Jim Fan· @DrJimFan · X·· 2025-08-05精选
AI 导读
物理AI评估无法靠实车碰撞测试完成，传统游戏引擎（sim 1.0）也难以覆盖所有边缘情况。基于神经网络的sim 2.0由数据驱动，随车队规模扩展。Tesla已应用多年，用于生成近正面碰撞等罕见危险场景的训练数据，补充800万辆实车难以采集的极端案例。
推荐理由
Jim Fan 指出物理 AI 评估难题，提出神经网络驱动的 Sim 2.0 数据飞轮方案
正文 · 原文
Evaluation is the hardest problem for physical AI systems: do you crash test cars every time you debug a new FSD build? Traditional game engine (sim 1.0) is an alternative, but it's not possible to hard-code all edge cases. A neural net-based sim 2.0 is purely programmed by data, grows more capable with data, and scales as the fleet data flywheel scales.
@DrJimFan Tesla has had this for a few years. Used for creating unusual training examples (eg near head-on collisions), where even 8 million vehicles in the field need supplemental data, especially as our cars get safer and dangerous situations become very rare.在 X 查看被引用的帖子
来源：Jim Fan · x.com