A press conference will closely link "family" and "technology", ideally calling it the car brand that best understands the consumption concept of middle-class families.
On June 17, at the first Home Technology Day press conference of Li Auto, Li Auto's 400,000th mass production rolled off the production line. At the same time, Ideal also showed the latest progress in high-voltage pure electricity, intelligent space and intelligent driving.
From the first ideal ONE rolled off the production line in 2019 to the 400,000th vehicle rolled off the production line today, this new force has taken less than four years, and the sales growth rate is still increasing rapidly.
While sales are growing, ideals are also sorting out the next stage of thinking: what to attract more users?
At this Home Tech Day, the keywords we heard the most were "AI", "big model" and "intelligence", which also indicates that the ideal is changing to an artificial intelligence company.
But what remains unchanged is that these new functions are inseparable from a carrier - home, bringing more convenient travel methods for home users.
01. For home users, what is the ideal to do right?
Starting with the ideal ONE, ideal created the "family car" market.
To some extent, this can also be understood as scene-based car building.
Focusing on the market segment, although there are already many excellent products in the domestic SUV market, and the space has been adjusted for the needs of domestic users, the product homogenization is too serious - there is no model on the market that is really aimed at the family car scene, and many products that advertise "family" still stay in more space.
In most cases, these models do not take into account the travel experience of other passengers on board.
"If I can only own one car, why don't I expect it to do a little more?"
This is a very rustic consumer voice.
Reflected in new energy vehicles, car companies have insight into user needs and function presentation, ideally being the best among them.
In terms of marketing and product configuration, Ideal models all target the needs of family users, create a "daddy car" atmosphere, and define their own products as "mobile homes".
As you can see, the full-size six-seat SUV ideal L9 is "180 square meters large three-bedroom", the medium and large six-seat SUV ideal L8 is "120 square three-bedroom", and the medium and large five-seat SUV ideal L7 is "120 square large two bedrooms".
In addition, the ideal has also been ridiculed by the outside world as "refrigerators, color TVs, and large sofas", which seems to be a kind of debasement, but in fact, because these configurations are high-quality enough, they directly hit the hearts of home users.
Li Xiang once answered the product standards that conform to humanistic brands in an interview:
"I think if it is a company with some rational verification for sentiment first, the essence of everything is to exceed the needs of users, if it can always exceed the needs of users, it does not need to build multiple brands to do things, Nike does not need, Nintendo does not need it, Apple does not need it."
To put it simply, the ideal idea is to think about solving users' concerns about buying a car through the model itself and technology from the perspective of product, experience and technology-ride comfort, more intelligent multi-dimensional interaction and entertainment functions, to achieve dimensionality reduction against competing products.
For example, if Xpeng's full-scene voice technology led the industry by one position in 2020, then after more than a year, Ideal not only made up for the shortcomings of voice, but even launched a new intelligent cockpit space blueprint: in addition to voice, gesture interaction is integrated to shape the three-dimensional interactive experience of space.
This also shows that Ideal is striving to pull the focus of the competition of new cars from the intelligent hardware arms race back to the human-vehicle interaction in the cockpit space.
This is the most perceptible part of the consumer market, and it is also the place where the ideal product is most focused.
From the ideal ONE to the L8, the ideal rolled up in the family 6-seater car market for three years, and from the L7, the ideal officially rolled into the 5-seater SUV market.
This is the largest segment in the domestic market, and even with a focus on the segment of medium and large SUVs, the market capacity of 693,500 units (sold in 2022) is considerable.
The ideal product play is very clever, through the different price coverage of L7, L8, L9, to build a product line and a product network.
Taking the ideal L7 and L8 as an example, the biggest difference between the two is only the number of seats, L7 users tend to be three/two families, L8 users are mainly two-child, or three-generation families.
Simply understood, 89% of China's automobile market of more than 200,000 yuan is home users, and the ideal product logic is to cover different families from top to bottom.
From the perspective of sales performance, the label of ideal "family" has been successfully implanted in the minds of target groups.
After all, for mass consumers, a car that sells well enough has a stronger appeal.
That's why the new SUV sales are so good that they are now updating their sales rankings on a weekly basis.
