The new competition has begun with AI large model

Since ChatGPT has been widely known, major technology companies have published their own model that is similar to ChatGPT. Among them, AI’s large model has become a focal point in the area of intelligent vehicles and industry which is full of competition.

Recently, Apollo announced that their explore partners of API intelligent cabin are GWM and ECAR X. According to the official announcement, the collaboration will fully utilize the ERNIE large model to explore practical applications in auto interaction scenarios, validating multiple innovative features on mass-produced vehicle platforms. In the future, it is expected to be initially implemented on Great Wall, Lynk & Co., and Smart.

Top companies start to develop large model

After the electrification, the vehicle field has changed to AI large model slowly.

Besides the official announcement of Baidu, Li Auto also announced its large model named MindGPT. SenseTime and HAOMO published their large model as well. Other companies choose to cooperate with others. Xpeng and Powered by AliOS choose to cooperate with Alibaba’s large model.

At present, the current mass production of intelligent driving technology primarily utilizes a modular architecture. In simpler terms, the earlier concept of intelligent driving used a “stacking” method with multiple smaller models. The recognition principle involves an initial observation followed by a comparison to existing knowledge. However, accurate identification can be challenging if this information has not been previously acquired.

The self-learning algorithm capability of large models is crucial in auto-driving, especially at perception and decision-making levels. It enhances reasoning ability when dealing with complex road scenarios and human-like cognitive processes.

The CEO of Li Auto, Li Xiang, presents the importance of AI advancement by stating that intelligent electric vehicles are currently undergoing an AI revolution. During the software 2.0 phase characterized by cognitive learning abilities, cars start perceiving and interpreting their surroundings similar to human observation in order to make well-informed judgments.

In the meantime, it is predicted that intelligent driving will ascend as the popular scenery of application for AI. Based on iiMedia Research, China's self-driving car industry is expected to experience a remarkable surge from 3.05 billion RMB in 2015 to an approximate value of 26.76 billion RMB by 2025.

First begin with intelligent cabin

During the time of technology upgrades, a large model has entered the implementation phase of the industry. It is widely recognized that promoting their industrialization and usage will play a key role in driving further progress in this field. In the auto industry, both new energy vehicles and larger models have technological attributes. The primary advantage of integrating them is learning from each other, which is strengthening manufacturing capabilities and consumer appeal for larger models, while simultaneously enhancing electronic functionalities and technology property.

In addition to intelligent driving, the development of AI virtual assistants is equally crucial. The integration of AI large models significantly enhances the intelligence of intelligent cabins. With the support of a large model, future in-car voice assistants are expected to transform into companions capable of engaging in casual conversations with users.

According to incomplete statistics, several car companies such as Changan, JIDU, Geely, Voyah, Hongqi, GWM, NISSAN and Leapmotor announced to use ERNIE Bot. Changan Yida has become one of the mass-production vehicles that carry ERNIE Bot through OTA type in a new car. Huawei published its newest Harmony intelligent cabin in HDC.Together which is built on the HarmonyOS 4.0 architecture and incorporates the recently released Pangu 3.0 model. In SparkDesk large model, it published “1+N” architecture and related products. Officially, after SparkDesk large model integrates with intelligent cabin, it can achieve free communication across different industries and scenarios.

The leading Chinese domestic companies keep a highly consistent pace and unwavering focus on the application of large models in the auto industry. Furthermore, when dealing with these models, they prioritize vertical scenarios, effectively accentuating 'practicality' through the swift implementation of specific products and realization of their commercial value. The integration of AI voice interaction with these models can be like a 'race', enabling users to experience the product at an earlier stage.

About the importance of large models, one opinion that has been widely approved is that a large model is an operating system in the age of AI. What seems clear is that it’s more closely between large models and the combination of auto companies and vehicles. In the future, every driver will have an assistant that good at other fields; while the technology continues to iterate, the assistant can be a primary driver and even be a team that can cope with various scenery.

Translator：Wei Xiong                                          

Reviser：Yan Luo