities, covering 90% of the cities where MAX version owners are located, with an available mileage of 5690,000 kilometers, note that it does not contain LCC and ACC water injection, which is "the absolute first in China".
It is not said that Xiaopeng only carries the flag on its own, but it is undeniable that every step of China's intelligent driving forward cannot bypass Xiaopeng.
Why can Xpeng?
He Xiaopeng repeatedly asked himself and answered this question, and finally boiled down to two points:
The first is persistence, even in the darkest moments, Xpeng's investment in intelligent driving has never been reduced by a point.
For example, in the second half of 2019, it was the most difficult time for Xpeng, either in the ICU or sitting at the door of the ICU. All the new car-making forces in China have no money. Xiaopeng has subtracted many places, and even at that time, it has not cut intelligent driving.
Second, we still insist and never be crushed by accidents. For example, during the special period of 2022, it was particularly difficult to do tests in Guangzhou, so the team took their luggage and the team to wander in other places for testing.
This intelligent iron army of thousands of people has been fighting a tough battle. Xpeng's opening logic is to ensure that every open road section in each city has been tested, checked, and clouded**. Originally, last year's goal was to keep 50 and fight for 100, but my peers were particularly aggressive, so in the past year, I had to hold back 150 hard rolls, and accidentally ended up with 243. Someone is really good at tough battles, only familiar with it.
If there is a third, I am afraid that it is still insisting, insisting on believing that the future of travel will be changed without any basis in reality.
The beginning was painful. In 2014, Xpeng tried autonomous driving for the first time, and it couldn't even make an unprotected left turn. He Xiaopeng himself is the first key to injecting faith, until he officially joined in 2017, Xiaopeng had less than 400 people, and the first thing he did after joining was to go to the United States to recruit technical talents.
At that time, the first stop was Stanford University, and Xpeng only had a G3 10 production version, mediocre. However, He Xiaopeng said: Xiaopeng Motors will not build electric SUVs in the future, but build young Internet electric vehicles that meet the mentality of China's first- and second-tier cities, and will put L3 autonomous driving into it, and will design different means of transportation to change the car from private assets to public assets, and then have L4 autonomous driving capabilities, which can be used in limited scenarios.
Until yesterday, He Xiaopeng was still saying, "I don't know what year it is, but I believe it will happen." ”
The process was extremely painful. High investment and low return have always been a problem for Xpeng. Do Chinese consumers really need smart driving? Are you willing to pay a premium for Smart Driving? Is it possible for urban intelligent driving to be delivered on full scales?
Last year, the problem could not be said to be solved, but at least the balance was seen commercially.
Yesterday was also the first anniversary of Ms. Wang Fengying's joining Xpeng. Previously, she led the continuous organizational structure adjustment, and while improving efficiency and performance, she did not shake the label that intelligent driving is still Xiaopeng.
According to "Late LatePost", in December 2022, in order to persuade Wang Fengying to join, He Xiaopeng invited her to take a ride in Xpeng Motors three times, until Wang Fengying believed that "smart cars can be driven in the city in a relaxed way" and agreed that autonomous driving is a good marketing point.
After joining, she advocated "focusing on and strengthening users' perception of Xpeng intelligent driving tags" and "to be the first in the industry in the next few years". For example, last year's smart product G6, which had the most market power, was instructed to focus on intelligence in publicity, and there was no second article.
In terms of sales, it also cooperates with the strategy, and uses the Nappa leather seats with strong perception to guide the MAX version to guide the options. At the same time, training sales conveys intelligent selling points and guides users to test drive and experience intelligent driving. Facts have proved that there is no intelligent driving that cannot be sold, only a strategy that cannot be sold.
According to statistics, in 2023, the penetration rate of high-speed intelligent driving of Xpeng users will be 947%, and the penetration rate of urban intelligent driving is 832%, the weekly average penetration rate of XNGP is 71%, and the weekly average mileage penetration rate is 40% (including high-speed and urban scenarios).
Only by selling it can it be useful.
This year, Xpeng will continue to emphasize the cost of the first chain, and its plan is to reduce the cost of autonomous driving BOM by half, and the cost of vehicle hardware (including power system) by about 25%, after which Xpeng will realize the standard configuration of hardware, bringing high-end intelligent driving to less than 200,000. At the press conference at the beginning of the year, He Xiaopeng revealed that one of the new cars planned this year is Mona, which is in cooperation with Didi, at the level of 150,000 yuan.
With a grasp of ability, experience, and cost, cost, and cost, Xpeng is expected to return to the sales track. Whether you need it or not, maybe it's never a problem, whether you can hit the ** or not, I don't mind, yes.
