AI-Driven Autonomy Vision at XPeng | Generated by AI
Question: Who is Liu Xianming (刘先明) at XPeng?
Answer:
Liu Xianming joined XPeng in early 2024 after working on machine learning and computer vision at Cruise (General Motors’ autonomous driving division). He’s now the chief driver of XPeng’s AI strategy.
The Key Insight:
Liu was struck by ChatGPT’s architecture — a simple model + massive scale + sufficient data and compute yielding emergent intelligence. He realized autonomous driving could follow the same path: skip hand-coding thousands of rules, skip mounting 5 lidars and trimming street trees, and instead let end-to-end learning emerge from data.
What He’s Built:
Starting March 2024, XPeng shipped its second-generation VLA (Vision-Language-Action model) based on this logic. The demand was so high that test drive vehicles with the new software were being rented out at 8,000 RMB/day.
His Role:
Liu previously held senior positions at Meta and Cruise and is considered a representative figure in the international AI industry. Within XPeng, he’s led the North America AI division, foundation models, and the autonomous driving center, and promoted foundation models in autonomous driving.
Why This Matters to You:
Liu’s bet is exactly aligned with your interests — scaling foundation models (not traditional software engineering) to solve autonomous driving. He’s betting that a VLA (vision + language + action) approach can generalize across driving scenarios the way LLMs generalize across language tasks. No hand-tuned rules, no lidar gymnastics. Just model scaling.
References:
- XPeng Interview with Liu Xianming (晚点Auto)
- Zhou Shuchang Joins XPeng Under Liu’s Leadership (Sina Finance)