Waymo Challenges Human Driving Norms with Data-Driven Argument
TL;DR. Waymo states that human-driven miles are not a suitable benchmark for autonomous vehicle training data. - Waymo's head of safety suggests human driving includes irrelevant data for AV development. - The company's models are trained on specific scenarios relevant to autonomous operation. - This approach aims for safer and more efficient AV development compared to human metrics.
- Waymo's head of safety, Trent Victor, argues against using human driving miles as a benchmark for AV training.
- Victor believes that human driving data contains too much 'junk' irrelevant to autonomous system development.
- Waymo focuses on training its AI models with relevant and critical scenarios, emphasizing quality over raw mileage.
- The company's methodology suggests a paradigm shift in how autonomous driving safety and efficacy are measured.