Defining the Next-Gen Smart Cockpit in the "No Man's Land"

This year, many newly released car models have placed a significant emphasis on in-car intelligence and smart cockpits. During interviews, it was understood that smart cockpits have become an important focus for automakers to create differentiated automotive experiences in a fiercely competitive market environment. Currently, the AI large model track is heating up, and large models are bringing disruptive changes to smart cars, continuously breaking through the "upper limits" of smart cockpits.

"Smart cockpits have seen explosive growth in the past five years, but problems soon emerged. How should we deal with the fact that the product series in this field has developed so quickly that the entire chain of other systems cannot keep up? In other words, the cockpit has now entered the 'no man's land', where domestic development is faster than abroad... Before the outbreak of the COVID-19 pandemic, domestic cockpit products were basically one to two generations behind foreign ones, but since last year, domestic products have been one to two generations ahead of foreign ones." Professor Cao Dongpu from the National Key Laboratory of Intelligent Green Vehicles and Transportation at Tsinghua University said that in recent years, the industry has been quite confused about how to define the next generation of smart cockpits, and leading cockpit companies have encountered bottlenecks. However, since the release of large models at the end of 2022, they have played a significant role in promoting cockpits.

Advertisement

Li Tao, General Manager of Baidu Apollo's Smart Cockpit Business, also said in an interview with reporters that the large model's entry into vehicles is not a future tense, but a present continuous tense. The overlap of the great wave of automotive intelligence and the great wave of large models is an inevitable part of history. To achieve the ultimate perceptual form of AI cars, it is necessary to use strong logical reasoning, memory, and generation capabilities, which happen to be the areas where large models excel.

Smart cockpits have become a selling point for differentiation. Early driving cockpits were mainly composed of mechanical dials and simple information entertainment systems, including car radios, music playback functions, navigation functions, etc.

With the rapid iteration of hardware and software technology levels such as automotive chips, human-computer interaction, and automotive systems, automotive cockpits have begun to fully enter the intelligent stage. Intelligent hardware continues to expand and upgrade, liquid crystal instruments begin to replace mechanical instruments, central control screens, and multi-screens gradually become standard configurations, HUD (Head-up Display, head-up display) is accelerating its promotion, and at the same time, the cockpit entertainment system is continuously enriched. Navigation, games, lifestyle, and other applications are gradually equipped on the in-car system, gradually shifting from physical buttons to complete touch control, and voice interaction technology has matured and developed.

KPMG China's "Focus on the Second Half of Electrification, Smart Cockpit White Paper" released in April this year pointed out that the development of high-level intelligent driving is lower than expected, and manufacturers are investing more R&D resources into smart cockpits. The automotive information entertainment systems and cockpit domain controllers involved in smart cockpits need to meet the automotive functional safety levels, which are lower than those involved in intelligent driving, such as the driving domain and chassis domain, and are more easily perceived by people inside the car (drivers and passengers). Therefore, they have become the main competitive point for differentiation at this stage, and each manufacturer's new models promote one or several functions of smart cockpits as selling points for market sales.

Many configurations of smart cockpits have basically achieved standardization. Data provided by the Auto Home Research Institute shows that the penetration rate of configurations such as APP remote control, large central control screens, voice recognition, and vehicle networking has reached 90%.

Domestically, smart cockpits are currently divided into five levels: L0 functional cockpits, L1 perceptual smart cockpits, L2 partial cognitive smart cockpits, L3 high-level cognitive smart cockpits, and L4 comprehensive cognitive smart cockpits. The higher the level, the stronger the active interaction ability of the smart cockpit, the richer the service content, and the wider the execution scenarios."Currently, some car cockpits are at the L2 level, while most are still at the L1 level." Cao Dongpu believes that by 2027, the form of L3 high-level cognitive intelligent cockpits will emerge, and it is expected that by 2032, the form of L4 fully cognitive intelligent cockpits will appear.

Li Tao told the reporter in an interview that, looking at the current stage, voice interaction is the most frequently used application in intelligent cockpits, with as many as 70% of new cars produced this year equipped with voice interaction features. "The frequency of people using the cockpit has increased from the earliest single-digit number of 3-5 times a day to today's double-digit number, and in some models with leading voice interaction technology, it has approached the triple-digit number."

Intelligent cockpits are an "enticing" track. According to KPMG's calculation, the scale of the intelligent cockpit market in China will reach 212.7 billion yuan in 2026, with a compound annual growth rate of over 17%. Among them, the value per car will rise from 8,100 yuan in 2022 to 9,650 yuan in 2026, and the penetration rate of intelligent cockpits will increase from 59% to 82% during the same period.

