GPT-BI is launched at China FAW, and large model technology is integrated into digital and intelligent transformation

Yun Zhongfa comes from Ao Fei Si
Qubit | Public account QbitAI

On January 22, the large-scale model application GPT-BI created by China FAW and Alibaba Cloud Tongyi Qianwen was first launched, adding new vitality to China FAW's digital transformation and upgrading. The application can receive natural language queries and automatically generate analysis charts based on corporate data, currently reaching an accuracy of nearly 90%. What's more worth mentioning is that, compared with the "fixed question and answer" of traditional BI (Business Intelligence) , it can achieve any combination of questions and answers, and data can be penetrated at any time, so that "questions and answers are insights."

Since China FAW started its digital transformation and upgrading in 2022, it has built a modern enterprise management system based on cloud-native business unit twins and created a digital transformation path for traditional enterprises with FAW characteristics.

China FAW has top-level planning, adheres to a blueprint to draw it to the end, establishes the "1164 overall strategy" for digital transformation, establishes nine business IT integration teams around 6 vertical and 3 horizontal business lines, and builds the FAW Digital Intelligent Operation System (DIOS) to achieve " "Double 100" goal, 100% business digital twins, 100% improvement in operational efficiency, and continuous construction of "four systems": transformation technology system, transformation method system, IT product operation system, and digital intelligence capability system.

According to reports, China FAW has achieved a series of phased results through transformation practices in recent years.

• Apply architectural thinking to deconstruct the business to form a new model of management innovation;

• Realizing information exchange and business collaboration have become new ways of creating value;

• Deepen data application and digital twins drive new changes in business models;

• Design intelligent models and accumulate knowledge to promote new improvements in business efficiency;

• Continuous agile iteration and user participation inspire new performance of IT products;

• Realize workbench operations and start a new track for new business operations.

China FAW's business unit twins have accumulated massive amounts of data, and the emergence of large model technology has injected new momentum into improving corporate efficiency.

China FAW is actively exploring the "GPT+" large model innovation paradigm. GPT-BI is not only China FAW's first large model implementation case, but also the first large model BI application in the automotive industry.

BI is one of the core systems of an enterprise and the key to digital decision-making and data governance.

Behind BI is a complex data governance process. China FAW created a "five-stage and sixteen-step" method for indicator data governance to ensure data accuracy, decompose indicators into indicator objects, dimensions and measures, and realize digital twins of indicators.

Relying on Alibaba Cloud, based on the initial corpus of 468 managed indicators, 60,000 pieces of evaluation data were formed to build the five capabilities of indicator design, indicator disassembly, data sourcing, data modeling and data analysis of the FAW data large model. Through continuous After fine-tuning the badcase review, the overall accuracy of the model increased from the initial 3.2% to 90%, which has exceeded the average level of manual management. The FAW Data Big Model transforms data governance in core areas driven by demand into data governance in all areas covering the enterprise.

GPT-BI not only greatly shortens the delivery cycle of report design and data modeling for BI analysis, but also can exhaust all indicators, models and reports in the limited domain of the enterprise. After the user inputs a question, the large model identifies the intention of the question, analyzes the decision variables, and generates SQL data query statements match enterprise real-time data, automatically generate optimal decision-making solutions, meet users' more flexible and intelligent data needs, realize "question and answer is insight", and bring about a decision-making revolution based on dynamic factors and real-time data.

For example, when asking "Why is the production of a certain model less than expected?", the future large model can first compare the expected production with the actual production, and after obtaining the difference, it will not only analyze the explicit variables (for example: production due to equipment abnormal production shutdown for 20 minutes, abnormal quality of certain model parts, etc.), but analyze all variables involved (such as fluctuations in raw material supply, energy consumption and supply stability).

By investigating the data, we can finally find out the most relevant reasons and generate visual reports.

Men Xin, Vice President of China FAW Hongqi Brand Operations Committee, believes:

The large model is the cornerstone of future production relations. We need to redo all FAW Group's businesses using the GPT large model. Business unit twins have accumulated massive amounts of data. FAW Group’s large-scale model innovation path is AI+ business unit, which will bring longer-term and deeper change capabilities and allow enterprises to take advantage of AI. In the future, we will also export FAW Group’s digital transformation methodology and tool platform to provide SaaS services to the industry.

Based on the exploration of real business scenarios such as GPT-BI, China FAW will also use a large number of high-quality data assets to create a one-stop large model application development platform for R&D, manufacturing, after-sales services and other fields on the "Alibaba Cloud Bailian" Vertical large models allow large models to be effective in multiple links such as production and sales.

At the same time, as the world's leading cloud computing and artificial intelligence service provider, Alibaba Cloud is promoting the application of Tongyi Qianwen's large model in different industries such as aviation and automobiles, and jointly building an industrial ecosystem.

Li Qiang, vice president of Alibaba Cloud Intelligence and general manager of the automotive energy industry, said:

Alibaba Cloud aims to be the most open cloud in the AI ​​era, and has already provided computing power support to 80% of technology companies and half of large model companies in the country. In the automotive field, Alibaba Cloud will continue to promote the implementation of AI in specific industry application scenarios based on the Tongyi large model to better serve customers' business innovation. "

Click here ???? Follow me and remember to star~