From research needs to product definition, Ideal has established a set of methodologies, but the "smart home car" label it expects has not yet taken shape in the minds of users, which is also the shortcoming that Ideal needs to make up in the next stage.
02, Mind GPT, commuter NOA, product transformation into the "new era of AI"
In order to improve the ability of intelligence, Li Xiang once said that to strive to catch up with the head of the industry in 2023, it is necessary to take two steps.
- Step 1: Create an "ideal classmate" who learns and thinks like a person, expresses and interacts like a person;
- Step 2: Open NOA for uncharted cities, and NOA for commuter without picking sections.
Let's look at the first step first.
As a self-developed speech engine, ideal students will be fully iteratively upgraded, and the core goal is to actively perceive the environment and perceive people like people; Learn, recognize, reason and remember like a person, and think, express and interact like a human.
Specifically, start from three directions:
(1) First, use the hardware of digital silicon microphone, 3D ToF camera, and IR sensor covering the multi-sound zone of the whole car to improve the perception ability of ideal students;
(2) Next, multi-modal perception data interaction is carried out to improve the "ears, eyes and brain" ability of ideal students, including visual perception technology, human voice separation algorithm and speech perception algorithm;
(3) Finally, the introduction of the Mind GPT large model, you can become a companion "teacher" who travels all over the world, a professional car "housekeeper", and an "expert" who teaches drawing and programming.
It can be seen that the ideal is abandoning the cockpit mode of directly moving refrigerators and multi-screen color TVs into the car, but polishing the cockpit experience from the deep interaction between man and machine.
This is an ideal new attempt when the AI model is on the rise.
But the best embodiment of the ideal pursuit of AI is probably the active expansion of urban intelligent driving.
In the context that "high-speed NOA" has entered the first year of mass production in 2022, some leading companies in the industry will launch urban NOA this year, and Xpeng and Huawei have successively announced the "100 Cities Plan".
Not only will we reach 100 cities by the end of the year, but we will also deliver the city NOA function that does not rely on HD maps to beta users in Beijing and Shanghai this month.
At the Shanghai Auto Show in April this year, Lang Xianpeng, vice president of ideal car intelligent driving, has introduced the large model algorithm behind the ideal city NOA:
Static BEV network algorithms, dynamic BEV network algorithms, and Occupancy network algorithms, and enhance Occupancy network algorithms with NeRF technology for greater accuracy and detail.
However, due to the large span of urban road entrances and the large number of passing vehicles, the sensor's field of view is easily occluded, resulting in the loss of some local information as a result of vehicle-end perception.
In response to this situation, the ideal is to develop a self-developed neural prior network, referred to as NPN network (NeuralPriorNet, neural prior network).
Lang Xianpeng said that through this network, NPN feature extraction and storage can be carried out in advance at complex intersections.
When the vehicle drives to the intersection again, the previously extracted NPN features can be taken out and fused with the BEV feature layer of the vehicle-end perception large model to obtain better perception results. As NPN features continue to be automatically generated in the cloud, the available range of city NOAs will continue to expand.
At the same time, Ideal trained TrafficIntentionNet (TIN, signal intent model) by learning a lot about the behavior of human drivers in response to changes in traffic lights at real intersections, and got rid of the dependence on map rules to deal with various forms of signal light styles across the country.
Simply understood, the training data of the TIN network is based on the video of the driver passing through this intersection + brake throttle opening and closing degree and steering wheel angle data, and the ideal internal team has extracted a large number of driver intersection data from 600 million kilometers of training samples, and achieved fully automated labeling.
Overall, the characteristics of the ideal city NOA technology architecture can be summarized into three points:
- Use NPN features and TIN networks to enhance the BEV large model to get rid of HD maps and identify everything;
- Use imitation learning to make regulatory algorithms make more anthropomorphic decisions;
- The fully automated, fully closed-loop training platform supports the continuous evolution of large models.
While continuously promoting the implementation of urban NOA, Ideal also released a commuter NOA, users can set a commuter route by themselves, learn NPN characteristics by themselves, and use the NOA function on the commuter route after learning.