The reporter learned in the interview that the future development of intelligent cockpits will focus on the deep integration of technology, the improvement of personalized services, and the integration with the external ecosystem. With the support of multimodal interaction, AI large models, 5G, and the Internet of Vehicles, intelligent cockpits will not only provide drivers and passengers with more convenient, safe, and intelligent driving experiences but will also gradually become an important connection point for life inside and outside the car.

"The ultimate form of the cockpit will be a car-end intelligent entity, with its own understanding, cognition, learning ability, and the ability to automatically provide users with situational awareness and services. In the future, a new era cockpit that can know the user's context, know what the user wants at this moment, and automatically generate a panoramic execution plan is the ultimate solution." Li Tao said.

Can large models reshape intelligent cockpits?

While intelligent cockpits bring differentiated competitive strength to automotive products, due to various factors, the intelligent cockpits on the market have shown a trend of homogenization, and the products launched by major manufacturers are becoming more and more similar.

The reporter learned that, in terms of technology, because technologies such as voice recognition, screen interaction, navigation systems, and entertainment systems have been relatively mature after years of development, the implementation paths and solutions for these technologies are relatively fixed, leading to a convergence in technology applications among major manufacturers. The low technical threshold, concentration of the supply chain, consistent market demand, and the cost and risk of innovation have also led to the homogenization of intelligent cockpits.

The industry believes that large models can solve the homogenization problem of intelligent cockpits. In the future, intelligent cockpits will use AI large models to provide personalized human-computer interaction services for thousands of people, promoting the deep evolution of intelligent cockpit functions.

According to the "AI Automotive Development Report (2024)" recently released by the China EV100 Forum, at present, intelligent cockpits are in the transition phase from 2.0 to 3.0, and large models are the key to accelerating this transition process.Several industry insiders told reporters that large models, trained through deep learning and massive amounts of data, possess powerful data analysis and understanding capabilities. They can provide personalized services based on each user's driving habits, preferences, daily behavior patterns, and more. This customized experience allows users to truly experience the "thousand faces for a thousand people" service of smart cars, breaking the limitations of traditional cockpit functions being uniform. At the same time, large models support multimodal inputs such as voice, image, and gesture, enabling more natural and intelligent interactions with users, enhancing the experience differentiation of the cockpit. In addition, large models can not only optimize the in-car experience but also achieve the linkage of smart devices inside and outside the vehicle through cross-platform and cross-device ecosystem integration. Different manufacturers can combine their own ecosystems to build unique smart ecosystems based on large models, thereby achieving functional differentiation.

"Our application in the cockpit is called NOMI. In the past two years, NOMI seemed a bit clumsy, but it wasn't clumsy, it just wasn't growing fast enough, reflecting the bottlenecks encountered in the development of artificial intelligence technology. Without GPT, without large models, it would be difficult for NOMI to become smarter. Thanks to the development of large language models in the past two years, we have rebuilt the underlying framework of NOMI with the GPT large language model. The AI large model has greatly improved our entire cockpit experience. Now, NOMI knows everything from astronomy to geography, and there's no problem chatting about anything." Recently, Li Bin, founder, chairman, and CEO of NIO, said at the Global Intelligent Vehicle Industry Conference (GIV2024) high-level forum.

"The cockpit is the best scene for large models to land," Yang Liwei, General Manager of the automotive industry at Volcano Engine, told reporters. In the future, the greatest value of large models in the cockpit will be to help the cockpit integrate a good ecosystem and connect various terminals, which will be a more imaginative scene.

Wang Min, Vice President of Alipay and General Manager of the Digital Travel and Tourism Business Division, told reporters in an interview that with large models, a large number of services may be reshaped. In the past, people only thought of parking fees, but in the future, services that can be completed on mobile Alipay can also be implemented on the vehicle side.

The deep learning ability of large models allows cars of different brands to create unique cockpit experiences. Wu Yongqiao, President of the Intelligent Driving and Control Systems Division of Bosch China, said that in terms of the cockpit, the future direction is to create an AI cockpit. The current cockpit is only based on ChatGPT to improve natural semantic interaction capabilities. Bosch is researching with several leading manufacturers how to deploy a single-sided large model on the cockpit, making its computing power exceed 300 TOPS of autonomous driving. This will make the operation of the smart cockpit more efficient, smooth, and intelligent.

In terms of smart cockpits, large models are not only displayed on the vehicle but have already moved towards commercial use. Li Tao revealed to reporters that Ji Yue 01 has fully equipped Baidu's Wenxin Yayan large model. This car has now entered the mass production stage, and its "cockpit and driving dual excellence" intelligent positioning has gradually been recognized by the market and consumers. At the same time, the new Buick GL8 Lux PHEV equipped with Baidu's large model will soon be mass-produced. In addition, the sixth-generation unmanned car of Luobo Kuai will also be equipped with Baidu's super cockpit intelligent body, and more new technologies will be released on the vehicle one after another.