Large Language Models in the Automotive Industry: Overview, Impact, and Practical Exploration

The automotive industry is rapidly adopting large language models (LLMs) like the GPT series to accelerate digital transformation, improve in‑vehicle intelligent system interactions, optimize customer service, and enhance product development and marketing strategies.

LLM Overview – After the 2012 breakthrough with AlexNet, deep learning dominated image tasks, but natural language processing lagged until the 2017 introduction of the Transformer architecture, which enabled parallel processing and self‑supervised learning, leading to encoder‑only (e.g., BERT), decoder‑only (e.g., GPT), and encoder‑decoder models.

Impact on Automotive – LLMs provide superior language understanding, generation, planning, and evaluation capabilities that can be applied across the automotive supply chain, including voice assistants, smart cabins, customer support, sales and marketing analytics, vehicle design assistance, internal knowledge services, and autonomous driving scenario simulation.

Practical Exploration – Experiments covered FAQ bots, long‑text Q&A, text classification, report summarization, AI agents for natural‑language database queries, and RAG‑augmented generation. Results showed that combining LLM‑generated similar questions with semantic similarity boosted FAQ answer accuracy to 94%.

Data Analyst Requirements – With LLMs, analysts need stronger prompt‑engineering skills, the ability to define and decompose problems, code quality assessment, testing and debugging of generated code, model evaluation, resource budgeting, and judgment on when LLMs are appropriate.

Q&A Highlights – Discussed controllability via RAG, limited recommendation system use in automotive after‑sales, the effect of fine‑tuning on 70B models, SQL generation for database queries, and hardware requirements (e.g., INT4‑quantized 70B model needing ~43 GB VRAM, supporting ~4 concurrent requests on dual A100 40 GB GPUs).