Author | Myths
Edit | Zhihao
From early adopters to commonly used, how can large models better support smart cars?
Previously, a number of industry executives in the industry said that generative AI is the new productive force for the next round of economic development and enterprise competition.
From a global perspective, in the next 10-20 years or even longer, AI will be one of the resonant technology industry trends at home and abroad. The large model is one of the core tools for AI to set off technological change.
Especially in the current automotive industry, large models are accelerating their integration at an unprecedented speed and becoming an important engine for industrial upgrading.
A front-line executive pointed out that from autonomous driving, intelligent cockpit, automobile marketing and sales to enterprise digitalization, the in-depth application of large models not only improves the convenience and intelligence of intelligent driving, but also redefines the relationship between people and vehicles, opening a new era of automotive intelligence.
The slogan of large models on the car is getting louder and louder, but whether it is in intelligent driving, intelligent cockpit and other intelligent configurations, or in the entire ecological chain of the automotive industry, the large model has not yet achieved the most extensive landing, and there is a gap between the effect of getting on the car and the ideal state.
How to give full play to the true value of large models is a challenge for the entire automotive industry.
The ability of large models to be put on the car needs to be further developed, and the automotive industry needs large model products with better performance and easier to land, so as to promote the large model from the early stage to the landing period.
At present, the model products represented by the bean bag model are rapidly opening up the market, and in May this year, the bean bag model independently developed by ByteDance officially provided services to the outside world through the volcano engine.
As of September this year, the average daily usage of Tokens of the Doubao model has exceeded 1.3 trillion, and the overall growth of Tokens has exceeded 10 times in four months.
The huge usage proves that the bean bag model has successfully passed the test of easy use and ease of use in the market.
Not only that, Volcano Engine is also continuing to open up its large-scale model capabilities and co-create industry solutions with car companies and AI technology enthusiasts.
During the Volcano Engine AI Creator Contest hosted by Volcano Engine, Lynk & Co Auto and Intel, more than 1,000 AI developers and enthusiasts used the Bean Bao model and the professional version of the button to create 631 intelligent twins applications in the automotive industry with different forms and functions.
These award-winning bots have developed intelligent twin solutions with practical application value for the real business scenarios of Lynk & Co cars, bringing more possibilities for the implementation of automotive intelligence.
How can large models better drive the upgrading of the automobile industry? Zhidong found the answer in the exchange with Yang Liwei, general manager of Volcano Engine Automobile business, Wen Yuzhen, director of brand public relations department of Lynk & Co Automobile, and other senior executives.
First, the large-scale model race upgrade market demand torture intensified
At present, many car companies have taken the large model on the car as an important selling point, but the speed and effect of the large model on the car still need to be improved.
For example, the intelligent cockpit is one of the easiest footholds for most car companies to access the large model, but at present, the main functional points of the large model are concentrated in voice interaction, entertainment functions, knowledge base and other modules, and there is still more room for improvement in emotional perception, intelligent assistant, emotional dialogue, multi-modal interaction, etc.
In addition, the single interactive content, weak semantic comprehension ability, and the contempt for personalization in entertainment recommendations also reduce the user's experience.
In this regard, the current consensus in the industry is that in the future, large models will reshape the cockpit, and the intelligent cockpit will develop in the direction of agents, which can understand scene information, naturally understand user needs, generate scenario-based solutions and complete implementation.
Yang Liwei, general manager of Volcano Engine Automotive Industry, further pointed out that the ideal state of the future large-scale model power cockpit is to meet the personalized needs of users at a lower cost and faster speed, so that users can enjoy the car more "lazy".
But now most large-scale products are still in their infancy in the implementation of these cockpit functions.
In addition to the cockpit putting forward more requirements for large models, there are also more extensive and profound needs for large models in manufacturing, R&D and design, marketing and after-sales services.
For example, in terms of distribution, with the rapid changes in the industry environment, terminal sales personnel may have problems such as insufficient product knowledge, lack of customer service skills, outdated sales strategies and insufficient use of technical tools, which not only affect performance, but also affect brand competitiveness.
In this regard, Lynk & Co cooperated with Volcano Engine to fully integrate the newly launched SalesCopilot of the bean bag model into Lynk & Co's direct sales system, thus creating a more intelligent sales AI assistant and improving the business capabilities and service levels of sales consultants.
With the help of the Doubao model, Lynk & Co has also realized a number of model-based capabilities such as smart invitation seats, AI training & internal training, smart business reports and AI car manuals, which can realize services such as real-time data analysis, customer behavior prediction, customer training and evaluation summary, and personalized sales strategy suggestions, helping sales consultants improve their ability to respond to problems and comprehensively improve their sales capabilities.
The exploration of Lynk & Co cars and Volcano engines has provided a reference for the industry, but it will take time for more players in the car circle to integrate large models into the whole chain of automobile production and sales.
In this regard, Yang Liwei said: "The large model is not like electricity, compared with electricity, the large model still lacks a fulcrum, and it has become a productive force that really causes social change, and I think we need to wait for this moment." ”
Although large-scale model products continue to emerge, the evolution of large-scale models into productivity will also go through the process of slow penetration, which may not immediately set off a revolution in the intelligent automotive industry, but will continue to become a propeller.