Next
The new decade begins in 2024, and He Xiaopeng believes that the logo will be smart cars in city-level intelligent driving, which will definitely completely subvert the past travel logic in the new decade.
For example, security. In 2023, the average accident rate per million kilometers of Xpeng users will be about one-tenth of that of manual driving when intelligent driving is turned on.
He Xiaopeng pointed out that the number of traffic accident deaths in the world is about 2 million every year, and if autonomous driving develops at this rate, this number can be reduced to less than 200,000 (1%) in ten years. Drunk driving is the most common criminal offense in China today, accounting for 50 percent, and in a few years it may be gone.
In He Xiaopeng's view, although the goal is huge, the future is not far away. Next, Xiaopeng has to do a few things:
First, we will aim to be able to use it in all areas, reduce costs on a large scale, and improve capabilities. In 2024, XNGP will achieve full coverage of the road network in major cities across the country, and the number of Kaecheng is expected to double or double this year and next year, including more western and northern cities.
Expand the current first-class highways, second-class highways, and some first-class highways to a wider area of first-class, fourth-class, and fifth-class highways, covering alleys, internal roads, residential areas, and parking lots, and achieve point-to-point in the whole range of China.
It's not a big deal. In the past year, Xpeng's ability to open cities and generalize has improved dramatically. In the past, the team had to spend a few months to open a city, make high-precision maps and generalization logic; Now the speed is 384 times higher, and it only takes 1-2 days to open a city.
By 2025, Xpeng will develop XNGP for the world and export it to the world, so that overseas autonomous driving can be used as well as China's. Let Xpeng represent the pinnacle of China's AI company's engineering capabilities and global efficiency, and go global.
Second, the takeover rate has further declined, and in core areas such as Beijing, Shanghai, Guangzhou, Shenzhen and other medium-sized and large cities, the number of passive takeovers per 1,000 kilometers is less than 1, improving traffic efficiency, experiencing the benchmark Waymo, and making the efficiency of intelligent driving infinitely close to that of human driving.
An average person drives 70 kilometers a day to commute and 600-700 kilometers a week, which means that in addition to getting on and off the car, you don't have to take over once, and 99% of the time you don't have to drive, which will bring about a huge change.
Now in the high-speed scenario, a new generation of high-speed NGP passively takes over 05 times, that is, 2000 km to take over; A 13% increase in average speed is almost always slightly shorter than what the AutoNavi map shows, meaning that the traffic efficiency is improved.
Although the XNGP city test drive has not been launched for a long time, from the internal test to the full release, a total of 271 internal iterations have been completed in 70 days, with an average of 387 version, the outside world can only see the update once a season. In fact, He Xiaopeng's own car is often the pioneer of the white mouse car, and the functions that are still there during the day change at night, and everything is rolling and iterating rapidly.
Third, improve the processing ability of XNGP in various abnormal traffic scenarios and combination environments, such as snow in the north and snow on the road, driving should be different from the situation without snow. XNGP requires greater intelligence to make judgments.
Fourth, Xpeng will achieve an end-to-end large-scale model on the car. AI will bring huge opportunities. Until the beginning of last year, the easiest way to optimize the experience was to do a good job of generalization; But today, large models can turn generalization** into simple end-to-end models, making perception, behavior, and control smarter.
For example, the text prompt on the road, "Please change lanes for ETC maintenance ahead", or some intersections in Chengdu will have "drive to the lane to be transferred after 8 seconds", which can be understood by cars in the near future.
To this end, Xpeng's AI R&D data operation team with autonomous driving as the core has reached 3,000+ people, accounting for nearly 40% of Xpeng's R&D personnel. The annual investment scale is 3.5 billion yuan, which can rank 100 in the R&D investment of Chinese private enterprises, while Xpeng's investment in AI alone has reached this magnitude.
With vigorous investment in research, Xpeng can lead China and the world to become stronger AI capabilities. He Xiaopeng is very confident and believes that this is a high probability event, "Today's part of the large model that has been commercialized is chips and companies such as OpenAI and Microsoft. ”
The Chinese market is half the battle
He Xiaopeng said that five years ago, no one thought that new energy would have such a large proportion today, and similarly, no one would have thought about what fully autonomous driving would look like five years later.
The world's brightest minds are working together to create a new era.
Although not all of them are looking to the same brand new. Last December, the city of San Francisco filed a lawsuit against the Public Utilities Commission, which approved Waymo to operate around the clock in the area, and the basic requirement of the lawsuit was to re-examine whether the decision was legal.
In contrast, China has a loose policy environment and very aggressive consumers. Xpeng's dream of trying to lead the world is not a pipe dream. On the way to Niu B, the scenery is so bad that people just want to swear, but what you really need to see is in the distance.