特斯拉面臨145億美元Autopilot訴訟

🔑 增強重點摘要

• •The $14.5B figure represents a cumulative estimate of potential damages across multiple consolidated class actions, including punitive damages sought by plaintiffs alleging Tesla knowingly marketed 'Full Self-Driving' as a finished product despite internal engineering warnings.
• •The NHTSA investigation into the 3.2 million vehicles has expanded to include a specific review of Tesla's 'Vision-only' sensor suite, questioning whether the lack of LiDAR or radar hardware inherently limits the system's ability to detect low-contrast obstacles in adverse weather conditions.
• •Legal filings indicate that Tesla's defense strategy has shifted from arguing 'driver negligence' to emphasizing the 'beta' nature of the software, a pivot that plaintiffs argue contradicts the company's public marketing campaigns and CEO statements regarding imminent autonomy.

🛠️ 技術深入

• •Tesla's 'Vision-only' architecture relies on a deep neural network (HydraNet) that processes raw camera data to create a 3D vector space representation of the environment.
• •The system utilizes occupancy networks to predict the probability of space being occupied by obstacles, which has been a focal point of the NHTSA investigation regarding low-visibility detection failures.
• •The transition from legacy 'Autopilot' code to the current 'FSD' stack involves an end-to-end neural network approach, moving away from explicit C++ hard-coded rules for object detection and path planning.

🔮 前景展望基於引用來源的 AI 分析

⏳ 時間線