This requires users to "try it first", and on this basis, the large model can be used on a large scale through data precipitation, and the model performance can be continuously iteratively upgraded, so that users feel that it is easy to use, love to use, and then commonly used.
Second, grasp the actual demand for large models to build a moat for car companies
In order to allow more users to use the large model and continue to improve the effect of getting on the car, how should the large model enterprises lay out the "landing"?
Grasping the automotive demand scenario and meeting the diverse needs of users is the key direction for large-scale model enterprises to break through.
A few days ago, the "Go! Look at the stars! Look at the moon! "The agent won the first prize in the AI cockpit track of the Volcano Engine AI Creator Competition, and its creative team used the bean bag model and the professional version of the button to meet the user's interest in getting close to astronomy.
At present, in order to alleviate the increasingly crowded inner anxiety and depression that have nowhere to release, more and more users have begun to pay attention to the starry sky and want to experience the fun of stargazing, which has gradually evolved from a niche hobby to a popular form of tourism.
However, urban life has led to fewer opportunities for users to watch the stars and the moon, and most users also lack relevant knowledge and skills, which has led to a single experience and a lack of immersion in stargazing and moon viewing.
Lynk & Co has realized the panoramic large sunroof on the car, and is committed to developing more practical scenarios based on the characteristics of existing models.
In order to improve the stargazing and moon-watching experience of Lynk & Co users, with the support of the large model of the bean bag and the professional version of the button, the contestants combined the data of authoritative institutions such as Shanghai Planetarium, NASA and SkySafari to develop intelligent twins.
The agent can not only provide users with AR pointing commentary, play astronomical music, satisfy users' curiosity, but also recommend stargazing locations for them, plan routes, and help them "chase the stars" journey, and further meet the needs of users to record and share, assist them in taking photos of the starry sky, and record and share stargazing experiences.
Wen Yuzhen, Director of Brand Public Relations of Lynk & Co Auto, emphasized that Lynk & Co has been actively exploring industry trends, and Lynk & Co's users are also a group of customers who embrace technology and are willing to try new things.
In addition, a number of intelligent twin solutions developed according to the actual needs of users have also emerged.
On the one hand, it proves the accuracy and high-speed use experience of the bean bag model and the flexibility and ease of use of the professional version of the button.
On the other hand, it also shows that the volcano engine can create an intelligent twin solution that meets the personalized needs of users, improve the efficiency and experience of users in car scenarios, and create an intelligent moat for enterprises.
Third, the large model should be easy to use and easy to use, not only economical, but also landing
In the future, large models will continue to change the ecology of the automotive industry.
In the car model race, the volcano engine provides a powerful bean bag model series for the car rim, as well as complete application development tools such as the buckle professional version, and the bean bag model has also become a fertile soil for the growth of AI native applications.
How does the bean bag model do this?
Yang Liwei said that the volcano engine requires a large model, not only economical, but also more practical.
It is not only necessary to optimize the cost-effectiveness of the bean bag model, but also to make it easier for more people to use and make it easier for users to use it.
Since its release in May this year, the bean bag model has set off a price war for automobile models with a pricing strategy lower than the average price of the industry, and has moved towards the goal of better models and easier implementation.
With the practice verification of ByteDance's internal 50+ business scenarios and the continuous polishing of the daily trillion-level tokens usage, the model capability of the Doubao large model is stronger, and the use effect continues to evolve iteratively.
According to the public and internal evaluation collections, the comprehensive ability of the latest version of the Doubao large language model has increased by 20.3% compared with the initial release.
In addition, the model has also been enhanced to varying degrees in terms of long essay tasks, mathematics, professional knowledge, and code ability.
In order to further lower the threshold for the implementation of the bean bag large model and adapt to the needs of more business scenarios, Volcano Engine has newly upgraded the plug-in ecosystem and AI native application development services.
Through related actions, Volcano Engine further improves the all-round functions and services such as fine-tuning, reasoning, and evaluation of the Volcano Ark model, helping enterprises customize large models to meet their diverse needs.
In addition to these, Volcano Engine has also made more attempts with 30+ manufacturers in the automobile model ecological alliance.
Volcano Engine and these manufacturers have worked closely with these manufacturers on the definition of the new generation of AI in the automotive industry, the release of industry evaluation standards, and cross-terminal industry exchanges and co-creation, so as to reduce the cost of large model development and user use, and continue to promote the commercialization of the bean bag model for the AI industry.
Based on the implementation in multiple scenarios of thousands of industries, the bean bag model has accumulated a huge amount of use, which has further fed back to itself, forming a flywheel effect, and promoting its performance enhancement, price reduction, and faster landing.
Conclusion: The wave of AI for all is coming
Large models not only accelerate the automotive industry's move towards intelligence, but are also expected to reshape the entire industry and become a key supporting technology for new quality productivity.
To achieve these visions, solving real problems for users is key.
In this regard, while accelerating the help of thousands of industries, the bean bag model also continues to feed itself to promote technology iteration and performance improvement, and expand the new realm of large model application.
With the continuous explosion of large models, the wave of AI for all is also coming. With the help of large-scale model products such as the bean bag model, more powerful and innovative agent applications will continue to appear.