Exploring the Application of AI Large Models in the Automotive Industry

1. The Evolution of AI Large Models Artificial intelligence emerged in 1956, followed by machine learning, deep learning, and the advent of pre‑trained large models in 2021. The release of ChatGPT in 2023 sparked widespread interest and deeper exploration of these models.

2. What Is a Large Model? Large models are built on transformer architectures, contain billions of parameters, and are trained on massive datasets, enabling them to capture extensive knowledge from the internet.

3. Early‑Stage Limitations Despite strong capabilities, large models are still in early development; industry applications face issues such as high costs, difficulty in vertical domain adaptation, privacy concerns, and compliance with regulations.

High deployment and training costs, especially for private solutions.

Vertical use‑cases require extensive data preparation and iterative refinement.

Privacy and security risks when handling sensitive data.

4. Efficiency Revolution for the Industry Large models drive cost reduction and productivity gains across the automotive value chain, including marketing, sales, service, research, production, and supply. They enable AI‑generated content, intelligent assistants, and broader, deeper scenario coverage.

Stronger reasoning and inference abilities.

Higher efficiency through AI‑generated content and digital employees.

Wider scenario applicability, from smart cabins to end‑to‑end workflow automation.

5. Applications in Marketing, Sales, and Service

Marketing: AI models assist with sentiment matching, media placement keyword expansion, and AIGC‑driven content creation for campaigns such as vehicle launches.

Sales: AI sales assistants provide lead insights, dialogue strategies, training simulations, and call summarization.

Service: AI‑powered knowledge bases answer customer queries, while diagnostic AI leverages vehicle telemetry to build specialized troubleshooting knowledge.

6. Data Management Enhancements AI‑driven data retrieval (ChatBI) simplifies BI reporting, assists in metric definition, and supports metadata governance, improving data consistency and accessibility.

7. Infrastructure Strategies Building AI large‑model infrastructure can use public cloud instances, private clouds, or hybrid approaches with vector databases. Private deployments are costly; shared APIs or open‑source models offer lower‑cost alternatives.

8. Challenges and Evaluation Key difficulties include rapidly evolving technology, talent scarcity, high upfront investment, lack of mature toolchains, and evaluating model impact. Evaluation should consider accuracy, human‑likeness, relevance, safety compliance, and business value.

9. Future Outlook Model adoption will progress through three stages: human‑AI collaboration, partial automation (e.g., content generation, training), and full automation where decision‑making and execution are entirely model‑driven.

10. Q&A Highlights

Q1: Cabin model use‑cases include image generation, wallpapers, and voice assistants. Q2: Large models are still in early experimental phases for large‑scale industry deployment. Q3: Fine‑tuning ChatBI is challenging due to complex data dependencies. Q4: Models can support metric attribution by leveraging knowledge bases.

Overall, the session outlines the current state, practical applications, infrastructure considerations, and future trajectory of AI large models within the automotive sector.