Simple sections can be trained in one week, and complex sections take two to three weeks, and the NOA function of the corresponding section can be activated after training.
It is not difficult to imagine that the launch of the commuter NOA will greatly motivate the ideal car owner.
As more and more car owners join the team of training commuter NOA, the coverage of the ideal self-built NPN network will also expand rapidly, thereby accelerating the popularity of ideal city NOA.
03. There are 7/8-seat versions of pure electric MPVs, which are rushing to land on a new highland in the pure electric market
Looking at the development of ideals at the product level, we can see three different stages:
The first stage: Ideal ONE helped Ideal complete the development stage from 0 to 1, and its explosive case also made a sample for the model after Ideal.
The second stage: In the past year, three models of the ideal L series were released one after another, which not only initially completed the systematization of products and supply chains, but also prioritized product power to focus on the most perceptible parts of the consumer market, and strengthened the core selling points of "family".
The third stage: entering the AI era, transforming into an artificial intelligence company, the landing of ideal AD Max 3.0, urban NOA and Mind GPT, ideal product power is switching to a new track.
To some extent, behind the three different stages, it is more like the measures made by Ideal to cope with the "three-year (2023-2025) knockout competition of smart electric vehicles".
Of course, in order to seize a fuller market, Ideal will also launch its first pure electric model - MEGA at the end of the year, which Ideal calls "super flagship", with luxury comfort exceeding any L series and is expected to cost more than 500,000 yuan.
"MGEA is available in seven-seat and eight-seat versions, and eight seats offer all the features and comfort of a VIP seat. The eight-seater version was specially developed for myself," Li Xiang also revealed some details about the MGEA model on social platforms.
From last year to this year, new energy has gradually become a siege of traditional fuel vehicles, and almost all categories have new energy models in sales in the top three or even won the first place, but high-end MPVs seem to be a little "out of place" - high unit price, stable brand pattern, and perennial Alpha "dominated" by the god Alpha, is a difficult bone.
Since it is a "small and difficult" market segment, why did Ideal bet on the MPV track with its first pure electric product?
From the perspective of market and brand positioning, some clues may be found.
First, high-end MPVs are a fierce battleground because their strategic position is critical.
No matter which brand, as long as it wins the segment of high-end MPV, it is equivalent to winning the brand advantage of "high product price and high customer net worth", forming a reputation with a high gold content, and then forming a dive on other categories of the same brand, and winning the high ground in the passenger car market in one fell swoop, Toyota Alpha is one of the typical examples.
Secondly, whether it is a large sedan or a full-size SUV, the play of cockpit space is always difficult to show its strength, and the biggest use scenario of MPV is the family, coupled with the ideal insistence on building cars for family users, it is natural to need an MPV model to maximize the label of "smart family car".
Ideal MEGA spy photos (source network)
Borrowing technological means and caring for family users, maximizing the deterrent power of the ideal MGEA to achieve the subversion of the traditional luxury market, which may be the strength of the ideal to make MGEA the first in sales of more than 500,000 passenger cars.
From such a state, it is not difficult to see that the ideal is trying to integrate artificial intelligence with the traditional automobile manufacturing industry, rather than simply using the ruler of traditional car companies to measure the ideal.
At the spring media conference in March this year, Li Xiang announced the ideal target plan:
- 35% of the passenger car market of more than 200,000 yuan in 2027;
- Become the world's leading artificial intelligence company by 2030, competing with Tesla and Apple.
If the L series models have won market recognition and are the credit of "home positioning", then the upcoming MGEA will more reflect the power of technology, the car is only a shell, and artificial intelligence is the greatest competitiveness of the ideal future.
As far as the current state of the industry is concerned, the first to introduce "big models" in the fields of on-board space and assisted driving is not ideal, but at the industrialization level, ideal is one of the best car companies.
Excellent industrialization capabilities have laid the foundation for the development of artificial intelligence, and also created the possibility of ideal use of science and technology to innovate the automotive industry.
Ideal MGEA can win the championship of more than 500,000 sales, just the appetizer of this industry competition, I believe that it will not be long before "artificial intelligence" will become the core competitiveness of the automotive industry.