Li Auto challenges Tesla with end-to-end neural network
Undoubtedly, advanced driving has become the forefront of competition for many automakers.
Last year, automakers frenziedly competed in the number of 'open cities' and AEB braking distances. This year, everyone is competing to see who can be the first to achieve a true 'End To End' smart driving model. The one who has recently surged ahead is Li Auto, which has just crossed the milestone of a million deliveries.
Lang Xianpeng, Vice President of Intelligent Driving at Li Auto, recently revealed that Li is determined to become the global leader in artificial intelligence by 2030. Regarding the takeover rate of smart driving, the plan is to gradually achieve passive intervention frequencies of less than once every 500 kilometres, 1000 kilometres, and 2500 kilometres by the end of 2025 to 2026. This also means that Li Auto's smart driving can achieve 10 times better driving safety than humans.
Since the beginning of this year, this emerging force known for 'refrigerators, colour TVs, and large sofas' has been frantically catching up in smart driving, aiming to enter the first echelon of smart driving and compete on par with Tesla, Huawei, and others.
At a strategic meeting last autumn, Li Xiang, the director of Li Auto, announced that 'by 2024, we aim to be the absolute leader in smart driving.' At the same time, Li Auto is aggressively recruiting smart driving talents within the industry, especially those who like to move from Huawei's Car BU, and the annual salary package can double directly.
Also, due to the rapid iteration, from the NPN solution to the image-less solution, Li Auto only took 4 months to complete the switch. Li Auto's "end-to-end+VLM" solution completed three generations of iterations in about a little over a year.
After intense catch-up lessons, Li auto released a brand-new autonomous driving technology architecture based on end-to-end model, VLM visual language model, and world model in Jul, claiming to be the first domestic automaker to implement the one model solution; subsequently, during the Chengdu Auto Show at the end of Aug, Li Auto announced the recruitment of a large number of participants to experience the new generation of ideal intelligent driving based on end-to-end and VLM visual language model.
While developing the autonomous driving software, Li Auto inc also perceived the trend of integration of software and hardware in the intelligent driving industry. Huawei has strong hardware capabilities, Nio announced its own autonomous driving chip last year, and in Aug this year, Xpeng also announced the successful mass production of its autonomous driving chip.
In this regard, Li Auto inc is also stepping up its pace. Recently, Li Aut plans to establish a chip R&D office in Hong Kong and is recruiting strategic talents related to chips and training clusters in Hong Kong, where some of these chips will also serve the intelligent driving scenarios.
However, for users, the technical iterations can only be perceived through specific functions. Therefore, with the addition of the end-to-end+Visual Language Model (VLM), the driving experience of ideal intelligent driving is further enhanced.
Firstly, the end-to-end brings about functional enhancements.
For example, in underground parking scenarios, the system can navigate directly to the destination parking space without the need for additional navigation settings; if the car is parked on the roadside, the system can also start directly on the side of the road with a double-click on the lever, without the need to start within the lane.
The iteration of the Visual Language Model makes the capabilities of this system more comprehensive.
The system can recognize speed bumps on the road, provide advance voice + in-car EID reminders, and decelerate to pass through; it can also recognize tidal lanes, as well as permissible and forbidden status signs, and choose the correct lane to pass through.
In reality, there are countless driving scenarios, and at this time, the end-to-end representation of generalization ability demonstrates powerful vitality.
Nowadays, among the factors influencing consumer car purchases, the importance of intelligent driving performance has also become a key consideration. In terms of ideal scenarios, Wall Street News has learned that after the NOA feature push, test drives doubled in two months. Among models priced above RMB 300k, AD Max (advanced intelligent driving edition) accounts for 70%, with L9 AD Max even exceeding 90%.
This means that intelligent driving is expected to become the next competitive advantage for Li Auto, especially as the high-end market shows a stronger willingness to pay for intelligent driving, which can also drive the optimization of its sales structure.
In the first few years when the new forces began deliveries, Li Auto was not at the top of the sales charts. However, later on, with the delivery of the L series, Li Auto quickly crossed one integer milestone after another, eventually becoming one of the new forces to achieve a million deliveries. In the competition of intelligent driving, the story of a comeback is once again written, with Li Auto starting late but handing in the paper early.
However, the automotive industry is long and complex, and on the road to becoming a giant, Li Auto still needs to face many challenges.
Next year, Li Auto will also join the pure electric camp. Having suffered losses with the pure electric MPV MEGA before, Li Auto's ability to handle pure electric SUVs will depend on the market's level of acceptance at that time.
Disclaimer: Community is offered by Moomoo Technologies Inc. and is for educational purposes only.
Read more
Comment
Sign in to post a comment
Moo Contributor
Follow me to get useful analysis, learn TA and make profitable trades.
Maximise profit 利润最大化
8030Followers
48Following
27KVisitors
Follow
10baggerbamm : I have many friends that own Teslas and they routinely drive with no human intervention at all. right Tesla's have billions of Miles accumulated and Blackwell is going to catapult their technology light years ahead and that's what no one is talking about because Blackwell incorporates video. all of the experiences that are caught on cameras every time you drive down the road are part of the learning curve for a Tesla to achieve FSD. I read as much as I can difference opinions people that are experts in the field regular Joe investors that have experience behind the wheel of different vehicles and if lidar is superior right now to Tesla it's going to be leapfrogged this up and coming year exponentially. it's going to go parabolic Tesla's use of Blackwell because of the video because of their Superior AI their data collection is going to take everything that happens and in millions of a second determine corrective course of action far quicker than a human. so if lidar right now has a slight edge it's going to become abundantly clear that within a year of today it's way behind the curve of technology evolution.